1 2012-05-31 Whamcloud, Inc.
4 2.6.18-308.4.1.el5 (RHEL 5)
5 2.6.32-220.17.1.el6 (RHEL 6) - patchless client only
6 * Recommended e2fsprogs version: 1.41.90.wc4
8 See http://wiki.whamcloud.com/display/PUB/Changelog+1.8 for more information.
10 -------------------------------------------------------------------------------
12 2011-10-31 Whamcloud, Inc.
14 * Support for kernels:
15 2.6.18-274.3.1.el5 (RHEL 5)
16 2.6.32-131.12.1.el6 (RHEL 6) - patchless client only
17 * Recommended e2fsprogs version: 1.41.90.wc3
19 See http://wiki.whamcloud.com/display/PUB/Changelog+1.8 for more information.
21 -------------------------------------------------------------------------------
23 2011-06-30 Whamcloud, Inc.
25 * Support for kernels:
26 2.6.18-238.12.1.el5 (RHEL 5)
27 2.6.32-131.2.1.el6 (RHEL 6) - patchless client only
28 * Recommended e2fsprogs version: 1.41.90.wc3
29 * Use ext4-based ldiskfs as default for RHEL5
30 * Add support for 24TB LUN (RHEL5 server only)
32 Severity : enhancement
34 Description: Add support for RHEL5.6 (2.6.18-238.12.1.el5)
36 Severity : enhancement
37 Jira : LU-62, LU-73, LU-402, LU-130
38 Description: Add RHEL6 client support
42 Description: Fix client crash introduced in 1.8.5
43 Details : The patch from bugzilla ticket 18213 landed for 1.8.5 can cause
44 a crash in mdc_exit_request() when the process is interrupted.
48 Description: ls -l reports wrong file size
49 Details : Always update LVB from disk when a glimpse callback returns an
52 Severity : major (2.6.32-based kernels only)
54 Description: Add workaround for a race causing an assertion failure in
55 clear_inode() on lustre clients.
59 Description: Fix interoperability issue between 1.8 clients and 2.x servers
60 Details : The ldlm hash handling in Lustre 2.x is incompatible with the one
61 used in Lustre 1.8. Add support for 64-bit dir name hash to the
66 Description: slow IO with read-intense application
67 Details : Align the readahead extent by 1M after when it is trimed by
72 Description: Skip quotacheck on administratively disabled OSTs
76 Description: Some autogen improvements
78 Severity : enhancement
79 Jira : LU-123, LU-296, LU-230
80 Description: Add support for yaml data logging and auster script.
82 Severity : enhancement
84 Description: Allow OSTs to be created with no primary node
85 Details : Add a --servicenode parameter for mkfs.lustre to treat all
86 service nodes equally.
88 -------------------------------------------------------------------------------
90 2011-05-12 Oracle, Inc.
92 * Support for kernels:
93 2.6.16.60-0.42.8 (SLES 10),
94 2.6.27.39-0.3.1 (SLES 11),
95 2.6.18-194.3.1.el5 (RHEL 5)
96 2.6.18-194.3.1.0.1.el5 (OEL 5)
97 * Client support for unpatched kernels:
98 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
99 2.6.16 - 2.6.32 vanilla (kernel.org)
100 * Recommended e2fsprogs version: 1.41.10-sun2
101 * The async journal commit feature (bug 19128) and the cancel
102 lock before replay feature (bug 16774) are disabled by default.
106 Description: llap pages are moved to head of LRU list instead of tail
110 Description: disable statahead by default due to important races found in the
111 statahead implementation.
115 Description: release request ref through ptlrpc_server_drop_request()
116 Details : make sure the request is protected by rq_refcount while
117 being processed in ptlrpc_server_handle_req_in().
121 Description: correct check for transno value in filter_finish_transno()
125 Description: Set body->eadatasize in mdc_getattr_pack()
129 Description: Handle unsent requests with rq_net_err in ptlrpc_check_set()
133 Description: apps stuck in ptlrpc_check_set() during direct I/O
134 Details : start async bulk unregistering at the same time at reply unlink
138 Description: lfs setstripe --pool broken in 1.8.4
139 Details : fix llapi_search_fsname() to handle relative pathnames
143 Description: Only force the mode change if we're changing the size as well
144 Details : Fix regression introduced in 1.8.0. The offending code was added
145 by commit 77ba4b2141d04180211efa8a75c11ab0abf7fafb to remove
146 setgid/setuid bits when do_truncate() is called on the file.
147 We should only force the change when that occurs, similarly to
148 ll_setattr() in lustre/llite/llite_lib.c
150 Severity : enhancement
152 Description: add procfs tunable to enable/disable lockless direct I/O
153 Details : llite.lustre-*.lockless_direct_io=0 will disable default semantics
154 of direct I/O that forces it to be lockless. lockless_direct_io
155 value, however, will be ignored if per-file LL_FILE_LOCKED_DIRECTIO
160 Description: Lustre servers crashing with NULL pointer errors
161 Details : Make sure the request is protected by rq_refcount while being
162 processed in ptlrpc_server_handle_req_in().
166 Description: kernel BUG at drivers/scsi/sd.c
167 Details : Remove sd iostats patch from sles11 patch series.
171 Description: Modified value of at_min is not taken into account
175 Description: lustre grants flock exclusive locks to two fd's in the same process
179 Description: LBUG in the mfs_verify_child()
183 Description: ASSERTION(atomic_read(&client_stat->nid_exp_ref_count) == 0)
184 Details : Reconnecting from a different nid causes old per-nid stats not to be
189 Description: create proper macro check for bdi interface.
191 -------------------------------------------------------------------------------
193 2010-10-29 Oracle, Inc.
195 * Support for kernels:
196 2.6.16.60-0.69.1 (SLES 10),
197 2.6.32.19-0.2.1 (SLES11),
198 2.6.18-194.17.1.el5 (RHEL 5)
199 2.6.18-194.17.1.0.1.el5 (OEL 5)
200 * Client support for unpatched kernels:
201 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
202 2.6.16 - 2.6.32 vanilla (kernel.org)
203 * Recommended e2fsprogs version: 1.41.10-sun2
204 * The async journal commit feature (bug 19128) and the cancel
205 lock before replay feature (bug 16774) are disabled by default.
209 Description: atime is not properly updated on an MDS
211 Severity : enhancement
213 Description: Update to RHEL5.5 kernel 2.6.19-194.17.1.el5.
214 Update to OEL5.5 kernel 2.6.19-194.17.1.0.1.el5.
216 Severity : enhancement
218 Description: Update to SLES10 SP3 kernel 2.6.16.60-0.69.1.
221 Frequency : only with SLES10
223 Description: Use OFED "KMP" provided by Novell
224 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
225 Lustre is now built against this package. That means you need to
226 install the ofed-kmp package from Novell for the patchless client
227 and from our download site for the server. Note that the ofed-kmp
228 that Novell ships may not exactly match the kernel version but
229 should still be compatible.
231 Severity : enhancement
233 Description: Update SLES11 SP1 kernel to 2.6.32.19-0.2.1.
237 Description: Enabling quotas fails with non-consecutive OST numbering.
241 Description: Fix kernel warning due to lookup_on_len() called without i_mutex
246 Description: Account direct i/o inflight rpcs separately from non-direct i/o so
247 that direct i/o, which is limited by max_rpcs_in_flight, should not
248 block non-direct i/o, which is not limited by max_rpcs_in_flight.
252 Description: Fix per-NID reporting on outstanding writes
256 Description: Reduce stack pressure by uninlining some mds and ptlrpc functions.
260 Description: Remove LASSERT in lprocfs_rd_conn_uuid() since conn == NULL is a
265 Description: fix obdo leak issue in ll_setattr_raw()
269 Description: limit MMP interval
271 Severity : enhancement
273 Description: add several lfs ost enhancements
277 Description: Too many default ACLs break directory access on new directories
281 Description: Lustre inode size is not coherent across nodes.
282 Details : Update lvbo from disk when AST fails with EINVAL. Lvbo is updated
283 on EINVAL error in ldlm_handle_ast_error(). The updates in
284 filter_intent_policy() and ldlm_cb_interpret() have been removed as
289 Description: Oops at __percpu_counter_add+0x1b
290 Details : Use bdi_init()/bdi_destroy() to proper initialize backing_dev_info
295 Description: add mount option to generate 32bit ino, this can be used for 32bit
296 application compatibility.
300 Description: keep reference count for "lli_sai" to prevent it to be released
301 when "statahead_enter()"
305 Description: allow quotacheck over OSTs with sparse indices
309 Description: Objects not getting deleted for files which have been removed
310 Details : ll_have_md_lock() should differentiate between CR and CW OPEN
315 Description: pin object's inode in memory to avoid certain timeouts
319 Description: fix LBUG when obdfilter-survey is interrupted.
321 -------------------------------------------------------------------------------
323 2010-07-31 Oracle, Inc.
325 * Support for kernels:
326 2.6.16.60-0.42.8 (SLES 10),
327 2.6.27.39-0.3.1 (SLES11),
328 2.6.18-194.3.1.el5 (RHEL 5)
329 2.6.18-194.3.1.0.1.el5 (OEL 5)
330 * Client support for unpatched kernels:
331 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
332 2.6.16 - 2.6.30 vanilla (kernel.org)
333 * Recommended e2fsprogs version: 1.41.10-sun2
334 * The async journal commit feature (bug 19128) and the cancel
335 lock before replay feature (bug 16774) are disabled by default.
337 Severity : enhancement
339 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
340 2.6.18-194.8.1.0.1.el5.
342 Severity : enhancement
344 Description: using inkernel OFED stack for rhel5 & oel5.
346 Severity : enhancement
348 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
349 Details : lfs_migrate does a "poor man's" migration of files from their
350 current OST layout to a new OST layout as chosen by the MDS.
354 Description: mds_orphan_add_link()) error linking orphan to PENDING
355 Details : quota limits might disallow linking orphans to PENDING
356 when unlinking a file - temporary raise threads' privileges
357 when processing unlinks.
359 Severity : enhancement
361 Description: add conf-param -d option to remove permanent settings.
362 Details : Add the ability to remove permanent lctl conf_param settings.
363 (Previously conf_param settings could only be changed, not
364 removed.) This also provides a method to change failover
365 nid locations. Improve lctl man page.
367 Severity : enhancement
369 Description: add list_param to b1_8 and add "-R" option to list params
372 Severity : enhancement
374 Description: lfs quota output is not very convenient for awk/sed-parsing
375 Details : Some positions in lfs quota output table could be empty or
376 non-empty which made it hard to parse it with scripts, now a dash
377 is put instead of space where there is not supposed to be any data.
379 Severity : enhancement
381 Description: fix obdfilter-survey script to work properly with remote oss-s
383 Severity : enhancement
385 Description: add new OBDFILTER_SURVEY test suite
387 Severity : enhancement
389 Description: add new multiple mount protection (MMP) test suite
391 Severity : enhancement
393 Description: add support for async journal commit in echo client
395 Severity : enhancement
397 Description: allow userland programs to include <lustre/lustre_idl.h>
398 from stardard include directories
400 Severity : enhancement
402 Description: The prune-icache-use-trylock is no longer needed now that
403 the patch from bug 20008 is landed.
407 Description: The shrink grant feature is still active on the client although the
408 connect flag is not set.
412 Description: Don't leak grant space if the write failed with quota exceeded.
416 Description: Don't consume grant space twice on recoverable resent.
420 Description: a race condition could lead to SIGBUS being sent to an
421 application using mmap-ped files from Lustre
422 Details : truncate_complete_page implementation for the patchless
423 client could arbitrarily unset PG_Uptodate flag for a
424 page being kicked from the page cache, an uptodate check
425 right after a readpage call in filemap_fault could fail
426 because of that as though the page read had been unsuccessful.
430 Description: dlm lock slab shrinking is not efficient
431 Details : The dlm_locks slab can grow significantly and consumes a lot of
432 memory on the server. Set a hardlimit to grant_plan.
436 Description: Lustre does not do 1MB IOs to HW RAID
437 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
438 This is what we do already for SLES kernels.
442 Description: bump maximum number of phys/hw segments in the SLES11 kernel
443 until s/g chaining works properly.
448 Description: LSI Fusion MPT driver hacks to improve performance
449 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
451 Severity : enhancement
453 Description: increase default md stripe_cache_size to 16k
456 Bugzilla : 15587/21439
457 Description: don't handle security.capability xattr
458 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
459 This results in additional getxattr calls, causing VBR test
460 failures as well as a preformance drop when writing.
464 Description: obdfilter-survey is no longer working
465 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
466 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
470 Description: ll_shrink_cache does not handle __GFP_FS properly
474 Description: lfs getstripe shows wrong info for directories
475 Details : Set correct LOVEA default values for filesystem-wide.
479 Description: FSX checksum false positves due to mmap IO
480 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
481 checksum errors, if the flag is set on a request.
485 Description: file operations after eviction have successful return values
486 Details : use vfs ->flush callback to return any pending async errors
491 Description: mdsrate fails to write after 1.3+M files opened
492 Details : decrease memory usage on clients by recycling dentries and
497 Description: obdfilter-survey gives unreasonably high numbers
498 Details : Wait for all threads to complete when running test_brw.
502 Description: do not set lustre read_only device when server umount and keep
503 client records for recoverable ones
507 Description: move sync_on_lock_cancel tunable to the obdfilter layer
508 Details : move the tunable to trigger a journal flush on lock cancel from
509 the ost layer to the obdfilter layer. This tunable is useful
510 when using the async journal commit feature.
514 Description: exp->exp_nid_stats == NULL in filter_tally()
515 Details : fix race with per-nid stats by delaying procfs cleanup until
520 Description: extent lock cancellation on client can keep the cpu busy for too
525 Description: Do not fail OST activation when a llog is not found, just
526 issue an error message.
530 Description: Don't enable extents by default for MDT.
534 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
535 the AT code can access it concurrently while sending early
540 Description: Disable lockless truncate by default since it is sometimes flawed
541 and causes the write_disjoint test to fail.
545 Description: OSSs which don't have the patch from bug 20278 can trigger an
548 Severity : enhancement
550 Description: don't print message to the console when we have not managed to
555 Description: The MDS fails to synchronize OSTs which registered with the MGS
556 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
557 raised too early and thus discarded by the MDT stack.
558 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
563 Description: Fix race when the ping evictor and a service thread execute
564 target_recovery_check_and_stop() concurrently.
568 Description: quota broadcast can trigger a LBUG on the MDT if there are
573 Description: Resetting the lov_objid values to last_id reported by the OST
574 during orphan recovery is incorrect and can cause the same
575 objects to be allocated twice.
577 Severity : enhancement
579 Description: "weak-modules" support
580 Details : Implement "weak-modules" support which enables kernel modules
581 to be used with any kernel that implements the same kABI. In
582 order to achieve this modules are now installed in
583 /lib/modules/$(uname -r)/updates/kernel on all distributions.
585 Severity : enhancement
587 Description: add writeconf as mount option
589 Severity : enhancement
591 Description: produce debuginfo packages for SLES.
593 Severity : enhancement
595 Description: add failover nidlist to the import proc file.
597 Severity : enhancement
599 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
600 clients and 2.0 servers.
602 Severity : enhancement
604 Description: lfs find -s does not work correctly because of a bug in
609 Description: ll_read_ahead_page() must validate the dlm lock before using
614 Description: Prevent failover nids from registering with MGS first.
618 Description: fix lock inversion in ll_setattr_raw().
622 Description: object allocation is not balanced across OSTs.
623 Details : osc_precreate() should return 0, if there are enough objects left.
625 -------------------------------------------------------------------------------
627 2010-04-30 Oracle, Inc.
629 * Support for kernels:
630 2.6.16.60-0.42.8 (SLES 10),
631 2.6.27.39-0.3.1 (SLES11),
632 2.6.18-164.11.1.el5 (RHEL 5)
633 2.6.18-164.11.1.0.1.el5 (OEL 5)
634 * Client support for unpatched kernels:
635 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
636 2.6.16 - 2.6.30 vanilla (kernel.org)
637 * Recommended e2fsprogs version: 1.41.10-sun2
638 * The async journal commit feature (bug 19128) and the cancel
639 lock before replay feature (bug 16774) are disabled by default.
643 Description: open-unlinked directories trigger MDS LBUG
644 Details : Fix regression introduced by the patch from bug 19640.
645 ext3_inc_count() can reset nlink to 1 when the directory
646 is indexed and inode->i_nlink == 2. Work around the problem
647 by incrementing nlink by 2 instead of 1.
651 Description: Reconnects are not throttled
652 Details : Don't wake up pinger on reconnect failures and rely on regular
653 pings to trigger the next reconnection. Please note that the
654 pinger already uses a smaller interval if the import is
658 Frequency : only with NFS export
660 Description: Console flooded with error message from ll_inode_from_lock()
662 Details : in mds_open, initialize the child_res_id before enqueuing
663 the OPEN lock for the child inode, then to avoid senting
664 wrong ldlm_res_id to client.
668 Description: allow multiple instances of the same nid in NID hash
669 Details : Case of multiple separate clients from the same NID (as
670 with liblustre) is legitimate and so we should allow
671 multiple instances of the same NID in nid hash.
675 Description: the readahead code can sleep on a semaphore while holding a
677 Details : in ras_update, "lov_get_info" could be called during increasing
678 readahead windows, which tries to get the mutex lock "lov_lock"
679 while holding the spin_lock "ras_lock", then causes system
684 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
685 Details : fix assertion failure in the grant code.
689 Description: Use CNETERR (which is rate limited) in specific places in
690 the portal's LNET driver to avoid flooding the console.
694 Description: include last created object in precreate slow case
698 Description: don't do rep-ack if not created anything
699 Details : mds_open currently always put a lock into a rep-ack regardless
700 if something was created or not. This is pointless and only
701 creates needless contention. In fact the entire idea was to do
702 this for real creates as a recovery protection.
706 Description: Spurious error messages from smp_processor_id() on preemptible
708 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
709 first. The function fs_trace_get_tcd() was moved up.
713 Description: interval_erase() fix
714 Details : interval_erase() calls update_maxhigh() properly when child
719 Description: "lfs df" does not print stats for all mountpoints
720 Details : Print all mounted lustre filesystems with "lfs df"
724 Description: lfs setstripe -p not longer work with a relative pathname
725 Details : Use realpath() to provide absolute pathname.
729 Description: fix for truncated reply buffer
730 Details : reply buffer could be referred by reply_in_callback after released
735 Description: lustre.lov error when backing up symlinks with extended attributes
736 Details : Improved logic in ll_listxattr()
740 Description: properly handle null value for setattr -n lustre.lov
741 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
742 in ll_setxattr() due to no checking if "value" is NULL.
743 This command now resets to the default striping when executed
748 Description: stack overflow on lock cancellation due to fsync call
749 Details : sync_on_lock_cancel is needed for recovery when async journal
750 is enabled, but we actually just need to make sure that
751 metadata blocks have hit the journal, so doing a fs sync
752 should be enough and should consume less stack (just create an
753 empty handle and commmit it).
757 Description: using current->journal_info to store per-thread data leads
758 to problem under memory pressure
759 Details : disable the per-thread data (current->journal_info) containing
760 the lock info during I/O to work around the issue for short tem
764 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
765 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
766 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
771 Description: Cannot send after transport shutdown
772 Details : Clear imp_vbr_failed flag upon eviction
776 Description: soft lock in request set code during recovery
777 Details : during recovery, uses req->rq_set itself to replay the request
778 instead of ptlrpcd_recovery_pc
782 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
783 Details : During an ll_readahead under ll_readpage, we have seen the the
784 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
785 a page that is already locked by the readahead code.
790 Description: stop waitting for next replay transno if shutdown
791 Details : if the system is shutting down, wake up service thread blocked
792 to wait for next replay transno during recovery, then all the
793 references held by queued requests can be dropped and device
798 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
799 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
803 Description: per-nid stats should not access lustre-hash internal structures
808 Description: mount.lustre fails to pass some options to mount()
813 Description: ext4 extent allocation is slower than in ext3
814 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
815 enlarge ext4 preallocation table for 2048 4K blocks extents
820 Description: incorrect triggering of synchronous IO
821 Details : The OSC can mistakenly fall back to synchronous IO when the
822 max_dirty_mb limit is reached and no write requests have yet
823 been issued. This can occur when the dirty pages are spread
824 over many files all of which are below the optimal request size.
828 Description: Optimize quota_ctl operations by sending requests in parallel
829 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
830 Compiled from two attachments in the ticket.
834 Description: ordering issue between transaction start & i_mutex
835 Details : start the transaction earlier in llog_lvfs_destroy to get
836 transaction start and inode mutex lock nested properly.
840 Description: lru resize SLV can get stuck
841 Details : calculate SLV with a greater precision to not lose small
842 changes due to interger math truncation; round up SLV only
843 if the amount of granted locks less than the limit to not
844 get stuck with this SLV
848 Description: avoid divide-by-zero in lprocfs_rd_import()
852 Description: lfs quota failed when OSTs are down
853 Details : really return approximate block/inode usage when OSTs are down
857 Description: llobdstat fix and enhancement
858 Details : add a counter to set a limit to how many samples will be returned
859 fix a wildcard in the path to limit to obdfilter stats only
863 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
865 Details : abort lock enqueue processing sooner when we umount is in progress.
867 Severity : enhancement
869 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
870 2.6.18-164.11.1.0.1.el5.
872 Severity : enhancement
874 Description: error message improvements
875 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
876 Simplify MDT/OST service start message.
877 Suppress "changing the import ..." warning.
880 Bugzilla : 21961/17914
881 Description: ignore trailing -mdc when determining index number
884 Frequency : only with SLES11
886 Description: fix for a race condition in linux quotas implementation
887 Details : dq_flags(struct dquot) access is not properly locked which could
888 lead to certain inconsistencies when accessing it using non-atomic
889 bit operations like __set_bit in do_set_dqblk.
893 Description: return any pending async errors in close(2) using flush callback
895 -------------------------------------------------------------------------------
897 2010-01-29 Sun Microsystems, Inc.
899 * Support for kernels:
900 2.6.16.60-0.42.8 (SLES 10),
901 2.6.27.39-0.3.1 (SLES11),
902 2.6.18-164.11.1.el5 (RHEL 5)
903 2.6.18-164.6.1.0.1.el5 (OEL 5)
904 * Client support for unpatched kernels:
905 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
906 2.6.16 - 2.6.30 vanilla (kernel.org)
907 * Recommended e2fsprogs version: 1.41.6.sun1
908 * The async journal commit feature (bug 19128) and the cancel
909 lock before replay feature (bug 16774) are disabled by default.
911 Severity : enhancement
913 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
914 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
916 Severity : enhancement
917 Bugzilla : 21511/19848
918 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
920 Severity : enhancement
922 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
924 Severity : enhancement
926 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
927 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
929 Severity : enhancement
931 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
934 Severity : enhancement
936 Description: Vanilla kernel 2.6.30 patchless client support.
941 Description: bad entry in directory xxx: inode out of bounds
942 Details : fix locking issue in the rename path which could race with any
943 other operations updating the same directory.
945 Severity : enhancement
947 Description: Make watchdog timer messages to be more clear and descriptive.
951 Description: cp -p command does not preserve the dates and timestamp
952 Details : mtime could be spoiled by a write callback
956 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
960 Description: Allow non-root access for "lfs check".
961 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
963 Severity : enhancement
965 Description: quotacheck performance/scaling issues
966 Details : reduce quotacheck time on empty filesystem by skipping uninit
969 Severity : enhancement
971 Description: Enhancement for lfs(1) command to use numeric uid/gid.
973 Severity : enhancement
975 Description: Adjust locks' extents on their first enqueue, so that at the time
976 they get granted, there is no need for another pass through the
977 queues since they are already shaped into the proper forms.
981 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
982 arguments and prevent LBUG() in lustre_shrink_reply_v2().
986 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
987 exactly the specified mount options are used. Leaving off
988 any "mandatory" mount options is an error. Leaving off any
989 default mount options causes a warning, but is allowed.
990 Change errors=remount-ro from mandatory to default. Sanitize
991 the mount string before storing it. Update man pages accordingly.
995 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
996 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
999 Severity : enhancement
1001 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
1003 Severity : enhancement
1005 Description: Clients can replay thousands of unused locks during recovery
1006 Details : Don't replay unused locks (only read locks for now) during
1007 recovery. This feature is disabled by default and can be
1008 enabled by running the following command on the clients:
1009 lctl get_param ldlm.cancel_unused_locks_before_replay
1013 Description: can't stat file in some situation.
1014 Details : improve initialize osc date when target is added to mds and
1015 ability to resend too big getattr request is client isn't have info
1020 Description: Prevent inconsistences between linux and lustre mount structures.
1021 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1022 Make the sleep interruptible.
1024 Severity : enhancement
1026 Description: Communicate OST degraded/readonly state via statfs to MDS
1027 Details : Flags in the statfs returned from OSTs indicate whether the
1028 OST is in a degraded RAID state, or if the filesystem has
1029 turned read-only after a filesystem error is detected.
1034 Description: don't panic if EPROTO was hit when reading symlink
1035 Details : correctly handling request reference in error cases.
1040 Description: open sometimes returns ENOENT instead of EACCES
1041 Details : checking permission should be part of open part of mds_open, not
1042 lookup part. so server should be set DISP_OPEN_OPEN disposition
1043 before starting permission check.
1044 Also not need revalidate dentry if client already have LOOKUP lock.
1048 Description: enable client interface failover
1049 Frequency : on servers with multiple network interfaces
1050 Details : When a child reconnects from another NID, properly update export
1051 nid hash position and ldlm reverse import.
1053 Severity : enhancement
1055 Description: implemented direct I/O with arbitrary (nonaligned) memory
1056 addresses and file offsets.
1058 Severity : enhancement
1060 Description: added more recovery timeout options.
1062 Severity : enhancement
1064 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
1069 Description: Avoid deadlock for local client writes
1070 Frequency : only on systems with clients writing to an OST on the same node
1071 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1072 memory freeing context. This allows OST threads to set the
1073 PF_MEMALLOC flag on task structures in order to allocate memory
1074 from reserved pools and complete IO.
1075 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1076 so that the OST threads generate memory pressure and allow
1077 inactive pages to be reclaimed.
1082 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1083 Details : move ldlm namespace creation in setup phase to avoid grab
1084 _lprocfs_lock with cli_sem held.
1088 Frequency : only during format of test systems
1089 Description: Unable to run several mkfs.lustre on loop devices at the same time
1090 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1091 formatting. The solution is to proper handle the error.
1093 Severity : enhancement
1095 Description: implement async create (obd_async_create) method for osc, to avoid
1096 too long waiting new ost objects with holding ldlm lock.
1100 Frequency : occasionally during network problems
1101 Description: client not allowed to reconnect to OST because of active request
1102 Details : abort bulk requests received by the OST once the client has timed
1103 out since the client will resend the request anyway. The client
1104 also now retries to reconnect to the same server if a connect
1105 request failed with EBUSY or -EAGAIN.
1108 Frequency : rare, if used wide striped file and one ost in down.
1110 Descriptoin: don't return error if we created a subset of objects for file.
1111 Details : lov_update_create_set() uses set->set_success as index for created
1112 objects, so if some requests failed, they will have hole at end of
1113 array and we can use qos_shrink_lsm for allocate correct lsm.
1117 Description: Slow stale export processing during normal start up
1118 Details : The global mgc lock prevents OST setup to be run in parallel.
1119 Replace the global lock with a per-config_llog_data semaphore.
1123 Description: Out or order replies might be lost on replay
1124 Details : In ptlrpc_retain_replayable_request if we cannot find retained
1125 request with tid smaller then one currently being added, add it
1126 to the start, not end of the list.
1130 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
1131 Details : add cond_resched() calls to avoid hogging the cpu for too long
1132 in the hash code. Make also lustre_hash_for_each_empty() more
1135 Severity : enhancement
1137 Description: Performance improvements for debug messages with D_RPCTRACE,
1138 D_LDLM, D_QUOTA options.
1141 Frequency : only with NFS export
1143 Description: (lov_merge.c:74:lov_merge_lvb())
1144 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
1145 Details : Fix a race in the nfs export code by populating inode
1146 info while the new inode is still locked
1148 Severity : enhancement
1150 Description: add a new file in procfs called force_lbug. Writting to this
1151 file triggers a LBUG. Only for test purpose.
1155 Description: OOM killer causes node hang
1156 Details : really interrupt the sleep in osc_enter_cache on signals
1160 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1161 Details : fix race during quota release on the slave.
1163 Severity : enhancement
1165 Description: smaller hash bucket sizes, cleanups
1166 Details : increase hash table sizes and enabled rehashing for pools, quota,
1167 uuid, nid & per-nid stats.
1169 Severity : enhancement
1171 Description: Add ldiskfs maxdirsize mount option
1172 Details : add max_dir size mount option
1176 Description: panic in ll_statahead_thread
1177 Details : prevent parent thread to be killed before its child
1180 Frequency : only with 16TB device
1182 Description: unable to perform "mount -t lustre" of 16TB OST device
1183 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
1187 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
1188 Details : unregistering should be zero if no RPC inflight.
1192 Description: hyperion: Oops during metabench
1193 Details : Correct the refcount of lov_request_set
1195 Severity : enhancement
1197 Description: Add mptlinux and nxge drivers to Lustre builds
1199 Severity : enhancement
1201 Description: Fix watchdog timer message to be more clear
1202 Details : Make watchdog timer messages more clear and descriptive.
1206 Description: LNET soft lockups in socknal_cd thread
1207 Details : don't hog CPU for active-connecting if another connd is
1208 accepting connecting-requst from the same peer
1212 Description: recovery-small test_17 hang
1213 Details : Land several AT improvements & fixes.
1217 Description: MDS panic and hanging client processes
1218 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
1222 Description: OSS stuck in recovery.
1223 Details : fix race during recovery. class_unlink_export,
1224 class_set_export_delayed and target_queue_last_replay_reply
1225 may race while increasing/decreasing obd_recoverable_clients
1226 and obd_delayed_clients, causing recovery to wait forever.
1228 Severity : enhancement
1230 Description: add cascading_rw.c to lustre/tests
1234 Description: filter_last_id() NULL deref
1235 Details : lprocfs_filter_rd_last_id() should check for the fully
1236 setup obd device, before proceeding further.
1238 Severity : enhancement
1240 Description: Loadgen improvements
1241 Details : stacksize and locking fixes for loadgen
1245 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
1246 Details : The atomic_read() allowing the atomic_inc() are not covered
1247 by a lock. Thus they may safely race and trip this CERROR()
1248 unless we add in a small fudge factor (+1).
1250 Severity : enhancement
1252 Description: shrink_slab: nr=-9223362083340912175
1253 Details : fix spurious message from shrink_slab reporing negative nr
1257 Description: Quiet bogus previously committed transno error
1258 Details : suppress the "server went back in time" error message which
1259 is always printed even in the common case after a client eviction
1261 Severity : enhancement
1263 Description: Parallel statfs() calls result in client eviction
1264 Details : cache statfs data for 1s.
1268 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
1270 Details : fix serveral issues in pinger code causing clients not to ping
1271 servers for too long, resulting in evictions.
1275 Description: e2fsck should warn when MMP update interval is extended
1276 Details : print mmp_check_interval and make it possible to abort mount
1277 operation in case it takes too long.
1281 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
1282 Details : fix bug in the RHEL5's jbd2 callback patch.
1286 Description: drop number of active requests when queued for recovery
1287 Details : Now that we take a reference on the original request instead of
1288 making a copy of it for recovery. We need to drop the number of
1289 active requests or the queued requests will prevent all request
1290 processing when they exceed (srv->srv_threads_running - 1).
1292 Severity : enhancement
1294 Description: refuse to invalidate operational quota files when they are in use
1295 Details : an attempt to invalidate operational quota files on the quota
1296 master is not actually permitted by VFS (returning -EPERM), but we
1297 should not depend on that and should return the error earlier.
1301 Description: Applications stuck in jbd2_log_wait_commit during exit
1302 Details : fix deadlock between kjournald2 trying to acquire the page lock
1303 owned by an ost_io thread waiting for journal commit.
1305 -------------------------------------------------------------------------------
1307 2009-10-16 Sun Microsystems, Inc.
1309 * Support for kernels:
1310 2.6.16.60-0.42.4 (SLES 10),
1311 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1312 2.6.18-128.7.1.el5 (RHEL 5),
1313 2.6.18-128.7.1.el5 (OEL 5),
1314 * Client support for unpatched kernels:
1315 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1316 2.6.16 - 2.6.27 vanilla (kernel.org)
1317 * Recommended e2fsprogs version: 1.41.6.sun1
1318 * ext4 support for RHEL5 is experimental and thus should not be
1321 Severity : enhancement
1323 Description: Add OEL5 support.
1325 Severity : enhancement
1327 Description: Update kernel to SLES11 2.6.27.29-0.1.
1331 Description: File checksum failures with OST read cache on
1332 Details : Disable page poisoning when the bulk transfer has to be aborted
1333 because the client got evicted.
1337 Description: Don't allow make backward step on assiging osc next id.
1338 Details : race between allocation next id and ll_sync thread can be cause
1339 of set wrong osc next id and can be kill valid ost objects.
1341 Severity : enhancement
1343 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1345 Severity : enhancement
1347 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1351 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1355 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1359 Description: Error handling in osc_statfs_interpret() has been improved.
1360 Details : Check in osc_statfs_interpret() for EBADR.
1364 Description: Do not send statfs() requests to OSTs disabled by administrator.
1365 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1369 Description: Do not update ctime for the deleted inode.
1370 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1374 Description: Increase of the size of the LDLM resource hash.
1375 Details : Bump up RES_HASH_BITS=12.
1379 Description: correctly send lsm on open replay
1380 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1385 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1386 Details : filter_destroy() does not hold the DLM lock over the whole
1387 operation. If the DLM lock is dropped, filter_commitrw() can go
1388 through, causing the deadlock between page lock and i_mutex.
1389 The i_alloc_sem should also be hold in filter_destroy() while
1390 truncating the file.
1394 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1395 Details : ldiskfs_truncate calls grab_cache_page which may start page
1396 allocation under an open transaction. This may lead to
1397 calling prune_icache with consequent lustre reentrance.
1400 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1401 still up and when the OST pool feature is used
1403 Description: interop testing got LBUG when run dd with OST pool
1404 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1405 Details : down/upgrading the MDS to a version that doesn't/does support OST
1406 pool can cause clients to crash because the lsm has changed
1411 Description: missing tree_status on 1.8.1 RPM build
1412 Details : make rpms failed due because the tree_status file is missing.
1416 Description: continuing LustreError "mds adjust qunit failed!"
1417 Details : don't print message on the console when ->adjust_qunit fails.
1421 Description: don't increase ldlm timeout if previous client was evicted
1422 Details : if a client doesn't respond to a blocking callback within the
1423 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1424 when the lock is next granted.
1428 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1429 affects recovery negatively.
1430 Details : make sure all exports have been properly destroyed by the zombie
1431 thread processed before stopping the target.
1435 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1436 Details : disable write barrier for ext4/SLES11.
1440 Description: Kernel panic - not syncing: Out of memory and no killable
1441 processes... on OSS when iozone
1442 Details : fix memory leak in the journal checksum patch.
1446 Description: group quota "too many blocks" OSS crashes
1447 Details : we should keep the same uid/gid for lquota_chkquota() and
1448 lquota_pending_commit()
1452 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1453 Details : don't LBUG on release quota error. Just a workaround until the
1454 problem is understood.
1456 ------------------------------------------------------------------------------
1457 2009-07-31 Sun Microsystems, Inc.
1459 * Support for kernels:
1460 2.6.16.60-0.39.3 (SLES 10),
1461 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1462 2.6.18-128.1.14.el5 (RHEL 5),
1463 * Client support for unpatched kernels:
1464 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1465 2.6.16 - 2.6.27 vanilla (kernel.org)
1466 * Recommended e2fsprogs version: 1.41.6.sun1
1467 * File join has been disabled in this release, refer to Bugzilla 16929.
1468 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1469 of Lustre filesystem with 4K stack may cause a stack overflow. For
1470 more information, please refer to bugzilla 17630.
1471 * ext4 support for RHEL5 is experimental and thus should not be
1474 Severity : enhancement
1476 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1479 Frequency : with 1.8 server and 1.6 clients
1481 Description: correctly shrink reply for avoid send too big message to client.
1482 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1483 some problems with sending this reply to 1.6 client.
1487 Description: Repeated atomic allocation failures.
1488 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1489 to generate memory pressure and allow reclaiming of inactive pages.
1490 At the same time, do not allow to exhaust emergency pools.
1491 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1493 Severity : enhancement
1494 Bugzilla : 19846, 18289
1495 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1497 Severity : enhancement
1498 Bugzilla : 19625, 16893, 18668, 19848
1499 Description: Add support for SLES11 2.6.27.23-0.1.
1501 Severity : enhancement
1503 Description: Update client support to vanila kernels up to 2.6.27.
1505 Severity : enhancement
1507 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1509 Severity : enhancement
1511 Description: Compile with -Werror by default for i686 and x86_64.
1515 Description: resolve race between obd_disconnect and class_disconnect_exports
1516 Details : if obd_disconnect will be called to already disconnected export he
1517 forget release one reference and osc module can't unloaded.
1519 Severity : enhancement
1521 Description: move AT tunable parameters for more consistent usage
1522 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1526 Descriptoin: correctly skip time estimate if in recovery
1527 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1531 Descriptoin: OSS DeadLock
1532 Details : Use trylock to prevent deadlock when shrink icache.
1534 Severity : enhancement
1536 Description: Allow tuning service thread via /proc
1537 Details : For each service a new
1538 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1539 created that can be used to set min/max thread counts, and get the
1540 current number of running threads.
1542 Severity : enhancement
1544 Description: Add state history info file, enhance import info file
1545 Details : Track import connection state changes in a new osc/mdc proc file;
1546 add overview-type data to the osc/mdc import proc file.
1550 Description: Reduce small size read RPC
1551 Details : Set read-ahead limite for every file and only do read-ahead when
1552 available read-ahead pages are bigger than 1M to avoid small size
1557 Description: free_entry erroneously used groups_free instead of put_group_info
1559 Severity : enhancement
1561 Description: Make read-ahead stripe size aligned.
1563 Severity : enhancement
1565 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1568 Frequency : rare, connect and disconnect target at same time
1570 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1571 Details : don't call obd_disconnect under lov_lock. this long time
1572 operation and can block ptlrpcd which answer to connect request.
1575 Frequency : start MDS on uncleanly shutdowned MDS device
1577 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1578 Details : stay in waiting mds<>ost recovery finished produce random bugs
1579 due race between two ll_sync thread for one lov target. send
1580 ACTIVATE event only if connect realy finished and import have
1584 Frequency : start MDS on uncleanly shutdowned MDS device
1586 Descriptoin: aborting recovery hang on MDS
1587 Details : don't throttle destroy RPCs for the MDT.
1591 Description: Slow reads beyond 8Tb offsets.
1592 Details : Page index integer overflow in ll_read_ahead_page
1596 Description: Soft lockup on OSS after MDS failover
1597 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1598 As a conseqence, the patch from bug 18224 is not operational
1599 and the MDS export cannot be reused on the OSTs until it gets
1603 Frequency : rare, only if using MMP with Linux RAID
1605 Description: MMP doesn't work with Linux RAID
1606 Details : While using HA for Lustre servers with Linux RAID, it is possible
1607 that MMP will not detect multiple mounts. To make this work we
1608 need to unplug the device queue in RAID when the MMP block is being
1609 written. Also while reading the MMP block, we should read it from
1610 disk and not the cached one.
1613 Frequency : rare, during recovery
1615 Description: Assertion failure in ldlm_lock_put
1616 Details : Do not put cancelled locks into replay list, hold references on
1617 locks in replay list
1621 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1622 Details : create_count always drops to the min value (=32) because grow_count
1623 is being changed before the precreate RPC completes.
1626 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1629 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1630 Details : add proper locking for creation of jbd_4k slab cache
1634 Description: MMP check in ext3_remount() fails without displaying any error
1635 Details : When multiple mount protection fails during remount, proper error
1640 Description: Rare Client crash on resend if the file was deleted.
1641 Details : When file is opened, but open reply is lost and file is
1642 subsequently deleted before resend, resend processing logic
1643 breaks trying to open the file again, should not try to open.
1647 Description: add check for >8TB ldiskfs filesystems
1648 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1649 Don't allow >8TB ldiskfs filesystems to be mounted without
1650 force_over_8tb mount option
1654 Description: Client locked up when running multiple instances of an app. on
1655 multiple mount points
1656 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1657 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1661 Description: Cannot acces an NFS-mounted Lustre filesystem
1662 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1663 from a Lustre-client exporting the Lustre filesystem via NFS.
1667 Description: panic in ll_statahead_thread
1668 Details : grab dentry reference in parent process.
1671 -------------------------------------------------------------------------------
1673 tbd Sun Microsystems, Inc.
1675 * Support for kernels:
1676 2.6.16.60-0.31 (SLES 10),
1677 2.6.18-128.1.6.el5 (RHEL 5),
1678 2.6.22.14 vanilla (kernel.org)
1679 * Client support for unpatched kernels:
1680 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1681 2.6.16 - 2.6.22 vanilla (kernel.org)
1682 * Recommended e2fsprogs version: 1.40.11-sun1
1683 * File join has been disabled, refer to Bugzilla 16929.
1684 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1685 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1686 Lustre filesystem with 4K stack may cause a stack overflow. For more
1687 information, please refer to bugzilla 17630.
1689 Severity : enhancement
1691 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1693 Severity : enhancement
1695 Description: Update OFED release to 1.4.1 RC4
1697 Severity : major, only with big OST
1699 Description: Very poor metadata performance on Infiniband lustre configuration
1700 Details : OST object precreation becomes very slow on big OSTs. This is due
1701 to the ialloc patch spending too much time scanning groups.
1704 Frequency : during recovery
1706 Description: don't mix llog inodes with normal.
1707 Details : allocate inodes for log in last inode group
1712 Description: fix lqs' reference which won't be put in some situations
1713 Details : This patch fixes:
1714 1. In quota_check_common(), this function will check quota
1715 for user and group, but only send one return via "pending".
1716 In most cases, the pendings should be same. But that is not
1718 2. If quotaoff runs between lquota_chkquota() and
1719 lquota_pending_commit(), the same thing will happen too.
1720 That is why it comes:
1721 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1724 -------------------------------------------------------------------------------
1726 2008-12-31 Sun Microsystems, Inc.
1728 * Support for kernels:
1729 2.6.5-7.314 (SLES 9),
1730 2.6.9-67.0.22.EL (RHEL 4),
1731 2.6.16.60-0.31 (SLES 10),
1732 2.6.18-92.1.17.el5 (RHEL 5),
1733 2.6.22.14 vanilla (kernel.org)
1734 * Client support for unpatched kernels:
1735 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1736 2.6.16 - 2.6.22 vanilla (kernel.org)
1737 * Client support for unpatched kernels:
1738 we do not recommend using patchless RHEL4 clients with kernels
1739 prior to 2.6.9-55EL (RHEL4U5).
1740 * Recommended e2fsprogs version: 1.40.11-sun1
1741 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1742 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1743 removed cwd "./" (refer to Bugzilla 14399).
1744 * A new quota file format has been introduced in 1.6.5.
1745 The format conversion from prior releases is handled transparently,
1746 but releases older than 1.4.12/1.6.5 don't understand this new
1747 format. The automatic format conversion can be avoided by running
1748 the following command on the MDS:
1749 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1750 For more information, please refer to bugzilla 13904.
1751 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1752 The format conversion from prior releases is handled transparently,
1753 but releases older than 1.6.6/1.8.0 don't understand this new
1754 format. The automatic format conversion can be avoided by running
1755 the following commands on the MDS and OSS servers (for
1756 pre 1.4.12-1.6.5 quota files):
1757 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1758 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1759 or (for 1.4.12/1.6.5 quota files)
1760 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1761 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1762 For more information, please refer to bugzilla 13904.
1763 * Output of lfs quota has been made less detailed by default,
1764 old (verbose) output can be obtained by using -v option.
1765 * File join has been disabled in this release, refer to Bugzilla 16929.
1766 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1767 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1768 Lustre filesystem with 4K stack may cause a stack overflow. For more
1769 information, please refer to bugzilla 17630.
1771 Severity : enhancement
1773 Description: Caching OSS
1774 Details : introduce data caching on the OSS. The OSS now relies on the linux
1775 kernel page cache to keep recently accessed data in memory.
1776 It is worth noting that all write requests are still flushed
1777 synchronously as in lustre 1.6.
1779 Severity : enhancement
1781 Description: Version based recovery
1782 Details : introduce finer grained recovery able to detect transaction
1783 dependencies and can deal with transaction gaps caused by clients
1784 failing at the same time as the server.
1786 Severity : enhancement
1788 Description: Enable adaptive timeouts by default
1789 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1790 managed dynamically based on server load and should not need
1791 to be tuned manually based on cluster size. This allows Lustre
1792 to work under a wider variety of system sizes and loads, without
1793 unnecessarily causing lengthy recovery times.
1795 Severity : enhancement
1797 Description: Add OST Pools support
1798 Details : File striping can now be set to use an arbitrary pool of OSTs
1800 Severity : enhancement
1802 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1803 Details : allow skip disconnected ost for send statfs request and hide error
1807 Frequency : rare, on llog test 6
1809 Descriptoin: don't allow connect to already connected import
1810 Details : allowing connect to already connected import is hide connecting problem.
1813 Frequency : rare, on failed llog setup
1815 Descriptoin: don't leak obd reference on failed llog setup
1816 Details : for failed llog setup - mgc forget call class_destroy_import for
1817 client import, move destroy import to more generic place.
1822 Descriptoin: allow kill process which wait statahead result
1823 Details : for some reasons 'ls' can stick in waiting result from statahead,
1824 in this case need way for kill this process.
1827 Frequency : rare, at shutdown
1829 Descriptoin: panic at umount
1830 Details : llap_shrinker can be raced with killing super block from list and
1831 this produce panic with access to already freeded pointer
1836 Descriptoin: don't lose wakeup for imp_recovery_waitq
1837 Details : recover_import_no_retry or invalidate_import and import_close can
1838 both sleep on imp_recovery_waitq, but we was send only one wakeup
1844 Descriptoin: panic in mds_open
1845 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1850 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1852 Details : release lock refernce only after add page to pages list.
1855 Frequency : always with long access acl
1857 Descriptoin: mds can't pack reply with long acl.
1858 Details : mds don't control size of acl but they limited by reint/getattr
1863 Frequency : on remount
1865 Description: external journal device not working after the remount
1866 Details : clear dev_rdonly flag for external journal devices in
1872 Description: shutdown vs evict race
1873 Details : client_disconnect_export vs connect request race.
1874 if client will evicted at this time - we start invalidate
1875 thread without referece to import and import can be freed
1881 Description: shrink LOV EAs before replying
1882 Details : correctly adjust LOV EA buffer for reply.
1887 Description: don't skip ost target if they assigned to file
1888 Details : Drop slow OSCs if we can, but not for requested start idx.
1889 This means "if OSC is slow and it is not the requested
1890 start OST, then it can be skipped, otherwise skip it only
1891 if it is inactive/recovering/out-of-space.
1893 Severity : enhancement
1895 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1898 Frequency : rare, need acl's on inode.
1900 Description: client can't handle ost additional correctly
1901 Details : if ost was added after client connected to mds client can have
1902 hit lnet_try_match_md ... to big messages to wide striped files.
1903 in this case need teach client to handle config events about add
1904 lov target and update client max ea size at that event.
1906 Severity : enhancement
1908 Description: Update to sles9 kernel-2.6.5-7.314.
1910 Severity : enhancement
1912 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1915 Frequency : Create a symlink file with a very long name
1917 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1918 Details : If there is no extra space in the request for early cancels,
1919 ldlm_req_handles_avail() returns 0 instead of a negative value.
1924 Description: mds is deadlocked
1925 Details : in rare cases, inode in catalog can have i_no less than have parent
1926 i_no, this produce wrong order for locking during open, and
1927 parallel unlink can be lock open. this need teach mds_open to grab
1928 locks in resource id order, not at parent -> child order.
1930 Severity : enhancement
1932 Description: Add /proc entry for import status
1933 Details : The mdc, osc, and mgc import directories now have
1934 an import directory that contains useful import data for debugging
1935 connection problems.
1937 Severity : enhancement
1939 Description: Re-disable certain /proc logging
1940 Details : Enable and disable client's offset_stats, extents_stats and
1941 extents_stats_per_process stats logging on the fly.
1944 Frequency : Only on FC kernels 2.6.22+
1946 Description: oops in statahead
1947 Details : Do not drop reference count for the dentry from VFS when lookup,
1948 VFS will do that by itself.
1950 Severity : enhancement
1952 Description: Generic /proc file permissions
1953 Details : Set /Proc file permissions in a more generic way to enable non-
1954 root users operate on some /proc files.
1958 Description: Hitting mdc_commit_close() ASSERTION
1959 Details : Properly handle request reference release in
1960 ll_release_openhandle().
1964 Frequency : only patchless client
1965 Description: add workaround for race between add/remove dentry from hash
1967 Severity : enhancement
1969 Description: Allow OST glimpses to return PW locks
1973 Description: LBUG when llog conf file is full
1974 Details : When llog bitmap is full, ENOSPC should be returned for plain
1979 Description: Prevent import from entering FULL state when server in recovery
1983 Description: service mount cannot take device name with ":"
1984 Details : Only when device name contains ":/" will mount treat it as
1990 Description: replace ptlrpcd with the statahead thread to interpret the async
1991 statahead RPC callback
1995 Frequency : on recovery
1996 Description: I/O failures after umount during fail back
1997 Details : if client reconnected to restarted server we need join to recovery
1998 instead of find server handler is changed and process self eviction
1999 with cancel all locks.
2001 Severity : enhancement
2003 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
2007 Description: Kernel BUG tries to release flock
2008 Details : Lustre does not destroy flock lock before last reference goes
2009 away. So always drop flock locks when client is evicted and
2010 perform unlock regardless of successfulness of speaking to MDS.
2012 Severity : enhancement
2014 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
2016 Severity : enhancement
2018 Description: Upcall on Lustre log has been dumped
2019 Details : Allow for a user mode script to be called once a Lustre log has
2020 been dumped. It passes the filename of the dumped log to the
2021 script, the location of the script can be specified via
2022 /proc/sys/lnet/debug_log_upcall.
2027 Description: avoid messages about idr_remove called for id that is not allocated
2028 Details : Move assigment s_dev for clustered nfs to end of initialization,
2029 for avoid problem with error handling.
2034 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
2035 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
2036 CONN_USED_HASH and this prodice warning when put connection again
2042 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
2043 Details : release reference to stats when client disconnected, not
2044 when export destroyed for avoid races when client destroyed
2045 after main ost export.
2049 Description: more cleanup in mds_lov
2050 Details : add workaround for get valid ost count for avoid warnings about
2051 drop too big messages, not init llog cat under semphore which
2052 can be blocked on reconnect and break normal replay, fix access
2055 Severity : enhancement
2057 Description: Export bytes_read/bytes_write count on OSC/OST.
2061 Description: Early reply size mismatch, MGC loses connection
2062 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
2063 the connect flags are properly negotiated.
2067 Description: Properly propagate oinfo flags from lov to osc for statfs
2068 Details : restore missing copy oi_flags to lov requests.
2072 Description: exports in /proc are broken
2073 Details : recreate /proc entries for clients when they reconnect.
2075 Severity : enhancement
2077 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
2078 : l_getgroups(8), lst(8), routerstat(8)
2079 Details : included man pages for llobdstat(8), llstat(8),
2080 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
2082 Severity : enhancement
2084 Description: Implement lustre ll_show_options method.
2086 Severity : enhancement
2088 Description: Update to SLES9 kernel-2.6.5-7.312.
2090 Severity : enhancement
2092 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
2096 Description: exports in /proc are broken
2097 Details : recreate /proc entries for clients when they reconnect.
2101 Description: don't fail open with -ERANGE
2102 Details : if client connected until mds will be know about real ost count
2103 get LOV EA can be fail because mds not allocate enougth buffer
2108 Description: Resolve device initialization race
2109 Details : Prevent proc handler from accessing devices added to the
2110 obd_devs array but yet be intialized.
2112 Severity : enhancement
2114 Description: configure's --enable-quota should check the
2115 : kernel .config for CONFIG_QUOTA
2116 Details : configure is terminated if --enable-quota is passed but
2117 : no quota support is in kernel
2119 Severity : enhancement
2121 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
2123 Severity : enhancement
2125 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
2129 Frequency : rare, on PPC clients
2130 Description: don't swab ost objects in response about directory, because
2132 Details : bug similar bug 14856, but in different function.
2134 Severity : enhancement
2136 Description: lfs quota tool enhancement
2137 Details : added units specifiers support for setquota, default to
2138 current uid/gid for quota report, short quota stats by
2139 default, nonpositional parameters for setquota, added
2140 llapi_quotactl manual page.
2142 Severity : enhancement
2144 Description: *optional* service tags registration
2145 Details : if the "service tags" package is installed on a Lustre node
2146 When the filesystem is mounted, a local-node service tag will
2147 be created. See http://inventory.sun.com/ for more information
2148 about the Service Tags asset management system.
2152 Description: Client runs out of low memory
2153 Details : Consider only lowmem when counting initial number of llap pages
2156 Frequency : occasional
2158 Description: add refcount for osc callbacks, so avoid panic on shutdown
2160 Severity : enhancement
2162 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
2165 Frequency : testing only
2167 Description: sanity test 65a fails if stripecount of -1 is set
2168 Details : handle -1 striping on filesystem in ll_dirstripe_verify
2171 Frequency : only in unusual configurations
2173 Description: Kernel panic with find ost index.
2174 Details : lov_obd have panic if some OST's have sparse indexes.
2176 Severity : enhancement
2178 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
2181 Frequency : rarely, if filesystem is mounted with -o flock
2183 Description: do not process already freed flock
2184 Details : flock can possibly be freed by another thread before it reaches
2185 to ldlm_flock_completion_ast.
2188 Frequency : rarely, if filesystem is mounted with -o flock
2190 Description: LBUG during stress test
2191 Details : Need properly lock accesses the flock deadlock detection list.
2194 Frequency : rarely, if binaries are being run from Lustre
2196 Description: oops in page fault handler
2197 Details : kernel page fault handler can return two special 'pages' in
2198 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
2201 Frequency : rarely, during shutdown
2203 Description: timeout with invalidate import.
2204 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
2205 should be handled by ptlrpcd. This produce long age waiting and
2206 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
2212 Description: ASSERTION(CheckWriteback(page,cmd)) failed
2213 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
2217 Frequency : only with broken builds/installations
2219 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
2220 Details : just return an error to a user, put a console error message
2222 Severity : enhancement
2224 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
2226 Severity : enhancement
2228 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
2230 Severity : enhancement
2232 Description: enable MGS and MDT services start separately
2233 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
2234 a co-located MGS without starting the MGS, which is a complement
2235 to 'nosvc' mount option.
2238 Frequency : always, on big-endian systems
2240 Description: cleanup in ptlrpc code, related to PPC platform
2241 Details : store magic in native order avoid panic's in recovery on PPC
2242 node and forbid from this error in future. Also fix posibily
2243 of twice swab data. Fix get lov striping to userpace.
2246 Frequency : rarely, if replay get lost on server
2248 Description: server incorrectly drop resent replays lead to recovery failure.
2249 Details : do not drop replay according to msg flags, instead we check the
2250 per-export recovery request queue for duplication of transno.
2253 Frequency : after recovery
2255 Description: precreate to many object's after del orphan.
2256 Details : del orphan st in oscc last_id == next_id and this triger growing
2257 count of precreated objects. Set flag LOW to skip increase count
2258 of precreated objects.
2261 Frequency : rare, on clear nid stats
2263 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
2264 Details : when clean nid stats sometimes try destroy live entry,
2265 and this produce panic in free.
2268 Frequency : occasionally since 1.6.4
2270 Description: Stack overflow during MDS log replay
2271 Details : ease stack pressure by using a thread dealing llog_process.
2274 Frequency : very rare
2276 Description: MDT cannot be unmounted, reporting "Mount still busy"
2277 Details : Mountpoint references were being leaked during open reply
2278 reconstruction after an MDS restart. Drop mountpoint reference
2279 in reconstruct_open() and free dentry reference also.
2284 Description: wait until IO finished before start new when do lock cancel.
2285 Details : VM protocol want old IO finished before start new, in this case
2286 need wait until PG_writeback is cleared until check dirty flag
2287 and call writepages in lock cancel callback.
2292 Description: mds_mfd_close() ASSERTION(rc == 0)
2293 Details : In mds_mfd_close(), we need protect inode's writecount change
2294 within its orphan write semaphore to prevent possible races.
2297 Frequency : rare, on shutdown ost
2299 Description: don't hit live lock with umount ost.
2300 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2301 use shrink_dcache_sb instead.
2304 Frequency : only when echo_client is used
2306 Description: don't panic with use echo_client
2307 Details : echo client pass NULL as client nid pointer and this produce NULL
2308 pointer dereference.
2311 Frequency : Always on 32-bit PowerPC systems
2313 Description: fix build on PPC32
2314 Details : compile code with -m64 flag produce wrong object file for PPC32.
2319 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2320 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2321 by a client, and the data on disk are dangerous and incorrect.
2322 Change them with client eviction.
2324 Severity : enhancement
2326 Description: skiplist implementation simplification
2327 Details : skiplists are used to group compatible locks on granted list
2328 that was implemented as tracking first and last lock of each lock
2329 group the patch changes that to using doubly linked lists
2333 Description: delete compatibility for 32bit qdata
2334 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2335 qunit. That means servers of b1_4 and servers of b1_8 can't be
2336 used together if users want to use quota.
2339 Frequency : only with administrator action
2341 Description: mount failure if config log has invalid conf_param setting
2342 Details : If administrator specified an incorrect configuration parameter
2343 with "lctl conf_param" this would cause an error during future
2344 client mounts. Instead, ignore the bad configuration parameter.
2347 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2349 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2350 Details : If blocks per group is less than blocksize*8, set rest of the
2354 Frequency : Application do stride read on lustre
2356 Description: The read performance will drop a lot if the application does
2358 Details : Because the stride_start_offset are missing in stride read-ahead,
2359 it will cause clients read a lot of unused pages in read-ahead,
2360 then the read-performance drops.
2364 Description: more ldlm soft lockups
2365 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2366 starve other threads from the resource lock for a long time in
2367 case of long waiting queue, so change the debug level from
2368 D_OTHER to the less frequently used D_INFO.
2370 Severity : enhancement
2372 Description: add -gid, -group, -uid, -user options to lfs find
2374 Severity : enhancement
2376 Description: ll_recover_lost_found_objs - recover objects in lost+found
2377 Details : OST corruption and subsequent e2fsck can leave objects in the
2378 lost+found directory. Using the "ll_recover_lost_found_objs"
2379 tool, these objects can be retrieved and data can be salvaged
2380 by using the object ID saved in the fid EA on each object.
2385 Description: this bug _only_ happens when inode quota limitation is very low
2386 (less than 12), so that inode quota unit is 1 at initialization.
2387 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2388 is effective now. So least quota qunit should be 2.
2392 Description: Hung threads in invalidate_inode_pages2_range
2393 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2394 one is completed. As a result, some RPCs are stuck in the queue
2399 Description: Procfs and llog threads access destoryed import sometimes.
2400 Details : Sync the import destoryed process with procfs and llog threads by
2401 the import refcount and semaphore.
2405 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2406 Details : Sort source/child resource pair after updating child resource.
2411 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2412 Details : If insertion of an extent fails, then discard the inode
2413 preallocation and free data blocks else it can lead to duplicate
2418 Description: don't always update ctime in ext3_xattr_set_handle()
2419 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2420 In some cases the ctime should not be updated, for example for
2421 2.0->1.8 compatibility it is necessary to delete an xattr and it
2422 should not update the ctime.
2426 Description: add quota statistics
2427 Details : 1. sort out quota proc entries and proc code.
2428 2. add quota statistics
2433 Description: quotas are not honored with O_DIRECT
2434 Details : all writes with the flag O_DIRECT will use grants which leads to
2435 this problem. Now using OBD_BRW_SYNC to guard this.
2439 Bugzilla : 15713/16362
2440 Description: Assertion in iopen_connect_dentry in 1.6.3
2441 Details : looking up an inode via iopen with the wrong generation number can
2442 populate the dcache with a disconneced dentry while the inode
2443 number is in the process of being reallocated. This causes an
2444 assertion failure in iopen since the inode's dentry list contains
2445 both a connected and disconnected dentry.
2449 Description: assertion failure in ldlm_handle2lock()
2450 Details : fix a race between class_handle_unhash() and class_handle2object()
2451 introduced in lustre 1.6.5 by bug 13622.
2453 Severity : enhancement
2455 Description: superblock lock contention with many SMP cores on one client
2456 Details : several client filesystem locks were highly contended on SMP
2457 NUMA systems with 8 or more cores. Per-CPU datastructures
2458 and more efficient locking implemented to reduce contention.
2463 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2464 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2469 Description: oops in sd_iostats_seq_show()
2470 Details : unloading/reloading the scsi low level driver triggers a kernel
2471 bug when trying to access the sd iostat file.
2476 Description: Kernel panics during QLogic driver reload
2477 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2478 patch, causing memory corruption.
2483 Description: journal_dev option does not work in b1_6
2484 Details : pass mount option during pre-mount.
2486 Severity : enhancement
2488 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2489 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2490 extent information of a file. It can be used to map logical blocks
2491 in a file to physical blocks in the block device.
2494 Frequency : only with adaptive timeout enabled
2496 Description: DEBUG_REQ() bad paging request
2497 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2498 because it can be accessed by reply_in_callback() without the
2502 Frequency : only on Cray X2
2504 Description: X2 build failures
2505 Details : fix build failures on Cray X2.
2509 Description: xid & resent requests
2510 Details : Initialize RPC XID from clock at startup (randomly if clock is
2515 Description: quota recovery deadlock during mds failover
2516 Details : This patch includes att18982, att18236, att18237 in bz14840.
2518 1. fix osts hang when mds does failover with quotaon
2519 2. prevent watchdog storm when osts threads wait for the
2524 Description: kernel panic on racer
2525 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2527 Severity : enhancement
2529 Description: Add lustre_start utility to start or stop multiple Lustre servers
2534 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2535 Details : In case of memory pressure, list_del() can be called twice on
2536 req->rq_history_list, causing a kernel oops.
2540 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2541 Details : fix stack overflow in the distributed lock manager by defering
2542 export eviction after a failed ast to the elt thread instead of
2543 handling it in the dlm interpret routine.
2545 Severity : enhancement
2547 Description: More exported tunables for mballoc
2548 Details : Add support for tunable preallocation window and new tunables for
2549 large/small requests
2553 Description: Detect corruption of block bitmap and checking for preallocations
2554 Details : Checks validity of on-disk block bitmap. Also it does better
2555 checking of number of applied preallocations. When corruption is
2556 found, it turns filesystem readonly to prevent further corruptions.
2560 Frequency : only for big-endian servers
2561 Description: Check for big-endian system while mounting fs with extents feature
2562 Details : Mounting a filesystem with extents feature will fail on big-endian
2563 systems since ext3-based ldiskfs is not supported on big-endian
2564 systems. Can be over-ridden with "bigendian_extents" mount option.
2568 Description: Excessive recovery window
2569 Details : With AT enabled, the recovery window can be excessively long (6000+
2570 seconds). To address this problem, we no longer use
2571 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2572 timeout no longer depends on the service time, it is set to
2573 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2574 time via pb_service_time.
2577 Descriptoin: Don't sync journal after every i/o
2578 Details : Implement write RPC replay to allow server replies for write RPCs
2579 before data is on disk. However, this feature is disabled by
2580 default since some issues leading to data corruptions have been
2581 found during recovery (e.g. bug 19128). This feature can be enabled
2582 by running the following command on the OSSs:
2583 lctl set_param obdfilter.*.sync_journal=0
2587 Description: Watchdog triggered on MDS failover
2588 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2589 the OSTs know that the MDS "UUID" can be reused for the same export
2590 from a different NID, so we do not need to wait for the export to
2595 Descriptoin: Lustre detected file system corruption with inode out of bounds
2596 Details : don't update i_size on MDS_CLOSE for directories. This causes
2597 directory corruptions on the MDT.
2599 -------------------------------------------------------------------------------
2601 2008-05-26 Sun Microsystems, Inc.
2603 * Support for kernels:
2604 2.6.5-7.311 (SLES 9),
2605 2.6.9-67.0.7.EL (RHEL 4),
2606 2.6.16.54-0.2.5 (SLES 10),
2607 2.6.18-53.1.14.el5 (RHEL 5),
2608 2.6.22.14 vanilla (kernel.org)
2609 * Client support for unpatched kernels:
2610 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2611 2.6.16 - 2.6.22 vanilla (kernel.org)
2612 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2613 we do not recommend using patchless RHEL4 clients with kernels
2614 prior to 2.6.9-55EL (RHEL4U5).
2615 * Recommended e2fsprogs version: 1.40.7-sun1
2616 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2617 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2618 removed cwd "./" (refer to Bugzilla 14399).
2619 * A new quota file format has been introduced in 1.6.5.
2620 The format conversion from prior releases is handled transparently,
2621 but releases older than 1.4.12/1.6.5 will not understand this new
2622 format. The automatic format conversion can be avoided by running
2623 the following command on the MDS before upgrading:
2624 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2625 For more information, please refer to bugzilla 13904.
2629 Description: quota performance fix
2630 Details : quota data is written in journalled mode instead of ordered to
2631 increase performance
2635 Description: lfs support for human-readable quota grace time strings
2636 Details : lfs setquota -t and lfs quota -t represent quota grace times
2637 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2640 Frequency : always with o2ib 1.3 and sles10
2642 Description: fix build with SLES10 and o2ib v3.
2643 Details : sles10 uses diffrent name for Module.symver file but configure
2644 assume this file has same name on RHEL/SLES/vanila kernels.
2647 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2649 Description: MDS may lose file striping (and hence file data) in some cases
2650 Details : If there are additional extended attributes stored on the MDS,
2651 in particular ACLs, SELinux, or user attributes (if user_xattr
2652 is specified for the client mount options) then there is a risk
2653 of attribute loss. Additionally, the Lustre file striping
2654 needs to be larger than default (e.g. striped over all OSTs),
2655 and an additional attribute must be stored initially in the
2656 inode and then increase in size enough to be moved to the
2657 external attribute block (e.g. ACL growing in size) for file
2660 Severity : enhancement
2662 Description: add message levels for liblustreapi
2667 Description: MDT cannot be unmounted, reporting "Mount still busy"
2668 Details : Mountpoint references were being leaked during open reply
2669 reconstruction after an MDS restart. Drop mountpoint reference
2670 in reconstruct_open() and free dentry reference also.
2675 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2676 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2677 will result in an -EAGAIN and in the exit call path we call
2678 alloc_rr() which will with increasing aggressiveness attempt to
2679 aquire precreated objects on the minimum number of required OSCs.
2683 Description: Use old size assignment to avoid deadlock
2684 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2685 the scheduling while holding a spinlock. We do not need locking
2686 for size in ll_update_inode() because size is only updated from
2687 the MDS for directories or files without objects, so there is no
2688 other place to do the update, and concurrent access to such inodes
2689 are protected by the inode lock.
2693 Description: Use __u64 instead of int for valid bits
2697 Description: resolve "_IOWR redefined" build error on SLES10
2701 Description: dump the memory debugging after all modules are unloaded to
2702 suppress false negative in conf_sanity test 39
2706 Description: the recovery timer never expires
2707 Details : for new client connect request, the recovery timer should not be
2708 reset, otherwise recovery timer will never expired, if the old
2709 client never come. Only old client connect and first connection
2710 req should trigger recovery timer reset.
2714 Description: the min numbers of lproc stats are wrong
2715 Details : adding a new constant LC_MIN_INIT and use it for initialization
2719 Frequency : always with interactive lfs
2721 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2724 Frequency : with multiple concurrent readdir processes in same directory
2725 Bugzilla : 15406, 15169, 15175
2726 Description: misc fixes for directory readahead.
2727 Details : prevent previous statahead async RPC callback from processing the
2728 current "statahead_info", race condition between async RPC callback
2729 add dentry into dentry hash table and "ls" thread revalidate such
2730 dentry, statahead his/miss control for hidden items, and so on.
2732 Severity : enhancement
2734 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2738 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2739 Details : several u32 variables declarations are replaced with u64 declarations
2744 Description: fix SLES kernel versioning
2745 Details : the kernel version for our SLES 10 kernel did not include a "-"
2746 before the "smp" at the end. while this was not a problem in
2747 general, it did mean that software trying to use the kernel
2748 version to try to detect a vendor specific kernel would fail.
2749 this was most evident by the OFED build scripts.
2754 Description: Don't update lov_desc members until making sure they are valid
2755 Details : When updating lov_desc members via proc fs, need fix their
2756 validities before doing the real update.
2759 Frequency : very rare
2761 Description: don't put request into delay list while invalidate in flight.
2762 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2763 import in flight. this produce timeout for invalidate and sometimes
2764 can cause stale data.
2766 Severity : enhancement
2768 Description: Update kernel to SLES9 2.6.5-7.311.
2770 Severity : enhancement
2772 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2777 Frequency : on PPC only
2778 Description: not convert ost objects for directory because it's not exist.
2779 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2781 Severity : enhancement
2783 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2784 __u64 as usingied long long type.
2787 Frequency : rare, on shutdown
2789 Description: race process ast vs remove callback
2790 Details : removing callback before disconnect import open race with
2791 processing callback.
2793 Severity : enhancement
2795 Description: Update kernel to SLES9 2.6.5-7.311.
2797 Severity : enhancement
2799 Description: Files open for execute are not marked busy on SLES10
2800 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2802 Severity : enhancement
2804 Description: Add server support for vanilla-2.6.22.14.
2807 Frequency : occasional
2809 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2810 Details : If an OST is down the MDS will hang indefinitely in
2811 obd_statfs_rqset() waiting for the statfs data. While for
2812 MDS QOS usage of statfs, it should not stuck in waiting.
2814 Severity : enhancement
2816 Description: Disable adaptive timeouts by default
2819 Frequency : on network error
2821 Description: panic with double free request if network error
2822 Details : mdc_finish_enqueue is finish request if any network error occuring,
2823 but it's true only for synchronus enqueue, for async enqueue
2824 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2828 Frequency : rare, on recovery
2830 Description: read procfs can produce deadlock in some situation
2831 Details : Holding lprocfs lock with send rpc can produce block for destroy
2832 obd objects and this also block reconnect with -EALREADY.
2833 This isn't fix all lprocfs bugs - but make it rare.
2835 Severity : enhancement
2837 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2840 Frequency : frequent on X2 node
2842 Description: mdc_set_open_replay_data LBUG
2843 Details : Set replay data for requests that are eligible for replay.
2848 Description: lustre_mgs: operation 101 on unconnected MGS
2849 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2850 MGC, and late on MGC cannot successfully connect to MGS and a lot
2851 of the error messages complaining that MGS is not connected.
2854 Frequency : rare, depends on device drivers and load
2856 Description: MDS or OSS nodes crash due to stack overflow
2857 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2858 In some cases, in conjunction with device drivers that use a lot
2859 of stack, the MDS (or possibly OSS) service threads could overflow
2860 the stack. One change which was identified to consume additional
2861 stack has been reworked to avoid the extra stack usage.
2863 Severity : enhancement
2865 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2867 Severity : enhancement
2869 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2871 Severity : enhancement
2873 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2875 Serverity : enhancement
2877 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2880 Frequency : rare on shutdown OST
2882 Description: Don't allow skipping OSTs if index has been specified.
2883 Details : Don't allow skipping OSTs if index has been specified, make locking
2884 in internal create lots better.
2889 Description: ASSERTION(!PageDirty(page)) failed
2890 Details : Wrong check could lead to an assertion failure under specific
2896 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2897 Details : access to bitfield in structure is always rounded to long
2898 and this produce problem with not atomic change any bit.
2903 Description: Lustre make rpms failed.
2904 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2905 builds Lustre rpms from tarball.
2907 Severity : enhancement
2909 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2912 Frequency : rare on shutdown OST
2914 Description: If llog cancel was not send before clean_exports phase, this can
2915 produce deadlock in llog code.
2916 Details : If llog thread has last reference to obd and call class_import_put
2917 this produce deadlock because llog_cleanup_commit_master wait when
2918 last llog_commit_thread exited, but this never success because was
2919 called from llog_commit_thread.
2922 Frequency : only if OST index is skipped
2924 Description: NULL lov_tgts causing MDS oops
2925 Details : more safe checks for NULL lov_tgts for avoid oops.
2927 Severity : enhancement
2929 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2931 Severity : enhancement
2933 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2938 Description: make mgs_setparam() handle fsname containing dash
2939 Details : fsname containing a dash does not work with lctl conf_param
2941 Severity : enhancement
2943 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2946 Frequency : rare, in recovery and (or) destroy lovobjid file.
2948 Description: rewrite lov objid code.
2949 Details : Cleanup for lov objid code, remove scability problems and wrong
2950 locking. Fix sending last_id into ost.
2952 Severity : enhancement
2954 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2956 Severity : enhancement
2958 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2959 Details : Use d_move_locked instead of __d_move.
2962 Frequency : rare, at shutdown
2964 Description: access already free / zero obd_namespace.
2965 Details : if client_disconnect_export was called without force flag set,
2966 and exist connect request in flight, this can produce access to
2967 NULL pointer (or already free pointer) when connect_interpret
2968 store ocd flags in obd_namespace.
2971 Frequency : only at startup
2973 Description: not alloc memory with spinlock held.
2974 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2975 if any spinlock held.
2980 Description: lfs find does not continue on file error
2981 Details : Continue other files processing when a file/dir is absent.
2985 Description: Inconsistent usage of lustre_pack_reply()
2986 Details : Standardize the usage of lustre_pack_reply() such that it
2987 always generate a CERROR on failure.
2990 Frequency : very rare
2992 Description: Fix replay if there is an un-replied request and open
2993 Details : In some cases, older replay request will revert the
2994 mcd->mcd_last_xid on MDS which is used to record the client's
2995 latest sent request.
2997 Severity : enhancement
2999 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
3001 Severity : enhancement
3003 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
3005 Severity : enhancement
3007 Description: Update to SLES9 kernel-2.6.5-7.287.3.
3009 Severity : enhancement
3011 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
3013 Severity : enhancement
3015 Description: Build SLES10 patchless client fails
3016 Details : The configure was broken by run ./configure with
3017 --with-linux-obj=.... argument for patchless client. When the
3018 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
3019 can't search header adequately. Use absolute path such as
3020 -I($LINUX)/include instead.
3022 Severity : enhancement
3024 Description: Lustre Page Accounting
3025 Details : New macros for page alloc and free which enable accounting
3026 of page allocation of Lustre. Use percpu counters to store memory
3027 and page statistics.
3030 Frequency : only if debugging is disabled
3032 Description: LASSERT_{REQ,REP}SWAB macros are buggy
3033 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
3034 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
3035 these macros and replace them with their definitions instead.
3040 Description: interrupt oig_wait produce painc on resend.
3041 Details : brw_redo_request can be used for resend requests from ptlrpcd and
3042 private set, and this produce situation when rq_ptlrpcd_data not
3043 copyed to new allocated request and triggered LBUG on assert
3044 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
3045 ptlrpcd set if request is changed and can be safety changed to use
3048 Severity : enhancement
3050 Description: organize the server-side client stats on per-nid basis
3051 Details : Change the structure of stats under obdfilter and mds to
3059 The "uuid"s file would list the uuids of _active_ exports.
3060 And the clear entry is to clear all stats and stale nids.
3065 Description: Processes looping in ll_readdir() on Lustre clients finally causing
3066 a full node pseudo-hang
3067 Details : Concurrent access to the same directory from multiple clients with
3068 intensive file creation/removal can cause a client node to spin in
3069 ll_readdir(). i_version must be increased every time the lock is
3070 cancelled to ensure a revalidate is done.
3075 Description: touch file failed when fs is not full
3076 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
3077 otherwise we can get ENOSP while fs is not full.
3080 Frequency : only for Cray XT3
3081 Bugzilla : 12829/13455
3082 Description: Changing primary group doesn't change the group lustre assigns to
3084 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
3085 overridden with the primary group provided by the group upcall,
3086 whereas the supplied fsgid can be trusted if it is in the list of
3087 supplementary groups returned by the group upcall.
3089 Severity : enhancement
3091 Description: Root Squash Functionality
3092 Details : Implementation of NFS-like root squash capability. Specifically,
3093 don't allow someone with root access on a client node to be able
3094 to manipulate files owned by root on a server node.
3096 Severity : enhancement
3098 Description: Slow trucate/writes to huge files at high offsets.
3099 Details : Directly associate cached pages to lock that protect those pages,
3100 this allows us to quickly find what pages to write and remove
3101 once lock callback is received.
3106 Description: Too many locks accumulating on client during NFS usage
3107 Details : mds_open improperly used accmode to find out access mode to a
3108 file. Also mdc_intent_lock logic to find out if we already have
3109 lock similar to just received was flawed since introduction of
3110 skiplists - locks are now added to the front of the granted
3115 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
3117 Details : In obd setup failure handler, obd_stopping will not necessarily to
3118 be 1, and obd_set_up should also be checked to make sure whether
3119 obd is completely setup.
3121 Severity : enhancement
3123 Description: Allow masking D_WARNING, D_ERROR messages from console
3124 Details : Console messages can now be disabled via lnet.printk.
3129 Description: User code with malformed file open parameter crashes client node
3130 Details : Before packing join_file req, all the related reference should be
3131 checked carefully in case some malformed flags cause fake join_file
3137 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
3138 grain of quota doing harm to quota's accuracy
3139 Details : qunit size will be changed when quota limitation is too low/high;
3140 record the pending quota write in order to get more accureate
3141 quota; delete the patch for bug12588, which is unnecessary when
3142 this patch is landed. This bug also contains fixes for bug 14526,
3143 14299, 14601 and 13794.
3147 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
3148 Details : ldlm_completion_ast() assumes that a lock is granted when the req
3149 mode is equal to the granted mode. However, it should also check
3150 that LDLM_FL_CP_REQD is not set.
3154 Description: Heavy nfs access might result in deadlocks
3155 Details : After ELC code landed, it is now improper to enqueue any mds
3156 locks under och_sem, because enqueue might want to decide to
3157 cancel open locks for same inode we are holding och_sem for.
3161 Description: 35% write performance drop with ldiskfs2 when quotas are on
3162 Details : Enable ext3 journalled quota by default to improve performance
3163 when quotas are turned on.
3167 Description: Client eviction while running blogbench
3168 Details : A lot of unlink operations with concurrent I/O can lead to a
3169 deadlock causing evictions. To address the problem, the number of
3170 oustanding OST_DESTROY requests is now throttled to
3171 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
3172 callbacks are processed in priority.
3175 Frequency : RHEL4 only
3177 Description: mkfs is very slow on IA64/RHEL4
3178 Details : A performance regression has been discovered in the MPT Fusion
3179 driver between versions 3.02.73rh and 3.02.99.00rh. As a
3180 consequence, we have downgraded the MPT Fusion driver in the RHEL4
3181 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
3184 Frequency : PPC/PPC64 only
3186 Description: conflicts between asm-ppc64/types.h and lustre_types.h
3187 Details : fix duplicated definitions between asm-ppc64/types.h and
3188 lustre_types.h on PPC.
3191 Frequency : PPC/PPC64 only
3193 Description: asm-ppc/segment.h does not exist
3194 Details : fix compile issue on PPC.
3198 Description: data checksumming impacts single node performance
3199 Details : add support for several checksum algorithms. Currently, CRC32 and
3200 Adler-32 are supported. The checksum type can be changed on the fly
3201 through /proc/fs/lustre/osc/*/checksum_type.
3205 Description: use adler32 for page checksums
3206 Details : when available, use the Adler-32 algorithm instead of CRC32 for
3211 Description: better handle error messages in extents code
3213 Severity : enhancement
3215 Description: SNMP support enhancement
3216 Details : Adding total number of sampled request for an MDS node in snmp
3219 Severity : enhancement
3221 Description: Optimize ldlm waiting list processing for PR extent locks
3222 Details : When processing waiting list for read extent lock and meeting read
3223 lock that is same or wider to it that is not contended, skip
3224 processing rest of the list and immediatelly return current
3225 status of conflictness, since we are guaranteed there are no
3226 conflicting locks in the rest of the list.
3230 Description: Time out and refuse to reconnect
3231 Details : When the failover node is the primary node, it is possible
3232 to have two identical connections in imp_conn_list. We must
3233 compare not conn's pointers but NIDs, otherwise we
3234 can defeat connection throttling.
3238 Description: Client not clear own cache if answer to reconnect is lost.
3239 Details : Client gets evicted from server. Now client also thinks it is
3240 disconnected (or gets ENOTCONN on its operation) and decides to
3241 reconnect. Server receives reconnect message, but cannot find
3242 export. New export is created that is fully valid (new cookie!),
3243 but reply is lost and not reported to client. Client reconnects
3244 again and gets back a just-created connection, but it is not new
3245 so client thinks it was not evicted and does not do recovery.
3249 Description: Detect stride IO mode in read-ahead
3250 Details : When a client does stride read, read-ahead should detect that and
3251 read-ahead pages according to the detected stride pattern.
3255 Description: build for x2 fails
3256 Details : fix compile issue on Cray systems.
3258 Severity : enhancement
3260 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
3261 Details : This greatly improves speed of NFS writes on 2.6 kernels.
3264 Frequency : only on PPC/SLES10
3266 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
3267 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
3268 be defined. Add a hack in mkfs_lustre.c to work around this compile
3273 Description: LASSERT on MDS when client holding flock lock dies
3274 Details : ldlm pool logic depends on number of granted locks equal to
3275 number of released locks which is not true for flock locks, so
3276 just exclude such locks from consideration.
3280 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
3281 Details : Use fsfilt_sync() for both the whole filesystem sync and
3282 individual file sync to eliminate dangerous inode locking
3283 with I_LOCK that can lead to a deadlock.
3287 Description: Update an obsolete wirecheck.c generator
3288 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
3292 Description: Client can panic on open sometimes
3293 Details : It is possible that we try to free already freed request in
3294 ll_file_open in some error cases when we send request from
3299 Description: performance in 1.6.3
3300 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3301 sure that 1MB requests are not fragmented by the block layer.
3305 Description: LDLM soft lockups - improvement
3306 Details : It is be possible to send the lock handle along with each read
3307 or write request because the client is already doing a lock match
3308 itself so there isn't any reason the OST should have to re-do that
3314 Description: lfs quota fails with deactivated OSTS
3315 Details : With this patch, three improvements are included:
3316 1. detete the softlimit in mds and osts when use "lfs quota".
3317 2. display the inaccurate data in the output of "lfs quota".
3318 3. try to get quota info when "lfs quota" is executed.
3323 Description: Extent locks not granted with no conflicts sometimes.
3324 Details : When race occurs in glimpse handler and nothing is returned,
3325 we do not reprocess the queue after lock cancel, and that leads
3326 to a stall until next activity on a resource
3329 Frequency : failover with quotaon
3331 Description: during mds failovers with quota on, OSTs got into deadlock state
3332 and causing dumpstack.
3333 Details : for every quota slave, at any time, there is only one quota req
3334 is sent to quota master for every uid/gid. Before that quota req
3335 returns, all the thread relative to the same uid/gid will wait.
3336 So if the quota req is lost because mds failovers or any other
3337 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3338 quota reqs who time out.
3340 Severity : enhancement
3343 Description: when quota slave checks if quota is enough, there is an unnecessary
3345 Details : place this wait on necessary place instead of always waiting.
3347 --------------------------------------------------------------------------------
3349 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3351 * Support for kernels:
3352 2.6.5-7.286 (SLES 9),
3353 2.6.9-55.0.9.EL (RHEL 4),
3354 2.6.16.53-0.8 (SLES 10),
3355 2.6.18-8.1.14.el5 (RHEL 5),
3356 2.6.18.8 vanilla (kernel.org)
3357 * Client support for unpatched kernels:
3358 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3359 2.6.16 - 2.6.22 vanilla (kernel.org)
3360 * Due to recently discovered recovery problems, we do not recommend
3361 using patchless RHEL 4 clients with this or any earlier release.
3362 * Recommended e2fsprogs version: 1.40.2-cfs1
3363 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3366 Frequency : occasional
3368 Description: MDS hang or stay in waiting lock
3369 Details : If client receive lock with CBPENDING flag ldlm need send lock
3370 cancel as separate rpc, to avoid situation when cancel request
3371 can't processed due all i/o threads stay in wait lock.
3374 Frequency : occasional
3376 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3377 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3378 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3379 oscc_has_objects and see OSC has no objests and call
3380 oscc_internal_create to resend create request.
3382 Severity : enhancement
3384 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3385 Details : Update which_patch & target file for SLES10 latest kernel.
3387 Severity : enhancement
3389 Description: add --type and --size parameters to lfs find
3390 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3391 multiple OBDs can now be specified for the --obd option.
3393 Severity : enhancement
3395 Description: eliminate client locks in face of contention
3396 Details : file contention detection and lockless i/o implementation
3397 for contended files.
3399 Severity : enhancement
3401 Description: Remove client patches from SLES 10 kernel.
3402 Details : This causes SLES 10 clients to behave as patchless clients
3403 even on a Lustre-patched (server) kernel.
3405 Severity : enhancement
3407 Description: use i_size_read and i_size_write in 2.6 port
3408 Details : replace inode->i_size access with i_size_read/write()
3411 Frequency : when removing large files
3413 Description: scheduling issue during removal of large Lustre files
3414 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3415 It causes scheduling issues when removing large files (17TB in the
3421 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3422 Details : Impossible (EPROTO is returned) to access a directory that has a
3423 non-default striping and ACLs.
3426 Frequency : only on ppc
3428 Description: /proc/fs/lustre/devices broken on ppc
3429 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3430 We should make sure the type of 'index' is loff_t and then cast
3431 explicitly as needed below. Do not assign an explicitly cast
3435 Frequency : only for rhel5
3437 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3438 Details : Modify the target file & which_kernel.
3441 Frequency : if the uninit_groups feature is enabled on ldiskfs
3443 Description: e2fsck reports "invalid unused inodes count"
3444 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3445 feature and only a single inode is created in a group then the
3446 "bg_unused_inodes" count is incorrectly updated. Creating a
3447 second inode in that group would update it correctly.
3452 Description: buffer overruns could theoretically occur
3453 Details : llapi_semantic_traverse() modifies the "path" argument by
3454 appending values to the end of the origin string, and a buffer
3455 overrun may occur. Adding buffer overrun check in liblustreapi.
3457 Severity : enhancement
3459 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3462 Frequency : only if filesystem is inconsistent
3464 Description: handle "serious error: objid * already exists" more gracefully
3465 Details : If LAST_ID value on disk is smaller than the objects existing in
3466 the O/0/d* directories, it indicates disk corruption and causes an
3467 LBUG(). If the object is 0-length, then we should use the existing
3468 object. This will help to avoid a full fsck in most cases.
3470 Severity : enhancement
3472 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3473 Details : Modify vm-tunables-rhel4.patch.
3475 Severity : enhancement
3477 Description: Kernel config for 2.6.18-vanilla.
3478 Details : Modify targets/2.6-vanilla.target.in.
3479 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3480 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3481 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3482 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3485 Frequency : occasional
3487 Description: improve handling recoverable errors
3488 Details : If request processed with error which can be recoverable on server
3489 request should be resend, otherwise page released from cache and
3495 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3496 Details : Modify target file & which_patch.
3497 A flaw was found in the IA32 system call emulation provided
3498 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3499 value could be stored in the %RAX register, which could trigger an
3500 out-of-bounds system call table access. An untrusted local user
3501 could exploit this flaw to run code in the kernel
3502 (ie a root privilege escalation). (CVE-2007-4573).
3506 Description: change order of libsysio includes
3507 Details : '#include sysio.h' should always come before '#include xtio.h'
3509 Severity : enhancement
3511 Description: adapt the lustre_config script to support the upgrade case
3512 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3513 server targets to 1.6 in parallel.
3518 Description: To avoid grant space > avaible space when the disk is almost
3519 full. Without this patch you might see the error "grant XXXX >
3520 available" or some LBUG about grant, when the disk is almost
3522 Details : In filter_check_grant, for non_grant cache write, we should
3523 check the left space by if (*left > ungranted + bytes), instead
3524 of (*left > ungranted), because only we are sure the left space
3525 is enough for another "bytes", then the ungrant space should be
3526 increase. In client, we should update cl_avail_grant only there is
3527 OBD_MD_FLGRANT in the reply.
3532 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3533 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3534 described in CVE-2007-4573. This problem could allow untrusted
3535 local users to gain root access.
3538 Frequency : when using O_DIRECT and quotas
3540 Description: Incorrect file ownership on O_DIRECT output files
3541 Details : block usage reported by 'lfs quota' does not take into account
3542 files that have been written with O_DIRECT.
3546 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3547 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3548 window to 0. If it does hit cache miss, and read-pattern is still
3549 stride-io mode, does not reset the stride window, but also does not
3550 increase the stride window length in this case.
3552 --------------------------------------------------------------------------------
3554 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3556 * Support for kernels:
3557 2.6.5-7.286 (SLES 9),
3558 2.6.9-55.0.2.EL (RHEL 4),
3559 2.6.16.46-0.14 (SLES 10),
3560 2.6.18-8.1.8.el5 (RHEL 5),
3561 2.6.18.8 vanilla (kernel.org)
3562 * Client support for unpatched kernels:
3563 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3564 2.6.16 - 2.6.21 vanilla (kernel.org)
3565 * Due to recently discovered recovery problems, we do not recommend
3566 using patchless RHEL 4 clients with this or any earlier release.
3567 * Recommended e2fsprogs version: 1.40.2-cfs1
3568 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3572 Description: Fix errors in lfs documentation
3573 Details : Fixes man pages
3575 Severity : enhancement
3577 Description: Adaptive timeouts
3578 Details : RPC timeouts adapt to changing server load and network
3579 conditions to reduce resend attempts and improve recovery time.
3581 Severity : enhancement
3583 Description: llapi_file_create() does not allow some changes
3584 Details : add llapi_file_open() that allows specifying the file creation
3585 mode and open flags, and also returns an open file handle.
3587 Severity : enhancement
3589 Description: df doesn't work properly if diskfs blocksize != 4K
3590 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3593 Frequency : always on directories with default striping set
3595 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3596 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3598 Severity : enhancement
3600 Description: merge and cleanup kernel patches.
3601 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3603 Severity : enhancement
3605 Description: RedHat Update kernel for RHEL5
3606 Details : Kernel config file for RHEL5.
3608 Severity : enhancement
3610 Description: OSS needs mutliple precreate threads
3611 Details : Add ability to start more than one create thread per OSS.
3614 Frequency : only with quota on the root user
3616 Description: mds_obd_create error creating tmp object
3617 Details : When the user sets quota on root, llog will be affected and can't
3618 create files and write files.
3621 Frequency : Always on ia64 patchless client, and possibly others.
3623 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3624 Details : This allows the patchless client to be loaded on architectures
3625 without this export.
3630 Description: disorder of journal start and llog_add cause deadlock.
3631 Details : in llog_origin_connect, journal start should happen before
3632 llog_add keep the same order as other functions to avoid
3635 Severity : enhancement
3637 Description: RedHat Update kernel for RHEL5
3638 Details : Modify the kernel config file more closer RHEL5.
3640 Severity : enhancement
3642 Description: Build failure against Centos5 (RHEL5)
3643 Details : Define PAGE_SIZE when it isn't present.
3646 Frequency : occasionally when using NFS
3648 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3649 Details : with NFS, the anon dentry's parent was set to itself in
3650 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3651 corresponding dentry other than use rec->ur_name.
3653 Severity : enhancement
3655 Description: enable data checksumming by default
3656 Details : enable checksum by default, allow --disable-checksum
3657 configure option and "-o nochecksum" mount option. Checksums
3658 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3661 Frequency : Occasionally with failover
3663 Description: Client eviction due to failover config
3664 Details : after a connection loss, the lustre client should attempt to
3665 reconnect to the last active server first before trying the
3666 other potential connections.
3668 Severity : enhancement
3670 Description: client-side metadata stat-ahead during readdir(directory readahead)
3671 Details : perform client-side metadata stat-ahead when the client detects
3672 readdir and sequential stat of dir entries therein
3677 Description: when mds and osts use different quota unit(32bit and 64bit),
3678 quota will be released repeatly.
3679 Details : void sending multiple quota reqs to mds, which will keep the
3680 status between the reqs.
3683 Frequency : only with liblustre clients on XT3
3685 Description: evictions taking too long
3686 Details : allow llrd to evict clients directly on OSTs
3690 Description: osts not allocated evenly to files
3691 Details : change the condition to increase offset_idx
3693 Severity : enhancement
3695 Description: self-adjustable client's lru lists
3696 Details : use adaptive algorithm for managing client cached locks lru
3697 lists according to current server load, other client's work
3698 pattern, memory activities, etc. Both, server and client
3699 side namespaces provide number of proc tunables for controlling
3704 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3705 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3706 readdir code from ext3) from scratch preserving functionality.
3710 Description: Only those disconnect error should be returned by rq_status.
3711 Details : In open/enqueue processs, Some errors, which will cause client
3712 disconnected, should be returned by rq_status, while other
3713 errors should still be returned by intent, then mdc or llite will
3716 Severity : enhancement
3718 Description: Tune the kernel for good SCSI performance.
3719 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3720 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3724 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3725 Bugzilla : 13547, 13627
3726 Description: Data corruption for OSTs that are formatted larger than 2TB
3728 Details : When generating the bio request for lustre file writes the
3729 sector number would overflow a temporary variable before being
3730 used for the IO. The data reads correctly from Lustre (which
3731 will overflow in a similar manner) but other file data or
3732 filesystem metadata may be corrupted in some cases.
3736 Description: "lfs find -obd UUID" prints directories
3737 Details : "lfs find -obd UUID" will return all directory names instead
3738 of just file names. It is incorrect because the directories
3739 do not reside on the OSTs.
3742 Bugzilla : 13596, 13614
3743 Description: MDS hang after unclean shutdown of lots of clients
3744 Details : Never resend AST requests.
3748 Frequency : Always, for kernels after 2.6.16
3749 Description: Fix warning idr_remove called for id=.. which is not allocated.
3750 Details : Last kernels save old s_dev before kill super and not allow
3751 to restore from callback - restore it before call kill_anon_super.
3755 Description: Fix error on 'ls .' at the top of the Lustre mount.
3756 Details : Don't revalidate dentry if it is a root dentry.
3758 --------------------------------------------------------------------------------
3760 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3762 * Support for kernels:
3763 2.6.5-7.286 (SLES 9),
3764 2.6.9-55.0.2.EL (RHEL 4),
3765 2.6.16.46-0.14 (SLES 10),
3766 2.6.18-8.1.8.el5 (RHEL 5),
3767 2.6.18.8 vanilla (kernel.org)
3768 * Client support for unpatched kernels:
3769 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3770 2.6.16 - 2.6.22 vanilla (kernel.org)
3771 * Due to recently discovered recovery problems, we do not recommend
3772 using patchless RHEL 4 clients with this or any earlier release.
3773 * Recommended e2fsprogs version: 1.39.cfs8
3774 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3779 Description: block reactivating mgc import until all deactivates complete
3780 Details : Fix race when failing back MDT/MGS to itself (testing)
3782 Severity : enhancement
3784 Description: lfs setstripe enhancement
3785 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3788 Frequency : mds/oss recovery
3790 Description: llog ctxt is refrenced after it has been freed.
3791 Details : llog ctxt refcount was added to avoide the race between ctxt free
3792 and llog recovery process. Each llog user must hold ctxt refcount
3793 before it access the llog. And the llog ctxt can only be freed
3794 when its refcount is zero.
3796 Severity : enhancement
3798 Description: randomly memory allocation failure util
3799 Details : Make lustre randomly failed allocating memory for testing purpose.
3802 Frequency : only for SLES10
3804 Description: Update kernel patch for SLES10 SP1
3805 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3807 Severity : enhancement
3809 Description: omit set fsid for export NFS
3810 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3811 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3814 Frequency : after network failures
3816 Description: Add sync option to mount_lustre.c
3817 Details : Client loses data written to lustre after a network interruption.
3819 Severity : enhancement
3821 Description: Error message improvement.
3822 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3824 Severity : enhancement
3826 Description: don't use GFP_* in generic Lustre code.
3827 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3829 Severity : enhancement
3831 Description: obdclass is limited by single OBD_ALLOC(idarray)
3832 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3834 Severity : enhancement
3836 Description: updated patchess for new RHEL4 kernel
3837 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3839 Severity : enhancement
3841 Description: warnings with build patchless client with vanila 2.6.19 and up
3842 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3843 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3845 Severity : enhancement
3847 Description: O_DIRECT bypasses client statistics.
3848 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3849 read_ahead_stats were updated. Copied stats section from
3850 osc_send_oap_rpc() into async_internal().
3853 Frequency : only for Cray XT3
3855 Description: peer credits not enough on many OST per OSS systems.
3856 Details : Use new lnet way to add credits as we need those for pings and ASTs
3859 Frequency : only with liblustre
3861 Description: Liblustre is not releasing flock locks on file close.
3862 Details : Release flock locks on file close.
3865 Frequency : only for RHEL4
3867 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3868 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3869 Update target file and kernel config.
3873 Description: lustre support for RHEL5
3874 Details : Add support for RHEL5.
3879 Description: ASSERTION(export != NULL) failed in target_handle_connect
3880 Details : Assetion hit is result of rare race between disconnect and connet
3881 to same nid. target_handle_connect found old connect cockie and
3882 tried to reconnect, but can't find export for this cockie.
3887 Description: umount blocks forever on error
3888 Details : In result of wrong using obd_no_recov and obd_force flags client
3889 can hand if cancel or some other requests is lost.
3892 Frequency : Only for SLES
3894 Description: sanity_quota fail test_1
3895 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3896 file, which makes TSTID[2] inunique.
3898 Severity : enhancement
3900 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3901 Details : Update target/ChangeLog/which_patch .
3903 Severity : enhancement
3905 Description: jbd statistics
3906 Details : Port older jbd statistics patch for sles10
3911 Description: lvbo_init failed for resource with missing objects.
3912 Details : Fix returning error if we do stat for file with missing/corrupted
3913 objects and i_size set to all sum of size of all avaible objects.
3914 if we to truncate/write to missing object - it is recreated.
3919 Description: Oops in read and write path when failing to allocate lock.
3920 Details : Check if lock allocation failed and return error back.
3923 Frequency : When flocks are used.
3925 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3926 Details : Flock locks might destroy just granted lock if it could be merged
3927 with another existing flock, this is done in completion handler,
3928 so teach ldlm_cli_enquque_fini that this is a valid case for
3934 Description: reply_lock_interpret crash due to race with it and lock cancel.
3935 Details : Do not replay locks that are being cancelled. Do not reference
3936 locks by their address during replay, just by their handle.
3939 Frequency : only with deactivated OSTs
3941 Description: lstripe command fails for valid OST index
3942 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3943 instead of lov->desc.ld_active_tgt_count.
3945 Severity : enhancement
3947 Description: Build failure against Centos5 (RHEL5)
3948 Details : Use getpagesize() instead of PAGE_SIZE.
3950 --------------------------------------------------------------------------------
3952 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3954 * Support for kernels:
3955 2.6.5-7.283 (SLES 9),
3956 2.6.9-55.EL (RHEL 4),
3957 2.6.16.46-0.14 (SLES 10),
3958 2.6.18.8 vanilla (kernel.org)
3959 * Client support for unpatched kernels:
3960 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3961 2.6.16 - 2.6.22 vanilla (kernel.org)
3962 * Due to recently discovered recovery problems, we do not recommend
3963 using patchless RHEL 4 clients with this or any earlier release.
3964 * Recommended e2fsprogs version: 1.39.cfs8
3965 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3966 * Starting with this release, the ldiskfs backing filesystem required
3967 by Lustre is now in its own package, lustre-ldiskfs. This package
3968 should be installed. It is versioned separately from Lustre and
3969 may be released separately in future.
3974 Description: server LBUG when shutting down
3975 Details : Block umount forever until the mount refcount is zero rather
3976 than giving up after an arbitrary timeout.
3978 Severity : enhancement
3980 Description: add optional extra BUILD_VERSION info
3981 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3982 to override the lustre version.
3985 Frequency : 2.6.18 servers only
3987 Description: ll_kern_mount() doesn't release the module reference
3988 Details : The ldiskfs module reference count never drops down to 0
3989 because ll_kern_mount() doesn't release the module reference.
3994 Description: server LBUG when using old ost_num_threads parameter
3995 Details : Accept the old ost_num_threads parameter but warn that it
3996 is deprecated, and fix an off-by-one error that caused an LBUG.
4001 Description: Transient SCSI error results in persistent IO issue
4002 Details : iobuf->dr_error is not reinitialized to 0 between two
4006 Frequency : sometimes when underlying device returns I/O errors
4008 Description: OSTs not going read-only during write failures
4009 Details : OSTs are not remounted read-only when the journal commit threads
4010 get I/O errors because fsfilt_ext3 calls journal_start/stop()
4011 instead of the ext3 wrappers.
4014 Frequency : only with 10000 clients or more
4016 Description: poor connect scaling with increasing client count
4017 Details : Don't run filter_grant_sanity_check for more than 100 exports
4018 to improve scaling for large numbers of clients.
4021 Frequency : SLES10 only
4023 Description: sanity-quota.sh quotacheck failed: rc = -22
4024 Details : Quotas cannot be enabled on SLES10.
4027 Frequency : liblustre clients only
4029 Description: getdirentries does not give error when run on compute nodes
4030 Details : getdirentries does not fail when the size specified as an argument
4031 is too small to contain at least one entry
4033 Severity : enhancement
4035 Description: Add LNET router traceability for debug purposes
4036 Details : If a checksum failure occurs with a router as part of the
4037 IO path, the NID of the last router that forwarded the bulk data
4038 is printed so it can be identified.
4043 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
4044 Details : Due to a race condition, liblustre clients were occasionally
4045 evicted incorrectly.
4047 Severity : enhancement
4049 Description: lfs setstripe use optional parameters instead of postional
4052 Severity : enhancement
4054 Description: Nanosecond timestamp support for ldiskfs
4055 Details : The on-disk ldiskfs filesystem has added support for nanosecond
4056 resolution timestamps. There is not yet support for this at
4057 the Lustre filesystem level.
4060 Frequency : during server recovery
4062 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
4063 Details : request with rq_no_resend flag not awake l_wait_event if they get a
4067 Frequency : nfs export on patchless client
4069 Description: connectathon hang when test nfs export over patchless client
4070 Details : Disconnected dentry cannot be found with lookup, so we do not need
4071 to unhash it or make it invalid
4074 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
4075 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
4076 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
4081 Description: open req refcounting wrong on reconnect
4082 Details : If reconnect happened between getting open reply from server and
4083 call to mdc_set_replay_data in ll_file_open, we will schedule
4084 replay for unreferenced request that we are about to free.
4085 Subsequent close will crash in variety of ways.
4086 Check that request is still eligible for replay in
4087 mdc_set_replay_data().
4092 Description: disable writes to filesystem when reading health_check file
4093 Details : the default for reading the health_check proc file has changed
4094 to NOT do a journal transaction and write to disk, because this
4095 can cause reads of the /proc file to hang and block HA state
4096 checking on a healthy but otherwise heavily loaded system. It
4097 is possible to return to the previous behaviour during configure
4098 with --enable-health-write.
4100 Severity : enhancement
4102 Description: 64-bit inode version
4103 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
4104 to the inode. This will be required for version-based recovery.
4109 Description: MDS fails to start if a duplicate client export is detected
4110 Details : in some rare cases it was possible for a client to connect to
4111 an MDS multiple times. Upon recovery the MDS would detect this
4112 and fail during startup. Handle this more gracefully.
4114 Severity : enhancement
4116 Description: Add -o localflock option to simulate old noflock behaviour.
4117 Details : This will achieve local-only flock/fcntl locks coherentness.
4122 Description: log_commit_thread vs filter_destroy race leads to crash
4123 Details : Take import reference before releasing llog record semaphore
4128 Description: Wrong request locking in request set processing
4129 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
4130 imp_delayed_list, in this place should be used imp_lock.
4133 Frequency : when reconnecting
4135 Description: Grant leak when OSC reconnect to OST
4136 Details : When osc reconnect ost, OST(filter) should check whether it
4137 should grant more space to client by comparing fed_grant and
4138 cl_avail_grant, and return the granted space to client instead
4139 of "new granted" space, because client will call osc_init_grant
4140 to update the client grant space info.
4143 Frequency : when client reconnects to OST
4145 Description: Grant leak when OSC does a resend and replays bulk write
4146 Details : When osc reconnect to OST, OST(filter) should clear grant info of
4147 bulk write request, because the grant info while be sync between
4148 OSC and OST when reconnect, and we should ignore the grant info
4149 these of resend/replay write req.
4154 Description: Grant space more than avaiable space sometimes.
4155 Details : When then OST is about to be full, if two bulk writing from
4156 different clients came to OST. Accord the avaliable space of the
4157 OST, the first req should be permitted, and the second one
4158 should be denied by ENOSPC. But if the seconde arrived before
4159 the first one is commited. The OST might wrongly permit second
4160 writing, which will cause grant space > avaiable space.
4163 Frequency : when client is evicted
4165 Description: Grant might be wrongly erased when osc is evicted by OST
4166 Details : when the import is evicted by server, it will fork another
4167 thread ptlrpc_invalidate_import_thread to invalidate the
4168 import, where the grant will be set to 0. While the original
4169 thread will update the grant it got when connecting. So if
4170 the former happened latter, the grant will be wrongly errased
4171 because of this race.
4176 Description: Checking Stale with correct fid
4177 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
4178 to check whether it is stale, because sometimes, we want the
4179 enqueue happened anyway, and op_data.fid2 will not be initialized.
4181 Severity : enhancement
4183 Description: update patchless client
4184 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
4187 Frequency : only with 2.4 kernel
4189 Description: random memory corruption
4190 Details : size of struct ll_inode_info is to big for union inode.u and this
4191 can be cause of random memory corruption.
4196 Description: Memory leak in recovery
4197 Details : Lov_mds_md was not free in an error handler in mds_create_object.
4198 It should also check obd_fail before fsfilt_start, otherwise if
4199 fsfilt_start return -EROFS,(failover mds during mds recovery).
4200 then the req will return with repmsg->transno = 0 and rc = EROFS.
4201 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
4202 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
4203 be freed no matter whether fsfilt_commit success or not.
4206 Frequency : only with large numbers of cores on a single node
4208 Description: Prevents from taking the superblock lock in llap_from_page for
4210 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
4211 for llap_from_page call in ll_removepage() prevents from taking
4212 the superblock lock for a soon died page.
4217 Description: Not check open intent error before release open handle
4218 Details : in some rare cases, the open intent error is not checked before
4219 release open handle, which may cause
4220 ASSERTION(open_req->rq_transno != 0), because it tries to release
4221 the failed open handle.
4226 Description: Set cat log bitmap only after create log success.
4227 Details : in some rare cases, the cat log bitmap is set too early. and it
4228 should be set only after create log success.
4232 Description: Accessing a block bevice can re-enable I/O when Lustre is
4233 tearing down a device.
4234 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
4239 Description: service threads may hog cpus when there are a lot of requests
4240 Details : Insert cond_resched to give other threads a chance to use some CPU
4245 Description: the cat log was not initialized in recovery
4246 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
4247 unlink log on mds will not be initialized until mds post
4248 recovery. And also in mds post recovery, the unlink log will
4249 initialization will be done asynchronausly, so there will be race
4250 between add unlink log and unlink log initialization.
4254 Description: brw_stats were being printed incorrectly
4255 Details : brw_stats were being printed as log2 but all of them were not
4256 recorded as log2. Also remove some code duplication arising from
4257 filter_tally_{read,write}.
4261 Frequency : rare, only in recovery.
4262 Description: ASSERTION(req->rq_type != LI_POISON) failed
4263 Details : imp_lock should be held while iterating over imp_sending_list for
4264 prevent destroy request after get timeout in ptlrpc_queue_wait.
4268 Description: replay-single.sh test 52 fails
4269 Details : A lock's skiplist need to be cleanup when it being unlinked
4270 from its resource list.
4275 Description: Short directio read returns full requested size rather than
4277 Details : Direct I/O operations should return actual amount of bytes
4278 transferred rather than requested size.
4280 Severity : enhancement
4282 Description: metadata RPC reduction (e.g. for rm performance)
4283 Details : decrease the amount of synchronous RPC between clients and servers
4284 by canceling conflicing lock before the operation on the client
4285 and packing thier handles into the main operation RPC to server.
4287 Severity : enhancement
4289 Description: add #ifdef HAVE_KERNEL_CONFIG_H
4290 Details : kernels from 2.6.19 not need include linux/config.h, but add
4291 include linux/autoconf.h in commpiler command line.
4293 Severity : enhancement
4295 Description: patchless client support for 2.6.22 kernel
4296 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
4297 removed. In this case we need drop using os depended interface to
4298 kmem_cache and use cfs_mem_cache API.
4303 Description: fix mal-formatted messages
4304 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4308 Frequency : always in liblustre
4309 Description: wrong IS_ERR implementation in liblustre.h
4310 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4315 Description: Correct condition for output debug message.
4316 Details : inode i_nlink equal zero is not enough for output message about
4317 disk corruption, i_ctime and i_mode should be also checked.
4321 Frequency : always in patchless client
4322 Description: add configure check for truncate_complete_page
4323 Details : improve checks for exported symbols. This allow run check without
4324 sources, but with Module.symvers shipped with kernel distribution.
4325 add check for truncate_complete_page used by patchless client.
4329 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4330 Details : test_77h uses a file which was messed by other test case.
4334 Description: Not Check whether lov_tgts is NULL in some lov functions
4335 Details : Checking whether lov_tgts is NULL in some functions.
4339 Description: replace obdo_alloc() with OBDO_ALLOC macro
4340 Details : nothing special is done in obdo_alloc() function, and for
4341 debugging purpose, it needs to be replaced with macros.
4345 Description: bad return value and errno from fcntl call
4346 Details : In liblustre API, errno should be a negative value if error
4351 Description: ptlrpc_check_set() LBUG
4352 Details : In case of positive reply from server and failed client bulk
4353 callback after bulk transfer shouldn't LBUG, but process this
4354 request as erroneous.
4356 Severity : enhancement
4358 Description: VFS operations stats tool.
4359 Details : Tool which collects stats by tracking value written in pid,
4360 ppid, gid and uses llstat to generate output to plot graph using
4362 Updated lustre/utils/Makefile.am
4363 Added lustre/utils/ltrack_stats.c
4365 Severity : enhancement
4367 Description: 2.6.18 server support (lustre 1.6.1)
4368 Details : Support for 2.6.18 kernels on the server side.
4373 Description: ASSERTION(imp->imp_conn_current) failed
4374 Details : an assertion failure is hit if a client node boots and attempts to
4375 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4378 Frequency : only for i686
4380 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4381 Details : Fixed config variable for build.
4386 Description: Updated patchess for new RHEL4 kernel
4387 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4388 Updated patch jbd-stats-2.6.9.patch
4389 Updated patch qsnet-rhel4-2.6.patch
4390 Updated patch quota-deadlock-on-pagelock-core.patch
4391 Updated patch vfs_intent-2.6-rhel4.patch
4392 Updated patch vfs_races-2.6-rhel4.patch
4393 Updated series file 2.6-rhel4-titech.series
4394 Updated series file 2.6-rhel4.series
4395 Updated kernel config files
4400 Description: lquota slave complains LBUG when reconnecting with mds
4402 Details : quota slave depends qctxt->lqc_import to send its quota request.
4403 This pointer will be invalid if mds did failover or broke its
4404 connect to osts, which leads to LBUG.
4407 Frequency : when qunit size is too small(less than 20M)
4409 Description: write is stopped by improper -EDQUOT
4410 Details : If the master is busy and qunit size is small enough(let's say 1M),
4411 the slave can not get quota from master on time, which will lead
4412 slave to trigger a -EQUOTA to client.
4417 Description: Deadlock during metadata tests
4418 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4419 called with the per-dentry lock held.
4422 Frequency : SLES9 only
4424 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4425 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4427 Severity : enhancement
4429 Description: remove fs_prep_san_write operation and related patches
4430 Details : remove the ext3-san-jdike patches which are no longer useful.
4435 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4436 Details : a disk corruption can cause the mballoc code to assert on a
4437 double free or other extent corruptions. Handle these with
4438 ext3_error() instead of assertions.
4441 Frequency : only with mballoc3 code and deep extent trees
4443 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4444 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4445 can cause the filesystem to be remounted read-only.
4450 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4451 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4452 few external users of this ioctl number is lfsck's call to
4453 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4454 Add the old ioctl number to the handler so both old and new
4459 Description: parallel lock callbacks
4460 Details : Instead of sending blocking and completion callbacks as separated
4461 requests, adding them to a set and sending in parallel.
4465 Description: Disable most debugging by default
4466 Details : To improve performance, disable most logging (for debug purposes)
4467 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4468 default, and HA includes fewer messages.
4472 Frequency : only run on patchless client.
4473 Description: use do_facet on sanity.sh for test handling recoverables errors
4474 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4478 Description: hash tables for lists of nids, connections and uuids
4479 Details : Hash tables noticeably help when a lot of clients connect to a
4480 server, to faster identify duplicate connections or reconnects,
4481 also to faster find export to evict in manual eviction case.
4485 Description: Sometimes, when the server evict a client, and the client will
4486 not be evicted as soon as possible.
4487 Details : In enqueue req, the error was returned by intent, instead of
4488 rq_status which make ptlrpc layer not detect this error, and
4489 does not evict the client. So enqueue error should be returned
4493 Frequency : only at startup
4495 Description: Delay client connections to MDT until fist MDT->OST connect
4496 Details : If a client tried to create a new file before the MDT had
4497 connected to any OSTs, the create would return EIO. Now
4498 the client will simply block until the MDT connects to the
4499 first OST and the create can succeed.
4502 Frequency : only for SLES9
4504 Description: Routinely utilize latest Quadrics drivers in CFS releases
4505 Details : Update patch qsnet-suse-2.6.patch.
4508 Frequency : only for sles10
4510 Description: Update patches for SLES 10 SP1 kernel.
4511 Details : Update the patch vfs_intent-2.6-sles10.patch.
4516 Description: Routinely utilize latest Quadrics drivers in CFS releases
4517 Details : Update patch qsnet-rhel4-2.6.patch.
4520 Frequency : at statup only
4522 Description: mds_lov_synchronize race leads to various problems
4523 Details : simultaneous MDT->OST connections at startup can cause the
4524 sync to abort, leaving the OSC in a bad state.
4529 Description: Using wrong pointer in osc_brw_prep_request
4530 Details : Access to array[-1] can produce panic if kernel compiled with
4531 CONFIG_PAGE_ALLOC enabled
4533 Severity : enhancement
4535 Description: Async OSC create to avoid the blocking unnecessarily.
4536 Details : If a OST has no remain object, system will block on the creating
4537 when need to create a new object on this OST. Now, ways use
4538 pre-created objects when available, instead of blocking on an
4539 empty osc while others are not empty. If we must block, we block
4540 for the shortest possible period of time.
4544 Frequency : only in recovery
4545 Description: Mark OST as early accessible if his start SYNC.
4546 Details : osc_precreate return flag early accessible if oscc marked as
4547 OSCC_FLAG_SYNC_IN_PROGRESS.
4552 Description: Sometimes precreate code can triger create object on wrong ost
4553 Details : Wrong protected or not not restored variables aftre precreate loop
4554 can produce creation object on wrong ost.
4557 Frequency : oss recovery
4559 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4561 Details : llog_commit_thread_count should be synced between llog_commit
4562 start and cleanup, so new llog_commit thread should not be started
4563 when llog_commit threads being stopped to avoid accessing some
4566 Severity : enhancement
4568 Description: Add printing inode info into message about error in writepage.
4570 --------------------------------------------------------------------------------
4572 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4577 Frequency : on some architectures
4579 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4580 Details : Uninitialized flags sometimes cause configuration commands to
4584 Frequency : patchless clients only
4586 Description: missing __iget() symbol export
4587 Details : The __iget() symbol export is missing. To avoid the need for
4588 this on patchless clients the deathrow inode reaper is turned
4589 off, and we depend on the VM to clean up old inodes. This
4590 dependency was during via the fix for bug 12181.
4595 Description: sanity.sh fail: test_52b
4596 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4597 incorrect inode's flags to client.
4599 --------------------------------------------------------------------------------
4601 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4603 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4604 INTEROPERATE with older versions automatically. In many cases a
4605 special upgrade step is needed. Please read the
4606 user documentation before upgrading any part of a 1.4.x system.
4607 * WARNING: Lustre configuration and startup changes are required with
4608 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4610 * Support for kernels:
4611 2.4.21-47.0.1.EL (RHEL 3),
4612 2.6.5-7.283 (SLES 9),
4613 2.6.9-42.0.10.EL (RHEL 4),
4614 2.6.12.6 vanilla (kernel.org),
4615 2.6.16.27-0.9 (SLES10)
4616 * Client support for unpatched kernels:
4617 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4618 2.6.16 - 2.6.19 vanilla (kernel.org),
4619 2.6.9-42.0.8EL (RHEL 4)
4620 * Recommended e2fsprogs version: 1.39.cfs6
4621 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4624 Severity : enhancement
4626 Description: Async OSC create to avoid the blocking unnecessarily.
4627 Details : If a OST has no remain object, system will block on the creating
4628 when need to create a new object on this OST. Now, ways use
4629 pre-created objects when available, instead of blocking on an
4630 empty osc while others are not empty. If we must block, we block
4631 for the shortest possible period of time.
4633 Severity : enhancement
4635 Description: MountConf
4636 Details : Lustre configuration is now managed via mkfs and mount
4637 commands instead of lmc and lconf. New obd types (MGS, MGC)
4638 are added for dynamic configuration management. See
4639 https://mail.clusterfs.com/wikis/lustre/MountConf for
4642 Severity : enhancement
4644 Description: dynamic OST addition
4645 Details : OSTs can now be added to a live filesystem
4647 Severity : enhancement
4649 Description: startup order invariance
4650 Details : MDTs and OSTs can be started in any order. Clients only
4651 require the MDT to complete startup.
4653 Severity : enhancement
4655 Description: parallel, asynchronous orphan cleanup
4656 Details : orphan cleanup is now performed in separate threads for each
4657 OST, allowing parallel non-blocking operation.
4659 Severity : enhancement
4661 Description: optimized stripe assignment
4662 Details : stripe assignments are now made based on ost space available,
4663 ost previous usage, and OSS previous usage, in order to try
4664 to optimize storage space and networking resources.
4666 Severity : enhancement
4668 Description: Permanently set tunables
4669 Details : All writable /proc/fs/lustre tunables can now be permanently
4670 set on a per-server basis, at mkfs time or on a live
4673 Severity : enhancement
4675 Description: Lustre message v2
4676 Details : Add lustre message format v2.
4678 Severity : enhancement
4680 Description: client OST exclusion list
4681 Details : Clients can be started with a list of OSTs that should be
4682 declared "inactive" for known non-responsive OSTs.
4686 Description: ENOENT returned for valid filehandle during dbench.
4687 Details : Check if a directory has children when invalidating dentries
4688 associated with an inode during lock cancellation. This fixes
4689 an incorrect ENOENT sometimes seen for valid filehandles during
4690 testing with dbench.
4693 Frequency : SFS test only (otherwise harmless)
4695 Description: SPEC SFS validation failure on NFS v2 over lustre.
4696 Details : Changes the blocksize for regular files to be 2x RPC size,
4697 and not depend on stripe size.
4699 Severity : enhancement
4701 Description: fine-grained SMP locking inside DLM
4702 Details : Improve DLM performance on SMP systems by removing the single
4703 per-namespace lock and replace it with per-resource locks.
4705 Severity : enhancement
4707 Description: don't hold multiple extent locks at one time
4708 Details : To avoid client eviction during large writes, locks are not
4709 held on multiple stripes at one time or for very large writes.
4710 Otherwise, clients can block waiting for a lock on a failed OST
4711 while holding locks on other OSTs and be evicted.
4713 Severity : enhancement
4715 Description: Multiple MD RPCs in flight.
4716 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4717 To avoid overly-overloading MDT, introduce a limit on number of
4718 MDT RPCs in flight for a single client and add /proc controls
4719 to adjust this limit.
4721 Severity : enhancement
4723 Description: client read/write statistics
4724 Details : Add client read/write call usage stats for performance
4725 analysis of user processes.
4726 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4727 file access. extents_stats shows chunk size distribution.
4728 extents_stats_per_process show chunk size distribution per
4731 Severity : enhancement
4733 Description: per-client statistics on server
4734 Details : Add ldlm and operations statistics for each client in
4735 /proc/fs/lustre/mds|obdfilter/*/exports/
4737 Severity : enhancement
4739 Description: improved MDT statistics
4740 Details : Add detailed MDT operations statistics in
4741 /proc/fs/lustre/mds/*/stats
4743 Severity : enhancement
4745 Description: VFS operations stats
4746 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4747 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4752 Description: Fix client-side osc byte counters
4753 Details : The osc read/write byte counters in
4754 /proc/fs/lustre/osc/*/stats are now working
4757 Frequency : always as root on SLES
4759 Description: Failure of copying files with lustre special EAs.
4760 Details : Client side always return success for setxattr call for lustre
4761 special xattr (currently only "trusted.lov").
4766 Description: Refcount LNET uuids
4767 Details : The global LNET uuid list grew linearly with every startup;
4768 refcount repeated list entries instead of always adding to
4771 Severity : enhancement
4773 Description: Dynamic service threads
4774 Details : Within a small range, start extra service threads
4775 automatically when the request queue builds up.
4778 Frequency : mixed-endian client/server environments
4780 Description: mixed-endian crashes
4781 Details : The new msg_v2 system had some failures in mixed-endian
4784 Severity : enhancement
4786 Description: Easy OST removal
4787 Details : OSTs can be permanently deactivated with e.g. 'lctl
4788 conf_param lustre-OST0001.osc.active=0'
4790 Severity : enhancement
4792 Description: MGS proc entries
4793 Details : Added basic proc entries for the MGS showing what filesystems
4796 Severity : enhancement
4798 Description: provide MGS failover
4799 Details : Added config lock reacquisition after MGS server failover.
4801 Severity : enhancement
4803 Description: add Linux 2.4 support
4804 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4808 Description: a large application tries to do I/O to the same resource and dies
4809 in the middle of it.
4810 Details : Check the req->rq_arrival time after the call to
4811 ost_brw_lock_get(), but before we do anything about
4812 processing it & sending the BULK transfer request. This
4813 should help move old stale pending locks off the queue as
4814 quickly as obd_timeout.
4817 Frequency : when an incorrect nid is specified during startup
4819 Description: ptlrpc connect to non-existant node causes kernel crash
4820 Details : LNET can't be re-entered from an event callback, which
4821 happened when we expire a message after the export has been
4822 cleaned up. Instead, hand the zombie cleanup off to another
4825 Severity : enhancement
4827 Description: plain/inodebits lock performance improvement
4828 Details : Grouping plain/inodebits in granted list by their request modes
4829 and bits policy, thus improving the performance of search through
4833 Frequency : only if OST filesystem is corrupted
4835 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4836 Details : for a short time RPCs with bulk IO are in the replay list,
4837 but replay of bulk IOs is unimplemented. If the OST filesystem
4838 is corrupted due to disk cache incoherency and then replay is
4839 started it is possible to trip an assertion. Avoid putting
4840 committed RPCs into the replay list at all to avoid this issue.
4843 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4845 Description: System hang on startup
4846 Details : This bug allowed the liblustre (e.g. catamount) client to
4847 return to the app before handling all startup RPCs. This
4848 could leave the node unresponsive to lustre network traffic
4849 and manifested as a server ptllnd timeout.
4851 Severity : enhancement
4853 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4854 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4855 boolean to control whether to print peer debug info when a
4856 client's RPC times out.
4859 Frequency : only for kernels with patches from Lustre below 1.4.3
4861 Description: Remove old rdonly API
4862 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4865 Frequency : only for devices with external journals
4867 Description: Set external device read-only also
4868 Details : During a commanded failover stop, we set the disk device
4869 read-only while the server shuts down. We now also set any
4870 external journal device read-only at the same time.
4873 Frequency : when upgrading from 1.4 while trying to change parameters
4875 Description: The wrong (new) MDC name was used when setting parameters for
4876 upgraded MDT's. Also allows changing of OSC (and MDC)
4877 parameters if --writeconf is specified at tunefs upgrade time.
4880 Frequency : when setting specific ost indicies
4882 Description: QOS code breaks on skipped indicies
4883 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4884 created with --index need not be sequential.
4886 Severity : enhancement
4888 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4889 Details : The uninit_groups feature works in conjunction with the kernel
4890 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4891 up the pass1 processing of e2fsck. This is a read-only feature
4892 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4893 mount filesystems that have had this feature enabled.
4895 Severity : enhancement
4897 Description: Improve multi-block allocation algorithm to avoid fragmentation
4898 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4899 allocation locality and avoid filesystem fragmentation.
4901 ------------------------------------------------------------------------------
4903 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4905 * Support for kernels:
4906 2.4.21-47.0.1.EL (RHEL 3)
4907 2.6.5-7.283 (SLES 9)
4908 2.6.9-42.0.10.EL (RHEL 4)
4909 2.6.12.6 vanilla (kernel.org)
4910 2.6.16.27-0.9 (SLES 10)
4911 * Recommended e2fsprogs version: 1.39.cfs5
4913 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4917 Frequency : occasional, depends on client load and configuration
4918 Bugzilla : 12181, 12203
4919 Description: data loss for recently-modified files
4921 Details : In some cases it is possible that recently written or created
4922 files may not be written to disk in a timely manner (this should
4923 normally be within 30s unless client IO load is very high).
4924 The problem appears as zero-length files or files that are a
4925 multiple of 1MB in size after a client crash or client eviction
4926 that are missing data at the end of the file.
4928 This problem is more likely to be hit on clients where files are
4929 repeatedly created and unlinked in the same directory, clients
4930 have a large amount of RAM, have many CPUs, the filesystem has
4931 many OSTs, the clients are rebooted frequently, and/or the files
4932 are not accessed by other nodes after being written.
4934 The presence of the problem can be detected by looking at
4935 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4936 smaller than the second (nr_unused) then dirty files will not
4937 be flushed automatically to disk. "sync; sleep 10" should be
4938 run several times on the node before unmounting it to update
4939 Lustre (this is also safe to run on nodes without this problem).
4941 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4942 that can cause this same problem, so customers using that kernel
4943 also need to update the kernel in addition to Lustre. In order
4944 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4946 It is normal that files written just before a client crash (less
4947 than 30s) may not yet have been flushed to disk, even for local
4951 Frequency : frequent on thin XT3 nodes
4953 Description: UUID collision on thin XT3 Linux nodes
4954 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4955 randomly, since we relied on an insufficiently-seeded PRNG.
4960 Description: OSS hangs after "All ost request buffers busy"
4961 Details : A deadlock between quota and journal operations caused OSS
4962 hangs after printing "All ost request buffers busy."
4965 Frequency : always on liblustre builds
4967 Description: Cleanup compiler warnings on liblustre
4970 Frequency : always on liblustre builds on XT3
4972 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4976 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4977 > 32000 subdirectories
4978 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4979 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4980 subdirectory count crosses 32000. This will aid e2fsck to
4981 correctly handle more than 32000 subdirectories.
4984 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4986 Description: System hang on startup
4987 Details : This bug allowed the liblustre (e.g. catamount) client to
4988 return to the app before handling all startup RPCs. This
4989 could leave the node unresponsive to lustre network traffic
4990 and manifested as a server ptllnd timeout.
4992 Severity : enhancement
4994 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4995 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4996 boolean to control whether to print peer debug info when a
4997 client's RPC times out.
5002 Description: make O_SYNC working on 2.6 kernels
5003 Details : 2.6 kernels use different method for mark pages for write,
5004 so need add a code to lustre for O_SYNC work.
5009 Description: Failure to close file and release space on NFS
5010 Details : Put inode details into lock acquired in ll_intent_file_open.
5011 Use mdc_intent_lock in ll_intent_open to properly
5012 detect all kind of errors unhandled by mdc_enqueue.
5017 Description: proc file read during shutdown sometimes raced obd removal,
5019 Details : Add lock to prevent obd access after proc file removal.
5022 Frequency : Only for files larger than 4GB on 32-bit clients.
5024 Description: improperly doing page alignment of locks
5025 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
5026 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
5031 Description: RPCs being resent when they shouldn't be
5032 Details : Some RPCs that should not be resent are being resent. This
5033 can cause inconsistencies in the RPC state machine. Do not
5034 resend such requests.
5037 Frequency : rare, only with NFS export
5039 Description: Crash on NFS re-export node
5040 Details : under very unusual load conditions an assertion is hit in
5041 ll_intent_file_open()
5044 Frequency : only if OST filesystem is corrupted
5046 Description: client incorrectly hits assertion in ptlrpc_replay_req()
5047 Details : for a short time RPCs with bulk IO are in the replay list,
5048 but replay of bulk IOs is unimplemented. If the OST filesystem
5049 is corrupted due to disk cache incoherency and then replay is
5050 started it is possible to trip an assertion. Avoid putting
5051 committed RPCs into the replay list at all to avoid this issue.
5056 Description: large O_DIRECT requests fail under memory pressure/fragmentation
5057 Details : Large single O_DIRECT read and write calls can fail to allocate
5058 a sufficiently large buffer to process the request. In case of
5059 allocation failure the allocation is retried with a smaller
5060 buffer and broken into smaller requests.
5062 Severity : enhancement
5064 Description: Add -o localflock option to simulate old noflock behaviour.
5065 Details : This will achieve local-only flock/fcntl locks coherentness.
5070 Description: versioning check is incomplete
5071 Details : Checking the version difference of client vs. server, report
5072 error if the gap is too big.
5077 Description: add support PG_writeback bit
5078 Details : add support for PG_writeback bit for Lustre, for more carefull
5079 work with page cache in 2.6 kernel. This also fix some deadlocks
5080 and remove hack for work O_SYNC with 2.6 kernel.
5082 Severity : enhancement
5084 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
5085 Details : The uninit_groups feature works in conjunction with the kernel
5086 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
5087 up the pass1 processing of e2fsck. This is a read-only feature
5088 in ldiskfs2 only, so older kernels and current ldiskfs cannot
5089 mount filesystems that have had this feature enabled.
5091 Severity : enhancement
5093 Description: Improve multi-block allocation algorithm to avoid fragmentation
5094 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
5095 allocation locality and avoid filesystem fragmentation.
5097 ------------------------------------------------------------------------------
5099 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
5101 * Support for kernels:
5102 2.6.9-42.0.3.EL (RHEL 4)
5103 2.6.5-7.276 (SLES 9)
5104 2.4.21-47.0.1.EL (RHEL 3)
5105 2.6.12.6 vanilla (kernel.org)
5106 2.6.16.21-0.8 (SLES10)
5107 * Recommended e2fsprogs version: 1.39.cfs2-0
5109 * The backwards-compatible /proc/sys/portals symlink has been removed
5110 in this release. Before upgrading, please ensure that you change
5111 any configuration scripts or /etc/sysctl.conf files that access
5112 /proc/sys/portals/* or sysctl portals.* to use the corresponding
5113 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
5114 in advance of the upgrade on any system running Lustre 1.4.6 or
5115 newer, since /proc/sys/lnet was added in that version.
5116 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
5120 Frequency : only when quota is used
5122 Description: avoid scanning export list for quota master
5123 Details : Change the algorithms to avoid scanning export list in order
5124 to improve the efficiency.
5127 Frequency : MDS failover only, very rarely
5129 Description: "went back in time" messages on mds failover
5130 Details : The greatest transno may be lost when the current operation
5131 finishes with an error (transno==0) and the client's last_rcvd
5132 record is over-written. Save the greatest transno in the
5133 mds_last_transno for this case.
5136 Frequency : always for specific kernels and striping counts
5138 Description: client may get "Matching packet too big" without ACL support
5139 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
5140 when trying to access files in certain configurations. The
5141 clients should in fact be denied when mounting because they do
5142 not understand ACLs.
5145 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
5147 Description: many clients connecting with IO in progress causes connect timeouts
5148 Details : Avoid synchronous journal commits to avoid delays caused by many
5149 clients connecting/disconnecting when bulk IO is in progress.
5150 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
5151 OST_IO_PORTAL to avoid delays behind potentially many pending
5155 Frequency : occasionally with multiple writers to a single file
5157 Description: shared writes to file may result in wrong size reported by stat()
5158 Details : Allow growing of kms when extent lock is cancelled
5161 Frequency : always with random mmap IO to multi-striped file
5163 Description: mmap write might be lost if we are writing to a 'hole' in stripe
5164 Details : Only if the hole is at the end of OST object so that kms is too
5165 small. Fix is to increase kms accordingly in ll_nopage.
5168 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
5170 Description: writes to a missing object would leak memory on the OST
5171 Details : If there is an inconsistency between the MDS and OST filesystems,
5172 such that the MDS references an object that doesn't exist, writes
5173 to that object will leak memory due to incorrect cleanup in the
5174 error handling path, eventually running out of memory on the OST.
5179 Description: Creating too long symlink causes lustre errors
5180 Details : Check symlink and name lengths before sending requests to MDS.
5183 Frequency : only if flock is enabled (not on by default)
5185 Description: posix locks not released on fd closure on 2.6.9+
5186 Details : We failed to add posix locks to list of inode locks on 2.6.9+
5187 kernels, this caused such locks not to be released on fd close and
5188 then assertions on fs unmount about still used locks.
5191 Frequency : MDS failover only, very rarely
5193 Description: clients may get ASSERTION(granted_lock != NULL)
5194 Details : When request was taking a long time, and a client was resending
5195 a getattr by name lock request. The were multiple lock requests
5196 with the same client lock handle and
5197 mds_getattr_name->fixup_handle_for_resent_request found one of the
5198 lock handles but later failed with ASSERTION(granted_lock != NULL).
5203 Description: handle->h_buffer_credits > 0, assertion failure
5204 Details : h_buffer_credits is zero after truncate, causing assertion
5205 failure. This patch extends the transaction or creates a new
5209 Frequency : NFS re-export or patchless client
5210 Bugzilla : 11179, 10796
5211 Description: Crash on NFS re-export node (__d_move)
5212 Details : We do not want to hash the dentry if we don't have a lock.
5213 But if this dentry is later used in d_move, we'd hit uninitialised
5214 list head d_hash, so we just do this to init d_hash field but
5215 leave dentry unhashed.
5218 Frequency : NFS re-export or patchless client
5220 Description: NFS exports has problem with symbolic link
5221 Details : lustre client didn't properly install dentry when re-exported
5222 to NFS or running patchless client.
5225 Frequency : NFS re-export or patchless client
5227 Description: Various nfs/patchless fixes.
5228 Details : fixes reuse disconected alias for lookup process - this fixes
5229 warning "find_exported_dentry: npd != pd",
5230 fix permission error with open files at nfs.
5231 fix apply umask when do revalidate.
5234 Frequency : occasional
5236 Description: Crash on NFS re-export node
5237 Details : calling clear_page() on the wrong pointer triggered oops in
5238 generic_mapping_read().
5241 Frequency : rarely, using O_DIRECT IO
5243 Description: unaligned directio crashes client with LASSERT
5244 Details : check for unaligned buffers before trying any requests.
5247 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
5249 Description: stale data returned from RAID cache
5250 Details : If only a small amount of IO is done to the RAID device before
5251 reading it again it is possible to get stale data from the RAID
5252 cache instead of reading it from disk.
5255 Frequency : always for sles10 kernel
5257 Description: sles10 support
5258 Details : ll_follow_link: compile fixes and using of nd_set_link
5259 under newer kernels.
5262 Frequency : depends on arch, kernel and compiler version, always on sles10
5265 Description: recursive or deep enough symlinks cause stack overflow
5266 Details : getting rid of large stack-allocated variable in
5270 Frequency : depends on hardware
5272 Description: lustre write performance loss in the SLES10 kernel
5273 Details : the performance loss is caused by using of write barriers in the
5274 ext3 code. The SLES10 kernel turns barrier support on by
5275 default. The fix is to undo that change for ldiskfs.
5277 ------------------------------------------------------------------------------
5279 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
5281 * Support for kernels:
5282 2.6.9-42.0.3EL (RHEL 4)
5283 2.6.5-7.276 (SLES 9)
5284 2.4.21-47.0.1.EL (RHEL 3)
5285 2.6.12.6 vanilla (kernel.org)
5289 Frequency : quota enabled and large files being deleted
5291 Description: releasing more than 4GB of quota at once hangs OST
5292 Details : If a user deletes more than 4GB of files on a single OST it
5293 will cause the OST to spin in an infinite loop. Release
5294 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5299 Description: statfs data retrieved from /proc may be stale or zero
5300 Details : When reading per-device statfs data from /proc, in the
5301 {kbytes,files}_{total,free,avail} files, it may appear
5302 as zero or be out of date.
5305 Frequency : systems with MD RAID1 external journal devices
5307 Description: lconf's call to blkid is confused by RAID1 journal devices
5308 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5309 of returning all block devices that match the journal UUID.
5312 Frequency : always, for aggregate stripe size over 4GB
5314 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5315 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5316 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5319 Frequency : always if "lfs find" used on a local file/directory
5321 Description: "lfs find" segfaults if used on a local file/directory
5322 Details : The case where a directory component was not specified wasn't
5323 handled correctly. Handle this properly.
5326 Frequency : always on ppc64
5328 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5329 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5330 perform endianness conversion.
5333 Frequency : rarely (truncate to non-zero file size after write under load)
5334 Bugzilla : 10730, 10687
5335 Description: Files padded with zeros to next 4K multiple
5336 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5337 it is possible that files that are truncated to a non-zero size
5338 immediately after being written are filled with zero bytes beyond
5339 the truncated size. No file data is lost.
5341 Severity : enhancement
5343 Description: Allow recovery/failover for liblustre clients.
5344 Details : liblustre clients were unaware of failover configurations until
5347 Severity : enhancement
5349 Description: user file locks should fail when not mounting with flock option
5350 Details : Set up an error-returning stub in ll_file_operations.lock field
5351 to prevent incorrect behaviour when client is mounted without
5352 flock option. Also, set up properly f_op->flock field for
5356 Frequency : always on ia64
5358 Description: "lfs df" loops on printing out MDS statfs information
5359 Details : The obd_ioctl_data was not initialized and in some systems
5360 this caused a failure during the ioctl that did not return
5361 an error. Initialize the struct and return an error on failure.
5364 Frequency : SLES 9 only
5366 Description: Error of copying files with lustre special EAs as root
5367 Details : Client side always return success for setxattr call for lustre
5368 special xattr (currently only "trusted.lov").
5371 Frequency : rarely on clusters with both ia64+i386 clients
5373 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5374 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5375 clients are doing concurrent writes to the same file it is
5376 possible that the ia64 clients may LASSERT because the OST
5377 extent locks are not PAGE_SIZE aligned. Ensure that grown
5378 locks are always aligned on the request boundary.
5381 Frequency : specific use, occasional
5383 Description: Overwriting in use executable truncates on-disk binary image
5384 Details : If one node attempts to overwrite an executable in use by
5385 another node, we now correctly return ETXTBSY instead of
5386 truncating the file.
5388 Severity : enhancement
5390 Description: Async OSC create to avoid the blocking unnecessarily.
5391 Details : If a OST has no remain object, system will block on the creating
5392 when need to create a new object on this OST. Now, ways use
5393 pre-created objects when available, instead of blocking on an
5394 empty osc while others are not empty. If we must block, we block
5395 for the shortest possible period of time.
5400 Description: chmod on Lustre root is propagated to other clients
5401 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5402 invalid by the follow_mount time.
5405 Frequency : liblustre clients only
5407 Description: Race in 'instant cancel' lock handling could lead to such locks
5408 never to be granted in case of SMP MDS
5409 Details : Do not destroy not yet granted but cbpending locks in
5413 Frequency : replay/resend of open
5415 Description: non null lock assetion failure in mds_intent_policy
5416 Details : Trying to replay/resend lockless open requests resulted in
5417 mds_open() returning 0 with no lock. Now it sets a flag if
5418 a lock is going to be returned.
5420 Severity : enhancement
5422 Description: Checksum enhancements
5423 Details : New checksum enhancements allow for resending RPCs that failed
5426 Severity : enhancement
5428 Description: Tunables on number of dirty pages in cacche
5429 Details : Allow to set limit on number of dirty pages cached.
5434 Description: client crash on unmount - lock still has references
5435 Details : In some error handling cases it was possible to leak a lock
5436 reference on a client while accessing a file. This was not
5437 harmful to the client during operation, but would cause the
5438 client to crash when the filesystem is unmounted.
5441 Frequency : specific case, rare
5443 Description: ETXTBSY on mds though file not in use by client
5444 Details : ETXTBSY is no longer incorrectly returned when attempting to
5445 chmod or chown a directory that the user previously tried to
5446 execute or a currently-executing binary.
5449 Frequency : extremely rare except on liblustre-based clients
5451 Description: Lustre space not freed when files are deleted
5452 Details : Clean up open-unlinked files after client eviction. Previously
5453 the unlink was skipped and the files remained as orphans.
5458 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5459 Details : In some cases it was possible to send a blocking callback to
5460 a client doing a glimpse, even though that client didn't get
5461 a lock granted. When the glimpse lock is cancelled on the OST
5462 the freed lock is left on the waiting list and corrupted the list.
5465 Frequency : all core dumps
5467 Description: Broke core dumps to lustre
5468 Details : Negative dentry may be unhashed if parent does not have UPDATE
5469 lock, but some callers, e.g. do_coredump, expect dentry to be
5470 hashed after successful create, hash it in ll_create_it.
5472 ------------------------------------------------------------------------------
5474 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5476 * Support for kernels:
5477 2.6.9-42.0.2.EL (RHEL 4)
5478 2.6.5-7.276 (SLES 9)
5479 2.4.21-40.EL (RHEL 3)
5480 2.6.12.6 vanilla (kernel.org)
5484 Frequency : always on RHEL 3
5486 Description: Number of open files grows over time
5487 Details : The number of open files grows over time, whether or not
5488 Lustre is started. This was due to a filp leak introduced
5489 by one of our kernel patches.
5491 ------------------------------------------------------------------------------
5493 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5495 * Support for kernels:
5496 2.6.9-42.EL (RHEL 4)
5497 2.6.5-7.267 (SLES 9)
5498 2.4.21-40.EL (RHEL 3)
5499 2.6.12.6 vanilla (kernel.org)
5504 Bugzilla : 5719, 9635, 9792, 9684
5505 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5506 Details : If a server is under heavy load and cannot reply to new
5507 connection requests before the client resends the (re)connect,
5508 the connection handling code can behave badly if two service
5509 threads are concurrently handing separate (re)connections from
5510 the same client. Add better locking to the connection handling
5511 code, and ensure that only a single connection will be processed
5512 for a given client UUID, even if the lock is dropped.
5514 Severity : enhancement
5516 Description: add TCP zero-copy support to kernel
5517 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5518 sends if the hardware supports scatter-gather and checksumming.
5519 This allows socklnd to do client-write and server-read more
5520 efficiently and reduce CPU utilization from skbuf copying.
5523 Frequency : only if NFS exporting from client
5525 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5526 Details : If setting attributes on a file created under NFS that had
5527 never been opened it would be possible to oops the client
5528 if the file had no objects.
5532 Bugzilla : 9326, 10402, 10897
5533 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5534 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5535 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5536 before ptlrpcd_wake() was called and the request was freed.
5539 Frequency : always for liblustre
5541 Description: liblustre client does MDS+OSTs setattr RPC for each write
5542 Details : When doing a write from a liblustre client, the client
5543 incorrectly issued an RPC to the MDS and each OST the file was
5544 striped over in order to update the timestamps. When writing
5545 with small chunks and many clients this could overwhelm the MDS
5546 with RPCs. In all cases it would slow down the write because
5547 these RPCs are unnecessary.
5549 Severity : enhancement
5551 Description: allow number of MDS service threads to be changed at module load
5552 Details : It is now possible to change the number of MDS service threads
5553 running. Adding "options mds mds_num_threads={N}" to the MDS's
5554 /etc/modprobe.conf will set the number of threads for the next
5555 time Lustre is restarted (assuming the "mds" module is also
5556 reloaded at that time). The default number of threads will
5557 stay the same, 32 for most systems.
5562 Description: OST crash if filesystem is unformatted or corrupt
5563 Details : If an OST is started on a device that has never been formatted
5564 or if the filesystem is corrupt and cannot even mount then the
5565 error handling cleanup routines would dereference a NULL pointer.
5570 Description: NULL pointer deref in llap_from_page.
5571 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5572 incorrect) mapping if the page was truncated/reclaimed while it was
5573 searched for. Check for this condition and skip such pages when
5574 doing readahead. Introduce extra check to llap_from_page() to
5575 verify page->mapping->host is non-NULL (so page is not anonymous).
5578 Frequency : Sometimes when using sys_sendfile
5580 Description: "page not covered by a lock" warnings from ll_readpage
5581 Details : sendfile called ll_readpage without right page locks present.
5582 Now we introduced ll_file_sendfile that does necessary locking
5583 around call to generic_file_sendfile() much like we do in
5587 Frequency : with certain MDS communication failures at client mount time
5589 Description: NULL pointer deref after failed client mount
5590 Details : a client connection request may delayed by the network layer
5591 and not be sent until after the PTLRPC layer has timed out the
5592 request. If the client fails the mount immediately it will try
5593 to clean up before the network times out the request. Add a
5594 reference from the request import to the obd device and delay
5595 the cleanup until the network drops the request.
5598 Frequency : occasionally during client (re)connect
5600 Description: assertion failure during client (re)connect
5601 Details : processing a client connection request may be delayed by the
5602 client or server longer than the client connect timeout. This
5603 causes the client to resend the connection request. If the
5604 original connection request is replied in this interval, the
5605 client may trip an assertion failure in ptlrpc_connect_interpret()
5606 which thought it would be the only running connect process.
5609 Frequency : only with obd_echo servers and clients that are rebooted
5611 Description: kernel BUG accessing uninitialized data structure
5612 Details : When running an obd_echo server it did not start the ping_evictor
5613 thread, and when a client was evicted an uninitialized data
5614 structure was accessed. Start the ping_evictor in the RPC
5615 service startup instead of the OBD startup.
5617 Severity : enhancement
5618 Bugzilla : 10193 (patchless)
5619 Description: Remove dependency on various unexported kernel interfaces.
5620 Details : No longer need reparent_to_init, exit_mm, exit_files,
5621 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5624 Frequency : rare (only users of deprecated and unsupported LDAP config)
5626 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5627 Details : LDAP apparently contains 'lustreName' attributes instead of
5628 'name'. A simple remapping of the name is sufficient.
5631 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5633 Description: waiting_locks_callback trips kernel BUG if client is evicted
5634 Details : Running with the dump_on_timeout debug flag turned on makes
5635 it possible that the waiting_locks_callback() can try to dump
5636 the Lustre kernel debug logs from an interrupt handler. Defer
5637 this log dumping to the expired_lock_main() thread.
5639 Severity : enhancement
5641 Description: Support NFS exporting on 2.6 kernels.
5642 Details : Implement non-rawops metadata methods for NFS server to use without
5643 changing NFS server code.
5646 Frequency : very rare (synthetic metadata workload only)
5648 Description: two racing renames might cause an MDS thread to deadlock
5649 Details : Running the "racer" program may cause one MDS thread to rename
5650 a file from being the source of a rename to being the target of
5651 a rename at exactly the same time that another thread is doing
5652 so, and the second thread has already enqueued these locks after
5653 doing a lookup of the target and is trying to relock them in
5654 order. Ensure that we don't try to re-lock the same resource.
5657 Frequency : only very large systems with liblustre clients
5659 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5660 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5661 all OSTs in parallel. This allows the eviction of stale liblustre
5662 clients to proceed much faster than if they were done in series,
5663 and also offers similar improvements for other set_info RPCs.
5668 Description: excessive CPU usage during initial read phase on client
5669 Details : During the initial read phase on a client, it would agressively
5670 retry readahead on the file, consuming too much CPU and impacting
5671 performance (since 1.4.5.8). Improve the readahead algorithm
5672 to avoid this, and also improve some other common cases (read
5673 of small files in particular, where "small" is files smaller than
5674 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5679 Description: MDS crash when receiving packet with unknown intent.
5680 Details : Do not LBUG in unknown intent case, just return -EFAULT
5682 Severity : enhancement
5683 Bugzilla : 9293, 9385
5684 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5685 Details : Do not serialize getattr (non-intent version) and statfs.
5688 Frequency : occasional, when OST network is overloaded/intermittent
5690 Description: client evicted by OST after bulk IO timeout
5691 Details : If a client sends a bulk IO request (read or write) the OST
5692 may evict the client if it is unresposive to its data GET/PUT
5693 request. This is incorrect if the network is overloaded (takes
5694 too long to transfer the RPC data) or dropped the OST GET/PUT
5695 request. There is no need to evict the client at all, since
5696 the pinger and/or lock callbacks will handle this, and the
5697 client can restart the bulk request.
5700 Frequency : Always when mmapping file with no objects
5702 Description: client crashes when mmapping file with no objects
5703 Details : Check that we actually have objects in a file before doing any
5704 operations on objects in ll_vm_open, ll_vm_close and
5710 Description: Request leak when working with deleted CWD
5711 Details : Introduce advanced request refcount tracking for requests
5712 referenced from lustre intent.
5714 Severity : Enhancement
5716 Description: Cache open file handles on client.
5717 Details : MDS now will return special lock along with openhandle, if
5718 requested and client is allowed to hold openhandle, even if unused,
5719 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5720 closing files for every read/write openration.
5722 Severity : Enhancement
5724 Description: Cache open negative dentries on client when possible.
5725 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5726 negative dentries on lock revocation.
5731 Description: Remounting a client read-only wasn't possible with a zconf mount
5732 Details : It wasn't possible to remount a client read-only with llmount.
5734 Severity : enhancement
5735 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5736 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5737 MPICH is included to add Lustre-specific ADIO interfaces.
5738 This is based closely on the UFS ADIO layer and only differs
5739 in file creation, in order to allow the OST striping to be set.
5740 This is user-contributed code and not supported by CFS.
5745 Description: extended inode attributes (immutable, append-only) work improperly
5746 when 2.4 and 2.6 kernels are used on client/server or vice versa
5747 Details : Introduce kernel-independent values for these flags.
5749 Severity : enhancement
5752 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5753 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5754 granularity. Now, Fractional MB granularity is be supported,
5755 this is very useful for low memory system.
5757 Severity : enhancement
5759 Description: Getattr by fid
5760 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5761 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5762 have locks and avoids getattr rpc requests for it on every stat.
5765 Frequency : Always, for filesystems larger than 2TB
5767 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5768 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5769 filesystems larger than 2TB to be mounted with mballoc enabled.
5772 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5774 Description: filesystem corruption for non-standard kernels and very large OSTs
5775 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5776 filesystems larger than 2TB is mounted then the kernel will
5777 silently corrupt the start of the filesystem. CONFIG_LBD is
5778 enabled for all CFS-supported kernels, but the possibility of
5779 this happening with a modified kernel config exists.
5781 Severity : enhancement
5783 Description: add client O_DIRECT support for 2.6 kernels
5784 Details : It is now possible to do O_DIRECT reads and writes to files
5785 in the Lustre client mountpoint on 2.6 kernel clients.
5787 Severity : enhancement
5789 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5790 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5791 parallel, not waiting for response from every OST before sending
5792 a rpc to the next OST.
5797 Description: setattr vs write race when updating file timestamps
5798 Details : Client processes that update a file timestamp into the past
5799 right after writing to the file (e.g. tar) it is possible that
5800 the updated file modification time can be reset to the current
5801 time due to a race between processing the setattr and write RPC.
5803 Severity : enhancement
5805 Description: Bring 'lfs find' closer in line with regular Linux find.
5806 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5807 -print0 options and obtains all the needed info through the lustre
5810 Severity : enhancement
5812 Description: support up to 1024 configured devices on one node
5813 Details : change obd_dev array from statically allocated to dynamically
5814 allocated structs as they are first used to reduce memory usage
5819 Description: Flush dirty partially truncated pages during truncate
5820 Details : Immediatelly flush partially truncated pages in filter_setattr,
5821 this way we completely avoid having any pages in page cache on OST
5822 and can retire ugly workarounds during writes to flush such pages.
5827 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5828 Details : protect inode from truncation within vfs_unlink() context
5829 just take a reference before calling vfs_unlink() and release it
5830 when parent's i_sem is free.
5833 Frequency : always, if extents are used on OSTs
5835 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5836 Details : OSTs using the extents format would not zero the high 16 bits of
5837 the index physical block number. This is not a problem for any
5838 OST filesystems smaller than 16TB, and no kernels support ext3
5839 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5840 new/modified files) and can be fixed for existing filesystems
5841 with e2fsprogs-1.39-cfs1.
5846 Description: import connection selection may be incorrect if timer wraps
5847 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5848 import connections to be ignored if the 32-bit jiffies counter
5849 wraps. Use a 64-bit jiffies counter.
5852 Frequency : during server recovery
5854 Description: crash after server is denying duplicate export
5855 Details : If clients are resending connect requests to the server, the
5856 server refuses to allow a client to connect multiple times.
5857 Fixed a bug in the handling of this case.
5860 Frequency : very large clusters immediately after boot
5862 Description: LNET request buffers exhausted under heavy short-term load
5863 Details : If a large number of client requests are generated on a service
5864 that has previously never seen so many requests it is possible
5865 that the request buffer growth cannot keep up with the spike in
5866 demand. Instead of dropping incoming requests, they are held in
5867 the LND until the RPC service can accept more requests.
5870 Frequency : Sometimes during replay
5872 Description: Assertion failure in ll_local_open after replay.
5873 Details : If replay happened on an open request reply before we were able
5874 to set replay handler, reply will become not swabbed tripping the
5875 assertion in ll_local_open. Now we set the handler right after
5876 recognising of open request
5879 Frequency : very rare
5881 Description: kernel reports "badness in vsnprintf"
5882 Details : Reading from the "recovery_status" /proc file in small chunks
5883 may cause a negative length in lprocfs_obd_rd_recovery_status()
5884 call to vsnprintf() (which is otherwise harmless). Exit early
5885 if there is no more space in the output buffer.
5887 Severity : enhancement
5889 Description: clear OBD RPC statistics by writing to them
5890 Details : It is now possible to clear the OBD RPC statistics by writing
5891 to the "stats" file.
5896 Description: Client mtime is not the same on different clients after utimes
5897 Details : In some cases, the client was using the utimes() syscall on
5898 a file cached on another node. The clients now validate the
5899 ctime from the MDS + OSTs to determine which one is right.
5904 Description: Inability to activate failout mode
5905 Details : lconf script incorrectly assumed that in python string's numeric
5906 value is used in comparisons.
5909 Frequency : always with multiple stripes per file
5911 Description: Inefficient object allocation for mutli-stripe files
5912 Details : When selecting which OSTs to stripe files over, for files with
5913 a stripe count that divides evenly into the number of OSTs,
5914 the MDS is always picking the same starting OST for each file.
5915 Return the OST selection heuristic to the original design.
5920 Description: mount failures may take full timeout to return an error
5921 Details : Under some heavy load conditions it is possible that a
5922 failed mount can wait for the full obd_timeout interval,
5923 possibly several minutes, before reporting an error.
5924 Instead return an error as soon as the status is known.
5926 ------------------------------------------------------------------------------
5928 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5930 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5931 INTEROPERATE with older versions automatically. Please read the
5932 user documentation before upgrading any part of a live system.
5933 * WARNING: Lustre networking configuration changes are required with
5934 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5937 * Support for kernels:
5938 2.6.9-22.0.2.EL (RHEL 4)
5939 2.6.5-7.244 (SLES 9)
5940 2.6.12.6 vanilla (kernel.org)
5943 Severity : enhancement
5944 Bugzilla : 7981/8208
5945 Description: Introduced Lustre Networking (LNET)
5946 Details : LNET is new networking infrastructure for Lustre, it includes
5947 a reorganized network configuration mode (see the user
5948 documentation for full details) as well as support for routing
5949 between different network fabrics. Lustre Networking Devices
5950 (LNDs) for the supported network fabrics have also been
5951 created for this new infrastructure.
5953 Severity : enhancement
5954 Description: Introduced Access control lists
5955 Details : clients can set ACLs on files and directories in order to have
5956 more fine-grained permissions than the standard Unix UGO+RWX.
5957 The MDS must be started with the "-o acl" mount option.
5959 Severity : enhancement
5960 Description: Introduced filesystem quotas
5961 Details : Administrators may now establish per-user quotas on the
5964 Severity : enhancement
5966 Description: Configuration change for the XT3
5967 The PTLLND is now used to run Lustre over Portals on the XT3
5968 The configure option(s) --with-cray-portals are no longer used.
5969 Rather --with-portals=<path-to-portals-includes> is used to
5970 enable building on the XT3. In addition to enable XT3 specific
5971 features the option --enable-cray-xt3 must be used.
5976 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5977 Details : A race between allocating a new llcd and re-getting the llcd_lock
5978 allowed another thread to grab newly-allocated llcd.
5980 Severity : enhancement
5982 Description: 2.6 OST async journal commit and locking fix to improve performance
5983 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5984 2.6 kernels are now async as they already were in 2.4 kernels so
5985 that they can commit concurrently with the network bulk transfer.
5986 For block-allocated files the filter allocation semaphore is held
5987 to avoid filesystem fragmentation during allocation. BKL lock
5988 removed for 2.6 xattr operations where it is no longer needed.
5993 Description: lconf incorrectly determined whether two IP networks could talk
5994 Details : In some more complicated routing and multiple-network
5995 configurations, lconf will avoid trying to make a network
5996 connection to a disjoint part of the IP space. It was doing the
5997 math incorrectly for one set of cases.
6002 Description: Fix for potential infinite loop processing records in an llog.
6003 Details : If an llog record is corrupted/zeroed, it is possible to loop
6004 forever in llog_process(). Validate the llog record length
6005 and skip the remainder of the block on error.
6008 Frequency : occasional (liblustre only)
6010 Description: liblustre could not open files whose last component is a symlink
6011 Details : sysio_path_walk() would incorrectly pass the open intent to
6012 intermediate path components.
6015 Frequency : rare (liblustre only with non-standard tuning)
6016 Bugzilla : 7201 (7350)
6017 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
6018 Details : llu_lookup_finish_locks() tries to set lock data on a lock
6019 after it has been released, only do this for referenced locks
6021 Severity : enhancement
6023 Description: specifying an (invalid) directory default stripe_size of -1
6024 would reset the directory default striping
6025 Details : stripe_size -1 was used internally to signal directory stripe
6026 removal, now use "all default" to signal dir stripe removal
6027 as a directory striping of "all default" is not useful
6030 Frequency : common for large clusters running liblustre clients
6032 Description: doing an ls when liblustre clients are running is slow
6033 Details : sending a glimpse AST to a liblustre client waits for every AST
6034 to time out, as liblustre clients will not respond. Since they
6035 cannot cache data we refresh the OST lock LVB from disk instead.
6037 Severity : enhancement
6039 Description: doing an ls at the same time as file IO can be slow
6040 Details : enqueue and other "small" requests can be blocked behind many
6041 large IO requests. Create a new OST IO portal for non-IO
6042 requests so they can be processed faster.
6045 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
6047 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
6048 created file with mode 000
6049 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
6050 allow an ftruncate() as a non-root user to a file with mode 000.
6051 Lustre now respects this flag to disable mode checks when
6052 truncating a file owned by the user
6055 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
6058 Description: Revoking locks from clients that went dead or catatonic might take
6060 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
6061 cancellation of such locks instant on servers without waiting for
6062 any reply from clients. Clients drops these locks when cancel
6063 notification from server is received without replying.
6066 Frequency : liblustre-only, when liblustre client dies or becomes busy
6068 Description: Doing ls on Linux clients can take a long time with active
6070 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
6071 granting such locks to it in the first place if possible. Locks
6072 are taken by proxy on the OST during the read or write and
6073 dropped immediately afterward. Add connect flags handling, do
6074 not grant locks to liblustre clients for glimpse ASTs.
6076 Severity : enhancement
6078 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
6079 Details : Existing read-ahead algorithm is tuned for the case of streamlined
6080 sequential reads and behaves badly with applications doing random
6081 reads. Improve it by reading ahead at least read region, and
6082 avoiding excessive large RPC for small reads.
6084 Severity : enhancement
6086 Description: Creating more than 1000 files for a single job may cause a load
6087 imbalance on the OSTs if there are also a large number of OSTs.
6088 Details : qos_prep_create() uses an OST index reseed value that is an
6089 even multiple of the number of available OSTs so that if the
6090 reseed happens in the middle of the object allocation it will
6091 still utilize the OSTs as uniformly as possible.
6096 Description: OST or MDS may oops in ping_evictor_main()
6097 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
6098 but doesn't restart at beginning of obd_exports_timed list
6101 Severity : enhancement
6103 Description: improve by-nid export eviction on the MDS and OST
6104 Details : allow multiple exports with the same NID to be evicted at one
6105 time without re-searching the exports list.
6108 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
6110 Description: MDS may oops in groups_free()
6111 Details : in rare race conditions a newly allocated group_info struct is
6112 freed again, and this can be NULL. The 2.4 compatibility code
6113 for groups_free() checked for a NULL pointer, but 2.6 did not.
6116 Frequency : common for liblustre clients doing little filesystem IO
6117 Bugzilla : 9352, 7313
6118 Description: server may evict liblustre clients accessing contended locks
6119 Details : if a client is granted a lock or receives a completion AST
6120 with a blocking AST already set it would not reply to the AST
6121 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
6124 Frequency : lfs setstripe, only systems with more than 160 OSTs
6126 Description: unable to set striping with a starting offset beyond OST 160
6127 Details : llapi_create_file() incorrectly limited the starting stripe
6128 index to the maximum single-file stripe count.
6131 Frequency : LDAP users only
6133 Description: lconf did not handle in-kernel recovery with LDAP properly
6134 Details : lconf/LustreDB get_refs() is searching the wrong namespace
6136 Severity : enhancement
6138 Description: bind OST threads to NUMA nodes to improve performance
6139 Details : all OST threads are uniformly bound to CPUs on a single NUMA
6140 node and do their allocations there to localize memory access
6142 Severity : enhancement
6144 Description: llmount can determine client NID directly from Myrinet (GM)
6145 Details : the client NID code from gmnalnid was moved directly into
6146 llmount, removing the need to use this or specifying the
6147 client NID explicitly when mounting GM clients with zeroconf
6150 Frequency : if client is started with down MDS
6152 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
6153 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
6154 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
6155 ptlrpc_import_delayed_req() because we didn't check if the
6156 request was interrupted, and we also didn't break out of the
6157 event loop if there was no timeout
6162 Description: data loss during non-page-aligned writes to a single file from
6163 both multiple nodes and multiple threads on one node at same time
6164 Details : updates to KMS and lsm weren't protected by common lock. Resulting
6165 inconsistency led to false short-reads, that were cached and later
6166 used by ->prepare_write() to fill in partially written page,
6167 leading to data loss.
6170 Frequency : always, if lconf --abort_recovery used
6172 Description: lconf --abort_recovery fails with 'Operation not supported'
6173 Details : lconf was attempting to abort recovery on the MDT device and not
6176 Severity : enhancement
6178 Description: remove cleanup logs
6179 Details : replace lconf-generated cleanup logs with lustre internal
6180 cleanup routines. Eliminates the need for client-cleanup and
6183 Severity : enhancement
6185 Description: add support for EAs (user and system) on lustre filesystems
6186 Details : it is now possible to store extended attributes in the Lustre
6187 client filesystem, and with the user_xattr mount option it
6188 is possible to allow users to store EAs on their files also
6190 Severity : enhancement
6192 Description: Add possibility (config option) to show minimal available OST free
6194 Details : When compiled with --enable-mindf configure option, statfs(2)
6195 (and so, df) will return least minimal free space available from
6196 all OSTs as amount of free space on FS, instead of summary of
6197 free spaces of all OSTs.
6199 Severity : enhancement
6201 Description: do not expand extent locks acquired on OST-side
6202 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
6203 by server: they are not cached anyway.
6206 Frequency : when mmap is used/binaries executed from Lustre
6208 Description: Unmmap pages before throwing them away from read cache.
6209 Details : llap_shrink cache now attempts to unmap pages before discarding
6210 them (if unmapping failed - do not discard). SLES9 kernel has
6211 extra checks that trigger if this unmapping is not done first.
6216 Description: lconf didn't resolve symlinks before checking to see whether a
6217 given mountpoint was already in use
6220 Frequency : when migrating failover services
6221 Bugzilla : 6395, 9514
6222 Description: When migrating a subset of services from a node (e.g. failback
6223 from a failover service node) the remaining services would
6224 time out and evict clients.
6225 Details : lconf --force (implied by --failover) sets the global obd_timeout
6226 to 5 seconds in order to quickly disconnect, but this caused
6227 other RPCs to time out too quickly. Do not change the global
6228 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
6230 Severity : enhancement
6231 Frequency : if MDS is started with down OST
6232 Bugzilla : 9439,5706
6233 Description: Allow startup/shutdown of an MDS without depending on the
6234 availability of the OSTs.
6235 Details : Asynchronously call mds_lov_synchronize during MDS startup.
6236 Add appropriate locking and lov-osc refcounts for safe
6237 cleaning. Add osc abort_inflight calls in case the
6238 synchronize never started.
6241 Frequency : occasional (Cray XT3 only)
6243 Description: root not authorized to access files in CRAY_PORTALS environment
6244 Details : The client process capabilities were not honoured on the MDS in
6245 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
6246 been accessed by an authorized user then root was able to access
6247 the file on the local client also. The root user capabilities
6248 are now allowed on the MDS, as this environment has secure UID.
6251 Frequency : occasional
6253 Description: ldiskfs "too long searching" message happens too often
6254 Details : A debugging message (otherwise harmless) prints too often on
6255 the OST console. This has been reduced to only happen when
6256 there are fragmentation problems on the filesystem.
6261 Description: Division by zero in statfs when all OSCs are inactive
6262 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
6263 lov_statfs divides by value returned by lov_get_stripecnt().
6267 Bugzilla : 9489, 3273
6268 Description: First write from each client to each OST was only 4kB in size,
6269 to initialize client writeback cache, which caused sub-optimal
6270 RPCs and poor layout on disk for the first writen file.
6271 Details : Clients now request an initial cache grant at (re)connect time
6272 and so that they can start streaming writes to the cache right
6273 away and always do full-sized RPCs if there is enough data.
6274 If the OST is rebooted the client also re-establishes its grant
6275 so that client cached writes will be honoured under the grant.
6280 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
6281 Details : Now I/O RPCs go to different portal number and (presumably) fast
6282 lock requests (and glimses) and other RPCs get their own service
6283 threads pool that should be able to service those RPCs
6286 Severity : enhancement
6288 Description: Ability to exchange lustre version between client and servers and
6289 issue warnings at client side if client is too old. Also for
6290 liblustre clients there is ability to refuse connection of too old
6292 Details : New 'version' field is added to connect data structure that is
6293 filled with version info. That info is later checked by server and
6297 Frequency : rare, liblustre only.
6298 Bugzilla : 9296, 9581
6299 Description: Two simultaneous writes from liblustre at offset within same page
6300 might proceed at the same time overwriting eachother with stale
6302 Details : I/O lock withing llu_file_prwv was released too early, before data
6303 actually was hitting the wire. Extended lock-holding time until
6304 server acknowledges receiving data.
6307 Frequency : extremely rare. Never observed in practice.
6309 Description: avoid generating lustre_handle cookie of 0.
6310 Details : class_handle_hash() generates handle cookies by incrementing
6311 global counter, and can hit 0 occasionaly (this is unlikely, but
6312 not impossible, because initial value of cookie counter is
6313 selected randonly). Value of 0 is used as a sentinel meaning
6314 "unassigned handle" --- avoid it. Also coalesce two critical
6315 sections in this function into one.
6317 Severity : enhancement
6319 Description: allow liblustre clients to delegate truncate locking to OST
6320 Details : To avoid overhead of locking, liblustre client instructs OST to
6321 take extent lock in ost_punch() on client's behalf. New connection
6322 flag is added to handle backward compatibility.
6324 Severity : enhancement
6325 Bugzilla : 4928, 7341, 9758
6326 Description: allow number of OST service threads to be specified
6327 Details : a module parameter allows the number of OST service threads
6328 to be specified via "options ost ost_num_threads={N}" in the
6329 OSS's /etc/modules.conf or /etc/modprobe.conf.
6333 Bugzilla : 6146, 9635, 9895
6334 Description: servers crash with bad pointer in target_handle_connect()
6335 Details : In rare cases when a client is reconnecting it was possible that
6336 the connection request was the last reference for that export.
6337 We would temporarily drop the export reference and get a new
6338 one, but this may have been the last reference and the export
6339 was just destroyed. Get new reference before dropping old one.
6341 Severity : enhancement
6342 Frequency : if client is started with failover MDS
6344 Description: Allow multiple MDS hostnames in the mount command
6345 Details : Try to read the configuration from all specified MDS
6346 hostnames during a client mount in case the "primary"
6349 Severity : enhancement
6351 Description: Stop sending data to evicted clients as soon as possible.
6352 Details : Check if the client we are about to send or are sending data to
6353 was evicted already. (Check is done every second of waiting,
6354 for which l_wait_event interface was extended to allow checking
6355 of exit condition at specified intervals).
6358 Frequency : rare, normally only when NFS exporting is done from client
6360 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6362 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6365 Frequency : always, if srand() is called before liblustre initialization
6367 Description: Liblustre uses system PRNG disturbing its usage by user application
6368 Details : Introduce internal to lustre fast and high-quality PRNG for
6369 lustre usage and make liblustre and some other places in generic
6370 lustre code to use it.
6372 Severity : enhancement
6373 Bugzilla : 9477, 9557, 9870
6374 Description: Verify that the MDS configuration logs are updated when xml is
6375 Details : Check if the .xml configuration logs are newer than the config
6376 logs stored on the MDS and report an error if this is the case.
6377 Request --write-conf, or allow starting with --old_conf.
6379 Severity : enhancement
6381 Description: Handle symlinks in the path when checking if Lustre is mounted.
6382 Details : Resolve intermediate symlinks when checking if a client has
6383 mounted a filesystem to avoid duplicate client mounts.
6388 Description: lconf can hit an error exception but still return success.
6389 Details : The lconf command catches the Command error exception at the top
6390 level script context and will exit with the associated exit
6391 status, but doesn't ensure that this exit status is non-zero.
6396 Description: failure of ptlrpc thread startup can cause oops
6397 Details : Starting a ptlrpc service thread can fail if there are a large
6398 number of threads or the server memory is very fragmented.
6399 Handle this without oopsing.
6402 Frequency : always, only if liblustre and non-default acceptor port was used
6404 Description: liblustre cannot connect to servers with non-default acceptor port
6405 Details : tcpnal_set_default_params() was not called and was therefore
6406 ignoring the environment varaible TCPNAL_PORT, as well as other
6407 TCPNAL_ environment variables
6412 Description: two objects could be created on the same OST for a single file
6413 Details : If an OST is down, in some cases it was possible to create two
6414 objects on a single OST for a single file. No problems other
6415 than potential performance impact and spurious error messages.
6419 Bugzilla : 5681, 9562
6420 Description: Client may oops in ll_unhash_aliases
6421 Details : Client dcache may become inconsistent in race condition.
6422 In some cases "getcwd" can fail if the current directory is
6428 Description: Inode refcounting problems in NFS export code
6429 Details : link_raw functions used to call d_instantiate without obtaining
6430 extra inode reference first.
6434 Bugzilla : 9942, 9903
6435 Description: Referencing freed requests leading to crash, memleaks with NFS.
6436 Details : We used to require that call to ll_revalidate_it was always
6437 followed by ll_lookup_it. Also with revalidate_special() it is
6438 possible to call ll_revalidate_it() twice for the same dentry
6439 even if first occurence returned success. This fix changes semantic
6440 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6441 reference on a request referred from the intent.
6442 ll_intent_release() then releases such a request.
6445 Frequency : rare, normally benchmark loads only
6447 Description: unlinked inodes were kept in memory on the client
6448 Details : If a client is repeatedly creating and unlinking files it
6449 can accumulate a lot of stale inodes in the inode slab cache.
6450 If there is no other client load running this can cause the
6451 client node to run out of memory. Instead flush old inodes
6452 from client cache that have the same inode number as a new inode.
6455 Frequency : SLES9 2.6.5 kernel and long filenames only
6456 Bugzilla : 9969, 10379
6457 Description: utime reports stale NFS file handle
6458 Details : SLES9 uses out-of-dentry names in some cases, which confused
6459 the lustre dentry revalidation. Change it to always use the
6463 Frequency : rare, unless heavy write-truncate concurrency is continuous
6464 Bugzilla : 4180, 6984, 7171, 9963, 9331
6465 Description: OST becomes very slow and/or deadlocked during object unlink
6466 Details : filter_destroy() was holding onto the parent directory lock
6467 while truncating+unlinking objects. For very large objects this
6468 may block other threads for a long time and slow overall OST
6469 responsiveness. It may also be possible to get a lock ordering
6470 deadlock in this case, or run out of journal credits because of
6471 the combined truncate+unlink. Solution is to do object truncate
6472 first in one transaction without parent lock, and then do the
6473 final unlink in a new transaction with the parent lock. This
6474 reduces the lock hold time dramatically.
6477 Frequency : rare, 2.4 kernels only
6479 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6480 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6481 before unmounting their filesystem in order to avoid deadlock.
6482 The kernel_locked() function in 2.4 kernels only checks whether
6483 the kernel lock is held, not whether it is this process that is
6484 holding it as 2.6 kernels do.
6489 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6490 Details : The client ptlrpc code may be trying to reconnect to a down
6491 server before a previous connection attempt has timed out.
6492 Increase the reconnect interval to be longer than the connection
6493 timeout interval to avoid sending duplicate connections to
6497 Frequency : echo_client brw_test command
6499 Description: fix echo_client to work with OST preallocated code
6500 Details : OST preallocation code (5137) didn't take echo_client IO path
6501 into account: echo_client calls filter methods outside of any
6502 OST thread and, hence, there is no per-thread preallocated
6503 pages and buffers to use. Solution: hijack pga pages for IO. As
6504 a byproduct, this avoids unnecessary data copying.
6508 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6509 Description: Client can oops in mdc_commit_close() after open replay
6510 Details : It was possible for the MDS to return an open request with no
6511 transaction number in mds_finish_transno() if the client was
6512 evicted, but without actually returning an error. Clients
6513 would later try to replay that open and may trip an assertion
6514 Simplify the client close codepath, and always return an error
6515 from the MDS in case the open is not successful.
6518 Frequency : rare, 2.6 OSTs only
6520 Description: OST may deadlock under high load on fragmented files
6521 Details : If there was a heavy load and highly-fragmented OST filesystems
6522 it was possible to have all the OST threads deadlock waiting on
6523 allocation of biovecs, because the biovecs were not released
6524 until the entire RPC IO was completed. Instead, release biovecs
6525 as soon as they are complete to ensure forward IO progress.
6527 Severity : enhancement
6529 Description: Support for specifying external journal device at mount
6530 Details : If an OST or MDS device is formatted with an external journal
6531 device, this device major/minor is stored in the ext3 superblock
6532 and may not be valid for failover. Allow detecting and
6533 specifying the external journal at mount time.
6538 Description: Mounting an MDS with pending unlinked files may cause oops
6539 Details : target_finish_recovery() calls mds_postrecov() which returned
6540 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6541 considers this an error and immediately begins cleaning up the
6542 lov, just after starting the mds_lov process
6544 Severity : enhancement
6546 Description: Implement 'lfs df' to report actual free space on per-OST basis
6547 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6548 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6549 usage of objects; '-h' to report in human readable format.
6551 ------------------------------------------------------------------------------
6553 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6560 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6561 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6562 at startup during orphan unlinking. Instead, skip buddy update
6563 before setup, as it will be regenerated after recovery is complete.
6568 Description: If an OST is inactive, its locks might reference stale inodes.
6569 Details : lov_change_cbdata() must iterate over all namespaces, even if
6570 they are inactive to clear inode references from the lock.
6572 Severity : enhancement
6573 Frequency : occasional, if non-standard max_dirty_mb used
6575 Description: Client will block write RPCs if not enough grant
6576 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6577 then the client will block writes while waiting for another RPC
6578 to complete instead of consuming its dirty limit. With change
6579 we get improved performance when max_dirty_mb is small.
6581 Severity : enhancement
6582 Bugzilla : 3389, 6253
6583 Description: Add support for supplementary groups on the MDS.
6584 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6585 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6586 lookups for user supplementary groups into a cache.
6590 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6591 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6592 for liblustre O_EXCL is enforced by MDS.
6596 Description: readdir never returns NULL in liblustre.
6597 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6598 dentry in struct dirent.
6602 Frequency : liblustre only, depends on application IO pattern
6603 Description: liblustre clients evicted if not contacting servers
6604 Details : Don't put liblustre clients into the ping_evictor list, so
6605 they will not be evicted by the pinger ever.
6607 Severity : enhancement
6609 Description: Add ability to evict clients by NID from MDS.
6610 Details : By echoing "nid:$NID" string into
6611 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6612 $NID would be instantly evicted from this MDS and from all active
6613 OSTs connected to it.
6617 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6618 Details : lookup_it_finish() used to query file size from OSTs that was not
6623 Description: service threads change working directory to that of init
6624 Details : Starting lustre service threads may pin the working directory
6625 of the parent thread, making that filesystem busy. Threads
6626 now change to the working directory of init to avoid this.
6630 Frequency : during shutdown only
6631 Description: shutdown with a failed MDS or OST can cause unmount to hang
6632 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6637 Frequency : 2.6 only
6638 Description: chmod/chown may include an extra supplementary group
6639 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6640 supplementary group and if none is specified this is used.
6643 Bugzilla : 5479 (6816)
6645 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6646 Details : If lookup is in progress on a file that is unlinked we might try
6647 to revalidate the inode and fail in revalidate after lookup is
6648 complete and ll_file_open() enqueues the open again but
6649 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6652 Frequency : always, if lconf --abort_recovery used
6654 Description: lconf --abort_recovery fails with 'Operation not supported'
6655 Details : lconf was attempting to abort recovery on the MDT device and not
6658 ------------------------------------------------------------------------------
6660 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6665 Frequency : rare (only unsupported configurations with a node running as an
6667 Bugzilla : 6514, 5137
6668 Description: Mounting a Lustre file system on a node running as an OST could
6670 Details : OSTs now preallocates memory needed to write out data at
6671 startup, instead of when needed, to avoid having to
6672 allocate memory in possibly low memory situations.
6673 Specifically, if the file system is mounted on on OST,
6674 memory pressure could force it to try to write out data,
6675 which it needed to allocate memory to do. Due to the low
6676 memory, it would be unable to do so and the node would
6677 become unresponsive.
6679 Severity : enhancement
6681 Description: Addition of lconf --service command line option
6682 Details : lconf now accepts a '--service <arg>' option, which is
6683 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6685 Severity : enhancement
6687 Description: Failover mode is now the default for OSTs.
6688 Details : By default, OSTs will now run in failover mode. To return to
6689 the old behaviour, add '--failout' to the lmc line for OSTs.
6691 Severity : enhancement
6693 Description: Health checks are now provided for MDS and OSTs
6694 Details : Additional detailed health check information on MSD and OSTs
6695 is now provided through the procfs health_check value.
6698 Frequency : occasional, depends on IO load
6700 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6701 after numerous create/delete cycles
6702 Details : The ext3 inode allocation policy would not allocate new inodes
6703 very well on the OSTs because there are no new directories
6704 being created. Instead we look for groups with free space if
6705 the parent directories are nearly full.
6710 Description: Network or server problems during mount may cause partially
6711 mounted clients instead of returning an error.
6712 Details : The config llog parsing code may overwrite the error return
6713 code during mount error handling, returning success instead
6719 Description: MDS can fail to allocate large reply buffers
6720 Details : After long uptimes the MDS can fail to allocate large reply
6721 buffers (e.g. zconf client mount config records) due to memory
6722 fragmentation or consumption by the buffer cache. Preallocate
6723 some large reply buffers so that these replies can be sent even
6724 under memory pressure.
6728 Frequency : rare (liblustre)
6729 Description: fsx running with liblustre complained that using truncate() to
6730 extend the file doesn't work. This patch corrects that issue.
6731 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6732 ATTR_SIZE and lsm use in llu_setattr_raw.
6736 Frequency : rare, only 2.6 kernels
6737 Description: Unusual file access patterns on the MDS may result in inode
6738 data being lost in very rare circumstances.
6739 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6740 code in the 2.6 kernel caused the inode and/or EA data not to
6741 be read from disk, causing single-file corruption.
6745 Frequency : rare, only 2.6 filesystems using extents
6746 Description: Heavy concurrent write and delete load may cause data corruption.
6747 Details : It was possible under high-load situations to have an extent
6748 metadata block in the block device cache from a just-unlinked
6749 file overwrite a newly-allocated data block. We now unmap any
6750 metadata buffers that alias just-allocated data blocks.
6754 Frequency : filesystems with default stripe_count larger than 77
6755 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6756 Details : lconf specifies an inode size of 4096 bytes when the default
6757 stripe_count is larger than 77. This conflicts with the default
6758 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6759 this case to avoid pinning too much memory for large EAs.
6761 ------------------------------------------------------------------------------
6763 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6768 Frequency : rare (extremely heavy IO load with hundreds of clients)
6770 Description: Client is evicted, gets IO error writing to file
6771 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6772 caused clients to be evicted for AST timeouts. The fixes in
6773 bug 5192 mean we no longer need to have such short AST timeouts
6774 so ldlm_timeout has been increased.
6777 Frequency : occasional during --force or --failover shutdown under load
6778 Bugzilla : 5949, 4834
6779 Description: Server oops/LBUG if stopped with --force or --failover under load
6780 Details : a collection of import/export refcount and cleanup ordering
6781 issues fixed for safer force cleanup
6784 Frequency : only filesystems larger than 120 OSTs
6785 Bugzilla : 5990, 6223
6786 Description: lfs getstripe would oops on a very large filesystem
6787 Details : lov_getconfig used kfree on vmalloc'd memory
6790 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6791 Bugzilla : 6242, 6243
6792 Description: reading from files that had been truncated to a non-zero size
6793 but never opened returned no data
6794 Details : ll_file_read() reads zeros from no-object files to EOF
6799 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6800 Details : There's little harm in aborting MDS/OSS recovery and letting it
6801 try again, so I removed the LASSERT and return an error instead.
6803 Severity : enhancement
6805 Description: New debugging infrastructure for tracking down data corruption
6806 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6807 (b) cover more of the client-side code path, and (c) try to narrow
6808 down where problems occurred
6812 Bugzilla : 3819, 4364, 4397, 6313
6813 Description: Racing close and eviction MDS could cause assertion in mds_close
6814 Details : It was possible to get multiple mfd references during close and
6815 client eviction, leading to one thread referencing a freed mfd.
6817 Severity: : enhancement
6818 Bugzilla : 3262, 6359
6819 Description: Attempts to reconnect to servers are now more aggressive.
6820 Details : This builds on the enhanced upcall-less recovery that was added
6821 in 1.4.2. When trying to reconnect to servers, clients will
6822 now try each server in the failover group every 10 seconds. By
6823 default, clients would previously try one server every 25 seconds.
6828 Description: After recovery, certain operations trigger a failed
6829 assertion on a client.
6830 Details : Failing over an mds, using lconf -d --failover, while a
6831 client was doing a readdir() call would cause the client to
6832 LBUG after recovery completed and the readdir() was resent.
6834 Severity : enhancement
6836 Description: Default groups are now added by lconf
6837 Details : You can now run lconf --group <servicename> without having to
6838 manually add groups with lmc.
6841 Frequency : occasional
6843 Description: Nodes with an elan id of 0 trigger a failed assertion
6846 Frequency : always when accessing e.g. tty/console device nodes
6848 Description: tty and some other devices nodes cannot be used on lustre
6849 Details : file's private_data field is used by device data and lustre
6850 values in there got lost. New field was added to struct file to
6851 store fs-specific private data.
6854 Frequency : when exporting Lustre via NFS
6856 Description: NFSD failed occasionally when looking up a path component
6857 Details : NFSD is looking up ".." which was broken in ext3 directories
6858 that had grown large enough to become hashed.
6861 Frequency : Clusters with multiple interfaces not on the same subnet
6863 Description: Nodes will repeatedly try to reconnect to an interface which it
6864 cannot reach and report an error to the log.
6865 Details : Extra peer list entries will be created by lconf with some peers
6866 unreachable. lconf now validates the peer before adding it.
6869 Frequency : Only if a default stripe is set on the filesystem root.
6871 Description: Setting a default stripe on the filesystem root prevented the
6872 filesystem from being remounted.
6873 Details : The client was sending extra request flags in the root getattr
6874 request and did not allocate a reply buffer for the dir EA.
6877 Frequency : occasional, higher if lots of files are accessed by one client
6878 Bugzilla : 6159, 6097
6879 Description: Client trips assertion regarding lsm mismatch/magic
6880 Details : While revalidating inodes the VFS looks up inodes with ifind()
6881 and in rare cases can find an inode that is being freed.
6882 The ll_test_inode() code will free the lsm during ifind()
6883 when it finds an existing inode and then the VFS later attaches
6884 this free lsm to a new inode.
6888 Bugzilla : 6422, 7030
6889 Description: MDS deadlock between mkdir and client eviction
6890 Details : Creating a new file via mkdir or mknod (starting a transaction
6891 and getting the ns lock) can deadlock with client eviction
6892 (gets ns lock and trying to finish a synchronous transaction).
6895 Frequency : occasional
6896 Description: While starting a server, the fsfilt_ext3 module could not be
6898 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6899 kernels. Previously, lconf would still use the ext3 name
6900 when trying to load modules. Now, it will correctly use
6901 ext3 on 2.4 and ldiskfs on 2.6.
6903 Severity : enhancement
6904 Description: The default stripe count has been changed to 1
6905 Details : The interpretation of the default stripe count (0, to lfs
6906 or lmc) has been changed to mean striping across a single
6907 OST, rather than all available. For general usage we have
6908 found a stripe count of 1 or 2 works best.
6910 Severity : enhancement
6911 Description: Add support for compiling against Cray portals.
6912 Details : Conditional compiling for some areas that are different
6916 Frequency : occasional
6917 Bugzilla : 6409, 6834
6918 Description: Creating files with an explicit stripe count may lead to
6919 a failed assertion on the MDS
6920 Details : If some OSTs are full or unavailable, creating files may
6921 trigger a failed assertion on the MDS. Now, Lustre will
6922 try to use other servers or return an error to the
6926 Frequency : occasional
6928 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6929 caused lock timeout during readahead (since 1.4.2).
6930 Details : Processes doing readahead might match a lock that hasn't been
6931 granted yet if there are overlapping and conflicting lock
6932 requests. The readahead process waits on ungranted lock
6933 (original lock is CBPENDING), while OST waits for that process
6934 to cancel CBPENDING read lock and eventually evicts client.
6936 Severity : enhancement
6938 Description: Initial enabling of flock support for clients
6939 Details : Implements fcntl advisory locking and file status functions.
6940 This feature is provided as an optional mount flag (default
6941 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6942 locking are implemented yet, and those that are are not guaranteed
6943 to be completely correct in production environments.
6944 mount -t lustre -o [flock|noflock] ...
6947 Frequency : occasional
6949 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6950 assertion in the ext3 JBD (journaling) layer.
6951 Details : The b_committed_data struct is protected by the big kernel lock
6952 in 2.4 kernels, serializing journal_commit_transaction() and
6953 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6954 access to this struct. In 2.6 kernels there is finer grained
6955 locking to improve SMP performance of the JBD layer.
6959 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6961 -----------------------------------------------------------------------------
6963 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6965 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6966 versions of Lustre. Please update all servers and clients to
6967 version 1.4.2 or later at the same time. You must also run
6968 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6969 to update the configuration logs.
6971 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6972 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6973 - swab llog records of type '0' so we get proper header size/idx (5861)
6974 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6975 - fix rename of one directory over another leaking an inode (5953)
6976 - avoid SetPageDirty on 2.6 (5981)
6977 - don't re-add just-being-destroyed locks to the waiting list (5653)
6978 - when creating new directories, inherit the parent's custom
6979 striping settings if present parent (3048)
6980 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6981 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6982 - don't hold client locks on temporary worklist from l_lru (5666)
6983 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6984 - automatically evict dead clients (5921)
6985 - Update file size properly in create+truncate+fstat case (6196)
6986 - Do not unhash mountpoint dentries, do not allow removal of
6988 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6989 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6990 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6991 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6992 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6993 - Correctly handle reads of files with no objects (6243)
6994 - lctl recover will also mark a device active if deactivate used (5933)
6996 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6997 - allow --write-conf on an MDS with different nettype than client (5619)
6998 - don't write config llogs to MDS for mounts not from that MDS (5617)
6999 - lconf should create multiple TCP connections from a client (5201)
7000 - init scripts are now turned off by default; run chkconfig --on
7001 lustre and chkconfig --on lustrefs to use them
7002 - upcalls are no longer needed for clients to recover to failover
7004 - add --abort-recovery option to lconf to abort recovery on device
7006 - add support for an arbitrary number of OSTs (3026)
7007 - Quota support protocol changes.
7008 - forward compatibility changes to wire structs (6007)
7009 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
7010 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
7011 - improved "lustre status" script
7012 - initialize blocksize for non-regular files (6062)
7013 - added --disable-server and --disable-client configure options (5782)
7014 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
7015 - Vanilla 2.4.29 support
7016 - increase maximum number of obd devices to 520 (6242)
7017 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
7018 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
7019 SLES 9 2.6.5 kernels
7020 - limit stripes per file to 160 (the maximum EA size) (6093)
7022 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7025 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
7026 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
7027 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
7028 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
7029 - small ext3 extents cleanups and fixes (5733)
7030 - improved mballoc code, several small races and bugs fixed (5733, 5638)
7031 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
7032 - avoid needless client->OST connect, fix handle mismatch (5317)
7033 - fix DLM error path that led to out-of-sync client, long delays (5779)
7034 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
7035 - fix several locking issues related to i_size (5492,5624,5654,5672)
7036 - don't move pending lock onto export if it is already evicted (5683)
7037 - fix kernel oops when creating .foo in unlinked directory (5548)
7038 - fix deadlock in obdfilter statistics vs. object create (5811)
7039 - use time_{before,after} to avoid timer jiffies wrap (5882)
7040 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
7041 - Do not leak request if server was not able to process it (5154)
7042 - If mds_open unable to find parent dir, make that negative lookup(5154)
7043 - don't create new directories with extent-mapping (5909, 5936)
7045 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
7046 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
7047 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
7048 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
7049 - support for RHEL4 kernel on i686 (5773)
7050 - provide error messages when incompatible logs are encountered (5898)
7052 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
7053 * version 1.4.0.10 (1.4.1 release candidate 1)
7055 - don't keep a lock reference when lock is not granted (4238)
7056 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7057 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
7058 - drop import inflight refcount on signal_completed_replay error (5255)
7059 - unlock page after async write error during send (3677)
7060 - handle missing objects in filter_preprw_read properly (5265)
7061 - no transno return for symlink open, don't save no-trasno open (3440)
7062 - don't try to complete elan receive that already failed (4012)
7063 - free RPC server reply state on error (5406)
7064 - clean up thread from ptlrpc_start_thread() on error (5160)
7065 - readahead could read extra page into cache that wasn't ejected (5388)
7066 - prevent races in class_attach/setup/cleanup/detach (5260)
7067 - don't dereference de->d_inode after l_dput of de (5458)
7068 - use "int" for stripe value returned from lock_to_stripe (5544)
7069 - mballoc allocation and error-checking fixes in 2.6 (5504)
7070 - block device patches to fix I/O request sizes in 2.6 (5482)
7071 - look up hostnames for IB nals (5602)
7072 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
7073 - don't start multiple acceptors for the same port (5277)
7074 - fix incorrect LASSERT in mds_getattr_name (5635)
7075 - export a proc file for general "ping" checking (5628)
7076 - fix "lfs check" to not block when the MDS is down (5628)
7078 - service request history (4965)
7079 - put {ll,lov,osc}_async_page structs in a single slab (4699)
7080 - create an "evict_client" /proc entry on OSTs, like the MDS has
7081 - fix mount usage message, return errors per mount(8) (5168)
7082 - change grep [] to grep "[]" in tests so they work in more UMLs
7083 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
7084 - remove ancient LOV_MAGIC_V0 EA support (5047)
7085 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
7086 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
7087 - allow readahead allocations to fail when low on memory (5383)
7088 - mmap locking landed again, after considerable improvement (2828)
7089 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
7091 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
7094 - send OST transaction number in read/write reply to free req (4966)
7095 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7096 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7097 - the watchdog thread now runs as interruptible (5246)
7098 - flock/lockf fixes (but it's still disabled, pending 5135)
7099 - don't use EXT3 constants in llite code (5094)
7100 - memory shortage at startup could cause assertion (5176)
7102 - reorganization of lov code
7103 - single portals codebase
7105 - add extents/mballoc support (5025)
7106 - direct I/O reads in the obdfilter (4048)
7107 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
7109 tbd Cluster File Systems, Inc. <info@clusterfs.com>
7112 - send OST transaction number in read/write reply to free req (4966)
7113 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7114 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7115 - the watchdog thread now runs as interruptible (5246)
7116 - handle missing objects in filter_preprw_read properly (5265)
7117 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7118 - drop import inflight refcount on signal_completed_replay error (5255)
7119 - unlock page after async write error during send (3677)
7120 - return original error code on reconstructed replies (3761)
7121 - no transno return for symlink open, don't save no-trasno open (3440)
7123 - add pid to ldlm debugging output (4922)
7124 - bump the watchdog timeouts -- we can't handle 30sec yet
7125 - extra debugging for orphan dentry/inode bug (5259)
7127 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
7130 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
7131 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
7132 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
7133 - fix NULL dereference in /proc/sys/portals/routes (4827)
7134 - allow failed mdc_close() operations to be interrupted (4561)
7135 - stop precreate on OST before MDS would time out on it (4778)
7136 - don't send partial-page writes before EOF from client (4410)
7137 - discard client grant for sub-page writes on large-page clients (4520)
7138 - don't free dentries not owned by NFS code, check generation (4806)
7139 - fix lsm leak if mds_create_objects() fails (4801)
7140 - limit debug_daemon file size, always print CERROR messages (4789)
7141 - use transno after validating reply (3892)
7142 - process timed out requests if import state changes (3754)
7143 - update mtime on OST during writes, return in glimpse (4829)
7144 - add mkfsoptions to LDAP (4679)
7145 - use ->max_readahead method instead of zapping global ra (5039)
7146 - don't interrupt __l_wait_event() during strace
7148 - add software watchdogs to catch hung threads quickly (4941)
7149 - make lustrefs init script start after nfs is mounted
7150 - fix CWARN/ERROR duplication (4930)
7151 - return async write errors to application if possible (2248)
7152 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
7153 - print NAL number in %x format (4645)
7154 - update barely-supported suse-2.4.21-171 series (4842)
7155 - support for sles 9 %post scripts
7156 - support for building 2.6 kernel-source packages
7157 - support for sles km_* packages
7159 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
7162 - ignore -ENOENT errors in osc_destroy (3639)
7163 - notify osc create thread that OSC is being cleaned up (4600)
7164 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
7165 - reconstruct ost_handle() like mds_handle() (4657)
7166 - create a new thread to do import eviction to avoid deadlock (3969)
7167 - let lconf resolve symlinked-to devices (4629)
7168 - don't unlink "objects" from directory with default EA (4554)
7169 - hold socknal file ref over connect in case target is down (4394)
7170 - allow more than 32000 subdirectories in a single directory (3244)
7171 - fix blocks count for O_DIRECT writes (3751)
7172 - OST returns ENOSPC from object create when no space left (4539)
7173 - don't send truncate RPC if file size isn't changing (4410)
7174 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
7175 - bind to privileged port in socknal and tcpnal (3689)
7177 - rate limit CERROR/CWARN console message to avoid overload (4519)
7178 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
7179 - basic mmap support (3918)
7180 - kernel patch series update from b1_4 (4711)
7182 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
7185 - avoid crash during MDS cleanup with OST shut down (2775)
7186 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
7187 - don't use bad inodes on the MDS (3744)
7188 - dynamic object preallocation to improve recovery speed (4236)
7189 - don't hold spinlock over lock dumping or change debug flags (4401)
7190 - don't zero obd_dev when it is force cleaned (3651)
7191 - print grants to console if they go negative (4431)
7192 - "lctl deactivate" will stop automatic recovery attempts (3406)
7193 - look for existing locks in ldlm_handle_enqueue() (3764)
7194 - don't resolve lock handle twice in recovery avoiding race (4401)
7195 - revalidate should check working dir is a directory (4134)
7197 - don't always mark "slow" obdfilter messages as errors (4418)
7199 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
7202 - don't close LustreDB during write_conf until it is done (3860)
7203 - fix typo in lconf for_each_profile (3821)
7204 - allow dumping logs from multiple threads at one time (3820)
7205 - don't allow multiple threads in OSC recovery (3812)
7206 - fix debug_size parameters (3864)
7207 - fix mds_postrecov to initialize import for llog ctxt (3121)
7208 - replace config semaphore with spinlock (3306)
7209 - be sure to send a reply for a CANCEL rpc with bad export (3863)
7210 - don't allow enqueue to complete on a destroyed export (3822)
7211 - down write_lock before checking llog header bitmap (3825)
7212 - recover from lock replay timeout (3764)
7213 - up llog sem before sending rpc (3652)
7214 - reduce ns lock hold times when setting kms (3267)
7215 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
7216 - fix NULL deref and obd_dev leak on setup error (3312)
7217 - replace some LBUG about llog ops with error handling (3841)
7218 - don't match INVALID dentries from d_lookup and spin (3784)
7219 - hold dcache_lock while marking dentries INVALID and hashing (4255)
7220 - fix invalid assertion in ptlrpc_set_wait (3880)
7222 - add libwrap support for the TCP acceptor (3996)
7223 - add /proc/sys/portals/routes for non-root route listing (3994)
7224 - allow setting MDS UUID in .xml (2580)
7225 - print the stack of a process that LBUGs (4228)
7227 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
7230 - don't cleanup request in ll_file_open() on failed MDS open (3430)
7231 - make sure to unset replay flag from failed open requests (3440)
7232 - if default stripe count is 0, use OST count for inode size (3636)
7233 - update parent mtime/ctime on client for create/unlink (2611)
7234 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
7235 - free recovery state on server during a forced cleanup (3571)
7236 - unregister_reply for resent reqs (3063)
7237 - loop back devices mounting and status check on 2.6 (3563)
7238 - fix resource-creation race that can provoke i_size == 0 (3513)
7239 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
7240 - more debugging for page-accounting assertion (3746)
7241 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
7242 - avoid deadlock after precreation failure (3758)
7243 - fix race and lock order deadlock in orphan handling (3450, 3750)
7244 - add validity checks when grabbing inodes from l_ast_data (3599)
7246 - add /proc/.../recovery_status to obdfilter (3428)
7247 - lightweight CDEBUG infrastructure, debug daemon (3668)
7248 - change default OSC RPC parameters to be better on small clusters
7249 - turn off OST read cache for files smaller than 32MB
7250 - install man pages and include them in rpms (3100)
7251 - add new init script for (un)mounting lustre filesystems (2593)
7252 - run chkconfig in %post for init scripts (3701)
7253 - drop scimac NAL (unmaintained)
7255 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
7258 - clean kiobufs before and after use (3485)
7259 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
7260 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7261 - change init script to fail more clearly if not run as root (1528)
7262 - allow clients to reconnect during replay (1742)
7263 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
7264 - don't do DNS lookups on NIDs too small for IP addresses (3442)
7265 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
7266 - fix cond_resched (3554)
7267 - only evict unfinished clients after recovery (3515)
7268 - allow bulk resend, prevent data loss (3570)
7269 - dynamic ptlrpc request buffer allocation (2102)
7270 - don't allow unlinking open directory if it isn't empty (2904)
7271 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
7272 - remove extraneous obd dereference causing LASSERT failure (3334)
7273 - don't use get_cycles() when creating temp. files on the mds (3156)
7274 - hold i_sem when setting i_size in ll_extent_lock() (3564)
7275 - handle EEXIST for set-stripe, set proper directory name (3336)
7277 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
7278 - fix ksocknal_fmb_callback() error messages (2918)
7280 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
7283 - don't copy lvb into (possibly NULL) reply on error (2983)
7284 - don't deref dentry after dput, don't free lvb on error (2922)
7285 - use the kms to determine writeback rpc length (2947)
7286 - increment oti_logcookies when osc is inactive (2948)
7287 - update client's i_blocks count via lvb messages (2543)
7288 - handle intent open/close of special files properly (1557)
7289 - mount MDS with errors=remount-ro, like obdfilter (2009)
7290 - initialize lock handle to avoid ASSERT on error cleanup (3057)
7291 - don't use cancelling-locks' kms values (2947)
7292 - use highest lock extent for kms, not last one (2925)
7293 - don't dereference ERR_PTR() dentry in error handling path (3107)
7294 - fix thread race in portals_debug_dumplog() (3122)
7295 - create lprocfs device entries at setup instead of at attach (1519)
7296 - common AST error handler, don't evict client on completion race (3145)
7297 - zero nameidata in detach_mnt in 2.6 (3118)
7298 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7299 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7300 - initialize RPC timeout timer earlier for 2.6 (3219)
7301 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7302 - print nal/nid for unknown nid (3258)
7303 - additional checks for oscc recovery before doing precreate (3284)
7304 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7305 - don't crash in mdc_close for bad permissions on open (3285)
7306 - zero i_rdev for non-device files (3147)
7307 - clear page->private before handing to FS, better assertion (3119)
7308 - tune the read pipeline (3236)
7309 - fix incorrect decref of invalidated dentry (2350)
7310 - provide read-ahead stats and refine rpc in flight stats (3328)
7311 - don't hold journal transaction open across create RPC (3313)
7312 - update atime on MDS at close time (3265)
7313 - close LDAP connection when recovering to avoid server load (3315)
7314 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7315 - don't leak open file on MDS after open resend (3325)
7316 - serialize filter_precreate and filter_destroy_precreated (3329)
7317 - loop device shouldn't call sync_dev() for nul device (3092)
7318 - clear page cache after eviction (2766)
7319 - resynchronize MDS->OST in background (2824)
7320 - refuse to mount the same filesystem twice on same mountpoint (3394)
7321 - allow llmount to create routes for mounting behind routers (3320)
7322 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7323 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7324 - fix rare problem with rename on htree directories (3417)
7326 - allow default OST striping configuration per directory (1414)
7327 - fix compilation for qswnal for 2.6 kernels (3125)
7328 - increase maximum number of MDS request buffers for large systems
7329 - change liblustreapi to be useful for external progs like lfsck (3098)
7330 - increase local configuration timeout for slow disks (3353)
7331 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7333 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7336 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7337 - don't overwrite extent policy data in reply if lock was blocked (2901)
7338 - drop filter export grants atomically with removal from device (2663)
7339 - del obd_self_export from work_list in class_disconnect_exports (2908)
7340 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7341 - swab reply message in mdc_close, other PPC fixes (2464)
7342 - fix destroying of named logs (2325)
7343 - overwrite old logs when running lconf --write_conf (2264)
7344 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7345 - fix race in target_handle_connect (2898)
7346 - mds_reint_create() should take same inode create lock (2926)
7347 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7348 - don't close files for self_export to avoid uninitialized obd (2936)
7349 - allow MDS with the same name as client node (2939)
7350 - hold dentry reference for closed log files for unlink (2325)
7351 - reserve space for all logs during transactions (2059)
7352 - don't evict page beyond end of stripe extent (2925)
7353 - don't oops on a deleted current working directory (2399)
7354 - handle hard links to targets without a parent properly (2517)
7355 - don't dereference NULL lock when racing during eviction (2867)
7356 - don't grow lock extents when lots of conflicting locks (2919)
7358 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7361 - account for cache space usage on clients to avoid data loss (974)
7362 - lfsck support in lustre kernel code (2349)
7363 - reduce journal credits needed for BRW writes (2370)
7364 - orphan handling to avoid losing space on client/server crashes
7365 - ptlrpcd can be blocked, stopping ALL progress (2477)
7366 - use lock value blocks to assist in proper KMS, faster stat (1021)
7367 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7368 - recovery for initial connections (2355)
7369 - fixes for mds_cleanup_orphans (1934)
7370 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7371 - block all file creations until orphan recovery completes (1901)
7372 - client remove rq_connection from request struct (2423)
7373 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7374 - recovery timer race (2670)
7375 - mdc_close recovey bug (2532)
7376 - ptlrpc cleanup bug (2710)
7377 - mds timeout on local locks (2588)
7378 - namespace lock held during RPCs (2431)
7379 - handle interrupted sync write properly (2503)
7380 - don't try to handle a message that hasn't been replied to (2699)
7381 - client assert failure during cleanup after abort recovery (2701)
7382 - leak mdc device after failed mount (2712)
7383 - ptlrpc_check_set allows timedout requests to complete (2714)
7384 - wait for inflight reqs when ptlrpcd finishes (2710)
7385 - make sure unregistered services are removed from the srv_list
7386 - reset bulk XID's when resending them (caught by 1138 test)
7387 - unregister_bulk after timeout
7388 - fix lconf error (2694)
7389 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7390 - readahead locks pages, leaves pending causing memory pressure (2673)
7391 - increase OST request buffers to 4096 on large machines (2729)
7392 - fix up permission of existing directories in simple_mkdir (2661)
7393 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7394 - don't assign transno to errored transactions (2742)
7395 - don't delete objects on OST if given a bogus objid from MDS (2751)
7396 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7397 - if rq_no_resend, then timeout request after recovery (2432)
7398 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7399 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7400 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7401 - MDS recovery shouldn't delete objects that it has given out (2730)
7402 - if enqueue arrives after completion, don't clobber LVB (2819)
7403 - don't unlock pages twice when trigger_group_io returns error (2814)
7404 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7405 - don't write pages to disk if there was an error (1450)
7406 - don't ping imports that have recovery disabled (2676)
7407 - take buffered bytes into account when balancing socknal conn (2817)
7408 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7409 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7410 - remove little-endian swabbing of llog records (1987)
7411 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7412 - retry reposting request buffers if they fail (1191)
7413 - grow extent at grant time to avoid granting a revoked lock (2809)
7414 - lock revoke doesn't evict page if covered by a second lock (2765)
7415 - disable VM readahead to avoid reading outside lock extents (2805)
7417 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7418 - updated kernel patches for hp-2.4.20 kernel (2681)
7420 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7423 - fix truncated write corruption (2366)
7424 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7426 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7427 - ensure len(uuid) < 37 in lmc (1171)
7428 - fix ia64 OOPS in llog_test (2255)
7429 - zero end of page at obdfilter for partial page writes (2648)
7430 - don't leave stale dentries around after renames (bug 2428)
7431 - fix timeouts when evicting a client with a single lock held (2642)
7432 - set deadline for the initial HELLO message to drain (2634)
7433 - print out dotted-quad IP addresses in the socknal (2302)
7435 - additional debugging for MDS client eviction problem (2443)
7436 - fix mkfsoptions support for osts (2603, 2604)
7438 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7441 - add series for the vanilla 2.6.0 kernel
7442 - add series for the vanilla 2.4.24 kernel
7443 - add series for a cray x86/64 UL kernel drop
7444 - fix xattr patches for the vanilla 2.4.19 series
7446 - generate true UUIDs in lmc (1171)
7447 - have portals stack dumping break in UML (2466)
7448 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7449 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7450 - wake callback waiting threads on client eviction (2460)
7451 - Add --ptldebug and --subsystem to lmc (1719)
7452 - update assertion to allow safe interrupt allocation
7453 - set rq_no_resend for cancel requests (2432)
7454 - recalculate ptlrpcd timeout after resend (2494)
7455 - call vfs_rmdir when removing pending directories (2368)
7456 - fix renaming a file to itself (2429)
7457 - lmc creates a default one-stripe lov (2454)
7458 - expand procfs space to handle large clusters (2326)
7459 - increase UML stack to avoid overflow
7460 - update lconf's list of debug and subsystem masks
7461 - fix lfs find --obd (2510)
7462 - /proc tunable for disabling filter read caching (2591)
7463 - stop rpm packages from altering slapd.conf (2301)
7464 - disable nagle in the socknal under 0conf (2578)
7465 - choose mds inode size based on stripe count (2572)
7466 - fix kernel-source rpm problems (2516)
7468 - add --disable-doc to avoid pdf generation (2421)
7469 - update documentation, tests, type-os, comments
7470 - avoid format warnings on ia64
7471 - remove the TOE NAL
7472 - tiny code cleanups by removing unused fields
7474 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7477 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7478 - remove the most blatant lies from BUILDING file (2371)
7479 - change default debug level to reasonable production setting
7480 - reduce client side cache size to reduce cache flush time
7481 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7482 - make TCP zerocopy and pinger support enabled by default (2476)
7483 - sync writes completed after process exits caused crashes (2319)
7484 - maintain correct mount count on the MDS (2356)
7485 - backout 1557, because 2316 wasn't really fixed
7486 - better file I/O statistics gathering in /proc
7487 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7488 - another kernel patch to fix zero-copy TCP function export
7489 - don't take duplicate lock when processing re-sent getattr (2420)
7490 - lctl uses obd_self_export instead of creating new conn (2353)
7491 - MDS/OST recovery case which requires object creation asserted (2425)
7492 - move lfs from /usr/sbin to /usr/bin in packages
7493 - fix race between mds_client_add and mds_client_free (2417)
7494 - use kmalloc instead of slabs in portals (2430)
7495 - don't create duplicate records when a failover MDS is present (2442)
7496 - remove unnecessary mount age check (2332)
7497 - don't remove directory inodes from locks prematurely (2451)
7498 - don't break if MDS service name is the same as hostname (2103)
7499 - fix races in client write RPC generation when cache full (2482)
7501 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7504 - remove now-unused request->rq_obd (278)
7505 - if an allocation fails, print out how much memory we've used (1933)
7506 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7507 - add the "configurable stack size" patch to most series files (1256)
7508 - ability to write large log records, for 100+ OST configs (2306)
7509 - fix NULL deref when filter_prep fails (2314)
7510 - fix operator precedence error in filter_sync
7511 - dynamic allocation of socknal TX descriptors (2315)
7512 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7513 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7514 - ensure that CWARN messages in Portals always get to the syslog
7515 - __init/__exit are not for prototype decls (ldlm_init/exit)
7516 - x86-64 compile warning fixes
7517 - fix gateway LMC keyword conflict (2318)
7518 - fix MDS lock inversions in getattr/reint paths (1844)
7519 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7520 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7521 - DLM race condition prevented some lock evictions (2328)
7522 - ENOMEM detection and retry on socknal sends (2230)
7523 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7524 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7525 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7526 - better error messages when a client is rejected during recovery (1505)
7527 - avoid cancelling locks which were never granted, after failure (2330)
7528 - fix i_sem/journal inversion in mds_client_add (2333)
7529 - fix truncate/getattr lock cycle deadlock (2334)
7530 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7531 - fix two rare exit paths which could leak an l_lock() ref (2321)
7532 - fencepost error in MDS/OST orphan recovery (2226)
7533 - make log record alignment 8 bytes (1988)
7534 - lstripe now fails when requested offset > ost_count (2237)
7535 - ensure that all kernel series have a complete list.h (1607)
7536 - fix crashes in special-file operations (2316)
7537 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7539 - allow configurable automake binary, for testing new versions
7540 - small update to the lfs documentation
7542 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7544 * fix negative export reference count in fsfilt_sync (2312)
7546 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7547 * release candidate 0.9.1
7549 - orphans are moved into the PENDING directory for possible recovery
7550 - replayed opens now open by fid for orphan/rename safety (1042)
7551 - last close of an orphan inode generates a transno (683)
7552 - chdir() and mount() now pin the directory entry (1020)
7553 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7554 - discard very old requests without processing them (1502)
7555 - remove some common, well-understood CERRORs (1505)
7556 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7557 - clear "grant" flags in OST replies until OST grant code lands (1644)
7558 - fix read performance by not clobbering i_blksize on client (1598)
7559 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7560 - make LRU size a /proc tunable, clears locks when reduced (707)
7561 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7562 - server locks take references on exports now (1558)
7563 - build fixes for 2.4.20-rh trees (1663)
7564 - return an error from lov_create if all OSCs are inactive (1751)
7565 - fix import levels when a reconnect happens without a timeout (1597)
7566 - exit early from mds_open if we get a lookup error (1749)
7567 - partial page read at EOF wouldn't wait for disk before sending (1642)
7568 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7569 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7570 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7571 - write/truncate lock inversion (1639)
7572 - Don't auto-load obdclass, portals modules during cleanup (1495)
7573 - fix timestamps from jumping to "now" (1763)
7574 - extra journal assertions (1648)
7575 - add an extra multiunlink test (1771)
7576 - fix read_record/write_record API (1776)
7577 - fix leak of offset_extent, possible incorrect i_size later (1772)
7578 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7579 - Debugging for the kqswnal_get_idle_tx problems (1820)
7580 - Allow recovery to be attempted multiple times (1536)
7581 - Write out MDS last_rcvd file after it is first created (1600)
7582 - Fix tx_descriptor leak in failed transmit situations (1827)
7583 - ext3 journaling fixes for assertion failure after IO error (1871)
7584 - class_export_put() on freed export after completion AST error (1896)
7585 - Fix revalidate looping in VFS (1322)
7586 - Don't access a freed export during MDS_REINT timeout (1521)
7587 - Add open-unlink recovery support on the MDS (1673,1764)
7588 - Return an error if no MDS data was read from last_rcvd (1946)
7589 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7590 - Don't setup a disk device that doesn't match exported UUID (317)
7591 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7592 - avoid committing NULL handle in force close
7593 - local.sh is now a one-stripe LOV configuration
7594 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7595 - fix ext3 htree duplicate directory entry corruption (1516)
7596 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7597 - update to new LOV EA format (2097)
7598 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7600 2003-06-15 Phil Schwan <phil@clusterfs.com>
7603 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7604 - per-import/export recovery handling (958, 931, 959)
7605 - multiple last-rcvd slots, for serving multiple FSes (949)
7606 - connections are again shared between multiple imp/exports (963, 964)
7607 - "umount -f" would hang if any requests needed to be sent (393, 978)
7608 - avoid pinning large req buffer by copying for queued messages (989)
7609 - add "uuid" to "lctl device" command to help upcalls (991)
7610 - "open" RPCs with transnos would confuse recovery counters (1037)
7611 - do proper endian conversion of all wire messages (288, 340, 891)
7612 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7613 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7614 - fix ptlrpc_connection leak in target_handle_connect (1174)
7615 - fix import refcounting bug in OST and MDS cleanup (1134)
7616 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7617 - fix very unlikely obd_types race condition (501)
7618 - remove osc_open hack for echo_client (1187)
7619 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7620 - a failure in read_inode2 leads to deadlock (1139)
7621 - cancel ack-locks as soon as transaction is committed (1072)
7622 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7623 - make sure to commitrw after any preprw to avoid deadlock (1162)
7624 - failing to execute a file in a lustre FS would lock inode (1203)
7625 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7626 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7627 - obd_uuid2tgt would walk off the end of the list (1255)
7628 - on IA64 the osc would give portals incorrect bulk size (1258)
7629 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7630 - fix lock inversion caused by new llite matching code (1282)
7631 - limit the number of dirty pages on a client to 10MB (1286)
7632 - timed out locks were not being corrected cancelled (1289)
7633 - fix O_DIRECT above 4GB on IA-32 (1292)
7634 * major user-visible changes
7635 - fail out/fail over policy now controlled by the upcall (993)
7637 - add OBD_PING to check server availability and failure (954)
7638 - lustre messages are now sent in sending host order (288, 340, 891)
7639 - add eadatalen to MDS getattr reply (340)
7640 - OST read replies may contain second buffer, with per-page status (593)
7642 2003-03-11 Phil Schwan <phil@clusterfs.com>
7645 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7646 - failed open()s could cause deadlock; fixed (867, 869)
7647 - stop cancelling OST locks when files are closed (481)
7648 - overlapping XID spaces caused network corruption (851, 853)
7649 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7650 - setattr_raw added, to do single-RPC, server-side setattrs
7651 - lmc/lconf syntax change for OST UUIDs
7652 - fix crashy race condition between ptlrpc_free_req and osc_close
7653 - don't use request in mdc_enqueue if we hit a timeout (889)
7654 - don't set the inode i_size for regular files from the MDS (896)
7655 - handle out of order completion AST (842)
7656 - don't LBUG if a lock request times out after receiving AST (913)
7657 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7658 - if a bad lock AST arrives, send an error instead of dropping entirely
7659 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7660 - fix leak in bulk IO when only partially completed (899, 900, 926)
7661 - fix O_DIRECT for ia64 (55)
7662 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7663 - C-z after timeout could hang a process forever; fixed (977)
7665 - client-side I/O cache (678, 924, 929, 941, 970)
7667 - READPAGE and SETATTRs which don't take server-side locks get
7670 2003-02-11 Phil Schwan <phil@clusterfs.com>
7673 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7674 - Fix fsfilt_extN_readpage() to read a full page of directory
7675 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7676 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7677 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7678 - Most intent operations were removed, in favour of a new RPC mode
7679 that does a single RPC to the server and bypasses most of the VFS
7680 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7681 - Aggressively cancel local locks on DLM servers
7682 - mds_reint_unlink sends EA to the client if it's the last nlink.
7683 client uses that EA to unlink OST objects.
7684 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7685 - recursive symlinks were fixed (439)
7686 - fixed NULL deref in DEBUG_REQ
7687 - filter_update_lastobjid no longer calls sync, which annoyed extN
7688 - fixed multi-client small-writes to a single file problem (445)
7689 - fixed mtime updates during file writes (607)
7690 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7691 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7692 - fixed Linux OST size reporting problem (444, 656)
7693 - OST now updates object mtime with writes or setattr (607, 619)
7694 - client verifies file size before zeroing page past EOF (445)
7695 - OST now writes last allocated objid to disk with allocation (108)
7696 - LOV on echo now works (409)
7698 - mds_reint_unlink sends a new buffer, with the EA included. this
7699 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7700 set if a regular file was being unlinked, and it was the last link
7701 - use PtlGet from the target for bulk writes (315)
7702 - OST now updates object mtime with writes or setattr (607, 619)
7703 - LDLM now has a grant-time callback to revalidate locked items, if
7705 - Many MDS operations were reorganized to combat race conditions
7707 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7708 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7710 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7713 - Fully reactivate OST imports after reconnection (512, others)
7714 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7715 - More graceful error handling for truncating on dead OST (515)
7716 - Don't error out unless we're actually accessing dead stripes (474)
7717 - Fix garbage sizes when stripes are missing (410)
7718 - LRU counters were broken, causing constant lock purge (433, 432)
7719 - garbage on read from stripes with failed OSTs (441)
7720 - mark OSCs as active before reconnecting during recovery (438)
7721 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7722 - lfind did not preserve OST order in output (443)
7723 - symlinks cause hung clients, incorrect data (439)
7724 - stop dereferencing request after dropping refcount (457)
7725 - don't LASSERT(spin_is_locked) on non-SMP (455)
7726 - fixes for many rename() bugs
7727 - fstat didn't correctly synchronize attributes (399)
7728 - server must handle lock cancellation during blocking AST prep (487)
7729 - bulk descriptors were free()d too soon (511)
7730 - fix paths in lconf, which would load incorrect modules (451, 507)
7731 - fix confusing lconf 'host not found' error message (386)
7732 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7733 - fix race condition in mdc_blocking_ast() for inode access (526)
7734 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7735 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7736 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7737 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7738 - fix lstripe in several ways (kernel side) (527)
7739 - fix request leak in ldlm_cli_enqueue (262)
7740 - incorrect OSC was marked inactive after OST failure
7741 - call mds_fs_cleanup before unmounting filesystem (524)
7742 - fix races between taking ns_lock and ldlm_lock_change_resource
7743 - fix races updating LOV export open file list
7744 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7745 - fix recovery NULL deref in ldlm_cli_cancel_unused
7746 - fix some DLM races by using new hash table for lock handles (419)
7747 - permit the client to specify desired inodes, at replay
7748 - duplicate requests when we queue them for replay reintegration
7749 - fix last_rcvd offset calculation
7750 - sync after each recovered transaction, so we always make progress
7751 - never, not always, ERESTART requests without transnos
7752 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7754 - skip replay if the MDS says that the client is already connected
7755 - don't check for a recovery-enabled export to match lctl's UUID
7756 - don't INC_USE_COUNT for phantom exports
7757 - don't crash when cleaning up phantom exports (567)
7758 - don't double-finish or set replay data for errored mdc_open requests
7759 - abort requests when they time out, so we don't get old replies
7760 - send/receive replies for AST messages again
7761 - if the client says that it doesn't have the lock, cancel it on the
7763 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7764 mark it as destroyed, it will all work out when decref is called
7765 - fix module use counts (22, 581)
7767 - ASTs now expect a reply (server cancels lock on error reply)
7769 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7772 - fix many simultaneous client startup (392)
7773 - fix dentry->d_it clobbering
7774 - credentials weren't being shipped for readdir/getattr operations
7775 - remove invalid assertions triggered during some concurrent MD
7777 - proper Lustre versions added (336, 389)
7778 - fix memory leak for create error case (398)
7779 - fix LOV locking bug that would get cli/srv out of sync
7780 - fix echo client over LOV (409)
7781 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7782 - fix double-O_EXCL intent crash (424)
7783 - avoid sending multiple lock CANCELs (352)
7785 - MDS can do multi-client recovery (modulo bugs in new code)
7787 - many updates, edits, cleanups
7789 2002-11-18 Phil Schwan <phil@clusterfs.com>
7792 - fix null d_it dereference (346)
7793 - fix full OST/dbench hang (333)
7794 - fix permission problem with file removal (286)
7795 - fix removal of OSCs from LOV when they fail
7796 - fix NULL deref during bulk timeout (214)
7797 - fix problems related to multiple filesystems on one MDS (241)
7798 - fixed serious subtle metadata locking bugs
7799 - free locks on clients when inodes are removed due to memory
7801 - fix inode pointer in lock data (285)
7802 - partial support for multiple MDS on a single host (241)
7803 - data locks weren't cancelled at clear_inode time (290, 311)
7804 - intent locks could lead to unbounded lock growth (205)
7805 - added a maximum lock count, an LRU list, and a flusher
7806 - fix multiple rename (365)
7807 - properly abstracted the echo client
7808 - OSC locked 1 byte too many; fixed
7809 - rewrote brw callback code:
7810 - fixed recovery bugs related to LOVs (306)
7811 - fixed too-many-pages-in-one-write crash (191)
7812 - fixed (again) crash in sync_io_timeout (214)
7813 - probably fixed callback-related race (385)
7815 - Add capability to MDS protocol
7816 - LDLM cancellations and callbacks on different portals
7818 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7821 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7822 - defer open object destruction to close time (601981, 138)
7823 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7824 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7825 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7826 - don't call obd_setattr in ll_file_release for destroyed objects
7827 * protocol change to lustre_msg: move |version| and add |flags|
7828 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7829 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7830 * added replay of create, unlink, link and rename operations during
7831 MDS failover; recovery should be much more robust now
7832 * remove failed OSCs from LOVs (only lov_create uses this so far)
7833 * the lustre-HOWTO was brought (more) up to date (582544)
7835 2002-10-23 Phil Schwan <phil@clusterfs.com>
7838 - in-use dentries weren't being reused properly (617851)
7839 - prevent multiple LDLM setup (599178)
7840 - fix LOV size calculations for truncate (617853)
7841 - fix client handling of MDS intent errors (POSIX)
7842 - fix permission bug in lovstripe.c test (624321)
7843 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7844 - truncate past end of file could corrupt data
7845 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7846 - a race in recovery could return ETIMEDOUT to apps (623947)
7847 - building outside the source directory was fixed
7848 * the lustre-HOWTO was brought (more) up to date (582544)
7849 * major progress was made on recovery functionality
7851 2002-10-10 Phil Schwan <phil@clusterfs.com>
7854 - recovery deadlock fix
7855 - rm -rf causes LBUG fix (617817)
7856 - file open by multiple tasks fix (618962)
7857 - directory permissions bugs (602707 and 620007)
7858 - journal_stop fixed with locking (611313)
7859 - O_APPEND failures resolved (618273, perhaps 614459)
7860 - lconf PATH fix (619770)
7861 - IA64 build fix (621450)
7862 - RPC buffer sizes scale with amount of memory
7864 2002-10-01 Phil Schwan <phil@clusterfs.com>
7867 - locks would be cancelled without throwing away data pages,
7868 resulting in inconsistent data (605627)
7869 - inode attributes were not always being refreshed (605627, 612449)
7870 - lconf now continues to cleanup after lctl reports an error
7871 - MDS now enforces user permissions (602707)
7872 - lprocfs cleanup fixed, but not yet enabled (614157)
7873 - fixed infinite server hang, should a client not respond to an AST
7874 - avoid going into recovery if user calls readlink() with a buffer
7875 that's too small (613941)
7876 - AST RPCs no longer require replies (614867) -- this may be changed
7877 - don't crash server if client sends an IOV that's too big (611336)
7878 - fixed lock conversion deadlock (611892)
7879 - fixed the following of symlinks (614622)
7880 * recovery: the server can remove locks from a client that dies, other
7881 clients can make progress
7882 * more extN patch fixes
7883 * compile-time configurable ptlrpc buffer allocations
7885 - collaborative read cache document
7886 - Lustre Lite Performance CDR document-in-progress
7888 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7891 - fix typo in patch-2.4.18
7893 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7896 - clear ptlrpc request each time in handle_incoming_request()
7897 - unlink of files now destroys the object on the OST
7899 2002-09-19 Peter Braam <braam@clusterfs.com>
7901 * add hard link support
7902 * change obdfile creation method
7903 * kernel patch changed
7905 2002-09-19 Peter Braam <braam@clusterfs.com>
7908 - stack overflow bug in extN fixed
7910 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7912 * documentation updates
7913 - add man pages for config tools
7914 - update tests/README to describe testing with new config tools
7915 - finish metadata API descriptions
7916 * bug fixes and cleanups
7917 - statfs workaround for 16TB limit
7918 - LOV stripe allocation improved, can stripe on subset of OSTs
7919 - LOV file size/IO offset was wrong for files > 4GB in size
7920 - object EA data was being dropped, caused files to be unreadable
7921 - memory overflow with non-LOV OST caused memory corruption
7922 - fixed regression tests to work with new config tools, obdfilter
7923 - fixed bug when directory size became larger than 1 block
7924 - fixed bug (for single client case) when PWD was deleted
7925 - invalidate local directory pages when doing intent-based ops
7926 - avoid LDLM oops when lock callback contained bad data
7928 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7930 * documentation updates
7931 * bug fixes and cleanups
7932 - configuration tools
7935 - 64-bit compile warnings
7936 - 64-bit internal statfs data
7938 * test_brw on persistent OST devices
7940 * lprocfs (disabled)
7942 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7944 * documentation updates
7945 * bug fixes and cleanups
7946 * configuration tools
7948 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7951 * many small fixes to 0_5_4
7952 * io/network handling
7953 * thinkos in MDS operations
7955 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7958 * crucial basic fixes to 0.5.3
7959 * IOR, Iozone work over Elan
7962 2002-08-07 Phil Schwan <phil@clusterfs.com>
7963 * version 0_5_3, our first alpha
7964 * we use the new Portals iovs
7965 * documentation updates
7966 * bug fixes and cleanups
7967 * small changes in the DLM wire protocol
7969 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7970 * version 0_5_1 with some initial stability,
7971 * locking on MD and file I/O.
7972 * documentation updates
7973 * several bug fixes since 0.5.0
7974 * small changes in wire protocol
7976 2002-07-18 Phil Schwan <phil@clusterfs.com>
7978 * delivered as Lustre Light Alpha
7979 * fixed a crash after handling invalid MDS requests
7980 * fixed directory pages for architectures with non-4k pages sizes
7982 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7983 * release version v0_4_4
7984 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7986 * Fixes a number of interruption problems with OST operations.
7987 * Update documentation for portals header changes
7988 * Move all wire protocol structs/defines to lustre_idl.h
7989 * Fixes symlink length bug.
7990 * Add tcpdump to repository.
7992 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7993 * release version v0_4_3
7994 * Fixes statfs for inodes on extN.
7995 * Fixes bug in runtests which would delete /etc/hosts.
7996 * Use 64-bit object IDs wherever possible (not into VFS though)
7997 Remove ost_get_info, which is unused by lustre, and out of date.
7999 2002-07-03 Peter Braam <braam@clusterfs.com>
8000 * release version v0_4_2 Fixes a lookup error (type not passed)
8001 * move forward to head of Portals
8002 * move forward to latest Lustre kernel
8004 2002-06-25 Peter Braam <braam@clusterfs.com>
8005 * release version v0_4_1. Hopefully stable on single node use.