1 2013-02-28 Intel Corporation
4 2.6.18-348.3.1.el5 (RHEL 5)
5 2.6.32-358.2.1.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 2012-05-31 Whamcloud, Inc.
14 * Support for kernels:
15 2.6.18-308.11.1.el5 (RHEL 5)
16 2.6.32-279.5.1.el6 (RHEL 6) - patchless client only
17 * Recommended e2fsprogs version: 1.41.90.wc4
19 See http://wiki.whamcloud.com/display/PUB/Changelog+1.8 for more information.
21 -------------------------------------------------------------------------------
23 2011-10-31 Whamcloud, Inc.
25 * Support for kernels:
26 2.6.18-274.3.1.el5 (RHEL 5)
27 2.6.32-131.12.1.el6 (RHEL 6) - patchless client only
28 * Recommended e2fsprogs version: 1.41.90.wc3
30 See http://wiki.whamcloud.com/display/PUB/Changelog+1.8 for more information.
32 -------------------------------------------------------------------------------
34 2011-06-30 Whamcloud, Inc.
36 * Support for kernels:
37 2.6.18-238.12.1.el5 (RHEL 5)
38 2.6.32-131.2.1.el6 (RHEL 6) - patchless client only
39 * Recommended e2fsprogs version: 1.41.90.wc3
40 * Use ext4-based ldiskfs as default for RHEL5
41 * Add support for 24TB LUN (RHEL5 server only)
43 Severity : enhancement
45 Description: Add support for RHEL5.6 (2.6.18-238.12.1.el5)
47 Severity : enhancement
48 Jira : LU-62, LU-73, LU-402, LU-130
49 Description: Add RHEL6 client support
53 Description: Fix client crash introduced in 1.8.5
54 Details : The patch from bugzilla ticket 18213 landed for 1.8.5 can cause
55 a crash in mdc_exit_request() when the process is interrupted.
59 Description: ls -l reports wrong file size
60 Details : Always update LVB from disk when a glimpse callback returns an
63 Severity : major (2.6.32-based kernels only)
65 Description: Add workaround for a race causing an assertion failure in
66 clear_inode() on lustre clients.
70 Description: Fix interoperability issue between 1.8 clients and 2.x servers
71 Details : The ldlm hash handling in Lustre 2.x is incompatible with the one
72 used in Lustre 1.8. Add support for 64-bit dir name hash to the
77 Description: slow IO with read-intense application
78 Details : Align the readahead extent by 1M after when it is trimed by
83 Description: Skip quotacheck on administratively disabled OSTs
87 Description: Some autogen improvements
89 Severity : enhancement
90 Jira : LU-123, LU-296, LU-230
91 Description: Add support for yaml data logging and auster script.
93 Severity : enhancement
95 Description: Allow OSTs to be created with no primary node
96 Details : Add a --servicenode parameter for mkfs.lustre to treat all
97 service nodes equally.
99 -------------------------------------------------------------------------------
101 2011-05-12 Oracle, Inc.
103 * Support for kernels:
104 2.6.16.60-0.42.8 (SLES 10),
105 2.6.27.39-0.3.1 (SLES 11),
106 2.6.18-194.3.1.el5 (RHEL 5)
107 2.6.18-194.3.1.0.1.el5 (OEL 5)
108 * Client support for unpatched kernels:
109 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
110 2.6.16 - 2.6.32 vanilla (kernel.org)
111 * Recommended e2fsprogs version: 1.41.10-sun2
112 * The async journal commit feature (bug 19128) and the cancel
113 lock before replay feature (bug 16774) are disabled by default.
117 Description: llap pages are moved to head of LRU list instead of tail
121 Description: disable statahead by default due to important races found in the
122 statahead implementation.
126 Description: release request ref through ptlrpc_server_drop_request()
127 Details : make sure the request is protected by rq_refcount while
128 being processed in ptlrpc_server_handle_req_in().
132 Description: correct check for transno value in filter_finish_transno()
136 Description: Set body->eadatasize in mdc_getattr_pack()
140 Description: Handle unsent requests with rq_net_err in ptlrpc_check_set()
144 Description: apps stuck in ptlrpc_check_set() during direct I/O
145 Details : start async bulk unregistering at the same time at reply unlink
149 Description: lfs setstripe --pool broken in 1.8.4
150 Details : fix llapi_search_fsname() to handle relative pathnames
154 Description: Only force the mode change if we're changing the size as well
155 Details : Fix regression introduced in 1.8.0. The offending code was added
156 by commit 77ba4b2141d04180211efa8a75c11ab0abf7fafb to remove
157 setgid/setuid bits when do_truncate() is called on the file.
158 We should only force the change when that occurs, similarly to
159 ll_setattr() in lustre/llite/llite_lib.c
161 Severity : enhancement
163 Description: add procfs tunable to enable/disable lockless direct I/O
164 Details : llite.lustre-*.lockless_direct_io=0 will disable default semantics
165 of direct I/O that forces it to be lockless. lockless_direct_io
166 value, however, will be ignored if per-file LL_FILE_LOCKED_DIRECTIO
171 Description: Lustre servers crashing with NULL pointer errors
172 Details : Make sure the request is protected by rq_refcount while being
173 processed in ptlrpc_server_handle_req_in().
177 Description: kernel BUG at drivers/scsi/sd.c
178 Details : Remove sd iostats patch from sles11 patch series.
182 Description: Modified value of at_min is not taken into account
186 Description: lustre grants flock exclusive locks to two fd's in the same process
190 Description: LBUG in the mfs_verify_child()
194 Description: ASSERTION(atomic_read(&client_stat->nid_exp_ref_count) == 0)
195 Details : Reconnecting from a different nid causes old per-nid stats not to be
200 Description: create proper macro check for bdi interface.
202 -------------------------------------------------------------------------------
204 2010-10-29 Oracle, Inc.
206 * Support for kernels:
207 2.6.16.60-0.69.1 (SLES 10),
208 2.6.32.19-0.2.1 (SLES11),
209 2.6.18-194.17.1.el5 (RHEL 5)
210 2.6.18-194.17.1.0.1.el5 (OEL 5)
211 * Client support for unpatched kernels:
212 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
213 2.6.16 - 2.6.32 vanilla (kernel.org)
214 * Recommended e2fsprogs version: 1.41.10-sun2
215 * The async journal commit feature (bug 19128) and the cancel
216 lock before replay feature (bug 16774) are disabled by default.
220 Description: atime is not properly updated on an MDS
222 Severity : enhancement
224 Description: Update to RHEL5.5 kernel 2.6.19-194.17.1.el5.
225 Update to OEL5.5 kernel 2.6.19-194.17.1.0.1.el5.
227 Severity : enhancement
229 Description: Update to SLES10 SP3 kernel 2.6.16.60-0.69.1.
232 Frequency : only with SLES10
234 Description: Use OFED "KMP" provided by Novell
235 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
236 Lustre is now built against this package. That means you need to
237 install the ofed-kmp package from Novell for the patchless client
238 and from our download site for the server. Note that the ofed-kmp
239 that Novell ships may not exactly match the kernel version but
240 should still be compatible.
242 Severity : enhancement
244 Description: Update SLES11 SP1 kernel to 2.6.32.19-0.2.1.
248 Description: Enabling quotas fails with non-consecutive OST numbering.
252 Description: Fix kernel warning due to lookup_on_len() called without i_mutex
257 Description: Account direct i/o inflight rpcs separately from non-direct i/o so
258 that direct i/o, which is limited by max_rpcs_in_flight, should not
259 block non-direct i/o, which is not limited by max_rpcs_in_flight.
263 Description: Fix per-NID reporting on outstanding writes
267 Description: Reduce stack pressure by uninlining some mds and ptlrpc functions.
271 Description: Remove LASSERT in lprocfs_rd_conn_uuid() since conn == NULL is a
276 Description: fix obdo leak issue in ll_setattr_raw()
280 Description: limit MMP interval
282 Severity : enhancement
284 Description: add several lfs ost enhancements
288 Description: Too many default ACLs break directory access on new directories
292 Description: Lustre inode size is not coherent across nodes.
293 Details : Update lvbo from disk when AST fails with EINVAL. Lvbo is updated
294 on EINVAL error in ldlm_handle_ast_error(). The updates in
295 filter_intent_policy() and ldlm_cb_interpret() have been removed as
300 Description: Oops at __percpu_counter_add+0x1b
301 Details : Use bdi_init()/bdi_destroy() to proper initialize backing_dev_info
306 Description: add mount option to generate 32bit ino, this can be used for 32bit
307 application compatibility.
311 Description: keep reference count for "lli_sai" to prevent it to be released
312 when "statahead_enter()"
316 Description: allow quotacheck over OSTs with sparse indices
320 Description: Objects not getting deleted for files which have been removed
321 Details : ll_have_md_lock() should differentiate between CR and CW OPEN
326 Description: pin object's inode in memory to avoid certain timeouts
330 Description: fix LBUG when obdfilter-survey is interrupted.
332 -------------------------------------------------------------------------------
334 2010-07-31 Oracle, Inc.
336 * Support for kernels:
337 2.6.16.60-0.42.8 (SLES 10),
338 2.6.27.39-0.3.1 (SLES11),
339 2.6.18-194.3.1.el5 (RHEL 5)
340 2.6.18-194.3.1.0.1.el5 (OEL 5)
341 * Client support for unpatched kernels:
342 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
343 2.6.16 - 2.6.30 vanilla (kernel.org)
344 * Recommended e2fsprogs version: 1.41.10-sun2
345 * The async journal commit feature (bug 19128) and the cancel
346 lock before replay feature (bug 16774) are disabled by default.
348 Severity : enhancement
350 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
351 2.6.18-194.8.1.0.1.el5.
353 Severity : enhancement
355 Description: using inkernel OFED stack for rhel5 & oel5.
357 Severity : enhancement
359 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
360 Details : lfs_migrate does a "poor man's" migration of files from their
361 current OST layout to a new OST layout as chosen by the MDS.
365 Description: mds_orphan_add_link()) error linking orphan to PENDING
366 Details : quota limits might disallow linking orphans to PENDING
367 when unlinking a file - temporary raise threads' privileges
368 when processing unlinks.
370 Severity : enhancement
372 Description: add conf-param -d option to remove permanent settings.
373 Details : Add the ability to remove permanent lctl conf_param settings.
374 (Previously conf_param settings could only be changed, not
375 removed.) This also provides a method to change failover
376 nid locations. Improve lctl man page.
378 Severity : enhancement
380 Description: add list_param to b1_8 and add "-R" option to list params
383 Severity : enhancement
385 Description: lfs quota output is not very convenient for awk/sed-parsing
386 Details : Some positions in lfs quota output table could be empty or
387 non-empty which made it hard to parse it with scripts, now a dash
388 is put instead of space where there is not supposed to be any data.
390 Severity : enhancement
392 Description: fix obdfilter-survey script to work properly with remote oss-s
394 Severity : enhancement
396 Description: add new OBDFILTER_SURVEY test suite
398 Severity : enhancement
400 Description: add new multiple mount protection (MMP) test suite
402 Severity : enhancement
404 Description: add support for async journal commit in echo client
406 Severity : enhancement
408 Description: allow userland programs to include <lustre/lustre_idl.h>
409 from stardard include directories
411 Severity : enhancement
413 Description: The prune-icache-use-trylock is no longer needed now that
414 the patch from bug 20008 is landed.
418 Description: The shrink grant feature is still active on the client although the
419 connect flag is not set.
423 Description: Don't leak grant space if the write failed with quota exceeded.
427 Description: Don't consume grant space twice on recoverable resent.
431 Description: a race condition could lead to SIGBUS being sent to an
432 application using mmap-ped files from Lustre
433 Details : truncate_complete_page implementation for the patchless
434 client could arbitrarily unset PG_Uptodate flag for a
435 page being kicked from the page cache, an uptodate check
436 right after a readpage call in filemap_fault could fail
437 because of that as though the page read had been unsuccessful.
441 Description: dlm lock slab shrinking is not efficient
442 Details : The dlm_locks slab can grow significantly and consumes a lot of
443 memory on the server. Set a hardlimit to grant_plan.
447 Description: Lustre does not do 1MB IOs to HW RAID
448 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
449 This is what we do already for SLES kernels.
453 Description: bump maximum number of phys/hw segments in the SLES11 kernel
454 until s/g chaining works properly.
459 Description: LSI Fusion MPT driver hacks to improve performance
460 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
462 Severity : enhancement
464 Description: increase default md stripe_cache_size to 16k
467 Bugzilla : 15587/21439
468 Description: don't handle security.capability xattr
469 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
470 This results in additional getxattr calls, causing VBR test
471 failures as well as a preformance drop when writing.
475 Description: obdfilter-survey is no longer working
476 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
477 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
481 Description: ll_shrink_cache does not handle __GFP_FS properly
485 Description: lfs getstripe shows wrong info for directories
486 Details : Set correct LOVEA default values for filesystem-wide.
490 Description: FSX checksum false positves due to mmap IO
491 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
492 checksum errors, if the flag is set on a request.
496 Description: file operations after eviction have successful return values
497 Details : use vfs ->flush callback to return any pending async errors
502 Description: mdsrate fails to write after 1.3+M files opened
503 Details : decrease memory usage on clients by recycling dentries and
508 Description: obdfilter-survey gives unreasonably high numbers
509 Details : Wait for all threads to complete when running test_brw.
513 Description: do not set lustre read_only device when server umount and keep
514 client records for recoverable ones
518 Description: move sync_on_lock_cancel tunable to the obdfilter layer
519 Details : move the tunable to trigger a journal flush on lock cancel from
520 the ost layer to the obdfilter layer. This tunable is useful
521 when using the async journal commit feature.
525 Description: exp->exp_nid_stats == NULL in filter_tally()
526 Details : fix race with per-nid stats by delaying procfs cleanup until
531 Description: extent lock cancellation on client can keep the cpu busy for too
536 Description: Do not fail OST activation when a llog is not found, just
537 issue an error message.
541 Description: Don't enable extents by default for MDT.
545 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
546 the AT code can access it concurrently while sending early
551 Description: Disable lockless truncate by default since it is sometimes flawed
552 and causes the write_disjoint test to fail.
556 Description: OSSs which don't have the patch from bug 20278 can trigger an
559 Severity : enhancement
561 Description: don't print message to the console when we have not managed to
566 Description: The MDS fails to synchronize OSTs which registered with the MGS
567 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
568 raised too early and thus discarded by the MDT stack.
569 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
574 Description: Fix race when the ping evictor and a service thread execute
575 target_recovery_check_and_stop() concurrently.
579 Description: quota broadcast can trigger a LBUG on the MDT if there are
584 Description: Resetting the lov_objid values to last_id reported by the OST
585 during orphan recovery is incorrect and can cause the same
586 objects to be allocated twice.
588 Severity : enhancement
590 Description: "weak-modules" support
591 Details : Implement "weak-modules" support which enables kernel modules
592 to be used with any kernel that implements the same kABI. In
593 order to achieve this modules are now installed in
594 /lib/modules/$(uname -r)/updates/kernel on all distributions.
596 Severity : enhancement
598 Description: add writeconf as mount option
600 Severity : enhancement
602 Description: produce debuginfo packages for SLES.
604 Severity : enhancement
606 Description: add failover nidlist to the import proc file.
608 Severity : enhancement
610 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
611 clients and 2.0 servers.
613 Severity : enhancement
615 Description: lfs find -s does not work correctly because of a bug in
620 Description: ll_read_ahead_page() must validate the dlm lock before using
625 Description: Prevent failover nids from registering with MGS first.
629 Description: fix lock inversion in ll_setattr_raw().
633 Description: object allocation is not balanced across OSTs.
634 Details : osc_precreate() should return 0, if there are enough objects left.
636 -------------------------------------------------------------------------------
638 2010-04-30 Oracle, Inc.
640 * Support for kernels:
641 2.6.16.60-0.42.8 (SLES 10),
642 2.6.27.39-0.3.1 (SLES11),
643 2.6.18-164.11.1.el5 (RHEL 5)
644 2.6.18-164.11.1.0.1.el5 (OEL 5)
645 * Client support for unpatched kernels:
646 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
647 2.6.16 - 2.6.30 vanilla (kernel.org)
648 * Recommended e2fsprogs version: 1.41.10-sun2
649 * The async journal commit feature (bug 19128) and the cancel
650 lock before replay feature (bug 16774) are disabled by default.
654 Description: open-unlinked directories trigger MDS LBUG
655 Details : Fix regression introduced by the patch from bug 19640.
656 ext3_inc_count() can reset nlink to 1 when the directory
657 is indexed and inode->i_nlink == 2. Work around the problem
658 by incrementing nlink by 2 instead of 1.
662 Description: Reconnects are not throttled
663 Details : Don't wake up pinger on reconnect failures and rely on regular
664 pings to trigger the next reconnection. Please note that the
665 pinger already uses a smaller interval if the import is
669 Frequency : only with NFS export
671 Description: Console flooded with error message from ll_inode_from_lock()
673 Details : in mds_open, initialize the child_res_id before enqueuing
674 the OPEN lock for the child inode, then to avoid senting
675 wrong ldlm_res_id to client.
679 Description: allow multiple instances of the same nid in NID hash
680 Details : Case of multiple separate clients from the same NID (as
681 with liblustre) is legitimate and so we should allow
682 multiple instances of the same NID in nid hash.
686 Description: the readahead code can sleep on a semaphore while holding a
688 Details : in ras_update, "lov_get_info" could be called during increasing
689 readahead windows, which tries to get the mutex lock "lov_lock"
690 while holding the spin_lock "ras_lock", then causes system
695 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
696 Details : fix assertion failure in the grant code.
700 Description: Use CNETERR (which is rate limited) in specific places in
701 the portal's LNET driver to avoid flooding the console.
705 Description: include last created object in precreate slow case
709 Description: don't do rep-ack if not created anything
710 Details : mds_open currently always put a lock into a rep-ack regardless
711 if something was created or not. This is pointless and only
712 creates needless contention. In fact the entire idea was to do
713 this for real creates as a recovery protection.
717 Description: Spurious error messages from smp_processor_id() on preemptible
719 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
720 first. The function fs_trace_get_tcd() was moved up.
724 Description: interval_erase() fix
725 Details : interval_erase() calls update_maxhigh() properly when child
730 Description: "lfs df" does not print stats for all mountpoints
731 Details : Print all mounted lustre filesystems with "lfs df"
735 Description: lfs setstripe -p not longer work with a relative pathname
736 Details : Use realpath() to provide absolute pathname.
740 Description: fix for truncated reply buffer
741 Details : reply buffer could be referred by reply_in_callback after released
746 Description: lustre.lov error when backing up symlinks with extended attributes
747 Details : Improved logic in ll_listxattr()
751 Description: properly handle null value for setattr -n lustre.lov
752 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
753 in ll_setxattr() due to no checking if "value" is NULL.
754 This command now resets to the default striping when executed
759 Description: stack overflow on lock cancellation due to fsync call
760 Details : sync_on_lock_cancel is needed for recovery when async journal
761 is enabled, but we actually just need to make sure that
762 metadata blocks have hit the journal, so doing a fs sync
763 should be enough and should consume less stack (just create an
764 empty handle and commmit it).
768 Description: using current->journal_info to store per-thread data leads
769 to problem under memory pressure
770 Details : disable the per-thread data (current->journal_info) containing
771 the lock info during I/O to work around the issue for short tem
775 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
776 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
777 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
782 Description: Cannot send after transport shutdown
783 Details : Clear imp_vbr_failed flag upon eviction
787 Description: soft lock in request set code during recovery
788 Details : during recovery, uses req->rq_set itself to replay the request
789 instead of ptlrpcd_recovery_pc
793 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
794 Details : During an ll_readahead under ll_readpage, we have seen the the
795 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
796 a page that is already locked by the readahead code.
801 Description: stop waitting for next replay transno if shutdown
802 Details : if the system is shutting down, wake up service thread blocked
803 to wait for next replay transno during recovery, then all the
804 references held by queued requests can be dropped and device
809 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
810 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
814 Description: per-nid stats should not access lustre-hash internal structures
819 Description: mount.lustre fails to pass some options to mount()
824 Description: ext4 extent allocation is slower than in ext3
825 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
826 enlarge ext4 preallocation table for 2048 4K blocks extents
831 Description: incorrect triggering of synchronous IO
832 Details : The OSC can mistakenly fall back to synchronous IO when the
833 max_dirty_mb limit is reached and no write requests have yet
834 been issued. This can occur when the dirty pages are spread
835 over many files all of which are below the optimal request size.
839 Description: Optimize quota_ctl operations by sending requests in parallel
840 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
841 Compiled from two attachments in the ticket.
845 Description: ordering issue between transaction start & i_mutex
846 Details : start the transaction earlier in llog_lvfs_destroy to get
847 transaction start and inode mutex lock nested properly.
851 Description: lru resize SLV can get stuck
852 Details : calculate SLV with a greater precision to not lose small
853 changes due to interger math truncation; round up SLV only
854 if the amount of granted locks less than the limit to not
855 get stuck with this SLV
859 Description: avoid divide-by-zero in lprocfs_rd_import()
863 Description: lfs quota failed when OSTs are down
864 Details : really return approximate block/inode usage when OSTs are down
868 Description: llobdstat fix and enhancement
869 Details : add a counter to set a limit to how many samples will be returned
870 fix a wildcard in the path to limit to obdfilter stats only
874 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
876 Details : abort lock enqueue processing sooner when we umount is in progress.
878 Severity : enhancement
880 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
881 2.6.18-164.11.1.0.1.el5.
883 Severity : enhancement
885 Description: error message improvements
886 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
887 Simplify MDT/OST service start message.
888 Suppress "changing the import ..." warning.
891 Bugzilla : 21961/17914
892 Description: ignore trailing -mdc when determining index number
895 Frequency : only with SLES11
897 Description: fix for a race condition in linux quotas implementation
898 Details : dq_flags(struct dquot) access is not properly locked which could
899 lead to certain inconsistencies when accessing it using non-atomic
900 bit operations like __set_bit in do_set_dqblk.
904 Description: return any pending async errors in close(2) using flush callback
906 -------------------------------------------------------------------------------
908 2010-01-29 Sun Microsystems, Inc.
910 * Support for kernels:
911 2.6.16.60-0.42.8 (SLES 10),
912 2.6.27.39-0.3.1 (SLES11),
913 2.6.18-164.11.1.el5 (RHEL 5)
914 2.6.18-164.6.1.0.1.el5 (OEL 5)
915 * Client support for unpatched kernels:
916 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
917 2.6.16 - 2.6.30 vanilla (kernel.org)
918 * Recommended e2fsprogs version: 1.41.6.sun1
919 * The async journal commit feature (bug 19128) and the cancel
920 lock before replay feature (bug 16774) are disabled by default.
922 Severity : enhancement
924 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
925 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
927 Severity : enhancement
928 Bugzilla : 21511/19848
929 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
931 Severity : enhancement
933 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
935 Severity : enhancement
937 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
938 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
940 Severity : enhancement
942 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
945 Severity : enhancement
947 Description: Vanilla kernel 2.6.30 patchless client support.
952 Description: bad entry in directory xxx: inode out of bounds
953 Details : fix locking issue in the rename path which could race with any
954 other operations updating the same directory.
956 Severity : enhancement
958 Description: Make watchdog timer messages to be more clear and descriptive.
962 Description: cp -p command does not preserve the dates and timestamp
963 Details : mtime could be spoiled by a write callback
967 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
971 Description: Allow non-root access for "lfs check".
972 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
974 Severity : enhancement
976 Description: quotacheck performance/scaling issues
977 Details : reduce quotacheck time on empty filesystem by skipping uninit
980 Severity : enhancement
982 Description: Enhancement for lfs(1) command to use numeric uid/gid.
984 Severity : enhancement
986 Description: Adjust locks' extents on their first enqueue, so that at the time
987 they get granted, there is no need for another pass through the
988 queues since they are already shaped into the proper forms.
992 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
993 arguments and prevent LBUG() in lustre_shrink_reply_v2().
997 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
998 exactly the specified mount options are used. Leaving off
999 any "mandatory" mount options is an error. Leaving off any
1000 default mount options causes a warning, but is allowed.
1001 Change errors=remount-ro from mandatory to default. Sanitize
1002 the mount string before storing it. Update man pages accordingly.
1006 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
1007 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
1010 Severity : enhancement
1012 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
1014 Severity : enhancement
1016 Description: Clients can replay thousands of unused locks during recovery
1017 Details : Don't replay unused locks (only read locks for now) during
1018 recovery. This feature is disabled by default and can be
1019 enabled by running the following command on the clients:
1020 lctl get_param ldlm.cancel_unused_locks_before_replay
1024 Description: can't stat file in some situation.
1025 Details : improve initialize osc date when target is added to mds and
1026 ability to resend too big getattr request is client isn't have info
1031 Description: Prevent inconsistences between linux and lustre mount structures.
1032 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1033 Make the sleep interruptible.
1035 Severity : enhancement
1037 Description: Communicate OST degraded/readonly state via statfs to MDS
1038 Details : Flags in the statfs returned from OSTs indicate whether the
1039 OST is in a degraded RAID state, or if the filesystem has
1040 turned read-only after a filesystem error is detected.
1045 Description: don't panic if EPROTO was hit when reading symlink
1046 Details : correctly handling request reference in error cases.
1051 Description: open sometimes returns ENOENT instead of EACCES
1052 Details : checking permission should be part of open part of mds_open, not
1053 lookup part. so server should be set DISP_OPEN_OPEN disposition
1054 before starting permission check.
1055 Also not need revalidate dentry if client already have LOOKUP lock.
1059 Description: enable client interface failover
1060 Frequency : on servers with multiple network interfaces
1061 Details : When a child reconnects from another NID, properly update export
1062 nid hash position and ldlm reverse import.
1064 Severity : enhancement
1066 Description: implemented direct I/O with arbitrary (nonaligned) memory
1067 addresses and file offsets.
1069 Severity : enhancement
1071 Description: added more recovery timeout options.
1073 Severity : enhancement
1075 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
1080 Description: Avoid deadlock for local client writes
1081 Frequency : only on systems with clients writing to an OST on the same node
1082 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1083 memory freeing context. This allows OST threads to set the
1084 PF_MEMALLOC flag on task structures in order to allocate memory
1085 from reserved pools and complete IO.
1086 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1087 so that the OST threads generate memory pressure and allow
1088 inactive pages to be reclaimed.
1093 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1094 Details : move ldlm namespace creation in setup phase to avoid grab
1095 _lprocfs_lock with cli_sem held.
1099 Frequency : only during format of test systems
1100 Description: Unable to run several mkfs.lustre on loop devices at the same time
1101 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1102 formatting. The solution is to proper handle the error.
1104 Severity : enhancement
1106 Description: implement async create (obd_async_create) method for osc, to avoid
1107 too long waiting new ost objects with holding ldlm lock.
1111 Frequency : occasionally during network problems
1112 Description: client not allowed to reconnect to OST because of active request
1113 Details : abort bulk requests received by the OST once the client has timed
1114 out since the client will resend the request anyway. The client
1115 also now retries to reconnect to the same server if a connect
1116 request failed with EBUSY or -EAGAIN.
1119 Frequency : rare, if used wide striped file and one ost in down.
1121 Descriptoin: don't return error if we created a subset of objects for file.
1122 Details : lov_update_create_set() uses set->set_success as index for created
1123 objects, so if some requests failed, they will have hole at end of
1124 array and we can use qos_shrink_lsm for allocate correct lsm.
1128 Description: Slow stale export processing during normal start up
1129 Details : The global mgc lock prevents OST setup to be run in parallel.
1130 Replace the global lock with a per-config_llog_data semaphore.
1134 Description: Out or order replies might be lost on replay
1135 Details : In ptlrpc_retain_replayable_request if we cannot find retained
1136 request with tid smaller then one currently being added, add it
1137 to the start, not end of the list.
1141 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
1142 Details : add cond_resched() calls to avoid hogging the cpu for too long
1143 in the hash code. Make also lustre_hash_for_each_empty() more
1146 Severity : enhancement
1148 Description: Performance improvements for debug messages with D_RPCTRACE,
1149 D_LDLM, D_QUOTA options.
1152 Frequency : only with NFS export
1154 Description: (lov_merge.c:74:lov_merge_lvb())
1155 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
1156 Details : Fix a race in the nfs export code by populating inode
1157 info while the new inode is still locked
1159 Severity : enhancement
1161 Description: add a new file in procfs called force_lbug. Writting to this
1162 file triggers a LBUG. Only for test purpose.
1166 Description: OOM killer causes node hang
1167 Details : really interrupt the sleep in osc_enter_cache on signals
1171 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1172 Details : fix race during quota release on the slave.
1174 Severity : enhancement
1176 Description: smaller hash bucket sizes, cleanups
1177 Details : increase hash table sizes and enabled rehashing for pools, quota,
1178 uuid, nid & per-nid stats.
1180 Severity : enhancement
1182 Description: Add ldiskfs maxdirsize mount option
1183 Details : add max_dir size mount option
1187 Description: panic in ll_statahead_thread
1188 Details : prevent parent thread to be killed before its child
1191 Frequency : only with 16TB device
1193 Description: unable to perform "mount -t lustre" of 16TB OST device
1194 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
1198 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
1199 Details : unregistering should be zero if no RPC inflight.
1203 Description: hyperion: Oops during metabench
1204 Details : Correct the refcount of lov_request_set
1206 Severity : enhancement
1208 Description: Add mptlinux and nxge drivers to Lustre builds
1210 Severity : enhancement
1212 Description: Fix watchdog timer message to be more clear
1213 Details : Make watchdog timer messages more clear and descriptive.
1217 Description: LNET soft lockups in socknal_cd thread
1218 Details : don't hog CPU for active-connecting if another connd is
1219 accepting connecting-requst from the same peer
1223 Description: recovery-small test_17 hang
1224 Details : Land several AT improvements & fixes.
1228 Description: MDS panic and hanging client processes
1229 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
1233 Description: OSS stuck in recovery.
1234 Details : fix race during recovery. class_unlink_export,
1235 class_set_export_delayed and target_queue_last_replay_reply
1236 may race while increasing/decreasing obd_recoverable_clients
1237 and obd_delayed_clients, causing recovery to wait forever.
1239 Severity : enhancement
1241 Description: add cascading_rw.c to lustre/tests
1245 Description: filter_last_id() NULL deref
1246 Details : lprocfs_filter_rd_last_id() should check for the fully
1247 setup obd device, before proceeding further.
1249 Severity : enhancement
1251 Description: Loadgen improvements
1252 Details : stacksize and locking fixes for loadgen
1256 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
1257 Details : The atomic_read() allowing the atomic_inc() are not covered
1258 by a lock. Thus they may safely race and trip this CERROR()
1259 unless we add in a small fudge factor (+1).
1261 Severity : enhancement
1263 Description: shrink_slab: nr=-9223362083340912175
1264 Details : fix spurious message from shrink_slab reporing negative nr
1268 Description: Quiet bogus previously committed transno error
1269 Details : suppress the "server went back in time" error message which
1270 is always printed even in the common case after a client eviction
1272 Severity : enhancement
1274 Description: Parallel statfs() calls result in client eviction
1275 Details : cache statfs data for 1s.
1279 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
1281 Details : fix serveral issues in pinger code causing clients not to ping
1282 servers for too long, resulting in evictions.
1286 Description: e2fsck should warn when MMP update interval is extended
1287 Details : print mmp_check_interval and make it possible to abort mount
1288 operation in case it takes too long.
1292 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
1293 Details : fix bug in the RHEL5's jbd2 callback patch.
1297 Description: drop number of active requests when queued for recovery
1298 Details : Now that we take a reference on the original request instead of
1299 making a copy of it for recovery. We need to drop the number of
1300 active requests or the queued requests will prevent all request
1301 processing when they exceed (srv->srv_threads_running - 1).
1303 Severity : enhancement
1305 Description: refuse to invalidate operational quota files when they are in use
1306 Details : an attempt to invalidate operational quota files on the quota
1307 master is not actually permitted by VFS (returning -EPERM), but we
1308 should not depend on that and should return the error earlier.
1312 Description: Applications stuck in jbd2_log_wait_commit during exit
1313 Details : fix deadlock between kjournald2 trying to acquire the page lock
1314 owned by an ost_io thread waiting for journal commit.
1316 -------------------------------------------------------------------------------
1318 2009-10-16 Sun Microsystems, Inc.
1320 * Support for kernels:
1321 2.6.16.60-0.42.4 (SLES 10),
1322 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1323 2.6.18-128.7.1.el5 (RHEL 5),
1324 2.6.18-128.7.1.el5 (OEL 5),
1325 * Client support for unpatched kernels:
1326 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1327 2.6.16 - 2.6.27 vanilla (kernel.org)
1328 * Recommended e2fsprogs version: 1.41.6.sun1
1329 * ext4 support for RHEL5 is experimental and thus should not be
1332 Severity : enhancement
1334 Description: Add OEL5 support.
1336 Severity : enhancement
1338 Description: Update kernel to SLES11 2.6.27.29-0.1.
1342 Description: File checksum failures with OST read cache on
1343 Details : Disable page poisoning when the bulk transfer has to be aborted
1344 because the client got evicted.
1348 Description: Don't allow make backward step on assiging osc next id.
1349 Details : race between allocation next id and ll_sync thread can be cause
1350 of set wrong osc next id and can be kill valid ost objects.
1352 Severity : enhancement
1354 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1356 Severity : enhancement
1358 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1362 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1366 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1370 Description: Error handling in osc_statfs_interpret() has been improved.
1371 Details : Check in osc_statfs_interpret() for EBADR.
1375 Description: Do not send statfs() requests to OSTs disabled by administrator.
1376 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1380 Description: Do not update ctime for the deleted inode.
1381 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1385 Description: Increase of the size of the LDLM resource hash.
1386 Details : Bump up RES_HASH_BITS=12.
1390 Description: correctly send lsm on open replay
1391 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1396 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1397 Details : filter_destroy() does not hold the DLM lock over the whole
1398 operation. If the DLM lock is dropped, filter_commitrw() can go
1399 through, causing the deadlock between page lock and i_mutex.
1400 The i_alloc_sem should also be hold in filter_destroy() while
1401 truncating the file.
1405 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1406 Details : ldiskfs_truncate calls grab_cache_page which may start page
1407 allocation under an open transaction. This may lead to
1408 calling prune_icache with consequent lustre reentrance.
1411 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1412 still up and when the OST pool feature is used
1414 Description: interop testing got LBUG when run dd with OST pool
1415 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1416 Details : down/upgrading the MDS to a version that doesn't/does support OST
1417 pool can cause clients to crash because the lsm has changed
1422 Description: missing tree_status on 1.8.1 RPM build
1423 Details : make rpms failed due because the tree_status file is missing.
1427 Description: continuing LustreError "mds adjust qunit failed!"
1428 Details : don't print message on the console when ->adjust_qunit fails.
1432 Description: don't increase ldlm timeout if previous client was evicted
1433 Details : if a client doesn't respond to a blocking callback within the
1434 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1435 when the lock is next granted.
1439 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1440 affects recovery negatively.
1441 Details : make sure all exports have been properly destroyed by the zombie
1442 thread processed before stopping the target.
1446 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1447 Details : disable write barrier for ext4/SLES11.
1451 Description: Kernel panic - not syncing: Out of memory and no killable
1452 processes... on OSS when iozone
1453 Details : fix memory leak in the journal checksum patch.
1457 Description: group quota "too many blocks" OSS crashes
1458 Details : we should keep the same uid/gid for lquota_chkquota() and
1459 lquota_pending_commit()
1463 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1464 Details : don't LBUG on release quota error. Just a workaround until the
1465 problem is understood.
1467 ------------------------------------------------------------------------------
1468 2009-07-31 Sun Microsystems, Inc.
1470 * Support for kernels:
1471 2.6.16.60-0.39.3 (SLES 10),
1472 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1473 2.6.18-128.1.14.el5 (RHEL 5),
1474 * Client support for unpatched kernels:
1475 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1476 2.6.16 - 2.6.27 vanilla (kernel.org)
1477 * Recommended e2fsprogs version: 1.41.6.sun1
1478 * File join has been disabled in this release, refer to Bugzilla 16929.
1479 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1480 of Lustre filesystem with 4K stack may cause a stack overflow. For
1481 more information, please refer to bugzilla 17630.
1482 * ext4 support for RHEL5 is experimental and thus should not be
1485 Severity : enhancement
1487 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1490 Frequency : with 1.8 server and 1.6 clients
1492 Description: correctly shrink reply for avoid send too big message to client.
1493 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1494 some problems with sending this reply to 1.6 client.
1498 Description: Repeated atomic allocation failures.
1499 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1500 to generate memory pressure and allow reclaiming of inactive pages.
1501 At the same time, do not allow to exhaust emergency pools.
1502 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1504 Severity : enhancement
1505 Bugzilla : 19846, 18289
1506 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1508 Severity : enhancement
1509 Bugzilla : 19625, 16893, 18668, 19848
1510 Description: Add support for SLES11 2.6.27.23-0.1.
1512 Severity : enhancement
1514 Description: Update client support to vanila kernels up to 2.6.27.
1516 Severity : enhancement
1518 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1520 Severity : enhancement
1522 Description: Compile with -Werror by default for i686 and x86_64.
1526 Description: resolve race between obd_disconnect and class_disconnect_exports
1527 Details : if obd_disconnect will be called to already disconnected export he
1528 forget release one reference and osc module can't unloaded.
1530 Severity : enhancement
1532 Description: move AT tunable parameters for more consistent usage
1533 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1537 Descriptoin: correctly skip time estimate if in recovery
1538 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1542 Descriptoin: OSS DeadLock
1543 Details : Use trylock to prevent deadlock when shrink icache.
1545 Severity : enhancement
1547 Description: Allow tuning service thread via /proc
1548 Details : For each service a new
1549 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1550 created that can be used to set min/max thread counts, and get the
1551 current number of running threads.
1553 Severity : enhancement
1555 Description: Add state history info file, enhance import info file
1556 Details : Track import connection state changes in a new osc/mdc proc file;
1557 add overview-type data to the osc/mdc import proc file.
1561 Description: Reduce small size read RPC
1562 Details : Set read-ahead limite for every file and only do read-ahead when
1563 available read-ahead pages are bigger than 1M to avoid small size
1568 Description: free_entry erroneously used groups_free instead of put_group_info
1570 Severity : enhancement
1572 Description: Make read-ahead stripe size aligned.
1574 Severity : enhancement
1576 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1579 Frequency : rare, connect and disconnect target at same time
1581 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1582 Details : don't call obd_disconnect under lov_lock. this long time
1583 operation and can block ptlrpcd which answer to connect request.
1586 Frequency : start MDS on uncleanly shutdowned MDS device
1588 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1589 Details : stay in waiting mds<>ost recovery finished produce random bugs
1590 due race between two ll_sync thread for one lov target. send
1591 ACTIVATE event only if connect realy finished and import have
1595 Frequency : start MDS on uncleanly shutdowned MDS device
1597 Descriptoin: aborting recovery hang on MDS
1598 Details : don't throttle destroy RPCs for the MDT.
1602 Description: Slow reads beyond 8Tb offsets.
1603 Details : Page index integer overflow in ll_read_ahead_page
1607 Description: Soft lockup on OSS after MDS failover
1608 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1609 As a conseqence, the patch from bug 18224 is not operational
1610 and the MDS export cannot be reused on the OSTs until it gets
1614 Frequency : rare, only if using MMP with Linux RAID
1616 Description: MMP doesn't work with Linux RAID
1617 Details : While using HA for Lustre servers with Linux RAID, it is possible
1618 that MMP will not detect multiple mounts. To make this work we
1619 need to unplug the device queue in RAID when the MMP block is being
1620 written. Also while reading the MMP block, we should read it from
1621 disk and not the cached one.
1624 Frequency : rare, during recovery
1626 Description: Assertion failure in ldlm_lock_put
1627 Details : Do not put cancelled locks into replay list, hold references on
1628 locks in replay list
1632 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1633 Details : create_count always drops to the min value (=32) because grow_count
1634 is being changed before the precreate RPC completes.
1637 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1640 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1641 Details : add proper locking for creation of jbd_4k slab cache
1645 Description: MMP check in ext3_remount() fails without displaying any error
1646 Details : When multiple mount protection fails during remount, proper error
1651 Description: Rare Client crash on resend if the file was deleted.
1652 Details : When file is opened, but open reply is lost and file is
1653 subsequently deleted before resend, resend processing logic
1654 breaks trying to open the file again, should not try to open.
1658 Description: add check for >8TB ldiskfs filesystems
1659 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1660 Don't allow >8TB ldiskfs filesystems to be mounted without
1661 force_over_8tb mount option
1665 Description: Client locked up when running multiple instances of an app. on
1666 multiple mount points
1667 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1668 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1672 Description: Cannot acces an NFS-mounted Lustre filesystem
1673 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1674 from a Lustre-client exporting the Lustre filesystem via NFS.
1678 Description: panic in ll_statahead_thread
1679 Details : grab dentry reference in parent process.
1682 -------------------------------------------------------------------------------
1684 tbd Sun Microsystems, Inc.
1686 * Support for kernels:
1687 2.6.16.60-0.31 (SLES 10),
1688 2.6.18-128.1.6.el5 (RHEL 5),
1689 2.6.22.14 vanilla (kernel.org)
1690 * Client support for unpatched kernels:
1691 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1692 2.6.16 - 2.6.22 vanilla (kernel.org)
1693 * Recommended e2fsprogs version: 1.40.11-sun1
1694 * File join has been disabled, refer to Bugzilla 16929.
1695 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1696 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1697 Lustre filesystem with 4K stack may cause a stack overflow. For more
1698 information, please refer to bugzilla 17630.
1700 Severity : enhancement
1702 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1704 Severity : enhancement
1706 Description: Update OFED release to 1.4.1 RC4
1708 Severity : major, only with big OST
1710 Description: Very poor metadata performance on Infiniband lustre configuration
1711 Details : OST object precreation becomes very slow on big OSTs. This is due
1712 to the ialloc patch spending too much time scanning groups.
1715 Frequency : during recovery
1717 Description: don't mix llog inodes with normal.
1718 Details : allocate inodes for log in last inode group
1723 Description: fix lqs' reference which won't be put in some situations
1724 Details : This patch fixes:
1725 1. In quota_check_common(), this function will check quota
1726 for user and group, but only send one return via "pending".
1727 In most cases, the pendings should be same. But that is not
1729 2. If quotaoff runs between lquota_chkquota() and
1730 lquota_pending_commit(), the same thing will happen too.
1731 That is why it comes:
1732 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1735 -------------------------------------------------------------------------------
1737 2008-12-31 Sun Microsystems, Inc.
1739 * Support for kernels:
1740 2.6.5-7.314 (SLES 9),
1741 2.6.9-67.0.22.EL (RHEL 4),
1742 2.6.16.60-0.31 (SLES 10),
1743 2.6.18-92.1.17.el5 (RHEL 5),
1744 2.6.22.14 vanilla (kernel.org)
1745 * Client support for unpatched kernels:
1746 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1747 2.6.16 - 2.6.22 vanilla (kernel.org)
1748 * Client support for unpatched kernels:
1749 we do not recommend using patchless RHEL4 clients with kernels
1750 prior to 2.6.9-55EL (RHEL4U5).
1751 * Recommended e2fsprogs version: 1.40.11-sun1
1752 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1753 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1754 removed cwd "./" (refer to Bugzilla 14399).
1755 * A new quota file format has been introduced in 1.6.5.
1756 The format conversion from prior releases is handled transparently,
1757 but releases older than 1.4.12/1.6.5 don't understand this new
1758 format. The automatic format conversion can be avoided by running
1759 the following command on the MDS:
1760 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1761 For more information, please refer to bugzilla 13904.
1762 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1763 The format conversion from prior releases is handled transparently,
1764 but releases older than 1.6.6/1.8.0 don't understand this new
1765 format. The automatic format conversion can be avoided by running
1766 the following commands on the MDS and OSS servers (for
1767 pre 1.4.12-1.6.5 quota files):
1768 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1769 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1770 or (for 1.4.12/1.6.5 quota files)
1771 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1772 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1773 For more information, please refer to bugzilla 13904.
1774 * Output of lfs quota has been made less detailed by default,
1775 old (verbose) output can be obtained by using -v option.
1776 * File join has been disabled in this release, refer to Bugzilla 16929.
1777 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1778 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1779 Lustre filesystem with 4K stack may cause a stack overflow. For more
1780 information, please refer to bugzilla 17630.
1782 Severity : enhancement
1784 Description: Caching OSS
1785 Details : introduce data caching on the OSS. The OSS now relies on the linux
1786 kernel page cache to keep recently accessed data in memory.
1787 It is worth noting that all write requests are still flushed
1788 synchronously as in lustre 1.6.
1790 Severity : enhancement
1792 Description: Version based recovery
1793 Details : introduce finer grained recovery able to detect transaction
1794 dependencies and can deal with transaction gaps caused by clients
1795 failing at the same time as the server.
1797 Severity : enhancement
1799 Description: Enable adaptive timeouts by default
1800 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1801 managed dynamically based on server load and should not need
1802 to be tuned manually based on cluster size. This allows Lustre
1803 to work under a wider variety of system sizes and loads, without
1804 unnecessarily causing lengthy recovery times.
1806 Severity : enhancement
1808 Description: Add OST Pools support
1809 Details : File striping can now be set to use an arbitrary pool of OSTs
1811 Severity : enhancement
1813 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1814 Details : allow skip disconnected ost for send statfs request and hide error
1818 Frequency : rare, on llog test 6
1820 Descriptoin: don't allow connect to already connected import
1821 Details : allowing connect to already connected import is hide connecting problem.
1824 Frequency : rare, on failed llog setup
1826 Descriptoin: don't leak obd reference on failed llog setup
1827 Details : for failed llog setup - mgc forget call class_destroy_import for
1828 client import, move destroy import to more generic place.
1833 Descriptoin: allow kill process which wait statahead result
1834 Details : for some reasons 'ls' can stick in waiting result from statahead,
1835 in this case need way for kill this process.
1838 Frequency : rare, at shutdown
1840 Descriptoin: panic at umount
1841 Details : llap_shrinker can be raced with killing super block from list and
1842 this produce panic with access to already freeded pointer
1847 Descriptoin: don't lose wakeup for imp_recovery_waitq
1848 Details : recover_import_no_retry or invalidate_import and import_close can
1849 both sleep on imp_recovery_waitq, but we was send only one wakeup
1855 Descriptoin: panic in mds_open
1856 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1861 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1863 Details : release lock refernce only after add page to pages list.
1866 Frequency : always with long access acl
1868 Descriptoin: mds can't pack reply with long acl.
1869 Details : mds don't control size of acl but they limited by reint/getattr
1874 Frequency : on remount
1876 Description: external journal device not working after the remount
1877 Details : clear dev_rdonly flag for external journal devices in
1883 Description: shutdown vs evict race
1884 Details : client_disconnect_export vs connect request race.
1885 if client will evicted at this time - we start invalidate
1886 thread without referece to import and import can be freed
1892 Description: shrink LOV EAs before replying
1893 Details : correctly adjust LOV EA buffer for reply.
1898 Description: don't skip ost target if they assigned to file
1899 Details : Drop slow OSCs if we can, but not for requested start idx.
1900 This means "if OSC is slow and it is not the requested
1901 start OST, then it can be skipped, otherwise skip it only
1902 if it is inactive/recovering/out-of-space.
1904 Severity : enhancement
1906 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1909 Frequency : rare, need acl's on inode.
1911 Description: client can't handle ost additional correctly
1912 Details : if ost was added after client connected to mds client can have
1913 hit lnet_try_match_md ... to big messages to wide striped files.
1914 in this case need teach client to handle config events about add
1915 lov target and update client max ea size at that event.
1917 Severity : enhancement
1919 Description: Update to sles9 kernel-2.6.5-7.314.
1921 Severity : enhancement
1923 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1926 Frequency : Create a symlink file with a very long name
1928 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1929 Details : If there is no extra space in the request for early cancels,
1930 ldlm_req_handles_avail() returns 0 instead of a negative value.
1935 Description: mds is deadlocked
1936 Details : in rare cases, inode in catalog can have i_no less than have parent
1937 i_no, this produce wrong order for locking during open, and
1938 parallel unlink can be lock open. this need teach mds_open to grab
1939 locks in resource id order, not at parent -> child order.
1941 Severity : enhancement
1943 Description: Add /proc entry for import status
1944 Details : The mdc, osc, and mgc import directories now have
1945 an import directory that contains useful import data for debugging
1946 connection problems.
1948 Severity : enhancement
1950 Description: Re-disable certain /proc logging
1951 Details : Enable and disable client's offset_stats, extents_stats and
1952 extents_stats_per_process stats logging on the fly.
1955 Frequency : Only on FC kernels 2.6.22+
1957 Description: oops in statahead
1958 Details : Do not drop reference count for the dentry from VFS when lookup,
1959 VFS will do that by itself.
1961 Severity : enhancement
1963 Description: Generic /proc file permissions
1964 Details : Set /Proc file permissions in a more generic way to enable non-
1965 root users operate on some /proc files.
1969 Description: Hitting mdc_commit_close() ASSERTION
1970 Details : Properly handle request reference release in
1971 ll_release_openhandle().
1975 Frequency : only patchless client
1976 Description: add workaround for race between add/remove dentry from hash
1978 Severity : enhancement
1980 Description: Allow OST glimpses to return PW locks
1984 Description: LBUG when llog conf file is full
1985 Details : When llog bitmap is full, ENOSPC should be returned for plain
1990 Description: Prevent import from entering FULL state when server in recovery
1994 Description: service mount cannot take device name with ":"
1995 Details : Only when device name contains ":/" will mount treat it as
2001 Description: replace ptlrpcd with the statahead thread to interpret the async
2002 statahead RPC callback
2006 Frequency : on recovery
2007 Description: I/O failures after umount during fail back
2008 Details : if client reconnected to restarted server we need join to recovery
2009 instead of find server handler is changed and process self eviction
2010 with cancel all locks.
2012 Severity : enhancement
2014 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
2018 Description: Kernel BUG tries to release flock
2019 Details : Lustre does not destroy flock lock before last reference goes
2020 away. So always drop flock locks when client is evicted and
2021 perform unlock regardless of successfulness of speaking to MDS.
2023 Severity : enhancement
2025 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
2027 Severity : enhancement
2029 Description: Upcall on Lustre log has been dumped
2030 Details : Allow for a user mode script to be called once a Lustre log has
2031 been dumped. It passes the filename of the dumped log to the
2032 script, the location of the script can be specified via
2033 /proc/sys/lnet/debug_log_upcall.
2038 Description: avoid messages about idr_remove called for id that is not allocated
2039 Details : Move assigment s_dev for clustered nfs to end of initialization,
2040 for avoid problem with error handling.
2045 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
2046 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
2047 CONN_USED_HASH and this prodice warning when put connection again
2053 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
2054 Details : release reference to stats when client disconnected, not
2055 when export destroyed for avoid races when client destroyed
2056 after main ost export.
2060 Description: more cleanup in mds_lov
2061 Details : add workaround for get valid ost count for avoid warnings about
2062 drop too big messages, not init llog cat under semphore which
2063 can be blocked on reconnect and break normal replay, fix access
2066 Severity : enhancement
2068 Description: Export bytes_read/bytes_write count on OSC/OST.
2072 Description: Early reply size mismatch, MGC loses connection
2073 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
2074 the connect flags are properly negotiated.
2078 Description: Properly propagate oinfo flags from lov to osc for statfs
2079 Details : restore missing copy oi_flags to lov requests.
2083 Description: exports in /proc are broken
2084 Details : recreate /proc entries for clients when they reconnect.
2086 Severity : enhancement
2088 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
2089 : l_getgroups(8), lst(8), routerstat(8)
2090 Details : included man pages for llobdstat(8), llstat(8),
2091 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
2093 Severity : enhancement
2095 Description: Implement lustre ll_show_options method.
2097 Severity : enhancement
2099 Description: Update to SLES9 kernel-2.6.5-7.312.
2101 Severity : enhancement
2103 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
2107 Description: exports in /proc are broken
2108 Details : recreate /proc entries for clients when they reconnect.
2112 Description: don't fail open with -ERANGE
2113 Details : if client connected until mds will be know about real ost count
2114 get LOV EA can be fail because mds not allocate enougth buffer
2119 Description: Resolve device initialization race
2120 Details : Prevent proc handler from accessing devices added to the
2121 obd_devs array but yet be intialized.
2123 Severity : enhancement
2125 Description: configure's --enable-quota should check the
2126 : kernel .config for CONFIG_QUOTA
2127 Details : configure is terminated if --enable-quota is passed but
2128 : no quota support is in kernel
2130 Severity : enhancement
2132 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
2134 Severity : enhancement
2136 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
2140 Frequency : rare, on PPC clients
2141 Description: don't swab ost objects in response about directory, because
2143 Details : bug similar bug 14856, but in different function.
2145 Severity : enhancement
2147 Description: lfs quota tool enhancement
2148 Details : added units specifiers support for setquota, default to
2149 current uid/gid for quota report, short quota stats by
2150 default, nonpositional parameters for setquota, added
2151 llapi_quotactl manual page.
2153 Severity : enhancement
2155 Description: *optional* service tags registration
2156 Details : if the "service tags" package is installed on a Lustre node
2157 When the filesystem is mounted, a local-node service tag will
2158 be created. See http://inventory.sun.com/ for more information
2159 about the Service Tags asset management system.
2163 Description: Client runs out of low memory
2164 Details : Consider only lowmem when counting initial number of llap pages
2167 Frequency : occasional
2169 Description: add refcount for osc callbacks, so avoid panic on shutdown
2171 Severity : enhancement
2173 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
2176 Frequency : testing only
2178 Description: sanity test 65a fails if stripecount of -1 is set
2179 Details : handle -1 striping on filesystem in ll_dirstripe_verify
2182 Frequency : only in unusual configurations
2184 Description: Kernel panic with find ost index.
2185 Details : lov_obd have panic if some OST's have sparse indexes.
2187 Severity : enhancement
2189 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
2192 Frequency : rarely, if filesystem is mounted with -o flock
2194 Description: do not process already freed flock
2195 Details : flock can possibly be freed by another thread before it reaches
2196 to ldlm_flock_completion_ast.
2199 Frequency : rarely, if filesystem is mounted with -o flock
2201 Description: LBUG during stress test
2202 Details : Need properly lock accesses the flock deadlock detection list.
2205 Frequency : rarely, if binaries are being run from Lustre
2207 Description: oops in page fault handler
2208 Details : kernel page fault handler can return two special 'pages' in
2209 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
2212 Frequency : rarely, during shutdown
2214 Description: timeout with invalidate import.
2215 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
2216 should be handled by ptlrpcd. This produce long age waiting and
2217 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
2223 Description: ASSERTION(CheckWriteback(page,cmd)) failed
2224 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
2228 Frequency : only with broken builds/installations
2230 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
2231 Details : just return an error to a user, put a console error message
2233 Severity : enhancement
2235 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
2237 Severity : enhancement
2239 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
2241 Severity : enhancement
2243 Description: enable MGS and MDT services start separately
2244 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
2245 a co-located MGS without starting the MGS, which is a complement
2246 to 'nosvc' mount option.
2249 Frequency : always, on big-endian systems
2251 Description: cleanup in ptlrpc code, related to PPC platform
2252 Details : store magic in native order avoid panic's in recovery on PPC
2253 node and forbid from this error in future. Also fix posibily
2254 of twice swab data. Fix get lov striping to userpace.
2257 Frequency : rarely, if replay get lost on server
2259 Description: server incorrectly drop resent replays lead to recovery failure.
2260 Details : do not drop replay according to msg flags, instead we check the
2261 per-export recovery request queue for duplication of transno.
2264 Frequency : after recovery
2266 Description: precreate to many object's after del orphan.
2267 Details : del orphan st in oscc last_id == next_id and this triger growing
2268 count of precreated objects. Set flag LOW to skip increase count
2269 of precreated objects.
2272 Frequency : rare, on clear nid stats
2274 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
2275 Details : when clean nid stats sometimes try destroy live entry,
2276 and this produce panic in free.
2279 Frequency : occasionally since 1.6.4
2281 Description: Stack overflow during MDS log replay
2282 Details : ease stack pressure by using a thread dealing llog_process.
2285 Frequency : very rare
2287 Description: MDT cannot be unmounted, reporting "Mount still busy"
2288 Details : Mountpoint references were being leaked during open reply
2289 reconstruction after an MDS restart. Drop mountpoint reference
2290 in reconstruct_open() and free dentry reference also.
2295 Description: wait until IO finished before start new when do lock cancel.
2296 Details : VM protocol want old IO finished before start new, in this case
2297 need wait until PG_writeback is cleared until check dirty flag
2298 and call writepages in lock cancel callback.
2303 Description: mds_mfd_close() ASSERTION(rc == 0)
2304 Details : In mds_mfd_close(), we need protect inode's writecount change
2305 within its orphan write semaphore to prevent possible races.
2308 Frequency : rare, on shutdown ost
2310 Description: don't hit live lock with umount ost.
2311 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2312 use shrink_dcache_sb instead.
2315 Frequency : only when echo_client is used
2317 Description: don't panic with use echo_client
2318 Details : echo client pass NULL as client nid pointer and this produce NULL
2319 pointer dereference.
2322 Frequency : Always on 32-bit PowerPC systems
2324 Description: fix build on PPC32
2325 Details : compile code with -m64 flag produce wrong object file for PPC32.
2330 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2331 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2332 by a client, and the data on disk are dangerous and incorrect.
2333 Change them with client eviction.
2335 Severity : enhancement
2337 Description: skiplist implementation simplification
2338 Details : skiplists are used to group compatible locks on granted list
2339 that was implemented as tracking first and last lock of each lock
2340 group the patch changes that to using doubly linked lists
2344 Description: delete compatibility for 32bit qdata
2345 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2346 qunit. That means servers of b1_4 and servers of b1_8 can't be
2347 used together if users want to use quota.
2350 Frequency : only with administrator action
2352 Description: mount failure if config log has invalid conf_param setting
2353 Details : If administrator specified an incorrect configuration parameter
2354 with "lctl conf_param" this would cause an error during future
2355 client mounts. Instead, ignore the bad configuration parameter.
2358 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2360 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2361 Details : If blocks per group is less than blocksize*8, set rest of the
2365 Frequency : Application do stride read on lustre
2367 Description: The read performance will drop a lot if the application does
2369 Details : Because the stride_start_offset are missing in stride read-ahead,
2370 it will cause clients read a lot of unused pages in read-ahead,
2371 then the read-performance drops.
2375 Description: more ldlm soft lockups
2376 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2377 starve other threads from the resource lock for a long time in
2378 case of long waiting queue, so change the debug level from
2379 D_OTHER to the less frequently used D_INFO.
2381 Severity : enhancement
2383 Description: add -gid, -group, -uid, -user options to lfs find
2385 Severity : enhancement
2387 Description: ll_recover_lost_found_objs - recover objects in lost+found
2388 Details : OST corruption and subsequent e2fsck can leave objects in the
2389 lost+found directory. Using the "ll_recover_lost_found_objs"
2390 tool, these objects can be retrieved and data can be salvaged
2391 by using the object ID saved in the fid EA on each object.
2396 Description: this bug _only_ happens when inode quota limitation is very low
2397 (less than 12), so that inode quota unit is 1 at initialization.
2398 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2399 is effective now. So least quota qunit should be 2.
2403 Description: Hung threads in invalidate_inode_pages2_range
2404 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2405 one is completed. As a result, some RPCs are stuck in the queue
2410 Description: Procfs and llog threads access destoryed import sometimes.
2411 Details : Sync the import destoryed process with procfs and llog threads by
2412 the import refcount and semaphore.
2416 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2417 Details : Sort source/child resource pair after updating child resource.
2422 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2423 Details : If insertion of an extent fails, then discard the inode
2424 preallocation and free data blocks else it can lead to duplicate
2429 Description: don't always update ctime in ext3_xattr_set_handle()
2430 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2431 In some cases the ctime should not be updated, for example for
2432 2.0->1.8 compatibility it is necessary to delete an xattr and it
2433 should not update the ctime.
2437 Description: add quota statistics
2438 Details : 1. sort out quota proc entries and proc code.
2439 2. add quota statistics
2444 Description: quotas are not honored with O_DIRECT
2445 Details : all writes with the flag O_DIRECT will use grants which leads to
2446 this problem. Now using OBD_BRW_SYNC to guard this.
2450 Bugzilla : 15713/16362
2451 Description: Assertion in iopen_connect_dentry in 1.6.3
2452 Details : looking up an inode via iopen with the wrong generation number can
2453 populate the dcache with a disconneced dentry while the inode
2454 number is in the process of being reallocated. This causes an
2455 assertion failure in iopen since the inode's dentry list contains
2456 both a connected and disconnected dentry.
2460 Description: assertion failure in ldlm_handle2lock()
2461 Details : fix a race between class_handle_unhash() and class_handle2object()
2462 introduced in lustre 1.6.5 by bug 13622.
2464 Severity : enhancement
2466 Description: superblock lock contention with many SMP cores on one client
2467 Details : several client filesystem locks were highly contended on SMP
2468 NUMA systems with 8 or more cores. Per-CPU datastructures
2469 and more efficient locking implemented to reduce contention.
2474 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2475 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2480 Description: oops in sd_iostats_seq_show()
2481 Details : unloading/reloading the scsi low level driver triggers a kernel
2482 bug when trying to access the sd iostat file.
2487 Description: Kernel panics during QLogic driver reload
2488 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2489 patch, causing memory corruption.
2494 Description: journal_dev option does not work in b1_6
2495 Details : pass mount option during pre-mount.
2497 Severity : enhancement
2499 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2500 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2501 extent information of a file. It can be used to map logical blocks
2502 in a file to physical blocks in the block device.
2505 Frequency : only with adaptive timeout enabled
2507 Description: DEBUG_REQ() bad paging request
2508 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2509 because it can be accessed by reply_in_callback() without the
2513 Frequency : only on Cray X2
2515 Description: X2 build failures
2516 Details : fix build failures on Cray X2.
2520 Description: xid & resent requests
2521 Details : Initialize RPC XID from clock at startup (randomly if clock is
2526 Description: quota recovery deadlock during mds failover
2527 Details : This patch includes att18982, att18236, att18237 in bz14840.
2529 1. fix osts hang when mds does failover with quotaon
2530 2. prevent watchdog storm when osts threads wait for the
2535 Description: kernel panic on racer
2536 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2538 Severity : enhancement
2540 Description: Add lustre_start utility to start or stop multiple Lustre servers
2545 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2546 Details : In case of memory pressure, list_del() can be called twice on
2547 req->rq_history_list, causing a kernel oops.
2551 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2552 Details : fix stack overflow in the distributed lock manager by defering
2553 export eviction after a failed ast to the elt thread instead of
2554 handling it in the dlm interpret routine.
2556 Severity : enhancement
2558 Description: More exported tunables for mballoc
2559 Details : Add support for tunable preallocation window and new tunables for
2560 large/small requests
2564 Description: Detect corruption of block bitmap and checking for preallocations
2565 Details : Checks validity of on-disk block bitmap. Also it does better
2566 checking of number of applied preallocations. When corruption is
2567 found, it turns filesystem readonly to prevent further corruptions.
2571 Frequency : only for big-endian servers
2572 Description: Check for big-endian system while mounting fs with extents feature
2573 Details : Mounting a filesystem with extents feature will fail on big-endian
2574 systems since ext3-based ldiskfs is not supported on big-endian
2575 systems. Can be over-ridden with "bigendian_extents" mount option.
2579 Description: Excessive recovery window
2580 Details : With AT enabled, the recovery window can be excessively long (6000+
2581 seconds). To address this problem, we no longer use
2582 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2583 timeout no longer depends on the service time, it is set to
2584 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2585 time via pb_service_time.
2588 Descriptoin: Don't sync journal after every i/o
2589 Details : Implement write RPC replay to allow server replies for write RPCs
2590 before data is on disk. However, this feature is disabled by
2591 default since some issues leading to data corruptions have been
2592 found during recovery (e.g. bug 19128). This feature can be enabled
2593 by running the following command on the OSSs:
2594 lctl set_param obdfilter.*.sync_journal=0
2598 Description: Watchdog triggered on MDS failover
2599 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2600 the OSTs know that the MDS "UUID" can be reused for the same export
2601 from a different NID, so we do not need to wait for the export to
2606 Descriptoin: Lustre detected file system corruption with inode out of bounds
2607 Details : don't update i_size on MDS_CLOSE for directories. This causes
2608 directory corruptions on the MDT.
2610 -------------------------------------------------------------------------------
2612 2008-05-26 Sun Microsystems, Inc.
2614 * Support for kernels:
2615 2.6.5-7.311 (SLES 9),
2616 2.6.9-67.0.7.EL (RHEL 4),
2617 2.6.16.54-0.2.5 (SLES 10),
2618 2.6.18-53.1.14.el5 (RHEL 5),
2619 2.6.22.14 vanilla (kernel.org)
2620 * Client support for unpatched kernels:
2621 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2622 2.6.16 - 2.6.22 vanilla (kernel.org)
2623 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2624 we do not recommend using patchless RHEL4 clients with kernels
2625 prior to 2.6.9-55EL (RHEL4U5).
2626 * Recommended e2fsprogs version: 1.40.7-sun1
2627 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2628 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2629 removed cwd "./" (refer to Bugzilla 14399).
2630 * A new quota file format has been introduced in 1.6.5.
2631 The format conversion from prior releases is handled transparently,
2632 but releases older than 1.4.12/1.6.5 will not understand this new
2633 format. The automatic format conversion can be avoided by running
2634 the following command on the MDS before upgrading:
2635 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2636 For more information, please refer to bugzilla 13904.
2640 Description: quota performance fix
2641 Details : quota data is written in journalled mode instead of ordered to
2642 increase performance
2646 Description: lfs support for human-readable quota grace time strings
2647 Details : lfs setquota -t and lfs quota -t represent quota grace times
2648 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2651 Frequency : always with o2ib 1.3 and sles10
2653 Description: fix build with SLES10 and o2ib v3.
2654 Details : sles10 uses diffrent name for Module.symver file but configure
2655 assume this file has same name on RHEL/SLES/vanila kernels.
2658 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2660 Description: MDS may lose file striping (and hence file data) in some cases
2661 Details : If there are additional extended attributes stored on the MDS,
2662 in particular ACLs, SELinux, or user attributes (if user_xattr
2663 is specified for the client mount options) then there is a risk
2664 of attribute loss. Additionally, the Lustre file striping
2665 needs to be larger than default (e.g. striped over all OSTs),
2666 and an additional attribute must be stored initially in the
2667 inode and then increase in size enough to be moved to the
2668 external attribute block (e.g. ACL growing in size) for file
2671 Severity : enhancement
2673 Description: add message levels for liblustreapi
2678 Description: MDT cannot be unmounted, reporting "Mount still busy"
2679 Details : Mountpoint references were being leaked during open reply
2680 reconstruction after an MDS restart. Drop mountpoint reference
2681 in reconstruct_open() and free dentry reference also.
2686 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2687 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2688 will result in an -EAGAIN and in the exit call path we call
2689 alloc_rr() which will with increasing aggressiveness attempt to
2690 aquire precreated objects on the minimum number of required OSCs.
2694 Description: Use old size assignment to avoid deadlock
2695 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2696 the scheduling while holding a spinlock. We do not need locking
2697 for size in ll_update_inode() because size is only updated from
2698 the MDS for directories or files without objects, so there is no
2699 other place to do the update, and concurrent access to such inodes
2700 are protected by the inode lock.
2704 Description: Use __u64 instead of int for valid bits
2708 Description: resolve "_IOWR redefined" build error on SLES10
2712 Description: dump the memory debugging after all modules are unloaded to
2713 suppress false negative in conf_sanity test 39
2717 Description: the recovery timer never expires
2718 Details : for new client connect request, the recovery timer should not be
2719 reset, otherwise recovery timer will never expired, if the old
2720 client never come. Only old client connect and first connection
2721 req should trigger recovery timer reset.
2725 Description: the min numbers of lproc stats are wrong
2726 Details : adding a new constant LC_MIN_INIT and use it for initialization
2730 Frequency : always with interactive lfs
2732 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2735 Frequency : with multiple concurrent readdir processes in same directory
2736 Bugzilla : 15406, 15169, 15175
2737 Description: misc fixes for directory readahead.
2738 Details : prevent previous statahead async RPC callback from processing the
2739 current "statahead_info", race condition between async RPC callback
2740 add dentry into dentry hash table and "ls" thread revalidate such
2741 dentry, statahead his/miss control for hidden items, and so on.
2743 Severity : enhancement
2745 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2749 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2750 Details : several u32 variables declarations are replaced with u64 declarations
2755 Description: fix SLES kernel versioning
2756 Details : the kernel version for our SLES 10 kernel did not include a "-"
2757 before the "smp" at the end. while this was not a problem in
2758 general, it did mean that software trying to use the kernel
2759 version to try to detect a vendor specific kernel would fail.
2760 this was most evident by the OFED build scripts.
2765 Description: Don't update lov_desc members until making sure they are valid
2766 Details : When updating lov_desc members via proc fs, need fix their
2767 validities before doing the real update.
2770 Frequency : very rare
2772 Description: don't put request into delay list while invalidate in flight.
2773 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2774 import in flight. this produce timeout for invalidate and sometimes
2775 can cause stale data.
2777 Severity : enhancement
2779 Description: Update kernel to SLES9 2.6.5-7.311.
2781 Severity : enhancement
2783 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2788 Frequency : on PPC only
2789 Description: not convert ost objects for directory because it's not exist.
2790 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2792 Severity : enhancement
2794 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2795 __u64 as usingied long long type.
2798 Frequency : rare, on shutdown
2800 Description: race process ast vs remove callback
2801 Details : removing callback before disconnect import open race with
2802 processing callback.
2804 Severity : enhancement
2806 Description: Update kernel to SLES9 2.6.5-7.311.
2808 Severity : enhancement
2810 Description: Files open for execute are not marked busy on SLES10
2811 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2813 Severity : enhancement
2815 Description: Add server support for vanilla-2.6.22.14.
2818 Frequency : occasional
2820 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2821 Details : If an OST is down the MDS will hang indefinitely in
2822 obd_statfs_rqset() waiting for the statfs data. While for
2823 MDS QOS usage of statfs, it should not stuck in waiting.
2825 Severity : enhancement
2827 Description: Disable adaptive timeouts by default
2830 Frequency : on network error
2832 Description: panic with double free request if network error
2833 Details : mdc_finish_enqueue is finish request if any network error occuring,
2834 but it's true only for synchronus enqueue, for async enqueue
2835 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2839 Frequency : rare, on recovery
2841 Description: read procfs can produce deadlock in some situation
2842 Details : Holding lprocfs lock with send rpc can produce block for destroy
2843 obd objects and this also block reconnect with -EALREADY.
2844 This isn't fix all lprocfs bugs - but make it rare.
2846 Severity : enhancement
2848 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2851 Frequency : frequent on X2 node
2853 Description: mdc_set_open_replay_data LBUG
2854 Details : Set replay data for requests that are eligible for replay.
2859 Description: lustre_mgs: operation 101 on unconnected MGS
2860 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2861 MGC, and late on MGC cannot successfully connect to MGS and a lot
2862 of the error messages complaining that MGS is not connected.
2865 Frequency : rare, depends on device drivers and load
2867 Description: MDS or OSS nodes crash due to stack overflow
2868 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2869 In some cases, in conjunction with device drivers that use a lot
2870 of stack, the MDS (or possibly OSS) service threads could overflow
2871 the stack. One change which was identified to consume additional
2872 stack has been reworked to avoid the extra stack usage.
2874 Severity : enhancement
2876 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2878 Severity : enhancement
2880 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2882 Severity : enhancement
2884 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2886 Serverity : enhancement
2888 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2891 Frequency : rare on shutdown OST
2893 Description: Don't allow skipping OSTs if index has been specified.
2894 Details : Don't allow skipping OSTs if index has been specified, make locking
2895 in internal create lots better.
2900 Description: ASSERTION(!PageDirty(page)) failed
2901 Details : Wrong check could lead to an assertion failure under specific
2907 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2908 Details : access to bitfield in structure is always rounded to long
2909 and this produce problem with not atomic change any bit.
2914 Description: Lustre make rpms failed.
2915 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2916 builds Lustre rpms from tarball.
2918 Severity : enhancement
2920 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2923 Frequency : rare on shutdown OST
2925 Description: If llog cancel was not send before clean_exports phase, this can
2926 produce deadlock in llog code.
2927 Details : If llog thread has last reference to obd and call class_import_put
2928 this produce deadlock because llog_cleanup_commit_master wait when
2929 last llog_commit_thread exited, but this never success because was
2930 called from llog_commit_thread.
2933 Frequency : only if OST index is skipped
2935 Description: NULL lov_tgts causing MDS oops
2936 Details : more safe checks for NULL lov_tgts for avoid oops.
2938 Severity : enhancement
2940 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2942 Severity : enhancement
2944 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2949 Description: make mgs_setparam() handle fsname containing dash
2950 Details : fsname containing a dash does not work with lctl conf_param
2952 Severity : enhancement
2954 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2957 Frequency : rare, in recovery and (or) destroy lovobjid file.
2959 Description: rewrite lov objid code.
2960 Details : Cleanup for lov objid code, remove scability problems and wrong
2961 locking. Fix sending last_id into ost.
2963 Severity : enhancement
2965 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2967 Severity : enhancement
2969 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2970 Details : Use d_move_locked instead of __d_move.
2973 Frequency : rare, at shutdown
2975 Description: access already free / zero obd_namespace.
2976 Details : if client_disconnect_export was called without force flag set,
2977 and exist connect request in flight, this can produce access to
2978 NULL pointer (or already free pointer) when connect_interpret
2979 store ocd flags in obd_namespace.
2982 Frequency : only at startup
2984 Description: not alloc memory with spinlock held.
2985 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2986 if any spinlock held.
2991 Description: lfs find does not continue on file error
2992 Details : Continue other files processing when a file/dir is absent.
2996 Description: Inconsistent usage of lustre_pack_reply()
2997 Details : Standardize the usage of lustre_pack_reply() such that it
2998 always generate a CERROR on failure.
3001 Frequency : very rare
3003 Description: Fix replay if there is an un-replied request and open
3004 Details : In some cases, older replay request will revert the
3005 mcd->mcd_last_xid on MDS which is used to record the client's
3006 latest sent request.
3008 Severity : enhancement
3010 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
3012 Severity : enhancement
3014 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
3016 Severity : enhancement
3018 Description: Update to SLES9 kernel-2.6.5-7.287.3.
3020 Severity : enhancement
3022 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
3024 Severity : enhancement
3026 Description: Build SLES10 patchless client fails
3027 Details : The configure was broken by run ./configure with
3028 --with-linux-obj=.... argument for patchless client. When the
3029 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
3030 can't search header adequately. Use absolute path such as
3031 -I($LINUX)/include instead.
3033 Severity : enhancement
3035 Description: Lustre Page Accounting
3036 Details : New macros for page alloc and free which enable accounting
3037 of page allocation of Lustre. Use percpu counters to store memory
3038 and page statistics.
3041 Frequency : only if debugging is disabled
3043 Description: LASSERT_{REQ,REP}SWAB macros are buggy
3044 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
3045 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
3046 these macros and replace them with their definitions instead.
3051 Description: interrupt oig_wait produce painc on resend.
3052 Details : brw_redo_request can be used for resend requests from ptlrpcd and
3053 private set, and this produce situation when rq_ptlrpcd_data not
3054 copyed to new allocated request and triggered LBUG on assert
3055 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
3056 ptlrpcd set if request is changed and can be safety changed to use
3059 Severity : enhancement
3061 Description: organize the server-side client stats on per-nid basis
3062 Details : Change the structure of stats under obdfilter and mds to
3070 The "uuid"s file would list the uuids of _active_ exports.
3071 And the clear entry is to clear all stats and stale nids.
3076 Description: Processes looping in ll_readdir() on Lustre clients finally causing
3077 a full node pseudo-hang
3078 Details : Concurrent access to the same directory from multiple clients with
3079 intensive file creation/removal can cause a client node to spin in
3080 ll_readdir(). i_version must be increased every time the lock is
3081 cancelled to ensure a revalidate is done.
3086 Description: touch file failed when fs is not full
3087 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
3088 otherwise we can get ENOSP while fs is not full.
3091 Frequency : only for Cray XT3
3092 Bugzilla : 12829/13455
3093 Description: Changing primary group doesn't change the group lustre assigns to
3095 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
3096 overridden with the primary group provided by the group upcall,
3097 whereas the supplied fsgid can be trusted if it is in the list of
3098 supplementary groups returned by the group upcall.
3100 Severity : enhancement
3102 Description: Root Squash Functionality
3103 Details : Implementation of NFS-like root squash capability. Specifically,
3104 don't allow someone with root access on a client node to be able
3105 to manipulate files owned by root on a server node.
3107 Severity : enhancement
3109 Description: Slow trucate/writes to huge files at high offsets.
3110 Details : Directly associate cached pages to lock that protect those pages,
3111 this allows us to quickly find what pages to write and remove
3112 once lock callback is received.
3117 Description: Too many locks accumulating on client during NFS usage
3118 Details : mds_open improperly used accmode to find out access mode to a
3119 file. Also mdc_intent_lock logic to find out if we already have
3120 lock similar to just received was flawed since introduction of
3121 skiplists - locks are now added to the front of the granted
3126 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
3128 Details : In obd setup failure handler, obd_stopping will not necessarily to
3129 be 1, and obd_set_up should also be checked to make sure whether
3130 obd is completely setup.
3132 Severity : enhancement
3134 Description: Allow masking D_WARNING, D_ERROR messages from console
3135 Details : Console messages can now be disabled via lnet.printk.
3140 Description: User code with malformed file open parameter crashes client node
3141 Details : Before packing join_file req, all the related reference should be
3142 checked carefully in case some malformed flags cause fake join_file
3148 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
3149 grain of quota doing harm to quota's accuracy
3150 Details : qunit size will be changed when quota limitation is too low/high;
3151 record the pending quota write in order to get more accureate
3152 quota; delete the patch for bug12588, which is unnecessary when
3153 this patch is landed. This bug also contains fixes for bug 14526,
3154 14299, 14601 and 13794.
3158 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
3159 Details : ldlm_completion_ast() assumes that a lock is granted when the req
3160 mode is equal to the granted mode. However, it should also check
3161 that LDLM_FL_CP_REQD is not set.
3165 Description: Heavy nfs access might result in deadlocks
3166 Details : After ELC code landed, it is now improper to enqueue any mds
3167 locks under och_sem, because enqueue might want to decide to
3168 cancel open locks for same inode we are holding och_sem for.
3172 Description: 35% write performance drop with ldiskfs2 when quotas are on
3173 Details : Enable ext3 journalled quota by default to improve performance
3174 when quotas are turned on.
3178 Description: Client eviction while running blogbench
3179 Details : A lot of unlink operations with concurrent I/O can lead to a
3180 deadlock causing evictions. To address the problem, the number of
3181 oustanding OST_DESTROY requests is now throttled to
3182 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
3183 callbacks are processed in priority.
3186 Frequency : RHEL4 only
3188 Description: mkfs is very slow on IA64/RHEL4
3189 Details : A performance regression has been discovered in the MPT Fusion
3190 driver between versions 3.02.73rh and 3.02.99.00rh. As a
3191 consequence, we have downgraded the MPT Fusion driver in the RHEL4
3192 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
3195 Frequency : PPC/PPC64 only
3197 Description: conflicts between asm-ppc64/types.h and lustre_types.h
3198 Details : fix duplicated definitions between asm-ppc64/types.h and
3199 lustre_types.h on PPC.
3202 Frequency : PPC/PPC64 only
3204 Description: asm-ppc/segment.h does not exist
3205 Details : fix compile issue on PPC.
3209 Description: data checksumming impacts single node performance
3210 Details : add support for several checksum algorithms. Currently, CRC32 and
3211 Adler-32 are supported. The checksum type can be changed on the fly
3212 through /proc/fs/lustre/osc/*/checksum_type.
3216 Description: use adler32 for page checksums
3217 Details : when available, use the Adler-32 algorithm instead of CRC32 for
3222 Description: better handle error messages in extents code
3224 Severity : enhancement
3226 Description: SNMP support enhancement
3227 Details : Adding total number of sampled request for an MDS node in snmp
3230 Severity : enhancement
3232 Description: Optimize ldlm waiting list processing for PR extent locks
3233 Details : When processing waiting list for read extent lock and meeting read
3234 lock that is same or wider to it that is not contended, skip
3235 processing rest of the list and immediatelly return current
3236 status of conflictness, since we are guaranteed there are no
3237 conflicting locks in the rest of the list.
3241 Description: Time out and refuse to reconnect
3242 Details : When the failover node is the primary node, it is possible
3243 to have two identical connections in imp_conn_list. We must
3244 compare not conn's pointers but NIDs, otherwise we
3245 can defeat connection throttling.
3249 Description: Client not clear own cache if answer to reconnect is lost.
3250 Details : Client gets evicted from server. Now client also thinks it is
3251 disconnected (or gets ENOTCONN on its operation) and decides to
3252 reconnect. Server receives reconnect message, but cannot find
3253 export. New export is created that is fully valid (new cookie!),
3254 but reply is lost and not reported to client. Client reconnects
3255 again and gets back a just-created connection, but it is not new
3256 so client thinks it was not evicted and does not do recovery.
3260 Description: Detect stride IO mode in read-ahead
3261 Details : When a client does stride read, read-ahead should detect that and
3262 read-ahead pages according to the detected stride pattern.
3266 Description: build for x2 fails
3267 Details : fix compile issue on Cray systems.
3269 Severity : enhancement
3271 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
3272 Details : This greatly improves speed of NFS writes on 2.6 kernels.
3275 Frequency : only on PPC/SLES10
3277 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
3278 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
3279 be defined. Add a hack in mkfs_lustre.c to work around this compile
3284 Description: LASSERT on MDS when client holding flock lock dies
3285 Details : ldlm pool logic depends on number of granted locks equal to
3286 number of released locks which is not true for flock locks, so
3287 just exclude such locks from consideration.
3291 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
3292 Details : Use fsfilt_sync() for both the whole filesystem sync and
3293 individual file sync to eliminate dangerous inode locking
3294 with I_LOCK that can lead to a deadlock.
3298 Description: Update an obsolete wirecheck.c generator
3299 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
3303 Description: Client can panic on open sometimes
3304 Details : It is possible that we try to free already freed request in
3305 ll_file_open in some error cases when we send request from
3310 Description: performance in 1.6.3
3311 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3312 sure that 1MB requests are not fragmented by the block layer.
3316 Description: LDLM soft lockups - improvement
3317 Details : It is be possible to send the lock handle along with each read
3318 or write request because the client is already doing a lock match
3319 itself so there isn't any reason the OST should have to re-do that
3325 Description: lfs quota fails with deactivated OSTS
3326 Details : With this patch, three improvements are included:
3327 1. detete the softlimit in mds and osts when use "lfs quota".
3328 2. display the inaccurate data in the output of "lfs quota".
3329 3. try to get quota info when "lfs quota" is executed.
3334 Description: Extent locks not granted with no conflicts sometimes.
3335 Details : When race occurs in glimpse handler and nothing is returned,
3336 we do not reprocess the queue after lock cancel, and that leads
3337 to a stall until next activity on a resource
3340 Frequency : failover with quotaon
3342 Description: during mds failovers with quota on, OSTs got into deadlock state
3343 and causing dumpstack.
3344 Details : for every quota slave, at any time, there is only one quota req
3345 is sent to quota master for every uid/gid. Before that quota req
3346 returns, all the thread relative to the same uid/gid will wait.
3347 So if the quota req is lost because mds failovers or any other
3348 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3349 quota reqs who time out.
3351 Severity : enhancement
3354 Description: when quota slave checks if quota is enough, there is an unnecessary
3356 Details : place this wait on necessary place instead of always waiting.
3358 --------------------------------------------------------------------------------
3360 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3362 * Support for kernels:
3363 2.6.5-7.286 (SLES 9),
3364 2.6.9-55.0.9.EL (RHEL 4),
3365 2.6.16.53-0.8 (SLES 10),
3366 2.6.18-8.1.14.el5 (RHEL 5),
3367 2.6.18.8 vanilla (kernel.org)
3368 * Client support for unpatched kernels:
3369 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3370 2.6.16 - 2.6.22 vanilla (kernel.org)
3371 * Due to recently discovered recovery problems, we do not recommend
3372 using patchless RHEL 4 clients with this or any earlier release.
3373 * Recommended e2fsprogs version: 1.40.2-cfs1
3374 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3377 Frequency : occasional
3379 Description: MDS hang or stay in waiting lock
3380 Details : If client receive lock with CBPENDING flag ldlm need send lock
3381 cancel as separate rpc, to avoid situation when cancel request
3382 can't processed due all i/o threads stay in wait lock.
3385 Frequency : occasional
3387 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3388 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3389 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3390 oscc_has_objects and see OSC has no objests and call
3391 oscc_internal_create to resend create request.
3393 Severity : enhancement
3395 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3396 Details : Update which_patch & target file for SLES10 latest kernel.
3398 Severity : enhancement
3400 Description: add --type and --size parameters to lfs find
3401 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3402 multiple OBDs can now be specified for the --obd option.
3404 Severity : enhancement
3406 Description: eliminate client locks in face of contention
3407 Details : file contention detection and lockless i/o implementation
3408 for contended files.
3410 Severity : enhancement
3412 Description: Remove client patches from SLES 10 kernel.
3413 Details : This causes SLES 10 clients to behave as patchless clients
3414 even on a Lustre-patched (server) kernel.
3416 Severity : enhancement
3418 Description: use i_size_read and i_size_write in 2.6 port
3419 Details : replace inode->i_size access with i_size_read/write()
3422 Frequency : when removing large files
3424 Description: scheduling issue during removal of large Lustre files
3425 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3426 It causes scheduling issues when removing large files (17TB in the
3432 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3433 Details : Impossible (EPROTO is returned) to access a directory that has a
3434 non-default striping and ACLs.
3437 Frequency : only on ppc
3439 Description: /proc/fs/lustre/devices broken on ppc
3440 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3441 We should make sure the type of 'index' is loff_t and then cast
3442 explicitly as needed below. Do not assign an explicitly cast
3446 Frequency : only for rhel5
3448 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3449 Details : Modify the target file & which_kernel.
3452 Frequency : if the uninit_groups feature is enabled on ldiskfs
3454 Description: e2fsck reports "invalid unused inodes count"
3455 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3456 feature and only a single inode is created in a group then the
3457 "bg_unused_inodes" count is incorrectly updated. Creating a
3458 second inode in that group would update it correctly.
3463 Description: buffer overruns could theoretically occur
3464 Details : llapi_semantic_traverse() modifies the "path" argument by
3465 appending values to the end of the origin string, and a buffer
3466 overrun may occur. Adding buffer overrun check in liblustreapi.
3468 Severity : enhancement
3470 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3473 Frequency : only if filesystem is inconsistent
3475 Description: handle "serious error: objid * already exists" more gracefully
3476 Details : If LAST_ID value on disk is smaller than the objects existing in
3477 the O/0/d* directories, it indicates disk corruption and causes an
3478 LBUG(). If the object is 0-length, then we should use the existing
3479 object. This will help to avoid a full fsck in most cases.
3481 Severity : enhancement
3483 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3484 Details : Modify vm-tunables-rhel4.patch.
3486 Severity : enhancement
3488 Description: Kernel config for 2.6.18-vanilla.
3489 Details : Modify targets/2.6-vanilla.target.in.
3490 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3491 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3492 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3493 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3496 Frequency : occasional
3498 Description: improve handling recoverable errors
3499 Details : If request processed with error which can be recoverable on server
3500 request should be resend, otherwise page released from cache and
3506 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3507 Details : Modify target file & which_patch.
3508 A flaw was found in the IA32 system call emulation provided
3509 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3510 value could be stored in the %RAX register, which could trigger an
3511 out-of-bounds system call table access. An untrusted local user
3512 could exploit this flaw to run code in the kernel
3513 (ie a root privilege escalation). (CVE-2007-4573).
3517 Description: change order of libsysio includes
3518 Details : '#include sysio.h' should always come before '#include xtio.h'
3520 Severity : enhancement
3522 Description: adapt the lustre_config script to support the upgrade case
3523 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3524 server targets to 1.6 in parallel.
3529 Description: To avoid grant space > avaible space when the disk is almost
3530 full. Without this patch you might see the error "grant XXXX >
3531 available" or some LBUG about grant, when the disk is almost
3533 Details : In filter_check_grant, for non_grant cache write, we should
3534 check the left space by if (*left > ungranted + bytes), instead
3535 of (*left > ungranted), because only we are sure the left space
3536 is enough for another "bytes", then the ungrant space should be
3537 increase. In client, we should update cl_avail_grant only there is
3538 OBD_MD_FLGRANT in the reply.
3543 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3544 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3545 described in CVE-2007-4573. This problem could allow untrusted
3546 local users to gain root access.
3549 Frequency : when using O_DIRECT and quotas
3551 Description: Incorrect file ownership on O_DIRECT output files
3552 Details : block usage reported by 'lfs quota' does not take into account
3553 files that have been written with O_DIRECT.
3557 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3558 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3559 window to 0. If it does hit cache miss, and read-pattern is still
3560 stride-io mode, does not reset the stride window, but also does not
3561 increase the stride window length in this case.
3563 --------------------------------------------------------------------------------
3565 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3567 * Support for kernels:
3568 2.6.5-7.286 (SLES 9),
3569 2.6.9-55.0.2.EL (RHEL 4),
3570 2.6.16.46-0.14 (SLES 10),
3571 2.6.18-8.1.8.el5 (RHEL 5),
3572 2.6.18.8 vanilla (kernel.org)
3573 * Client support for unpatched kernels:
3574 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3575 2.6.16 - 2.6.21 vanilla (kernel.org)
3576 * Due to recently discovered recovery problems, we do not recommend
3577 using patchless RHEL 4 clients with this or any earlier release.
3578 * Recommended e2fsprogs version: 1.40.2-cfs1
3579 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3583 Description: Fix errors in lfs documentation
3584 Details : Fixes man pages
3586 Severity : enhancement
3588 Description: Adaptive timeouts
3589 Details : RPC timeouts adapt to changing server load and network
3590 conditions to reduce resend attempts and improve recovery time.
3592 Severity : enhancement
3594 Description: llapi_file_create() does not allow some changes
3595 Details : add llapi_file_open() that allows specifying the file creation
3596 mode and open flags, and also returns an open file handle.
3598 Severity : enhancement
3600 Description: df doesn't work properly if diskfs blocksize != 4K
3601 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3604 Frequency : always on directories with default striping set
3606 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3607 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3609 Severity : enhancement
3611 Description: merge and cleanup kernel patches.
3612 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3614 Severity : enhancement
3616 Description: RedHat Update kernel for RHEL5
3617 Details : Kernel config file for RHEL5.
3619 Severity : enhancement
3621 Description: OSS needs mutliple precreate threads
3622 Details : Add ability to start more than one create thread per OSS.
3625 Frequency : only with quota on the root user
3627 Description: mds_obd_create error creating tmp object
3628 Details : When the user sets quota on root, llog will be affected and can't
3629 create files and write files.
3632 Frequency : Always on ia64 patchless client, and possibly others.
3634 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3635 Details : This allows the patchless client to be loaded on architectures
3636 without this export.
3641 Description: disorder of journal start and llog_add cause deadlock.
3642 Details : in llog_origin_connect, journal start should happen before
3643 llog_add keep the same order as other functions to avoid
3646 Severity : enhancement
3648 Description: RedHat Update kernel for RHEL5
3649 Details : Modify the kernel config file more closer RHEL5.
3651 Severity : enhancement
3653 Description: Build failure against Centos5 (RHEL5)
3654 Details : Define PAGE_SIZE when it isn't present.
3657 Frequency : occasionally when using NFS
3659 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3660 Details : with NFS, the anon dentry's parent was set to itself in
3661 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3662 corresponding dentry other than use rec->ur_name.
3664 Severity : enhancement
3666 Description: enable data checksumming by default
3667 Details : enable checksum by default, allow --disable-checksum
3668 configure option and "-o nochecksum" mount option. Checksums
3669 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3672 Frequency : Occasionally with failover
3674 Description: Client eviction due to failover config
3675 Details : after a connection loss, the lustre client should attempt to
3676 reconnect to the last active server first before trying the
3677 other potential connections.
3679 Severity : enhancement
3681 Description: client-side metadata stat-ahead during readdir(directory readahead)
3682 Details : perform client-side metadata stat-ahead when the client detects
3683 readdir and sequential stat of dir entries therein
3688 Description: when mds and osts use different quota unit(32bit and 64bit),
3689 quota will be released repeatly.
3690 Details : void sending multiple quota reqs to mds, which will keep the
3691 status between the reqs.
3694 Frequency : only with liblustre clients on XT3
3696 Description: evictions taking too long
3697 Details : allow llrd to evict clients directly on OSTs
3701 Description: osts not allocated evenly to files
3702 Details : change the condition to increase offset_idx
3704 Severity : enhancement
3706 Description: self-adjustable client's lru lists
3707 Details : use adaptive algorithm for managing client cached locks lru
3708 lists according to current server load, other client's work
3709 pattern, memory activities, etc. Both, server and client
3710 side namespaces provide number of proc tunables for controlling
3715 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3716 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3717 readdir code from ext3) from scratch preserving functionality.
3721 Description: Only those disconnect error should be returned by rq_status.
3722 Details : In open/enqueue processs, Some errors, which will cause client
3723 disconnected, should be returned by rq_status, while other
3724 errors should still be returned by intent, then mdc or llite will
3727 Severity : enhancement
3729 Description: Tune the kernel for good SCSI performance.
3730 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3731 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3735 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3736 Bugzilla : 13547, 13627
3737 Description: Data corruption for OSTs that are formatted larger than 2TB
3739 Details : When generating the bio request for lustre file writes the
3740 sector number would overflow a temporary variable before being
3741 used for the IO. The data reads correctly from Lustre (which
3742 will overflow in a similar manner) but other file data or
3743 filesystem metadata may be corrupted in some cases.
3747 Description: "lfs find -obd UUID" prints directories
3748 Details : "lfs find -obd UUID" will return all directory names instead
3749 of just file names. It is incorrect because the directories
3750 do not reside on the OSTs.
3753 Bugzilla : 13596, 13614
3754 Description: MDS hang after unclean shutdown of lots of clients
3755 Details : Never resend AST requests.
3759 Frequency : Always, for kernels after 2.6.16
3760 Description: Fix warning idr_remove called for id=.. which is not allocated.
3761 Details : Last kernels save old s_dev before kill super and not allow
3762 to restore from callback - restore it before call kill_anon_super.
3766 Description: Fix error on 'ls .' at the top of the Lustre mount.
3767 Details : Don't revalidate dentry if it is a root dentry.
3769 --------------------------------------------------------------------------------
3771 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3773 * Support for kernels:
3774 2.6.5-7.286 (SLES 9),
3775 2.6.9-55.0.2.EL (RHEL 4),
3776 2.6.16.46-0.14 (SLES 10),
3777 2.6.18-8.1.8.el5 (RHEL 5),
3778 2.6.18.8 vanilla (kernel.org)
3779 * Client support for unpatched kernels:
3780 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3781 2.6.16 - 2.6.22 vanilla (kernel.org)
3782 * Due to recently discovered recovery problems, we do not recommend
3783 using patchless RHEL 4 clients with this or any earlier release.
3784 * Recommended e2fsprogs version: 1.39.cfs8
3785 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3790 Description: block reactivating mgc import until all deactivates complete
3791 Details : Fix race when failing back MDT/MGS to itself (testing)
3793 Severity : enhancement
3795 Description: lfs setstripe enhancement
3796 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3799 Frequency : mds/oss recovery
3801 Description: llog ctxt is refrenced after it has been freed.
3802 Details : llog ctxt refcount was added to avoide the race between ctxt free
3803 and llog recovery process. Each llog user must hold ctxt refcount
3804 before it access the llog. And the llog ctxt can only be freed
3805 when its refcount is zero.
3807 Severity : enhancement
3809 Description: randomly memory allocation failure util
3810 Details : Make lustre randomly failed allocating memory for testing purpose.
3813 Frequency : only for SLES10
3815 Description: Update kernel patch for SLES10 SP1
3816 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3818 Severity : enhancement
3820 Description: omit set fsid for export NFS
3821 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3822 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3825 Frequency : after network failures
3827 Description: Add sync option to mount_lustre.c
3828 Details : Client loses data written to lustre after a network interruption.
3830 Severity : enhancement
3832 Description: Error message improvement.
3833 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3835 Severity : enhancement
3837 Description: don't use GFP_* in generic Lustre code.
3838 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3840 Severity : enhancement
3842 Description: obdclass is limited by single OBD_ALLOC(idarray)
3843 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3845 Severity : enhancement
3847 Description: updated patchess for new RHEL4 kernel
3848 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3850 Severity : enhancement
3852 Description: warnings with build patchless client with vanila 2.6.19 and up
3853 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3854 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3856 Severity : enhancement
3858 Description: O_DIRECT bypasses client statistics.
3859 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3860 read_ahead_stats were updated. Copied stats section from
3861 osc_send_oap_rpc() into async_internal().
3864 Frequency : only for Cray XT3
3866 Description: peer credits not enough on many OST per OSS systems.
3867 Details : Use new lnet way to add credits as we need those for pings and ASTs
3870 Frequency : only with liblustre
3872 Description: Liblustre is not releasing flock locks on file close.
3873 Details : Release flock locks on file close.
3876 Frequency : only for RHEL4
3878 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3879 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3880 Update target file and kernel config.
3884 Description: lustre support for RHEL5
3885 Details : Add support for RHEL5.
3890 Description: ASSERTION(export != NULL) failed in target_handle_connect
3891 Details : Assetion hit is result of rare race between disconnect and connet
3892 to same nid. target_handle_connect found old connect cockie and
3893 tried to reconnect, but can't find export for this cockie.
3898 Description: umount blocks forever on error
3899 Details : In result of wrong using obd_no_recov and obd_force flags client
3900 can hand if cancel or some other requests is lost.
3903 Frequency : Only for SLES
3905 Description: sanity_quota fail test_1
3906 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3907 file, which makes TSTID[2] inunique.
3909 Severity : enhancement
3911 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3912 Details : Update target/ChangeLog/which_patch .
3914 Severity : enhancement
3916 Description: jbd statistics
3917 Details : Port older jbd statistics patch for sles10
3922 Description: lvbo_init failed for resource with missing objects.
3923 Details : Fix returning error if we do stat for file with missing/corrupted
3924 objects and i_size set to all sum of size of all avaible objects.
3925 if we to truncate/write to missing object - it is recreated.
3930 Description: Oops in read and write path when failing to allocate lock.
3931 Details : Check if lock allocation failed and return error back.
3934 Frequency : When flocks are used.
3936 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3937 Details : Flock locks might destroy just granted lock if it could be merged
3938 with another existing flock, this is done in completion handler,
3939 so teach ldlm_cli_enquque_fini that this is a valid case for
3945 Description: reply_lock_interpret crash due to race with it and lock cancel.
3946 Details : Do not replay locks that are being cancelled. Do not reference
3947 locks by their address during replay, just by their handle.
3950 Frequency : only with deactivated OSTs
3952 Description: lstripe command fails for valid OST index
3953 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3954 instead of lov->desc.ld_active_tgt_count.
3956 Severity : enhancement
3958 Description: Build failure against Centos5 (RHEL5)
3959 Details : Use getpagesize() instead of PAGE_SIZE.
3961 --------------------------------------------------------------------------------
3963 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3965 * Support for kernels:
3966 2.6.5-7.283 (SLES 9),
3967 2.6.9-55.EL (RHEL 4),
3968 2.6.16.46-0.14 (SLES 10),
3969 2.6.18.8 vanilla (kernel.org)
3970 * Client support for unpatched kernels:
3971 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3972 2.6.16 - 2.6.22 vanilla (kernel.org)
3973 * Due to recently discovered recovery problems, we do not recommend
3974 using patchless RHEL 4 clients with this or any earlier release.
3975 * Recommended e2fsprogs version: 1.39.cfs8
3976 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3977 * Starting with this release, the ldiskfs backing filesystem required
3978 by Lustre is now in its own package, lustre-ldiskfs. This package
3979 should be installed. It is versioned separately from Lustre and
3980 may be released separately in future.
3985 Description: server LBUG when shutting down
3986 Details : Block umount forever until the mount refcount is zero rather
3987 than giving up after an arbitrary timeout.
3989 Severity : enhancement
3991 Description: add optional extra BUILD_VERSION info
3992 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3993 to override the lustre version.
3996 Frequency : 2.6.18 servers only
3998 Description: ll_kern_mount() doesn't release the module reference
3999 Details : The ldiskfs module reference count never drops down to 0
4000 because ll_kern_mount() doesn't release the module reference.
4005 Description: server LBUG when using old ost_num_threads parameter
4006 Details : Accept the old ost_num_threads parameter but warn that it
4007 is deprecated, and fix an off-by-one error that caused an LBUG.
4012 Description: Transient SCSI error results in persistent IO issue
4013 Details : iobuf->dr_error is not reinitialized to 0 between two
4017 Frequency : sometimes when underlying device returns I/O errors
4019 Description: OSTs not going read-only during write failures
4020 Details : OSTs are not remounted read-only when the journal commit threads
4021 get I/O errors because fsfilt_ext3 calls journal_start/stop()
4022 instead of the ext3 wrappers.
4025 Frequency : only with 10000 clients or more
4027 Description: poor connect scaling with increasing client count
4028 Details : Don't run filter_grant_sanity_check for more than 100 exports
4029 to improve scaling for large numbers of clients.
4032 Frequency : SLES10 only
4034 Description: sanity-quota.sh quotacheck failed: rc = -22
4035 Details : Quotas cannot be enabled on SLES10.
4038 Frequency : liblustre clients only
4040 Description: getdirentries does not give error when run on compute nodes
4041 Details : getdirentries does not fail when the size specified as an argument
4042 is too small to contain at least one entry
4044 Severity : enhancement
4046 Description: Add LNET router traceability for debug purposes
4047 Details : If a checksum failure occurs with a router as part of the
4048 IO path, the NID of the last router that forwarded the bulk data
4049 is printed so it can be identified.
4054 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
4055 Details : Due to a race condition, liblustre clients were occasionally
4056 evicted incorrectly.
4058 Severity : enhancement
4060 Description: lfs setstripe use optional parameters instead of postional
4063 Severity : enhancement
4065 Description: Nanosecond timestamp support for ldiskfs
4066 Details : The on-disk ldiskfs filesystem has added support for nanosecond
4067 resolution timestamps. There is not yet support for this at
4068 the Lustre filesystem level.
4071 Frequency : during server recovery
4073 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
4074 Details : request with rq_no_resend flag not awake l_wait_event if they get a
4078 Frequency : nfs export on patchless client
4080 Description: connectathon hang when test nfs export over patchless client
4081 Details : Disconnected dentry cannot be found with lookup, so we do not need
4082 to unhash it or make it invalid
4085 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
4086 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
4087 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
4092 Description: open req refcounting wrong on reconnect
4093 Details : If reconnect happened between getting open reply from server and
4094 call to mdc_set_replay_data in ll_file_open, we will schedule
4095 replay for unreferenced request that we are about to free.
4096 Subsequent close will crash in variety of ways.
4097 Check that request is still eligible for replay in
4098 mdc_set_replay_data().
4103 Description: disable writes to filesystem when reading health_check file
4104 Details : the default for reading the health_check proc file has changed
4105 to NOT do a journal transaction and write to disk, because this
4106 can cause reads of the /proc file to hang and block HA state
4107 checking on a healthy but otherwise heavily loaded system. It
4108 is possible to return to the previous behaviour during configure
4109 with --enable-health-write.
4111 Severity : enhancement
4113 Description: 64-bit inode version
4114 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
4115 to the inode. This will be required for version-based recovery.
4120 Description: MDS fails to start if a duplicate client export is detected
4121 Details : in some rare cases it was possible for a client to connect to
4122 an MDS multiple times. Upon recovery the MDS would detect this
4123 and fail during startup. Handle this more gracefully.
4125 Severity : enhancement
4127 Description: Add -o localflock option to simulate old noflock behaviour.
4128 Details : This will achieve local-only flock/fcntl locks coherentness.
4133 Description: log_commit_thread vs filter_destroy race leads to crash
4134 Details : Take import reference before releasing llog record semaphore
4139 Description: Wrong request locking in request set processing
4140 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
4141 imp_delayed_list, in this place should be used imp_lock.
4144 Frequency : when reconnecting
4146 Description: Grant leak when OSC reconnect to OST
4147 Details : When osc reconnect ost, OST(filter) should check whether it
4148 should grant more space to client by comparing fed_grant and
4149 cl_avail_grant, and return the granted space to client instead
4150 of "new granted" space, because client will call osc_init_grant
4151 to update the client grant space info.
4154 Frequency : when client reconnects to OST
4156 Description: Grant leak when OSC does a resend and replays bulk write
4157 Details : When osc reconnect to OST, OST(filter) should clear grant info of
4158 bulk write request, because the grant info while be sync between
4159 OSC and OST when reconnect, and we should ignore the grant info
4160 these of resend/replay write req.
4165 Description: Grant space more than avaiable space sometimes.
4166 Details : When then OST is about to be full, if two bulk writing from
4167 different clients came to OST. Accord the avaliable space of the
4168 OST, the first req should be permitted, and the second one
4169 should be denied by ENOSPC. But if the seconde arrived before
4170 the first one is commited. The OST might wrongly permit second
4171 writing, which will cause grant space > avaiable space.
4174 Frequency : when client is evicted
4176 Description: Grant might be wrongly erased when osc is evicted by OST
4177 Details : when the import is evicted by server, it will fork another
4178 thread ptlrpc_invalidate_import_thread to invalidate the
4179 import, where the grant will be set to 0. While the original
4180 thread will update the grant it got when connecting. So if
4181 the former happened latter, the grant will be wrongly errased
4182 because of this race.
4187 Description: Checking Stale with correct fid
4188 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
4189 to check whether it is stale, because sometimes, we want the
4190 enqueue happened anyway, and op_data.fid2 will not be initialized.
4192 Severity : enhancement
4194 Description: update patchless client
4195 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
4198 Frequency : only with 2.4 kernel
4200 Description: random memory corruption
4201 Details : size of struct ll_inode_info is to big for union inode.u and this
4202 can be cause of random memory corruption.
4207 Description: Memory leak in recovery
4208 Details : Lov_mds_md was not free in an error handler in mds_create_object.
4209 It should also check obd_fail before fsfilt_start, otherwise if
4210 fsfilt_start return -EROFS,(failover mds during mds recovery).
4211 then the req will return with repmsg->transno = 0 and rc = EROFS.
4212 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
4213 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
4214 be freed no matter whether fsfilt_commit success or not.
4217 Frequency : only with large numbers of cores on a single node
4219 Description: Prevents from taking the superblock lock in llap_from_page for
4221 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
4222 for llap_from_page call in ll_removepage() prevents from taking
4223 the superblock lock for a soon died page.
4228 Description: Not check open intent error before release open handle
4229 Details : in some rare cases, the open intent error is not checked before
4230 release open handle, which may cause
4231 ASSERTION(open_req->rq_transno != 0), because it tries to release
4232 the failed open handle.
4237 Description: Set cat log bitmap only after create log success.
4238 Details : in some rare cases, the cat log bitmap is set too early. and it
4239 should be set only after create log success.
4243 Description: Accessing a block bevice can re-enable I/O when Lustre is
4244 tearing down a device.
4245 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
4250 Description: service threads may hog cpus when there are a lot of requests
4251 Details : Insert cond_resched to give other threads a chance to use some CPU
4256 Description: the cat log was not initialized in recovery
4257 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
4258 unlink log on mds will not be initialized until mds post
4259 recovery. And also in mds post recovery, the unlink log will
4260 initialization will be done asynchronausly, so there will be race
4261 between add unlink log and unlink log initialization.
4265 Description: brw_stats were being printed incorrectly
4266 Details : brw_stats were being printed as log2 but all of them were not
4267 recorded as log2. Also remove some code duplication arising from
4268 filter_tally_{read,write}.
4272 Frequency : rare, only in recovery.
4273 Description: ASSERTION(req->rq_type != LI_POISON) failed
4274 Details : imp_lock should be held while iterating over imp_sending_list for
4275 prevent destroy request after get timeout in ptlrpc_queue_wait.
4279 Description: replay-single.sh test 52 fails
4280 Details : A lock's skiplist need to be cleanup when it being unlinked
4281 from its resource list.
4286 Description: Short directio read returns full requested size rather than
4288 Details : Direct I/O operations should return actual amount of bytes
4289 transferred rather than requested size.
4291 Severity : enhancement
4293 Description: metadata RPC reduction (e.g. for rm performance)
4294 Details : decrease the amount of synchronous RPC between clients and servers
4295 by canceling conflicing lock before the operation on the client
4296 and packing thier handles into the main operation RPC to server.
4298 Severity : enhancement
4300 Description: add #ifdef HAVE_KERNEL_CONFIG_H
4301 Details : kernels from 2.6.19 not need include linux/config.h, but add
4302 include linux/autoconf.h in commpiler command line.
4304 Severity : enhancement
4306 Description: patchless client support for 2.6.22 kernel
4307 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
4308 removed. In this case we need drop using os depended interface to
4309 kmem_cache and use cfs_mem_cache API.
4314 Description: fix mal-formatted messages
4315 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4319 Frequency : always in liblustre
4320 Description: wrong IS_ERR implementation in liblustre.h
4321 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4326 Description: Correct condition for output debug message.
4327 Details : inode i_nlink equal zero is not enough for output message about
4328 disk corruption, i_ctime and i_mode should be also checked.
4332 Frequency : always in patchless client
4333 Description: add configure check for truncate_complete_page
4334 Details : improve checks for exported symbols. This allow run check without
4335 sources, but with Module.symvers shipped with kernel distribution.
4336 add check for truncate_complete_page used by patchless client.
4340 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4341 Details : test_77h uses a file which was messed by other test case.
4345 Description: Not Check whether lov_tgts is NULL in some lov functions
4346 Details : Checking whether lov_tgts is NULL in some functions.
4350 Description: replace obdo_alloc() with OBDO_ALLOC macro
4351 Details : nothing special is done in obdo_alloc() function, and for
4352 debugging purpose, it needs to be replaced with macros.
4356 Description: bad return value and errno from fcntl call
4357 Details : In liblustre API, errno should be a negative value if error
4362 Description: ptlrpc_check_set() LBUG
4363 Details : In case of positive reply from server and failed client bulk
4364 callback after bulk transfer shouldn't LBUG, but process this
4365 request as erroneous.
4367 Severity : enhancement
4369 Description: VFS operations stats tool.
4370 Details : Tool which collects stats by tracking value written in pid,
4371 ppid, gid and uses llstat to generate output to plot graph using
4373 Updated lustre/utils/Makefile.am
4374 Added lustre/utils/ltrack_stats.c
4376 Severity : enhancement
4378 Description: 2.6.18 server support (lustre 1.6.1)
4379 Details : Support for 2.6.18 kernels on the server side.
4384 Description: ASSERTION(imp->imp_conn_current) failed
4385 Details : an assertion failure is hit if a client node boots and attempts to
4386 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4389 Frequency : only for i686
4391 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4392 Details : Fixed config variable for build.
4397 Description: Updated patchess for new RHEL4 kernel
4398 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4399 Updated patch jbd-stats-2.6.9.patch
4400 Updated patch qsnet-rhel4-2.6.patch
4401 Updated patch quota-deadlock-on-pagelock-core.patch
4402 Updated patch vfs_intent-2.6-rhel4.patch
4403 Updated patch vfs_races-2.6-rhel4.patch
4404 Updated series file 2.6-rhel4-titech.series
4405 Updated series file 2.6-rhel4.series
4406 Updated kernel config files
4411 Description: lquota slave complains LBUG when reconnecting with mds
4413 Details : quota slave depends qctxt->lqc_import to send its quota request.
4414 This pointer will be invalid if mds did failover or broke its
4415 connect to osts, which leads to LBUG.
4418 Frequency : when qunit size is too small(less than 20M)
4420 Description: write is stopped by improper -EDQUOT
4421 Details : If the master is busy and qunit size is small enough(let's say 1M),
4422 the slave can not get quota from master on time, which will lead
4423 slave to trigger a -EQUOTA to client.
4428 Description: Deadlock during metadata tests
4429 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4430 called with the per-dentry lock held.
4433 Frequency : SLES9 only
4435 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4436 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4438 Severity : enhancement
4440 Description: remove fs_prep_san_write operation and related patches
4441 Details : remove the ext3-san-jdike patches which are no longer useful.
4446 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4447 Details : a disk corruption can cause the mballoc code to assert on a
4448 double free or other extent corruptions. Handle these with
4449 ext3_error() instead of assertions.
4452 Frequency : only with mballoc3 code and deep extent trees
4454 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4455 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4456 can cause the filesystem to be remounted read-only.
4461 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4462 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4463 few external users of this ioctl number is lfsck's call to
4464 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4465 Add the old ioctl number to the handler so both old and new
4470 Description: parallel lock callbacks
4471 Details : Instead of sending blocking and completion callbacks as separated
4472 requests, adding them to a set and sending in parallel.
4476 Description: Disable most debugging by default
4477 Details : To improve performance, disable most logging (for debug purposes)
4478 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4479 default, and HA includes fewer messages.
4483 Frequency : only run on patchless client.
4484 Description: use do_facet on sanity.sh for test handling recoverables errors
4485 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4489 Description: hash tables for lists of nids, connections and uuids
4490 Details : Hash tables noticeably help when a lot of clients connect to a
4491 server, to faster identify duplicate connections or reconnects,
4492 also to faster find export to evict in manual eviction case.
4496 Description: Sometimes, when the server evict a client, and the client will
4497 not be evicted as soon as possible.
4498 Details : In enqueue req, the error was returned by intent, instead of
4499 rq_status which make ptlrpc layer not detect this error, and
4500 does not evict the client. So enqueue error should be returned
4504 Frequency : only at startup
4506 Description: Delay client connections to MDT until fist MDT->OST connect
4507 Details : If a client tried to create a new file before the MDT had
4508 connected to any OSTs, the create would return EIO. Now
4509 the client will simply block until the MDT connects to the
4510 first OST and the create can succeed.
4513 Frequency : only for SLES9
4515 Description: Routinely utilize latest Quadrics drivers in CFS releases
4516 Details : Update patch qsnet-suse-2.6.patch.
4519 Frequency : only for sles10
4521 Description: Update patches for SLES 10 SP1 kernel.
4522 Details : Update the patch vfs_intent-2.6-sles10.patch.
4527 Description: Routinely utilize latest Quadrics drivers in CFS releases
4528 Details : Update patch qsnet-rhel4-2.6.patch.
4531 Frequency : at statup only
4533 Description: mds_lov_synchronize race leads to various problems
4534 Details : simultaneous MDT->OST connections at startup can cause the
4535 sync to abort, leaving the OSC in a bad state.
4540 Description: Using wrong pointer in osc_brw_prep_request
4541 Details : Access to array[-1] can produce panic if kernel compiled with
4542 CONFIG_PAGE_ALLOC enabled
4544 Severity : enhancement
4546 Description: Async OSC create to avoid the blocking unnecessarily.
4547 Details : If a OST has no remain object, system will block on the creating
4548 when need to create a new object on this OST. Now, ways use
4549 pre-created objects when available, instead of blocking on an
4550 empty osc while others are not empty. If we must block, we block
4551 for the shortest possible period of time.
4555 Frequency : only in recovery
4556 Description: Mark OST as early accessible if his start SYNC.
4557 Details : osc_precreate return flag early accessible if oscc marked as
4558 OSCC_FLAG_SYNC_IN_PROGRESS.
4563 Description: Sometimes precreate code can triger create object on wrong ost
4564 Details : Wrong protected or not not restored variables aftre precreate loop
4565 can produce creation object on wrong ost.
4568 Frequency : oss recovery
4570 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4572 Details : llog_commit_thread_count should be synced between llog_commit
4573 start and cleanup, so new llog_commit thread should not be started
4574 when llog_commit threads being stopped to avoid accessing some
4577 Severity : enhancement
4579 Description: Add printing inode info into message about error in writepage.
4581 --------------------------------------------------------------------------------
4583 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4588 Frequency : on some architectures
4590 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4591 Details : Uninitialized flags sometimes cause configuration commands to
4595 Frequency : patchless clients only
4597 Description: missing __iget() symbol export
4598 Details : The __iget() symbol export is missing. To avoid the need for
4599 this on patchless clients the deathrow inode reaper is turned
4600 off, and we depend on the VM to clean up old inodes. This
4601 dependency was during via the fix for bug 12181.
4606 Description: sanity.sh fail: test_52b
4607 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4608 incorrect inode's flags to client.
4610 --------------------------------------------------------------------------------
4612 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4614 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4615 INTEROPERATE with older versions automatically. In many cases a
4616 special upgrade step is needed. Please read the
4617 user documentation before upgrading any part of a 1.4.x system.
4618 * WARNING: Lustre configuration and startup changes are required with
4619 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4621 * Support for kernels:
4622 2.4.21-47.0.1.EL (RHEL 3),
4623 2.6.5-7.283 (SLES 9),
4624 2.6.9-42.0.10.EL (RHEL 4),
4625 2.6.12.6 vanilla (kernel.org),
4626 2.6.16.27-0.9 (SLES10)
4627 * Client support for unpatched kernels:
4628 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4629 2.6.16 - 2.6.19 vanilla (kernel.org),
4630 2.6.9-42.0.8EL (RHEL 4)
4631 * Recommended e2fsprogs version: 1.39.cfs6
4632 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4635 Severity : enhancement
4637 Description: Async OSC create to avoid the blocking unnecessarily.
4638 Details : If a OST has no remain object, system will block on the creating
4639 when need to create a new object on this OST. Now, ways use
4640 pre-created objects when available, instead of blocking on an
4641 empty osc while others are not empty. If we must block, we block
4642 for the shortest possible period of time.
4644 Severity : enhancement
4646 Description: MountConf
4647 Details : Lustre configuration is now managed via mkfs and mount
4648 commands instead of lmc and lconf. New obd types (MGS, MGC)
4649 are added for dynamic configuration management. See
4650 https://mail.clusterfs.com/wikis/lustre/MountConf for
4653 Severity : enhancement
4655 Description: dynamic OST addition
4656 Details : OSTs can now be added to a live filesystem
4658 Severity : enhancement
4660 Description: startup order invariance
4661 Details : MDTs and OSTs can be started in any order. Clients only
4662 require the MDT to complete startup.
4664 Severity : enhancement
4666 Description: parallel, asynchronous orphan cleanup
4667 Details : orphan cleanup is now performed in separate threads for each
4668 OST, allowing parallel non-blocking operation.
4670 Severity : enhancement
4672 Description: optimized stripe assignment
4673 Details : stripe assignments are now made based on ost space available,
4674 ost previous usage, and OSS previous usage, in order to try
4675 to optimize storage space and networking resources.
4677 Severity : enhancement
4679 Description: Permanently set tunables
4680 Details : All writable /proc/fs/lustre tunables can now be permanently
4681 set on a per-server basis, at mkfs time or on a live
4684 Severity : enhancement
4686 Description: Lustre message v2
4687 Details : Add lustre message format v2.
4689 Severity : enhancement
4691 Description: client OST exclusion list
4692 Details : Clients can be started with a list of OSTs that should be
4693 declared "inactive" for known non-responsive OSTs.
4697 Description: ENOENT returned for valid filehandle during dbench.
4698 Details : Check if a directory has children when invalidating dentries
4699 associated with an inode during lock cancellation. This fixes
4700 an incorrect ENOENT sometimes seen for valid filehandles during
4701 testing with dbench.
4704 Frequency : SFS test only (otherwise harmless)
4706 Description: SPEC SFS validation failure on NFS v2 over lustre.
4707 Details : Changes the blocksize for regular files to be 2x RPC size,
4708 and not depend on stripe size.
4710 Severity : enhancement
4712 Description: fine-grained SMP locking inside DLM
4713 Details : Improve DLM performance on SMP systems by removing the single
4714 per-namespace lock and replace it with per-resource locks.
4716 Severity : enhancement
4718 Description: don't hold multiple extent locks at one time
4719 Details : To avoid client eviction during large writes, locks are not
4720 held on multiple stripes at one time or for very large writes.
4721 Otherwise, clients can block waiting for a lock on a failed OST
4722 while holding locks on other OSTs and be evicted.
4724 Severity : enhancement
4726 Description: Multiple MD RPCs in flight.
4727 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4728 To avoid overly-overloading MDT, introduce a limit on number of
4729 MDT RPCs in flight for a single client and add /proc controls
4730 to adjust this limit.
4732 Severity : enhancement
4734 Description: client read/write statistics
4735 Details : Add client read/write call usage stats for performance
4736 analysis of user processes.
4737 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4738 file access. extents_stats shows chunk size distribution.
4739 extents_stats_per_process show chunk size distribution per
4742 Severity : enhancement
4744 Description: per-client statistics on server
4745 Details : Add ldlm and operations statistics for each client in
4746 /proc/fs/lustre/mds|obdfilter/*/exports/
4748 Severity : enhancement
4750 Description: improved MDT statistics
4751 Details : Add detailed MDT operations statistics in
4752 /proc/fs/lustre/mds/*/stats
4754 Severity : enhancement
4756 Description: VFS operations stats
4757 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4758 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4763 Description: Fix client-side osc byte counters
4764 Details : The osc read/write byte counters in
4765 /proc/fs/lustre/osc/*/stats are now working
4768 Frequency : always as root on SLES
4770 Description: Failure of copying files with lustre special EAs.
4771 Details : Client side always return success for setxattr call for lustre
4772 special xattr (currently only "trusted.lov").
4777 Description: Refcount LNET uuids
4778 Details : The global LNET uuid list grew linearly with every startup;
4779 refcount repeated list entries instead of always adding to
4782 Severity : enhancement
4784 Description: Dynamic service threads
4785 Details : Within a small range, start extra service threads
4786 automatically when the request queue builds up.
4789 Frequency : mixed-endian client/server environments
4791 Description: mixed-endian crashes
4792 Details : The new msg_v2 system had some failures in mixed-endian
4795 Severity : enhancement
4797 Description: Easy OST removal
4798 Details : OSTs can be permanently deactivated with e.g. 'lctl
4799 conf_param lustre-OST0001.osc.active=0'
4801 Severity : enhancement
4803 Description: MGS proc entries
4804 Details : Added basic proc entries for the MGS showing what filesystems
4807 Severity : enhancement
4809 Description: provide MGS failover
4810 Details : Added config lock reacquisition after MGS server failover.
4812 Severity : enhancement
4814 Description: add Linux 2.4 support
4815 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4819 Description: a large application tries to do I/O to the same resource and dies
4820 in the middle of it.
4821 Details : Check the req->rq_arrival time after the call to
4822 ost_brw_lock_get(), but before we do anything about
4823 processing it & sending the BULK transfer request. This
4824 should help move old stale pending locks off the queue as
4825 quickly as obd_timeout.
4828 Frequency : when an incorrect nid is specified during startup
4830 Description: ptlrpc connect to non-existant node causes kernel crash
4831 Details : LNET can't be re-entered from an event callback, which
4832 happened when we expire a message after the export has been
4833 cleaned up. Instead, hand the zombie cleanup off to another
4836 Severity : enhancement
4838 Description: plain/inodebits lock performance improvement
4839 Details : Grouping plain/inodebits in granted list by their request modes
4840 and bits policy, thus improving the performance of search through
4844 Frequency : only if OST filesystem is corrupted
4846 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4847 Details : for a short time RPCs with bulk IO are in the replay list,
4848 but replay of bulk IOs is unimplemented. If the OST filesystem
4849 is corrupted due to disk cache incoherency and then replay is
4850 started it is possible to trip an assertion. Avoid putting
4851 committed RPCs into the replay list at all to avoid this issue.
4854 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4856 Description: System hang on startup
4857 Details : This bug allowed the liblustre (e.g. catamount) client to
4858 return to the app before handling all startup RPCs. This
4859 could leave the node unresponsive to lustre network traffic
4860 and manifested as a server ptllnd timeout.
4862 Severity : enhancement
4864 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4865 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4866 boolean to control whether to print peer debug info when a
4867 client's RPC times out.
4870 Frequency : only for kernels with patches from Lustre below 1.4.3
4872 Description: Remove old rdonly API
4873 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4876 Frequency : only for devices with external journals
4878 Description: Set external device read-only also
4879 Details : During a commanded failover stop, we set the disk device
4880 read-only while the server shuts down. We now also set any
4881 external journal device read-only at the same time.
4884 Frequency : when upgrading from 1.4 while trying to change parameters
4886 Description: The wrong (new) MDC name was used when setting parameters for
4887 upgraded MDT's. Also allows changing of OSC (and MDC)
4888 parameters if --writeconf is specified at tunefs upgrade time.
4891 Frequency : when setting specific ost indicies
4893 Description: QOS code breaks on skipped indicies
4894 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4895 created with --index need not be sequential.
4897 Severity : enhancement
4899 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4900 Details : The uninit_groups feature works in conjunction with the kernel
4901 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4902 up the pass1 processing of e2fsck. This is a read-only feature
4903 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4904 mount filesystems that have had this feature enabled.
4906 Severity : enhancement
4908 Description: Improve multi-block allocation algorithm to avoid fragmentation
4909 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4910 allocation locality and avoid filesystem fragmentation.
4912 ------------------------------------------------------------------------------
4914 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4916 * Support for kernels:
4917 2.4.21-47.0.1.EL (RHEL 3)
4918 2.6.5-7.283 (SLES 9)
4919 2.6.9-42.0.10.EL (RHEL 4)
4920 2.6.12.6 vanilla (kernel.org)
4921 2.6.16.27-0.9 (SLES 10)
4922 * Recommended e2fsprogs version: 1.39.cfs5
4924 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4928 Frequency : occasional, depends on client load and configuration
4929 Bugzilla : 12181, 12203
4930 Description: data loss for recently-modified files
4932 Details : In some cases it is possible that recently written or created
4933 files may not be written to disk in a timely manner (this should
4934 normally be within 30s unless client IO load is very high).
4935 The problem appears as zero-length files or files that are a
4936 multiple of 1MB in size after a client crash or client eviction
4937 that are missing data at the end of the file.
4939 This problem is more likely to be hit on clients where files are
4940 repeatedly created and unlinked in the same directory, clients
4941 have a large amount of RAM, have many CPUs, the filesystem has
4942 many OSTs, the clients are rebooted frequently, and/or the files
4943 are not accessed by other nodes after being written.
4945 The presence of the problem can be detected by looking at
4946 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4947 smaller than the second (nr_unused) then dirty files will not
4948 be flushed automatically to disk. "sync; sleep 10" should be
4949 run several times on the node before unmounting it to update
4950 Lustre (this is also safe to run on nodes without this problem).
4952 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4953 that can cause this same problem, so customers using that kernel
4954 also need to update the kernel in addition to Lustre. In order
4955 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4957 It is normal that files written just before a client crash (less
4958 than 30s) may not yet have been flushed to disk, even for local
4962 Frequency : frequent on thin XT3 nodes
4964 Description: UUID collision on thin XT3 Linux nodes
4965 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4966 randomly, since we relied on an insufficiently-seeded PRNG.
4971 Description: OSS hangs after "All ost request buffers busy"
4972 Details : A deadlock between quota and journal operations caused OSS
4973 hangs after printing "All ost request buffers busy."
4976 Frequency : always on liblustre builds
4978 Description: Cleanup compiler warnings on liblustre
4981 Frequency : always on liblustre builds on XT3
4983 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4987 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4988 > 32000 subdirectories
4989 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4990 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4991 subdirectory count crosses 32000. This will aid e2fsck to
4992 correctly handle more than 32000 subdirectories.
4995 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4997 Description: System hang on startup
4998 Details : This bug allowed the liblustre (e.g. catamount) client to
4999 return to the app before handling all startup RPCs. This
5000 could leave the node unresponsive to lustre network traffic
5001 and manifested as a server ptllnd timeout.
5003 Severity : enhancement
5005 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
5006 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
5007 boolean to control whether to print peer debug info when a
5008 client's RPC times out.
5013 Description: make O_SYNC working on 2.6 kernels
5014 Details : 2.6 kernels use different method for mark pages for write,
5015 so need add a code to lustre for O_SYNC work.
5020 Description: Failure to close file and release space on NFS
5021 Details : Put inode details into lock acquired in ll_intent_file_open.
5022 Use mdc_intent_lock in ll_intent_open to properly
5023 detect all kind of errors unhandled by mdc_enqueue.
5028 Description: proc file read during shutdown sometimes raced obd removal,
5030 Details : Add lock to prevent obd access after proc file removal.
5033 Frequency : Only for files larger than 4GB on 32-bit clients.
5035 Description: improperly doing page alignment of locks
5036 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
5037 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
5042 Description: RPCs being resent when they shouldn't be
5043 Details : Some RPCs that should not be resent are being resent. This
5044 can cause inconsistencies in the RPC state machine. Do not
5045 resend such requests.
5048 Frequency : rare, only with NFS export
5050 Description: Crash on NFS re-export node
5051 Details : under very unusual load conditions an assertion is hit in
5052 ll_intent_file_open()
5055 Frequency : only if OST filesystem is corrupted
5057 Description: client incorrectly hits assertion in ptlrpc_replay_req()
5058 Details : for a short time RPCs with bulk IO are in the replay list,
5059 but replay of bulk IOs is unimplemented. If the OST filesystem
5060 is corrupted due to disk cache incoherency and then replay is
5061 started it is possible to trip an assertion. Avoid putting
5062 committed RPCs into the replay list at all to avoid this issue.
5067 Description: large O_DIRECT requests fail under memory pressure/fragmentation
5068 Details : Large single O_DIRECT read and write calls can fail to allocate
5069 a sufficiently large buffer to process the request. In case of
5070 allocation failure the allocation is retried with a smaller
5071 buffer and broken into smaller requests.
5073 Severity : enhancement
5075 Description: Add -o localflock option to simulate old noflock behaviour.
5076 Details : This will achieve local-only flock/fcntl locks coherentness.
5081 Description: versioning check is incomplete
5082 Details : Checking the version difference of client vs. server, report
5083 error if the gap is too big.
5088 Description: add support PG_writeback bit
5089 Details : add support for PG_writeback bit for Lustre, for more carefull
5090 work with page cache in 2.6 kernel. This also fix some deadlocks
5091 and remove hack for work O_SYNC with 2.6 kernel.
5093 Severity : enhancement
5095 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
5096 Details : The uninit_groups feature works in conjunction with the kernel
5097 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
5098 up the pass1 processing of e2fsck. This is a read-only feature
5099 in ldiskfs2 only, so older kernels and current ldiskfs cannot
5100 mount filesystems that have had this feature enabled.
5102 Severity : enhancement
5104 Description: Improve multi-block allocation algorithm to avoid fragmentation
5105 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
5106 allocation locality and avoid filesystem fragmentation.
5108 ------------------------------------------------------------------------------
5110 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
5112 * Support for kernels:
5113 2.6.9-42.0.3.EL (RHEL 4)
5114 2.6.5-7.276 (SLES 9)
5115 2.4.21-47.0.1.EL (RHEL 3)
5116 2.6.12.6 vanilla (kernel.org)
5117 2.6.16.21-0.8 (SLES10)
5118 * Recommended e2fsprogs version: 1.39.cfs2-0
5120 * The backwards-compatible /proc/sys/portals symlink has been removed
5121 in this release. Before upgrading, please ensure that you change
5122 any configuration scripts or /etc/sysctl.conf files that access
5123 /proc/sys/portals/* or sysctl portals.* to use the corresponding
5124 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
5125 in advance of the upgrade on any system running Lustre 1.4.6 or
5126 newer, since /proc/sys/lnet was added in that version.
5127 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
5131 Frequency : only when quota is used
5133 Description: avoid scanning export list for quota master
5134 Details : Change the algorithms to avoid scanning export list in order
5135 to improve the efficiency.
5138 Frequency : MDS failover only, very rarely
5140 Description: "went back in time" messages on mds failover
5141 Details : The greatest transno may be lost when the current operation
5142 finishes with an error (transno==0) and the client's last_rcvd
5143 record is over-written. Save the greatest transno in the
5144 mds_last_transno for this case.
5147 Frequency : always for specific kernels and striping counts
5149 Description: client may get "Matching packet too big" without ACL support
5150 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
5151 when trying to access files in certain configurations. The
5152 clients should in fact be denied when mounting because they do
5153 not understand ACLs.
5156 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
5158 Description: many clients connecting with IO in progress causes connect timeouts
5159 Details : Avoid synchronous journal commits to avoid delays caused by many
5160 clients connecting/disconnecting when bulk IO is in progress.
5161 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
5162 OST_IO_PORTAL to avoid delays behind potentially many pending
5166 Frequency : occasionally with multiple writers to a single file
5168 Description: shared writes to file may result in wrong size reported by stat()
5169 Details : Allow growing of kms when extent lock is cancelled
5172 Frequency : always with random mmap IO to multi-striped file
5174 Description: mmap write might be lost if we are writing to a 'hole' in stripe
5175 Details : Only if the hole is at the end of OST object so that kms is too
5176 small. Fix is to increase kms accordingly in ll_nopage.
5179 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
5181 Description: writes to a missing object would leak memory on the OST
5182 Details : If there is an inconsistency between the MDS and OST filesystems,
5183 such that the MDS references an object that doesn't exist, writes
5184 to that object will leak memory due to incorrect cleanup in the
5185 error handling path, eventually running out of memory on the OST.
5190 Description: Creating too long symlink causes lustre errors
5191 Details : Check symlink and name lengths before sending requests to MDS.
5194 Frequency : only if flock is enabled (not on by default)
5196 Description: posix locks not released on fd closure on 2.6.9+
5197 Details : We failed to add posix locks to list of inode locks on 2.6.9+
5198 kernels, this caused such locks not to be released on fd close and
5199 then assertions on fs unmount about still used locks.
5202 Frequency : MDS failover only, very rarely
5204 Description: clients may get ASSERTION(granted_lock != NULL)
5205 Details : When request was taking a long time, and a client was resending
5206 a getattr by name lock request. The were multiple lock requests
5207 with the same client lock handle and
5208 mds_getattr_name->fixup_handle_for_resent_request found one of the
5209 lock handles but later failed with ASSERTION(granted_lock != NULL).
5214 Description: handle->h_buffer_credits > 0, assertion failure
5215 Details : h_buffer_credits is zero after truncate, causing assertion
5216 failure. This patch extends the transaction or creates a new
5220 Frequency : NFS re-export or patchless client
5221 Bugzilla : 11179, 10796
5222 Description: Crash on NFS re-export node (__d_move)
5223 Details : We do not want to hash the dentry if we don't have a lock.
5224 But if this dentry is later used in d_move, we'd hit uninitialised
5225 list head d_hash, so we just do this to init d_hash field but
5226 leave dentry unhashed.
5229 Frequency : NFS re-export or patchless client
5231 Description: NFS exports has problem with symbolic link
5232 Details : lustre client didn't properly install dentry when re-exported
5233 to NFS or running patchless client.
5236 Frequency : NFS re-export or patchless client
5238 Description: Various nfs/patchless fixes.
5239 Details : fixes reuse disconected alias for lookup process - this fixes
5240 warning "find_exported_dentry: npd != pd",
5241 fix permission error with open files at nfs.
5242 fix apply umask when do revalidate.
5245 Frequency : occasional
5247 Description: Crash on NFS re-export node
5248 Details : calling clear_page() on the wrong pointer triggered oops in
5249 generic_mapping_read().
5252 Frequency : rarely, using O_DIRECT IO
5254 Description: unaligned directio crashes client with LASSERT
5255 Details : check for unaligned buffers before trying any requests.
5258 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
5260 Description: stale data returned from RAID cache
5261 Details : If only a small amount of IO is done to the RAID device before
5262 reading it again it is possible to get stale data from the RAID
5263 cache instead of reading it from disk.
5266 Frequency : always for sles10 kernel
5268 Description: sles10 support
5269 Details : ll_follow_link: compile fixes and using of nd_set_link
5270 under newer kernels.
5273 Frequency : depends on arch, kernel and compiler version, always on sles10
5276 Description: recursive or deep enough symlinks cause stack overflow
5277 Details : getting rid of large stack-allocated variable in
5281 Frequency : depends on hardware
5283 Description: lustre write performance loss in the SLES10 kernel
5284 Details : the performance loss is caused by using of write barriers in the
5285 ext3 code. The SLES10 kernel turns barrier support on by
5286 default. The fix is to undo that change for ldiskfs.
5288 ------------------------------------------------------------------------------
5290 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
5292 * Support for kernels:
5293 2.6.9-42.0.3EL (RHEL 4)
5294 2.6.5-7.276 (SLES 9)
5295 2.4.21-47.0.1.EL (RHEL 3)
5296 2.6.12.6 vanilla (kernel.org)
5300 Frequency : quota enabled and large files being deleted
5302 Description: releasing more than 4GB of quota at once hangs OST
5303 Details : If a user deletes more than 4GB of files on a single OST it
5304 will cause the OST to spin in an infinite loop. Release
5305 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5310 Description: statfs data retrieved from /proc may be stale or zero
5311 Details : When reading per-device statfs data from /proc, in the
5312 {kbytes,files}_{total,free,avail} files, it may appear
5313 as zero or be out of date.
5316 Frequency : systems with MD RAID1 external journal devices
5318 Description: lconf's call to blkid is confused by RAID1 journal devices
5319 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5320 of returning all block devices that match the journal UUID.
5323 Frequency : always, for aggregate stripe size over 4GB
5325 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5326 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5327 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5330 Frequency : always if "lfs find" used on a local file/directory
5332 Description: "lfs find" segfaults if used on a local file/directory
5333 Details : The case where a directory component was not specified wasn't
5334 handled correctly. Handle this properly.
5337 Frequency : always on ppc64
5339 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5340 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5341 perform endianness conversion.
5344 Frequency : rarely (truncate to non-zero file size after write under load)
5345 Bugzilla : 10730, 10687
5346 Description: Files padded with zeros to next 4K multiple
5347 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5348 it is possible that files that are truncated to a non-zero size
5349 immediately after being written are filled with zero bytes beyond
5350 the truncated size. No file data is lost.
5352 Severity : enhancement
5354 Description: Allow recovery/failover for liblustre clients.
5355 Details : liblustre clients were unaware of failover configurations until
5358 Severity : enhancement
5360 Description: user file locks should fail when not mounting with flock option
5361 Details : Set up an error-returning stub in ll_file_operations.lock field
5362 to prevent incorrect behaviour when client is mounted without
5363 flock option. Also, set up properly f_op->flock field for
5367 Frequency : always on ia64
5369 Description: "lfs df" loops on printing out MDS statfs information
5370 Details : The obd_ioctl_data was not initialized and in some systems
5371 this caused a failure during the ioctl that did not return
5372 an error. Initialize the struct and return an error on failure.
5375 Frequency : SLES 9 only
5377 Description: Error of copying files with lustre special EAs as root
5378 Details : Client side always return success for setxattr call for lustre
5379 special xattr (currently only "trusted.lov").
5382 Frequency : rarely on clusters with both ia64+i386 clients
5384 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5385 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5386 clients are doing concurrent writes to the same file it is
5387 possible that the ia64 clients may LASSERT because the OST
5388 extent locks are not PAGE_SIZE aligned. Ensure that grown
5389 locks are always aligned on the request boundary.
5392 Frequency : specific use, occasional
5394 Description: Overwriting in use executable truncates on-disk binary image
5395 Details : If one node attempts to overwrite an executable in use by
5396 another node, we now correctly return ETXTBSY instead of
5397 truncating the file.
5399 Severity : enhancement
5401 Description: Async OSC create to avoid the blocking unnecessarily.
5402 Details : If a OST has no remain object, system will block on the creating
5403 when need to create a new object on this OST. Now, ways use
5404 pre-created objects when available, instead of blocking on an
5405 empty osc while others are not empty. If we must block, we block
5406 for the shortest possible period of time.
5411 Description: chmod on Lustre root is propagated to other clients
5412 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5413 invalid by the follow_mount time.
5416 Frequency : liblustre clients only
5418 Description: Race in 'instant cancel' lock handling could lead to such locks
5419 never to be granted in case of SMP MDS
5420 Details : Do not destroy not yet granted but cbpending locks in
5424 Frequency : replay/resend of open
5426 Description: non null lock assetion failure in mds_intent_policy
5427 Details : Trying to replay/resend lockless open requests resulted in
5428 mds_open() returning 0 with no lock. Now it sets a flag if
5429 a lock is going to be returned.
5431 Severity : enhancement
5433 Description: Checksum enhancements
5434 Details : New checksum enhancements allow for resending RPCs that failed
5437 Severity : enhancement
5439 Description: Tunables on number of dirty pages in cacche
5440 Details : Allow to set limit on number of dirty pages cached.
5445 Description: client crash on unmount - lock still has references
5446 Details : In some error handling cases it was possible to leak a lock
5447 reference on a client while accessing a file. This was not
5448 harmful to the client during operation, but would cause the
5449 client to crash when the filesystem is unmounted.
5452 Frequency : specific case, rare
5454 Description: ETXTBSY on mds though file not in use by client
5455 Details : ETXTBSY is no longer incorrectly returned when attempting to
5456 chmod or chown a directory that the user previously tried to
5457 execute or a currently-executing binary.
5460 Frequency : extremely rare except on liblustre-based clients
5462 Description: Lustre space not freed when files are deleted
5463 Details : Clean up open-unlinked files after client eviction. Previously
5464 the unlink was skipped and the files remained as orphans.
5469 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5470 Details : In some cases it was possible to send a blocking callback to
5471 a client doing a glimpse, even though that client didn't get
5472 a lock granted. When the glimpse lock is cancelled on the OST
5473 the freed lock is left on the waiting list and corrupted the list.
5476 Frequency : all core dumps
5478 Description: Broke core dumps to lustre
5479 Details : Negative dentry may be unhashed if parent does not have UPDATE
5480 lock, but some callers, e.g. do_coredump, expect dentry to be
5481 hashed after successful create, hash it in ll_create_it.
5483 ------------------------------------------------------------------------------
5485 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5487 * Support for kernels:
5488 2.6.9-42.0.2.EL (RHEL 4)
5489 2.6.5-7.276 (SLES 9)
5490 2.4.21-40.EL (RHEL 3)
5491 2.6.12.6 vanilla (kernel.org)
5495 Frequency : always on RHEL 3
5497 Description: Number of open files grows over time
5498 Details : The number of open files grows over time, whether or not
5499 Lustre is started. This was due to a filp leak introduced
5500 by one of our kernel patches.
5502 ------------------------------------------------------------------------------
5504 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5506 * Support for kernels:
5507 2.6.9-42.EL (RHEL 4)
5508 2.6.5-7.267 (SLES 9)
5509 2.4.21-40.EL (RHEL 3)
5510 2.6.12.6 vanilla (kernel.org)
5515 Bugzilla : 5719, 9635, 9792, 9684
5516 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5517 Details : If a server is under heavy load and cannot reply to new
5518 connection requests before the client resends the (re)connect,
5519 the connection handling code can behave badly if two service
5520 threads are concurrently handing separate (re)connections from
5521 the same client. Add better locking to the connection handling
5522 code, and ensure that only a single connection will be processed
5523 for a given client UUID, even if the lock is dropped.
5525 Severity : enhancement
5527 Description: add TCP zero-copy support to kernel
5528 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5529 sends if the hardware supports scatter-gather and checksumming.
5530 This allows socklnd to do client-write and server-read more
5531 efficiently and reduce CPU utilization from skbuf copying.
5534 Frequency : only if NFS exporting from client
5536 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5537 Details : If setting attributes on a file created under NFS that had
5538 never been opened it would be possible to oops the client
5539 if the file had no objects.
5543 Bugzilla : 9326, 10402, 10897
5544 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5545 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5546 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5547 before ptlrpcd_wake() was called and the request was freed.
5550 Frequency : always for liblustre
5552 Description: liblustre client does MDS+OSTs setattr RPC for each write
5553 Details : When doing a write from a liblustre client, the client
5554 incorrectly issued an RPC to the MDS and each OST the file was
5555 striped over in order to update the timestamps. When writing
5556 with small chunks and many clients this could overwhelm the MDS
5557 with RPCs. In all cases it would slow down the write because
5558 these RPCs are unnecessary.
5560 Severity : enhancement
5562 Description: allow number of MDS service threads to be changed at module load
5563 Details : It is now possible to change the number of MDS service threads
5564 running. Adding "options mds mds_num_threads={N}" to the MDS's
5565 /etc/modprobe.conf will set the number of threads for the next
5566 time Lustre is restarted (assuming the "mds" module is also
5567 reloaded at that time). The default number of threads will
5568 stay the same, 32 for most systems.
5573 Description: OST crash if filesystem is unformatted or corrupt
5574 Details : If an OST is started on a device that has never been formatted
5575 or if the filesystem is corrupt and cannot even mount then the
5576 error handling cleanup routines would dereference a NULL pointer.
5581 Description: NULL pointer deref in llap_from_page.
5582 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5583 incorrect) mapping if the page was truncated/reclaimed while it was
5584 searched for. Check for this condition and skip such pages when
5585 doing readahead. Introduce extra check to llap_from_page() to
5586 verify page->mapping->host is non-NULL (so page is not anonymous).
5589 Frequency : Sometimes when using sys_sendfile
5591 Description: "page not covered by a lock" warnings from ll_readpage
5592 Details : sendfile called ll_readpage without right page locks present.
5593 Now we introduced ll_file_sendfile that does necessary locking
5594 around call to generic_file_sendfile() much like we do in
5598 Frequency : with certain MDS communication failures at client mount time
5600 Description: NULL pointer deref after failed client mount
5601 Details : a client connection request may delayed by the network layer
5602 and not be sent until after the PTLRPC layer has timed out the
5603 request. If the client fails the mount immediately it will try
5604 to clean up before the network times out the request. Add a
5605 reference from the request import to the obd device and delay
5606 the cleanup until the network drops the request.
5609 Frequency : occasionally during client (re)connect
5611 Description: assertion failure during client (re)connect
5612 Details : processing a client connection request may be delayed by the
5613 client or server longer than the client connect timeout. This
5614 causes the client to resend the connection request. If the
5615 original connection request is replied in this interval, the
5616 client may trip an assertion failure in ptlrpc_connect_interpret()
5617 which thought it would be the only running connect process.
5620 Frequency : only with obd_echo servers and clients that are rebooted
5622 Description: kernel BUG accessing uninitialized data structure
5623 Details : When running an obd_echo server it did not start the ping_evictor
5624 thread, and when a client was evicted an uninitialized data
5625 structure was accessed. Start the ping_evictor in the RPC
5626 service startup instead of the OBD startup.
5628 Severity : enhancement
5629 Bugzilla : 10193 (patchless)
5630 Description: Remove dependency on various unexported kernel interfaces.
5631 Details : No longer need reparent_to_init, exit_mm, exit_files,
5632 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5635 Frequency : rare (only users of deprecated and unsupported LDAP config)
5637 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5638 Details : LDAP apparently contains 'lustreName' attributes instead of
5639 'name'. A simple remapping of the name is sufficient.
5642 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5644 Description: waiting_locks_callback trips kernel BUG if client is evicted
5645 Details : Running with the dump_on_timeout debug flag turned on makes
5646 it possible that the waiting_locks_callback() can try to dump
5647 the Lustre kernel debug logs from an interrupt handler. Defer
5648 this log dumping to the expired_lock_main() thread.
5650 Severity : enhancement
5652 Description: Support NFS exporting on 2.6 kernels.
5653 Details : Implement non-rawops metadata methods for NFS server to use without
5654 changing NFS server code.
5657 Frequency : very rare (synthetic metadata workload only)
5659 Description: two racing renames might cause an MDS thread to deadlock
5660 Details : Running the "racer" program may cause one MDS thread to rename
5661 a file from being the source of a rename to being the target of
5662 a rename at exactly the same time that another thread is doing
5663 so, and the second thread has already enqueued these locks after
5664 doing a lookup of the target and is trying to relock them in
5665 order. Ensure that we don't try to re-lock the same resource.
5668 Frequency : only very large systems with liblustre clients
5670 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5671 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5672 all OSTs in parallel. This allows the eviction of stale liblustre
5673 clients to proceed much faster than if they were done in series,
5674 and also offers similar improvements for other set_info RPCs.
5679 Description: excessive CPU usage during initial read phase on client
5680 Details : During the initial read phase on a client, it would agressively
5681 retry readahead on the file, consuming too much CPU and impacting
5682 performance (since 1.4.5.8). Improve the readahead algorithm
5683 to avoid this, and also improve some other common cases (read
5684 of small files in particular, where "small" is files smaller than
5685 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5690 Description: MDS crash when receiving packet with unknown intent.
5691 Details : Do not LBUG in unknown intent case, just return -EFAULT
5693 Severity : enhancement
5694 Bugzilla : 9293, 9385
5695 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5696 Details : Do not serialize getattr (non-intent version) and statfs.
5699 Frequency : occasional, when OST network is overloaded/intermittent
5701 Description: client evicted by OST after bulk IO timeout
5702 Details : If a client sends a bulk IO request (read or write) the OST
5703 may evict the client if it is unresposive to its data GET/PUT
5704 request. This is incorrect if the network is overloaded (takes
5705 too long to transfer the RPC data) or dropped the OST GET/PUT
5706 request. There is no need to evict the client at all, since
5707 the pinger and/or lock callbacks will handle this, and the
5708 client can restart the bulk request.
5711 Frequency : Always when mmapping file with no objects
5713 Description: client crashes when mmapping file with no objects
5714 Details : Check that we actually have objects in a file before doing any
5715 operations on objects in ll_vm_open, ll_vm_close and
5721 Description: Request leak when working with deleted CWD
5722 Details : Introduce advanced request refcount tracking for requests
5723 referenced from lustre intent.
5725 Severity : Enhancement
5727 Description: Cache open file handles on client.
5728 Details : MDS now will return special lock along with openhandle, if
5729 requested and client is allowed to hold openhandle, even if unused,
5730 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5731 closing files for every read/write openration.
5733 Severity : Enhancement
5735 Description: Cache open negative dentries on client when possible.
5736 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5737 negative dentries on lock revocation.
5742 Description: Remounting a client read-only wasn't possible with a zconf mount
5743 Details : It wasn't possible to remount a client read-only with llmount.
5745 Severity : enhancement
5746 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5747 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5748 MPICH is included to add Lustre-specific ADIO interfaces.
5749 This is based closely on the UFS ADIO layer and only differs
5750 in file creation, in order to allow the OST striping to be set.
5751 This is user-contributed code and not supported by CFS.
5756 Description: extended inode attributes (immutable, append-only) work improperly
5757 when 2.4 and 2.6 kernels are used on client/server or vice versa
5758 Details : Introduce kernel-independent values for these flags.
5760 Severity : enhancement
5763 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5764 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5765 granularity. Now, Fractional MB granularity is be supported,
5766 this is very useful for low memory system.
5768 Severity : enhancement
5770 Description: Getattr by fid
5771 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5772 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5773 have locks and avoids getattr rpc requests for it on every stat.
5776 Frequency : Always, for filesystems larger than 2TB
5778 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5779 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5780 filesystems larger than 2TB to be mounted with mballoc enabled.
5783 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5785 Description: filesystem corruption for non-standard kernels and very large OSTs
5786 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5787 filesystems larger than 2TB is mounted then the kernel will
5788 silently corrupt the start of the filesystem. CONFIG_LBD is
5789 enabled for all CFS-supported kernels, but the possibility of
5790 this happening with a modified kernel config exists.
5792 Severity : enhancement
5794 Description: add client O_DIRECT support for 2.6 kernels
5795 Details : It is now possible to do O_DIRECT reads and writes to files
5796 in the Lustre client mountpoint on 2.6 kernel clients.
5798 Severity : enhancement
5800 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5801 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5802 parallel, not waiting for response from every OST before sending
5803 a rpc to the next OST.
5808 Description: setattr vs write race when updating file timestamps
5809 Details : Client processes that update a file timestamp into the past
5810 right after writing to the file (e.g. tar) it is possible that
5811 the updated file modification time can be reset to the current
5812 time due to a race between processing the setattr and write RPC.
5814 Severity : enhancement
5816 Description: Bring 'lfs find' closer in line with regular Linux find.
5817 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5818 -print0 options and obtains all the needed info through the lustre
5821 Severity : enhancement
5823 Description: support up to 1024 configured devices on one node
5824 Details : change obd_dev array from statically allocated to dynamically
5825 allocated structs as they are first used to reduce memory usage
5830 Description: Flush dirty partially truncated pages during truncate
5831 Details : Immediatelly flush partially truncated pages in filter_setattr,
5832 this way we completely avoid having any pages in page cache on OST
5833 and can retire ugly workarounds during writes to flush such pages.
5838 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5839 Details : protect inode from truncation within vfs_unlink() context
5840 just take a reference before calling vfs_unlink() and release it
5841 when parent's i_sem is free.
5844 Frequency : always, if extents are used on OSTs
5846 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5847 Details : OSTs using the extents format would not zero the high 16 bits of
5848 the index physical block number. This is not a problem for any
5849 OST filesystems smaller than 16TB, and no kernels support ext3
5850 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5851 new/modified files) and can be fixed for existing filesystems
5852 with e2fsprogs-1.39-cfs1.
5857 Description: import connection selection may be incorrect if timer wraps
5858 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5859 import connections to be ignored if the 32-bit jiffies counter
5860 wraps. Use a 64-bit jiffies counter.
5863 Frequency : during server recovery
5865 Description: crash after server is denying duplicate export
5866 Details : If clients are resending connect requests to the server, the
5867 server refuses to allow a client to connect multiple times.
5868 Fixed a bug in the handling of this case.
5871 Frequency : very large clusters immediately after boot
5873 Description: LNET request buffers exhausted under heavy short-term load
5874 Details : If a large number of client requests are generated on a service
5875 that has previously never seen so many requests it is possible
5876 that the request buffer growth cannot keep up with the spike in
5877 demand. Instead of dropping incoming requests, they are held in
5878 the LND until the RPC service can accept more requests.
5881 Frequency : Sometimes during replay
5883 Description: Assertion failure in ll_local_open after replay.
5884 Details : If replay happened on an open request reply before we were able
5885 to set replay handler, reply will become not swabbed tripping the
5886 assertion in ll_local_open. Now we set the handler right after
5887 recognising of open request
5890 Frequency : very rare
5892 Description: kernel reports "badness in vsnprintf"
5893 Details : Reading from the "recovery_status" /proc file in small chunks
5894 may cause a negative length in lprocfs_obd_rd_recovery_status()
5895 call to vsnprintf() (which is otherwise harmless). Exit early
5896 if there is no more space in the output buffer.
5898 Severity : enhancement
5900 Description: clear OBD RPC statistics by writing to them
5901 Details : It is now possible to clear the OBD RPC statistics by writing
5902 to the "stats" file.
5907 Description: Client mtime is not the same on different clients after utimes
5908 Details : In some cases, the client was using the utimes() syscall on
5909 a file cached on another node. The clients now validate the
5910 ctime from the MDS + OSTs to determine which one is right.
5915 Description: Inability to activate failout mode
5916 Details : lconf script incorrectly assumed that in python string's numeric
5917 value is used in comparisons.
5920 Frequency : always with multiple stripes per file
5922 Description: Inefficient object allocation for mutli-stripe files
5923 Details : When selecting which OSTs to stripe files over, for files with
5924 a stripe count that divides evenly into the number of OSTs,
5925 the MDS is always picking the same starting OST for each file.
5926 Return the OST selection heuristic to the original design.
5931 Description: mount failures may take full timeout to return an error
5932 Details : Under some heavy load conditions it is possible that a
5933 failed mount can wait for the full obd_timeout interval,
5934 possibly several minutes, before reporting an error.
5935 Instead return an error as soon as the status is known.
5937 ------------------------------------------------------------------------------
5939 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5941 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5942 INTEROPERATE with older versions automatically. Please read the
5943 user documentation before upgrading any part of a live system.
5944 * WARNING: Lustre networking configuration changes are required with
5945 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5948 * Support for kernels:
5949 2.6.9-22.0.2.EL (RHEL 4)
5950 2.6.5-7.244 (SLES 9)
5951 2.6.12.6 vanilla (kernel.org)
5954 Severity : enhancement
5955 Bugzilla : 7981/8208
5956 Description: Introduced Lustre Networking (LNET)
5957 Details : LNET is new networking infrastructure for Lustre, it includes
5958 a reorganized network configuration mode (see the user
5959 documentation for full details) as well as support for routing
5960 between different network fabrics. Lustre Networking Devices
5961 (LNDs) for the supported network fabrics have also been
5962 created for this new infrastructure.
5964 Severity : enhancement
5965 Description: Introduced Access control lists
5966 Details : clients can set ACLs on files and directories in order to have
5967 more fine-grained permissions than the standard Unix UGO+RWX.
5968 The MDS must be started with the "-o acl" mount option.
5970 Severity : enhancement
5971 Description: Introduced filesystem quotas
5972 Details : Administrators may now establish per-user quotas on the
5975 Severity : enhancement
5977 Description: Configuration change for the XT3
5978 The PTLLND is now used to run Lustre over Portals on the XT3
5979 The configure option(s) --with-cray-portals are no longer used.
5980 Rather --with-portals=<path-to-portals-includes> is used to
5981 enable building on the XT3. In addition to enable XT3 specific
5982 features the option --enable-cray-xt3 must be used.
5987 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5988 Details : A race between allocating a new llcd and re-getting the llcd_lock
5989 allowed another thread to grab newly-allocated llcd.
5991 Severity : enhancement
5993 Description: 2.6 OST async journal commit and locking fix to improve performance
5994 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5995 2.6 kernels are now async as they already were in 2.4 kernels so
5996 that they can commit concurrently with the network bulk transfer.
5997 For block-allocated files the filter allocation semaphore is held
5998 to avoid filesystem fragmentation during allocation. BKL lock
5999 removed for 2.6 xattr operations where it is no longer needed.
6004 Description: lconf incorrectly determined whether two IP networks could talk
6005 Details : In some more complicated routing and multiple-network
6006 configurations, lconf will avoid trying to make a network
6007 connection to a disjoint part of the IP space. It was doing the
6008 math incorrectly for one set of cases.
6013 Description: Fix for potential infinite loop processing records in an llog.
6014 Details : If an llog record is corrupted/zeroed, it is possible to loop
6015 forever in llog_process(). Validate the llog record length
6016 and skip the remainder of the block on error.
6019 Frequency : occasional (liblustre only)
6021 Description: liblustre could not open files whose last component is a symlink
6022 Details : sysio_path_walk() would incorrectly pass the open intent to
6023 intermediate path components.
6026 Frequency : rare (liblustre only with non-standard tuning)
6027 Bugzilla : 7201 (7350)
6028 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
6029 Details : llu_lookup_finish_locks() tries to set lock data on a lock
6030 after it has been released, only do this for referenced locks
6032 Severity : enhancement
6034 Description: specifying an (invalid) directory default stripe_size of -1
6035 would reset the directory default striping
6036 Details : stripe_size -1 was used internally to signal directory stripe
6037 removal, now use "all default" to signal dir stripe removal
6038 as a directory striping of "all default" is not useful
6041 Frequency : common for large clusters running liblustre clients
6043 Description: doing an ls when liblustre clients are running is slow
6044 Details : sending a glimpse AST to a liblustre client waits for every AST
6045 to time out, as liblustre clients will not respond. Since they
6046 cannot cache data we refresh the OST lock LVB from disk instead.
6048 Severity : enhancement
6050 Description: doing an ls at the same time as file IO can be slow
6051 Details : enqueue and other "small" requests can be blocked behind many
6052 large IO requests. Create a new OST IO portal for non-IO
6053 requests so they can be processed faster.
6056 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
6058 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
6059 created file with mode 000
6060 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
6061 allow an ftruncate() as a non-root user to a file with mode 000.
6062 Lustre now respects this flag to disable mode checks when
6063 truncating a file owned by the user
6066 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
6069 Description: Revoking locks from clients that went dead or catatonic might take
6071 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
6072 cancellation of such locks instant on servers without waiting for
6073 any reply from clients. Clients drops these locks when cancel
6074 notification from server is received without replying.
6077 Frequency : liblustre-only, when liblustre client dies or becomes busy
6079 Description: Doing ls on Linux clients can take a long time with active
6081 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
6082 granting such locks to it in the first place if possible. Locks
6083 are taken by proxy on the OST during the read or write and
6084 dropped immediately afterward. Add connect flags handling, do
6085 not grant locks to liblustre clients for glimpse ASTs.
6087 Severity : enhancement
6089 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
6090 Details : Existing read-ahead algorithm is tuned for the case of streamlined
6091 sequential reads and behaves badly with applications doing random
6092 reads. Improve it by reading ahead at least read region, and
6093 avoiding excessive large RPC for small reads.
6095 Severity : enhancement
6097 Description: Creating more than 1000 files for a single job may cause a load
6098 imbalance on the OSTs if there are also a large number of OSTs.
6099 Details : qos_prep_create() uses an OST index reseed value that is an
6100 even multiple of the number of available OSTs so that if the
6101 reseed happens in the middle of the object allocation it will
6102 still utilize the OSTs as uniformly as possible.
6107 Description: OST or MDS may oops in ping_evictor_main()
6108 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
6109 but doesn't restart at beginning of obd_exports_timed list
6112 Severity : enhancement
6114 Description: improve by-nid export eviction on the MDS and OST
6115 Details : allow multiple exports with the same NID to be evicted at one
6116 time without re-searching the exports list.
6119 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
6121 Description: MDS may oops in groups_free()
6122 Details : in rare race conditions a newly allocated group_info struct is
6123 freed again, and this can be NULL. The 2.4 compatibility code
6124 for groups_free() checked for a NULL pointer, but 2.6 did not.
6127 Frequency : common for liblustre clients doing little filesystem IO
6128 Bugzilla : 9352, 7313
6129 Description: server may evict liblustre clients accessing contended locks
6130 Details : if a client is granted a lock or receives a completion AST
6131 with a blocking AST already set it would not reply to the AST
6132 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
6135 Frequency : lfs setstripe, only systems with more than 160 OSTs
6137 Description: unable to set striping with a starting offset beyond OST 160
6138 Details : llapi_create_file() incorrectly limited the starting stripe
6139 index to the maximum single-file stripe count.
6142 Frequency : LDAP users only
6144 Description: lconf did not handle in-kernel recovery with LDAP properly
6145 Details : lconf/LustreDB get_refs() is searching the wrong namespace
6147 Severity : enhancement
6149 Description: bind OST threads to NUMA nodes to improve performance
6150 Details : all OST threads are uniformly bound to CPUs on a single NUMA
6151 node and do their allocations there to localize memory access
6153 Severity : enhancement
6155 Description: llmount can determine client NID directly from Myrinet (GM)
6156 Details : the client NID code from gmnalnid was moved directly into
6157 llmount, removing the need to use this or specifying the
6158 client NID explicitly when mounting GM clients with zeroconf
6161 Frequency : if client is started with down MDS
6163 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
6164 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
6165 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
6166 ptlrpc_import_delayed_req() because we didn't check if the
6167 request was interrupted, and we also didn't break out of the
6168 event loop if there was no timeout
6173 Description: data loss during non-page-aligned writes to a single file from
6174 both multiple nodes and multiple threads on one node at same time
6175 Details : updates to KMS and lsm weren't protected by common lock. Resulting
6176 inconsistency led to false short-reads, that were cached and later
6177 used by ->prepare_write() to fill in partially written page,
6178 leading to data loss.
6181 Frequency : always, if lconf --abort_recovery used
6183 Description: lconf --abort_recovery fails with 'Operation not supported'
6184 Details : lconf was attempting to abort recovery on the MDT device and not
6187 Severity : enhancement
6189 Description: remove cleanup logs
6190 Details : replace lconf-generated cleanup logs with lustre internal
6191 cleanup routines. Eliminates the need for client-cleanup and
6194 Severity : enhancement
6196 Description: add support for EAs (user and system) on lustre filesystems
6197 Details : it is now possible to store extended attributes in the Lustre
6198 client filesystem, and with the user_xattr mount option it
6199 is possible to allow users to store EAs on their files also
6201 Severity : enhancement
6203 Description: Add possibility (config option) to show minimal available OST free
6205 Details : When compiled with --enable-mindf configure option, statfs(2)
6206 (and so, df) will return least minimal free space available from
6207 all OSTs as amount of free space on FS, instead of summary of
6208 free spaces of all OSTs.
6210 Severity : enhancement
6212 Description: do not expand extent locks acquired on OST-side
6213 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
6214 by server: they are not cached anyway.
6217 Frequency : when mmap is used/binaries executed from Lustre
6219 Description: Unmmap pages before throwing them away from read cache.
6220 Details : llap_shrink cache now attempts to unmap pages before discarding
6221 them (if unmapping failed - do not discard). SLES9 kernel has
6222 extra checks that trigger if this unmapping is not done first.
6227 Description: lconf didn't resolve symlinks before checking to see whether a
6228 given mountpoint was already in use
6231 Frequency : when migrating failover services
6232 Bugzilla : 6395, 9514
6233 Description: When migrating a subset of services from a node (e.g. failback
6234 from a failover service node) the remaining services would
6235 time out and evict clients.
6236 Details : lconf --force (implied by --failover) sets the global obd_timeout
6237 to 5 seconds in order to quickly disconnect, but this caused
6238 other RPCs to time out too quickly. Do not change the global
6239 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
6241 Severity : enhancement
6242 Frequency : if MDS is started with down OST
6243 Bugzilla : 9439,5706
6244 Description: Allow startup/shutdown of an MDS without depending on the
6245 availability of the OSTs.
6246 Details : Asynchronously call mds_lov_synchronize during MDS startup.
6247 Add appropriate locking and lov-osc refcounts for safe
6248 cleaning. Add osc abort_inflight calls in case the
6249 synchronize never started.
6252 Frequency : occasional (Cray XT3 only)
6254 Description: root not authorized to access files in CRAY_PORTALS environment
6255 Details : The client process capabilities were not honoured on the MDS in
6256 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
6257 been accessed by an authorized user then root was able to access
6258 the file on the local client also. The root user capabilities
6259 are now allowed on the MDS, as this environment has secure UID.
6262 Frequency : occasional
6264 Description: ldiskfs "too long searching" message happens too often
6265 Details : A debugging message (otherwise harmless) prints too often on
6266 the OST console. This has been reduced to only happen when
6267 there are fragmentation problems on the filesystem.
6272 Description: Division by zero in statfs when all OSCs are inactive
6273 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
6274 lov_statfs divides by value returned by lov_get_stripecnt().
6278 Bugzilla : 9489, 3273
6279 Description: First write from each client to each OST was only 4kB in size,
6280 to initialize client writeback cache, which caused sub-optimal
6281 RPCs and poor layout on disk for the first writen file.
6282 Details : Clients now request an initial cache grant at (re)connect time
6283 and so that they can start streaming writes to the cache right
6284 away and always do full-sized RPCs if there is enough data.
6285 If the OST is rebooted the client also re-establishes its grant
6286 so that client cached writes will be honoured under the grant.
6291 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
6292 Details : Now I/O RPCs go to different portal number and (presumably) fast
6293 lock requests (and glimses) and other RPCs get their own service
6294 threads pool that should be able to service those RPCs
6297 Severity : enhancement
6299 Description: Ability to exchange lustre version between client and servers and
6300 issue warnings at client side if client is too old. Also for
6301 liblustre clients there is ability to refuse connection of too old
6303 Details : New 'version' field is added to connect data structure that is
6304 filled with version info. That info is later checked by server and
6308 Frequency : rare, liblustre only.
6309 Bugzilla : 9296, 9581
6310 Description: Two simultaneous writes from liblustre at offset within same page
6311 might proceed at the same time overwriting eachother with stale
6313 Details : I/O lock withing llu_file_prwv was released too early, before data
6314 actually was hitting the wire. Extended lock-holding time until
6315 server acknowledges receiving data.
6318 Frequency : extremely rare. Never observed in practice.
6320 Description: avoid generating lustre_handle cookie of 0.
6321 Details : class_handle_hash() generates handle cookies by incrementing
6322 global counter, and can hit 0 occasionaly (this is unlikely, but
6323 not impossible, because initial value of cookie counter is
6324 selected randonly). Value of 0 is used as a sentinel meaning
6325 "unassigned handle" --- avoid it. Also coalesce two critical
6326 sections in this function into one.
6328 Severity : enhancement
6330 Description: allow liblustre clients to delegate truncate locking to OST
6331 Details : To avoid overhead of locking, liblustre client instructs OST to
6332 take extent lock in ost_punch() on client's behalf. New connection
6333 flag is added to handle backward compatibility.
6335 Severity : enhancement
6336 Bugzilla : 4928, 7341, 9758
6337 Description: allow number of OST service threads to be specified
6338 Details : a module parameter allows the number of OST service threads
6339 to be specified via "options ost ost_num_threads={N}" in the
6340 OSS's /etc/modules.conf or /etc/modprobe.conf.
6344 Bugzilla : 6146, 9635, 9895
6345 Description: servers crash with bad pointer in target_handle_connect()
6346 Details : In rare cases when a client is reconnecting it was possible that
6347 the connection request was the last reference for that export.
6348 We would temporarily drop the export reference and get a new
6349 one, but this may have been the last reference and the export
6350 was just destroyed. Get new reference before dropping old one.
6352 Severity : enhancement
6353 Frequency : if client is started with failover MDS
6355 Description: Allow multiple MDS hostnames in the mount command
6356 Details : Try to read the configuration from all specified MDS
6357 hostnames during a client mount in case the "primary"
6360 Severity : enhancement
6362 Description: Stop sending data to evicted clients as soon as possible.
6363 Details : Check if the client we are about to send or are sending data to
6364 was evicted already. (Check is done every second of waiting,
6365 for which l_wait_event interface was extended to allow checking
6366 of exit condition at specified intervals).
6369 Frequency : rare, normally only when NFS exporting is done from client
6371 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6373 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6376 Frequency : always, if srand() is called before liblustre initialization
6378 Description: Liblustre uses system PRNG disturbing its usage by user application
6379 Details : Introduce internal to lustre fast and high-quality PRNG for
6380 lustre usage and make liblustre and some other places in generic
6381 lustre code to use it.
6383 Severity : enhancement
6384 Bugzilla : 9477, 9557, 9870
6385 Description: Verify that the MDS configuration logs are updated when xml is
6386 Details : Check if the .xml configuration logs are newer than the config
6387 logs stored on the MDS and report an error if this is the case.
6388 Request --write-conf, or allow starting with --old_conf.
6390 Severity : enhancement
6392 Description: Handle symlinks in the path when checking if Lustre is mounted.
6393 Details : Resolve intermediate symlinks when checking if a client has
6394 mounted a filesystem to avoid duplicate client mounts.
6399 Description: lconf can hit an error exception but still return success.
6400 Details : The lconf command catches the Command error exception at the top
6401 level script context and will exit with the associated exit
6402 status, but doesn't ensure that this exit status is non-zero.
6407 Description: failure of ptlrpc thread startup can cause oops
6408 Details : Starting a ptlrpc service thread can fail if there are a large
6409 number of threads or the server memory is very fragmented.
6410 Handle this without oopsing.
6413 Frequency : always, only if liblustre and non-default acceptor port was used
6415 Description: liblustre cannot connect to servers with non-default acceptor port
6416 Details : tcpnal_set_default_params() was not called and was therefore
6417 ignoring the environment varaible TCPNAL_PORT, as well as other
6418 TCPNAL_ environment variables
6423 Description: two objects could be created on the same OST for a single file
6424 Details : If an OST is down, in some cases it was possible to create two
6425 objects on a single OST for a single file. No problems other
6426 than potential performance impact and spurious error messages.
6430 Bugzilla : 5681, 9562
6431 Description: Client may oops in ll_unhash_aliases
6432 Details : Client dcache may become inconsistent in race condition.
6433 In some cases "getcwd" can fail if the current directory is
6439 Description: Inode refcounting problems in NFS export code
6440 Details : link_raw functions used to call d_instantiate without obtaining
6441 extra inode reference first.
6445 Bugzilla : 9942, 9903
6446 Description: Referencing freed requests leading to crash, memleaks with NFS.
6447 Details : We used to require that call to ll_revalidate_it was always
6448 followed by ll_lookup_it. Also with revalidate_special() it is
6449 possible to call ll_revalidate_it() twice for the same dentry
6450 even if first occurence returned success. This fix changes semantic
6451 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6452 reference on a request referred from the intent.
6453 ll_intent_release() then releases such a request.
6456 Frequency : rare, normally benchmark loads only
6458 Description: unlinked inodes were kept in memory on the client
6459 Details : If a client is repeatedly creating and unlinking files it
6460 can accumulate a lot of stale inodes in the inode slab cache.
6461 If there is no other client load running this can cause the
6462 client node to run out of memory. Instead flush old inodes
6463 from client cache that have the same inode number as a new inode.
6466 Frequency : SLES9 2.6.5 kernel and long filenames only
6467 Bugzilla : 9969, 10379
6468 Description: utime reports stale NFS file handle
6469 Details : SLES9 uses out-of-dentry names in some cases, which confused
6470 the lustre dentry revalidation. Change it to always use the
6474 Frequency : rare, unless heavy write-truncate concurrency is continuous
6475 Bugzilla : 4180, 6984, 7171, 9963, 9331
6476 Description: OST becomes very slow and/or deadlocked during object unlink
6477 Details : filter_destroy() was holding onto the parent directory lock
6478 while truncating+unlinking objects. For very large objects this
6479 may block other threads for a long time and slow overall OST
6480 responsiveness. It may also be possible to get a lock ordering
6481 deadlock in this case, or run out of journal credits because of
6482 the combined truncate+unlink. Solution is to do object truncate
6483 first in one transaction without parent lock, and then do the
6484 final unlink in a new transaction with the parent lock. This
6485 reduces the lock hold time dramatically.
6488 Frequency : rare, 2.4 kernels only
6490 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6491 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6492 before unmounting their filesystem in order to avoid deadlock.
6493 The kernel_locked() function in 2.4 kernels only checks whether
6494 the kernel lock is held, not whether it is this process that is
6495 holding it as 2.6 kernels do.
6500 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6501 Details : The client ptlrpc code may be trying to reconnect to a down
6502 server before a previous connection attempt has timed out.
6503 Increase the reconnect interval to be longer than the connection
6504 timeout interval to avoid sending duplicate connections to
6508 Frequency : echo_client brw_test command
6510 Description: fix echo_client to work with OST preallocated code
6511 Details : OST preallocation code (5137) didn't take echo_client IO path
6512 into account: echo_client calls filter methods outside of any
6513 OST thread and, hence, there is no per-thread preallocated
6514 pages and buffers to use. Solution: hijack pga pages for IO. As
6515 a byproduct, this avoids unnecessary data copying.
6519 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6520 Description: Client can oops in mdc_commit_close() after open replay
6521 Details : It was possible for the MDS to return an open request with no
6522 transaction number in mds_finish_transno() if the client was
6523 evicted, but without actually returning an error. Clients
6524 would later try to replay that open and may trip an assertion
6525 Simplify the client close codepath, and always return an error
6526 from the MDS in case the open is not successful.
6529 Frequency : rare, 2.6 OSTs only
6531 Description: OST may deadlock under high load on fragmented files
6532 Details : If there was a heavy load and highly-fragmented OST filesystems
6533 it was possible to have all the OST threads deadlock waiting on
6534 allocation of biovecs, because the biovecs were not released
6535 until the entire RPC IO was completed. Instead, release biovecs
6536 as soon as they are complete to ensure forward IO progress.
6538 Severity : enhancement
6540 Description: Support for specifying external journal device at mount
6541 Details : If an OST or MDS device is formatted with an external journal
6542 device, this device major/minor is stored in the ext3 superblock
6543 and may not be valid for failover. Allow detecting and
6544 specifying the external journal at mount time.
6549 Description: Mounting an MDS with pending unlinked files may cause oops
6550 Details : target_finish_recovery() calls mds_postrecov() which returned
6551 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6552 considers this an error and immediately begins cleaning up the
6553 lov, just after starting the mds_lov process
6555 Severity : enhancement
6557 Description: Implement 'lfs df' to report actual free space on per-OST basis
6558 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6559 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6560 usage of objects; '-h' to report in human readable format.
6562 ------------------------------------------------------------------------------
6564 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6571 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6572 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6573 at startup during orphan unlinking. Instead, skip buddy update
6574 before setup, as it will be regenerated after recovery is complete.
6579 Description: If an OST is inactive, its locks might reference stale inodes.
6580 Details : lov_change_cbdata() must iterate over all namespaces, even if
6581 they are inactive to clear inode references from the lock.
6583 Severity : enhancement
6584 Frequency : occasional, if non-standard max_dirty_mb used
6586 Description: Client will block write RPCs if not enough grant
6587 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6588 then the client will block writes while waiting for another RPC
6589 to complete instead of consuming its dirty limit. With change
6590 we get improved performance when max_dirty_mb is small.
6592 Severity : enhancement
6593 Bugzilla : 3389, 6253
6594 Description: Add support for supplementary groups on the MDS.
6595 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6596 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6597 lookups for user supplementary groups into a cache.
6601 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6602 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6603 for liblustre O_EXCL is enforced by MDS.
6607 Description: readdir never returns NULL in liblustre.
6608 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6609 dentry in struct dirent.
6613 Frequency : liblustre only, depends on application IO pattern
6614 Description: liblustre clients evicted if not contacting servers
6615 Details : Don't put liblustre clients into the ping_evictor list, so
6616 they will not be evicted by the pinger ever.
6618 Severity : enhancement
6620 Description: Add ability to evict clients by NID from MDS.
6621 Details : By echoing "nid:$NID" string into
6622 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6623 $NID would be instantly evicted from this MDS and from all active
6624 OSTs connected to it.
6628 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6629 Details : lookup_it_finish() used to query file size from OSTs that was not
6634 Description: service threads change working directory to that of init
6635 Details : Starting lustre service threads may pin the working directory
6636 of the parent thread, making that filesystem busy. Threads
6637 now change to the working directory of init to avoid this.
6641 Frequency : during shutdown only
6642 Description: shutdown with a failed MDS or OST can cause unmount to hang
6643 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6648 Frequency : 2.6 only
6649 Description: chmod/chown may include an extra supplementary group
6650 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6651 supplementary group and if none is specified this is used.
6654 Bugzilla : 5479 (6816)
6656 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6657 Details : If lookup is in progress on a file that is unlinked we might try
6658 to revalidate the inode and fail in revalidate after lookup is
6659 complete and ll_file_open() enqueues the open again but
6660 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6663 Frequency : always, if lconf --abort_recovery used
6665 Description: lconf --abort_recovery fails with 'Operation not supported'
6666 Details : lconf was attempting to abort recovery on the MDT device and not
6669 ------------------------------------------------------------------------------
6671 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6676 Frequency : rare (only unsupported configurations with a node running as an
6678 Bugzilla : 6514, 5137
6679 Description: Mounting a Lustre file system on a node running as an OST could
6681 Details : OSTs now preallocates memory needed to write out data at
6682 startup, instead of when needed, to avoid having to
6683 allocate memory in possibly low memory situations.
6684 Specifically, if the file system is mounted on on OST,
6685 memory pressure could force it to try to write out data,
6686 which it needed to allocate memory to do. Due to the low
6687 memory, it would be unable to do so and the node would
6688 become unresponsive.
6690 Severity : enhancement
6692 Description: Addition of lconf --service command line option
6693 Details : lconf now accepts a '--service <arg>' option, which is
6694 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6696 Severity : enhancement
6698 Description: Failover mode is now the default for OSTs.
6699 Details : By default, OSTs will now run in failover mode. To return to
6700 the old behaviour, add '--failout' to the lmc line for OSTs.
6702 Severity : enhancement
6704 Description: Health checks are now provided for MDS and OSTs
6705 Details : Additional detailed health check information on MSD and OSTs
6706 is now provided through the procfs health_check value.
6709 Frequency : occasional, depends on IO load
6711 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6712 after numerous create/delete cycles
6713 Details : The ext3 inode allocation policy would not allocate new inodes
6714 very well on the OSTs because there are no new directories
6715 being created. Instead we look for groups with free space if
6716 the parent directories are nearly full.
6721 Description: Network or server problems during mount may cause partially
6722 mounted clients instead of returning an error.
6723 Details : The config llog parsing code may overwrite the error return
6724 code during mount error handling, returning success instead
6730 Description: MDS can fail to allocate large reply buffers
6731 Details : After long uptimes the MDS can fail to allocate large reply
6732 buffers (e.g. zconf client mount config records) due to memory
6733 fragmentation or consumption by the buffer cache. Preallocate
6734 some large reply buffers so that these replies can be sent even
6735 under memory pressure.
6739 Frequency : rare (liblustre)
6740 Description: fsx running with liblustre complained that using truncate() to
6741 extend the file doesn't work. This patch corrects that issue.
6742 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6743 ATTR_SIZE and lsm use in llu_setattr_raw.
6747 Frequency : rare, only 2.6 kernels
6748 Description: Unusual file access patterns on the MDS may result in inode
6749 data being lost in very rare circumstances.
6750 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6751 code in the 2.6 kernel caused the inode and/or EA data not to
6752 be read from disk, causing single-file corruption.
6756 Frequency : rare, only 2.6 filesystems using extents
6757 Description: Heavy concurrent write and delete load may cause data corruption.
6758 Details : It was possible under high-load situations to have an extent
6759 metadata block in the block device cache from a just-unlinked
6760 file overwrite a newly-allocated data block. We now unmap any
6761 metadata buffers that alias just-allocated data blocks.
6765 Frequency : filesystems with default stripe_count larger than 77
6766 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6767 Details : lconf specifies an inode size of 4096 bytes when the default
6768 stripe_count is larger than 77. This conflicts with the default
6769 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6770 this case to avoid pinning too much memory for large EAs.
6772 ------------------------------------------------------------------------------
6774 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6779 Frequency : rare (extremely heavy IO load with hundreds of clients)
6781 Description: Client is evicted, gets IO error writing to file
6782 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6783 caused clients to be evicted for AST timeouts. The fixes in
6784 bug 5192 mean we no longer need to have such short AST timeouts
6785 so ldlm_timeout has been increased.
6788 Frequency : occasional during --force or --failover shutdown under load
6789 Bugzilla : 5949, 4834
6790 Description: Server oops/LBUG if stopped with --force or --failover under load
6791 Details : a collection of import/export refcount and cleanup ordering
6792 issues fixed for safer force cleanup
6795 Frequency : only filesystems larger than 120 OSTs
6796 Bugzilla : 5990, 6223
6797 Description: lfs getstripe would oops on a very large filesystem
6798 Details : lov_getconfig used kfree on vmalloc'd memory
6801 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6802 Bugzilla : 6242, 6243
6803 Description: reading from files that had been truncated to a non-zero size
6804 but never opened returned no data
6805 Details : ll_file_read() reads zeros from no-object files to EOF
6810 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6811 Details : There's little harm in aborting MDS/OSS recovery and letting it
6812 try again, so I removed the LASSERT and return an error instead.
6814 Severity : enhancement
6816 Description: New debugging infrastructure for tracking down data corruption
6817 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6818 (b) cover more of the client-side code path, and (c) try to narrow
6819 down where problems occurred
6823 Bugzilla : 3819, 4364, 4397, 6313
6824 Description: Racing close and eviction MDS could cause assertion in mds_close
6825 Details : It was possible to get multiple mfd references during close and
6826 client eviction, leading to one thread referencing a freed mfd.
6828 Severity: : enhancement
6829 Bugzilla : 3262, 6359
6830 Description: Attempts to reconnect to servers are now more aggressive.
6831 Details : This builds on the enhanced upcall-less recovery that was added
6832 in 1.4.2. When trying to reconnect to servers, clients will
6833 now try each server in the failover group every 10 seconds. By
6834 default, clients would previously try one server every 25 seconds.
6839 Description: After recovery, certain operations trigger a failed
6840 assertion on a client.
6841 Details : Failing over an mds, using lconf -d --failover, while a
6842 client was doing a readdir() call would cause the client to
6843 LBUG after recovery completed and the readdir() was resent.
6845 Severity : enhancement
6847 Description: Default groups are now added by lconf
6848 Details : You can now run lconf --group <servicename> without having to
6849 manually add groups with lmc.
6852 Frequency : occasional
6854 Description: Nodes with an elan id of 0 trigger a failed assertion
6857 Frequency : always when accessing e.g. tty/console device nodes
6859 Description: tty and some other devices nodes cannot be used on lustre
6860 Details : file's private_data field is used by device data and lustre
6861 values in there got lost. New field was added to struct file to
6862 store fs-specific private data.
6865 Frequency : when exporting Lustre via NFS
6867 Description: NFSD failed occasionally when looking up a path component
6868 Details : NFSD is looking up ".." which was broken in ext3 directories
6869 that had grown large enough to become hashed.
6872 Frequency : Clusters with multiple interfaces not on the same subnet
6874 Description: Nodes will repeatedly try to reconnect to an interface which it
6875 cannot reach and report an error to the log.
6876 Details : Extra peer list entries will be created by lconf with some peers
6877 unreachable. lconf now validates the peer before adding it.
6880 Frequency : Only if a default stripe is set on the filesystem root.
6882 Description: Setting a default stripe on the filesystem root prevented the
6883 filesystem from being remounted.
6884 Details : The client was sending extra request flags in the root getattr
6885 request and did not allocate a reply buffer for the dir EA.
6888 Frequency : occasional, higher if lots of files are accessed by one client
6889 Bugzilla : 6159, 6097
6890 Description: Client trips assertion regarding lsm mismatch/magic
6891 Details : While revalidating inodes the VFS looks up inodes with ifind()
6892 and in rare cases can find an inode that is being freed.
6893 The ll_test_inode() code will free the lsm during ifind()
6894 when it finds an existing inode and then the VFS later attaches
6895 this free lsm to a new inode.
6899 Bugzilla : 6422, 7030
6900 Description: MDS deadlock between mkdir and client eviction
6901 Details : Creating a new file via mkdir or mknod (starting a transaction
6902 and getting the ns lock) can deadlock with client eviction
6903 (gets ns lock and trying to finish a synchronous transaction).
6906 Frequency : occasional
6907 Description: While starting a server, the fsfilt_ext3 module could not be
6909 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6910 kernels. Previously, lconf would still use the ext3 name
6911 when trying to load modules. Now, it will correctly use
6912 ext3 on 2.4 and ldiskfs on 2.6.
6914 Severity : enhancement
6915 Description: The default stripe count has been changed to 1
6916 Details : The interpretation of the default stripe count (0, to lfs
6917 or lmc) has been changed to mean striping across a single
6918 OST, rather than all available. For general usage we have
6919 found a stripe count of 1 or 2 works best.
6921 Severity : enhancement
6922 Description: Add support for compiling against Cray portals.
6923 Details : Conditional compiling for some areas that are different
6927 Frequency : occasional
6928 Bugzilla : 6409, 6834
6929 Description: Creating files with an explicit stripe count may lead to
6930 a failed assertion on the MDS
6931 Details : If some OSTs are full or unavailable, creating files may
6932 trigger a failed assertion on the MDS. Now, Lustre will
6933 try to use other servers or return an error to the
6937 Frequency : occasional
6939 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6940 caused lock timeout during readahead (since 1.4.2).
6941 Details : Processes doing readahead might match a lock that hasn't been
6942 granted yet if there are overlapping and conflicting lock
6943 requests. The readahead process waits on ungranted lock
6944 (original lock is CBPENDING), while OST waits for that process
6945 to cancel CBPENDING read lock and eventually evicts client.
6947 Severity : enhancement
6949 Description: Initial enabling of flock support for clients
6950 Details : Implements fcntl advisory locking and file status functions.
6951 This feature is provided as an optional mount flag (default
6952 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6953 locking are implemented yet, and those that are are not guaranteed
6954 to be completely correct in production environments.
6955 mount -t lustre -o [flock|noflock] ...
6958 Frequency : occasional
6960 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6961 assertion in the ext3 JBD (journaling) layer.
6962 Details : The b_committed_data struct is protected by the big kernel lock
6963 in 2.4 kernels, serializing journal_commit_transaction() and
6964 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6965 access to this struct. In 2.6 kernels there is finer grained
6966 locking to improve SMP performance of the JBD layer.
6970 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6972 -----------------------------------------------------------------------------
6974 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6976 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6977 versions of Lustre. Please update all servers and clients to
6978 version 1.4.2 or later at the same time. You must also run
6979 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6980 to update the configuration logs.
6982 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6983 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6984 - swab llog records of type '0' so we get proper header size/idx (5861)
6985 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6986 - fix rename of one directory over another leaking an inode (5953)
6987 - avoid SetPageDirty on 2.6 (5981)
6988 - don't re-add just-being-destroyed locks to the waiting list (5653)
6989 - when creating new directories, inherit the parent's custom
6990 striping settings if present parent (3048)
6991 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6992 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6993 - don't hold client locks on temporary worklist from l_lru (5666)
6994 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6995 - automatically evict dead clients (5921)
6996 - Update file size properly in create+truncate+fstat case (6196)
6997 - Do not unhash mountpoint dentries, do not allow removal of
6999 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
7000 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
7001 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
7002 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
7003 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
7004 - Correctly handle reads of files with no objects (6243)
7005 - lctl recover will also mark a device active if deactivate used (5933)
7007 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
7008 - allow --write-conf on an MDS with different nettype than client (5619)
7009 - don't write config llogs to MDS for mounts not from that MDS (5617)
7010 - lconf should create multiple TCP connections from a client (5201)
7011 - init scripts are now turned off by default; run chkconfig --on
7012 lustre and chkconfig --on lustrefs to use them
7013 - upcalls are no longer needed for clients to recover to failover
7015 - add --abort-recovery option to lconf to abort recovery on device
7017 - add support for an arbitrary number of OSTs (3026)
7018 - Quota support protocol changes.
7019 - forward compatibility changes to wire structs (6007)
7020 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
7021 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
7022 - improved "lustre status" script
7023 - initialize blocksize for non-regular files (6062)
7024 - added --disable-server and --disable-client configure options (5782)
7025 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
7026 - Vanilla 2.4.29 support
7027 - increase maximum number of obd devices to 520 (6242)
7028 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
7029 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
7030 SLES 9 2.6.5 kernels
7031 - limit stripes per file to 160 (the maximum EA size) (6093)
7033 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7036 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
7037 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
7038 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
7039 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
7040 - small ext3 extents cleanups and fixes (5733)
7041 - improved mballoc code, several small races and bugs fixed (5733, 5638)
7042 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
7043 - avoid needless client->OST connect, fix handle mismatch (5317)
7044 - fix DLM error path that led to out-of-sync client, long delays (5779)
7045 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
7046 - fix several locking issues related to i_size (5492,5624,5654,5672)
7047 - don't move pending lock onto export if it is already evicted (5683)
7048 - fix kernel oops when creating .foo in unlinked directory (5548)
7049 - fix deadlock in obdfilter statistics vs. object create (5811)
7050 - use time_{before,after} to avoid timer jiffies wrap (5882)
7051 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
7052 - Do not leak request if server was not able to process it (5154)
7053 - If mds_open unable to find parent dir, make that negative lookup(5154)
7054 - don't create new directories with extent-mapping (5909, 5936)
7056 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
7057 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
7058 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
7059 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
7060 - support for RHEL4 kernel on i686 (5773)
7061 - provide error messages when incompatible logs are encountered (5898)
7063 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
7064 * version 1.4.0.10 (1.4.1 release candidate 1)
7066 - don't keep a lock reference when lock is not granted (4238)
7067 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7068 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
7069 - drop import inflight refcount on signal_completed_replay error (5255)
7070 - unlock page after async write error during send (3677)
7071 - handle missing objects in filter_preprw_read properly (5265)
7072 - no transno return for symlink open, don't save no-trasno open (3440)
7073 - don't try to complete elan receive that already failed (4012)
7074 - free RPC server reply state on error (5406)
7075 - clean up thread from ptlrpc_start_thread() on error (5160)
7076 - readahead could read extra page into cache that wasn't ejected (5388)
7077 - prevent races in class_attach/setup/cleanup/detach (5260)
7078 - don't dereference de->d_inode after l_dput of de (5458)
7079 - use "int" for stripe value returned from lock_to_stripe (5544)
7080 - mballoc allocation and error-checking fixes in 2.6 (5504)
7081 - block device patches to fix I/O request sizes in 2.6 (5482)
7082 - look up hostnames for IB nals (5602)
7083 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
7084 - don't start multiple acceptors for the same port (5277)
7085 - fix incorrect LASSERT in mds_getattr_name (5635)
7086 - export a proc file for general "ping" checking (5628)
7087 - fix "lfs check" to not block when the MDS is down (5628)
7089 - service request history (4965)
7090 - put {ll,lov,osc}_async_page structs in a single slab (4699)
7091 - create an "evict_client" /proc entry on OSTs, like the MDS has
7092 - fix mount usage message, return errors per mount(8) (5168)
7093 - change grep [] to grep "[]" in tests so they work in more UMLs
7094 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
7095 - remove ancient LOV_MAGIC_V0 EA support (5047)
7096 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
7097 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
7098 - allow readahead allocations to fail when low on memory (5383)
7099 - mmap locking landed again, after considerable improvement (2828)
7100 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
7102 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
7105 - send OST transaction number in read/write reply to free req (4966)
7106 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7107 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7108 - the watchdog thread now runs as interruptible (5246)
7109 - flock/lockf fixes (but it's still disabled, pending 5135)
7110 - don't use EXT3 constants in llite code (5094)
7111 - memory shortage at startup could cause assertion (5176)
7113 - reorganization of lov code
7114 - single portals codebase
7116 - add extents/mballoc support (5025)
7117 - direct I/O reads in the obdfilter (4048)
7118 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
7120 tbd Cluster File Systems, Inc. <info@clusterfs.com>
7123 - send OST transaction number in read/write reply to free req (4966)
7124 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7125 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7126 - the watchdog thread now runs as interruptible (5246)
7127 - handle missing objects in filter_preprw_read properly (5265)
7128 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7129 - drop import inflight refcount on signal_completed_replay error (5255)
7130 - unlock page after async write error during send (3677)
7131 - return original error code on reconstructed replies (3761)
7132 - no transno return for symlink open, don't save no-trasno open (3440)
7134 - add pid to ldlm debugging output (4922)
7135 - bump the watchdog timeouts -- we can't handle 30sec yet
7136 - extra debugging for orphan dentry/inode bug (5259)
7138 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
7141 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
7142 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
7143 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
7144 - fix NULL dereference in /proc/sys/portals/routes (4827)
7145 - allow failed mdc_close() operations to be interrupted (4561)
7146 - stop precreate on OST before MDS would time out on it (4778)
7147 - don't send partial-page writes before EOF from client (4410)
7148 - discard client grant for sub-page writes on large-page clients (4520)
7149 - don't free dentries not owned by NFS code, check generation (4806)
7150 - fix lsm leak if mds_create_objects() fails (4801)
7151 - limit debug_daemon file size, always print CERROR messages (4789)
7152 - use transno after validating reply (3892)
7153 - process timed out requests if import state changes (3754)
7154 - update mtime on OST during writes, return in glimpse (4829)
7155 - add mkfsoptions to LDAP (4679)
7156 - use ->max_readahead method instead of zapping global ra (5039)
7157 - don't interrupt __l_wait_event() during strace
7159 - add software watchdogs to catch hung threads quickly (4941)
7160 - make lustrefs init script start after nfs is mounted
7161 - fix CWARN/ERROR duplication (4930)
7162 - return async write errors to application if possible (2248)
7163 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
7164 - print NAL number in %x format (4645)
7165 - update barely-supported suse-2.4.21-171 series (4842)
7166 - support for sles 9 %post scripts
7167 - support for building 2.6 kernel-source packages
7168 - support for sles km_* packages
7170 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
7173 - ignore -ENOENT errors in osc_destroy (3639)
7174 - notify osc create thread that OSC is being cleaned up (4600)
7175 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
7176 - reconstruct ost_handle() like mds_handle() (4657)
7177 - create a new thread to do import eviction to avoid deadlock (3969)
7178 - let lconf resolve symlinked-to devices (4629)
7179 - don't unlink "objects" from directory with default EA (4554)
7180 - hold socknal file ref over connect in case target is down (4394)
7181 - allow more than 32000 subdirectories in a single directory (3244)
7182 - fix blocks count for O_DIRECT writes (3751)
7183 - OST returns ENOSPC from object create when no space left (4539)
7184 - don't send truncate RPC if file size isn't changing (4410)
7185 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
7186 - bind to privileged port in socknal and tcpnal (3689)
7188 - rate limit CERROR/CWARN console message to avoid overload (4519)
7189 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
7190 - basic mmap support (3918)
7191 - kernel patch series update from b1_4 (4711)
7193 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
7196 - avoid crash during MDS cleanup with OST shut down (2775)
7197 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
7198 - don't use bad inodes on the MDS (3744)
7199 - dynamic object preallocation to improve recovery speed (4236)
7200 - don't hold spinlock over lock dumping or change debug flags (4401)
7201 - don't zero obd_dev when it is force cleaned (3651)
7202 - print grants to console if they go negative (4431)
7203 - "lctl deactivate" will stop automatic recovery attempts (3406)
7204 - look for existing locks in ldlm_handle_enqueue() (3764)
7205 - don't resolve lock handle twice in recovery avoiding race (4401)
7206 - revalidate should check working dir is a directory (4134)
7208 - don't always mark "slow" obdfilter messages as errors (4418)
7210 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
7213 - don't close LustreDB during write_conf until it is done (3860)
7214 - fix typo in lconf for_each_profile (3821)
7215 - allow dumping logs from multiple threads at one time (3820)
7216 - don't allow multiple threads in OSC recovery (3812)
7217 - fix debug_size parameters (3864)
7218 - fix mds_postrecov to initialize import for llog ctxt (3121)
7219 - replace config semaphore with spinlock (3306)
7220 - be sure to send a reply for a CANCEL rpc with bad export (3863)
7221 - don't allow enqueue to complete on a destroyed export (3822)
7222 - down write_lock before checking llog header bitmap (3825)
7223 - recover from lock replay timeout (3764)
7224 - up llog sem before sending rpc (3652)
7225 - reduce ns lock hold times when setting kms (3267)
7226 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
7227 - fix NULL deref and obd_dev leak on setup error (3312)
7228 - replace some LBUG about llog ops with error handling (3841)
7229 - don't match INVALID dentries from d_lookup and spin (3784)
7230 - hold dcache_lock while marking dentries INVALID and hashing (4255)
7231 - fix invalid assertion in ptlrpc_set_wait (3880)
7233 - add libwrap support for the TCP acceptor (3996)
7234 - add /proc/sys/portals/routes for non-root route listing (3994)
7235 - allow setting MDS UUID in .xml (2580)
7236 - print the stack of a process that LBUGs (4228)
7238 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
7241 - don't cleanup request in ll_file_open() on failed MDS open (3430)
7242 - make sure to unset replay flag from failed open requests (3440)
7243 - if default stripe count is 0, use OST count for inode size (3636)
7244 - update parent mtime/ctime on client for create/unlink (2611)
7245 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
7246 - free recovery state on server during a forced cleanup (3571)
7247 - unregister_reply for resent reqs (3063)
7248 - loop back devices mounting and status check on 2.6 (3563)
7249 - fix resource-creation race that can provoke i_size == 0 (3513)
7250 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
7251 - more debugging for page-accounting assertion (3746)
7252 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
7253 - avoid deadlock after precreation failure (3758)
7254 - fix race and lock order deadlock in orphan handling (3450, 3750)
7255 - add validity checks when grabbing inodes from l_ast_data (3599)
7257 - add /proc/.../recovery_status to obdfilter (3428)
7258 - lightweight CDEBUG infrastructure, debug daemon (3668)
7259 - change default OSC RPC parameters to be better on small clusters
7260 - turn off OST read cache for files smaller than 32MB
7261 - install man pages and include them in rpms (3100)
7262 - add new init script for (un)mounting lustre filesystems (2593)
7263 - run chkconfig in %post for init scripts (3701)
7264 - drop scimac NAL (unmaintained)
7266 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
7269 - clean kiobufs before and after use (3485)
7270 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
7271 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7272 - change init script to fail more clearly if not run as root (1528)
7273 - allow clients to reconnect during replay (1742)
7274 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
7275 - don't do DNS lookups on NIDs too small for IP addresses (3442)
7276 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
7277 - fix cond_resched (3554)
7278 - only evict unfinished clients after recovery (3515)
7279 - allow bulk resend, prevent data loss (3570)
7280 - dynamic ptlrpc request buffer allocation (2102)
7281 - don't allow unlinking open directory if it isn't empty (2904)
7282 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
7283 - remove extraneous obd dereference causing LASSERT failure (3334)
7284 - don't use get_cycles() when creating temp. files on the mds (3156)
7285 - hold i_sem when setting i_size in ll_extent_lock() (3564)
7286 - handle EEXIST for set-stripe, set proper directory name (3336)
7288 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
7289 - fix ksocknal_fmb_callback() error messages (2918)
7291 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
7294 - don't copy lvb into (possibly NULL) reply on error (2983)
7295 - don't deref dentry after dput, don't free lvb on error (2922)
7296 - use the kms to determine writeback rpc length (2947)
7297 - increment oti_logcookies when osc is inactive (2948)
7298 - update client's i_blocks count via lvb messages (2543)
7299 - handle intent open/close of special files properly (1557)
7300 - mount MDS with errors=remount-ro, like obdfilter (2009)
7301 - initialize lock handle to avoid ASSERT on error cleanup (3057)
7302 - don't use cancelling-locks' kms values (2947)
7303 - use highest lock extent for kms, not last one (2925)
7304 - don't dereference ERR_PTR() dentry in error handling path (3107)
7305 - fix thread race in portals_debug_dumplog() (3122)
7306 - create lprocfs device entries at setup instead of at attach (1519)
7307 - common AST error handler, don't evict client on completion race (3145)
7308 - zero nameidata in detach_mnt in 2.6 (3118)
7309 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7310 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7311 - initialize RPC timeout timer earlier for 2.6 (3219)
7312 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7313 - print nal/nid for unknown nid (3258)
7314 - additional checks for oscc recovery before doing precreate (3284)
7315 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7316 - don't crash in mdc_close for bad permissions on open (3285)
7317 - zero i_rdev for non-device files (3147)
7318 - clear page->private before handing to FS, better assertion (3119)
7319 - tune the read pipeline (3236)
7320 - fix incorrect decref of invalidated dentry (2350)
7321 - provide read-ahead stats and refine rpc in flight stats (3328)
7322 - don't hold journal transaction open across create RPC (3313)
7323 - update atime on MDS at close time (3265)
7324 - close LDAP connection when recovering to avoid server load (3315)
7325 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7326 - don't leak open file on MDS after open resend (3325)
7327 - serialize filter_precreate and filter_destroy_precreated (3329)
7328 - loop device shouldn't call sync_dev() for nul device (3092)
7329 - clear page cache after eviction (2766)
7330 - resynchronize MDS->OST in background (2824)
7331 - refuse to mount the same filesystem twice on same mountpoint (3394)
7332 - allow llmount to create routes for mounting behind routers (3320)
7333 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7334 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7335 - fix rare problem with rename on htree directories (3417)
7337 - allow default OST striping configuration per directory (1414)
7338 - fix compilation for qswnal for 2.6 kernels (3125)
7339 - increase maximum number of MDS request buffers for large systems
7340 - change liblustreapi to be useful for external progs like lfsck (3098)
7341 - increase local configuration timeout for slow disks (3353)
7342 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7344 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7347 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7348 - don't overwrite extent policy data in reply if lock was blocked (2901)
7349 - drop filter export grants atomically with removal from device (2663)
7350 - del obd_self_export from work_list in class_disconnect_exports (2908)
7351 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7352 - swab reply message in mdc_close, other PPC fixes (2464)
7353 - fix destroying of named logs (2325)
7354 - overwrite old logs when running lconf --write_conf (2264)
7355 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7356 - fix race in target_handle_connect (2898)
7357 - mds_reint_create() should take same inode create lock (2926)
7358 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7359 - don't close files for self_export to avoid uninitialized obd (2936)
7360 - allow MDS with the same name as client node (2939)
7361 - hold dentry reference for closed log files for unlink (2325)
7362 - reserve space for all logs during transactions (2059)
7363 - don't evict page beyond end of stripe extent (2925)
7364 - don't oops on a deleted current working directory (2399)
7365 - handle hard links to targets without a parent properly (2517)
7366 - don't dereference NULL lock when racing during eviction (2867)
7367 - don't grow lock extents when lots of conflicting locks (2919)
7369 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7372 - account for cache space usage on clients to avoid data loss (974)
7373 - lfsck support in lustre kernel code (2349)
7374 - reduce journal credits needed for BRW writes (2370)
7375 - orphan handling to avoid losing space on client/server crashes
7376 - ptlrpcd can be blocked, stopping ALL progress (2477)
7377 - use lock value blocks to assist in proper KMS, faster stat (1021)
7378 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7379 - recovery for initial connections (2355)
7380 - fixes for mds_cleanup_orphans (1934)
7381 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7382 - block all file creations until orphan recovery completes (1901)
7383 - client remove rq_connection from request struct (2423)
7384 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7385 - recovery timer race (2670)
7386 - mdc_close recovey bug (2532)
7387 - ptlrpc cleanup bug (2710)
7388 - mds timeout on local locks (2588)
7389 - namespace lock held during RPCs (2431)
7390 - handle interrupted sync write properly (2503)
7391 - don't try to handle a message that hasn't been replied to (2699)
7392 - client assert failure during cleanup after abort recovery (2701)
7393 - leak mdc device after failed mount (2712)
7394 - ptlrpc_check_set allows timedout requests to complete (2714)
7395 - wait for inflight reqs when ptlrpcd finishes (2710)
7396 - make sure unregistered services are removed from the srv_list
7397 - reset bulk XID's when resending them (caught by 1138 test)
7398 - unregister_bulk after timeout
7399 - fix lconf error (2694)
7400 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7401 - readahead locks pages, leaves pending causing memory pressure (2673)
7402 - increase OST request buffers to 4096 on large machines (2729)
7403 - fix up permission of existing directories in simple_mkdir (2661)
7404 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7405 - don't assign transno to errored transactions (2742)
7406 - don't delete objects on OST if given a bogus objid from MDS (2751)
7407 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7408 - if rq_no_resend, then timeout request after recovery (2432)
7409 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7410 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7411 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7412 - MDS recovery shouldn't delete objects that it has given out (2730)
7413 - if enqueue arrives after completion, don't clobber LVB (2819)
7414 - don't unlock pages twice when trigger_group_io returns error (2814)
7415 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7416 - don't write pages to disk if there was an error (1450)
7417 - don't ping imports that have recovery disabled (2676)
7418 - take buffered bytes into account when balancing socknal conn (2817)
7419 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7420 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7421 - remove little-endian swabbing of llog records (1987)
7422 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7423 - retry reposting request buffers if they fail (1191)
7424 - grow extent at grant time to avoid granting a revoked lock (2809)
7425 - lock revoke doesn't evict page if covered by a second lock (2765)
7426 - disable VM readahead to avoid reading outside lock extents (2805)
7428 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7429 - updated kernel patches for hp-2.4.20 kernel (2681)
7431 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7434 - fix truncated write corruption (2366)
7435 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7437 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7438 - ensure len(uuid) < 37 in lmc (1171)
7439 - fix ia64 OOPS in llog_test (2255)
7440 - zero end of page at obdfilter for partial page writes (2648)
7441 - don't leave stale dentries around after renames (bug 2428)
7442 - fix timeouts when evicting a client with a single lock held (2642)
7443 - set deadline for the initial HELLO message to drain (2634)
7444 - print out dotted-quad IP addresses in the socknal (2302)
7446 - additional debugging for MDS client eviction problem (2443)
7447 - fix mkfsoptions support for osts (2603, 2604)
7449 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7452 - add series for the vanilla 2.6.0 kernel
7453 - add series for the vanilla 2.4.24 kernel
7454 - add series for a cray x86/64 UL kernel drop
7455 - fix xattr patches for the vanilla 2.4.19 series
7457 - generate true UUIDs in lmc (1171)
7458 - have portals stack dumping break in UML (2466)
7459 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7460 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7461 - wake callback waiting threads on client eviction (2460)
7462 - Add --ptldebug and --subsystem to lmc (1719)
7463 - update assertion to allow safe interrupt allocation
7464 - set rq_no_resend for cancel requests (2432)
7465 - recalculate ptlrpcd timeout after resend (2494)
7466 - call vfs_rmdir when removing pending directories (2368)
7467 - fix renaming a file to itself (2429)
7468 - lmc creates a default one-stripe lov (2454)
7469 - expand procfs space to handle large clusters (2326)
7470 - increase UML stack to avoid overflow
7471 - update lconf's list of debug and subsystem masks
7472 - fix lfs find --obd (2510)
7473 - /proc tunable for disabling filter read caching (2591)
7474 - stop rpm packages from altering slapd.conf (2301)
7475 - disable nagle in the socknal under 0conf (2578)
7476 - choose mds inode size based on stripe count (2572)
7477 - fix kernel-source rpm problems (2516)
7479 - add --disable-doc to avoid pdf generation (2421)
7480 - update documentation, tests, type-os, comments
7481 - avoid format warnings on ia64
7482 - remove the TOE NAL
7483 - tiny code cleanups by removing unused fields
7485 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7488 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7489 - remove the most blatant lies from BUILDING file (2371)
7490 - change default debug level to reasonable production setting
7491 - reduce client side cache size to reduce cache flush time
7492 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7493 - make TCP zerocopy and pinger support enabled by default (2476)
7494 - sync writes completed after process exits caused crashes (2319)
7495 - maintain correct mount count on the MDS (2356)
7496 - backout 1557, because 2316 wasn't really fixed
7497 - better file I/O statistics gathering in /proc
7498 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7499 - another kernel patch to fix zero-copy TCP function export
7500 - don't take duplicate lock when processing re-sent getattr (2420)
7501 - lctl uses obd_self_export instead of creating new conn (2353)
7502 - MDS/OST recovery case which requires object creation asserted (2425)
7503 - move lfs from /usr/sbin to /usr/bin in packages
7504 - fix race between mds_client_add and mds_client_free (2417)
7505 - use kmalloc instead of slabs in portals (2430)
7506 - don't create duplicate records when a failover MDS is present (2442)
7507 - remove unnecessary mount age check (2332)
7508 - don't remove directory inodes from locks prematurely (2451)
7509 - don't break if MDS service name is the same as hostname (2103)
7510 - fix races in client write RPC generation when cache full (2482)
7512 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7515 - remove now-unused request->rq_obd (278)
7516 - if an allocation fails, print out how much memory we've used (1933)
7517 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7518 - add the "configurable stack size" patch to most series files (1256)
7519 - ability to write large log records, for 100+ OST configs (2306)
7520 - fix NULL deref when filter_prep fails (2314)
7521 - fix operator precedence error in filter_sync
7522 - dynamic allocation of socknal TX descriptors (2315)
7523 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7524 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7525 - ensure that CWARN messages in Portals always get to the syslog
7526 - __init/__exit are not for prototype decls (ldlm_init/exit)
7527 - x86-64 compile warning fixes
7528 - fix gateway LMC keyword conflict (2318)
7529 - fix MDS lock inversions in getattr/reint paths (1844)
7530 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7531 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7532 - DLM race condition prevented some lock evictions (2328)
7533 - ENOMEM detection and retry on socknal sends (2230)
7534 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7535 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7536 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7537 - better error messages when a client is rejected during recovery (1505)
7538 - avoid cancelling locks which were never granted, after failure (2330)
7539 - fix i_sem/journal inversion in mds_client_add (2333)
7540 - fix truncate/getattr lock cycle deadlock (2334)
7541 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7542 - fix two rare exit paths which could leak an l_lock() ref (2321)
7543 - fencepost error in MDS/OST orphan recovery (2226)
7544 - make log record alignment 8 bytes (1988)
7545 - lstripe now fails when requested offset > ost_count (2237)
7546 - ensure that all kernel series have a complete list.h (1607)
7547 - fix crashes in special-file operations (2316)
7548 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7550 - allow configurable automake binary, for testing new versions
7551 - small update to the lfs documentation
7553 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7555 * fix negative export reference count in fsfilt_sync (2312)
7557 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7558 * release candidate 0.9.1
7560 - orphans are moved into the PENDING directory for possible recovery
7561 - replayed opens now open by fid for orphan/rename safety (1042)
7562 - last close of an orphan inode generates a transno (683)
7563 - chdir() and mount() now pin the directory entry (1020)
7564 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7565 - discard very old requests without processing them (1502)
7566 - remove some common, well-understood CERRORs (1505)
7567 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7568 - clear "grant" flags in OST replies until OST grant code lands (1644)
7569 - fix read performance by not clobbering i_blksize on client (1598)
7570 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7571 - make LRU size a /proc tunable, clears locks when reduced (707)
7572 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7573 - server locks take references on exports now (1558)
7574 - build fixes for 2.4.20-rh trees (1663)
7575 - return an error from lov_create if all OSCs are inactive (1751)
7576 - fix import levels when a reconnect happens without a timeout (1597)
7577 - exit early from mds_open if we get a lookup error (1749)
7578 - partial page read at EOF wouldn't wait for disk before sending (1642)
7579 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7580 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7581 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7582 - write/truncate lock inversion (1639)
7583 - Don't auto-load obdclass, portals modules during cleanup (1495)
7584 - fix timestamps from jumping to "now" (1763)
7585 - extra journal assertions (1648)
7586 - add an extra multiunlink test (1771)
7587 - fix read_record/write_record API (1776)
7588 - fix leak of offset_extent, possible incorrect i_size later (1772)
7589 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7590 - Debugging for the kqswnal_get_idle_tx problems (1820)
7591 - Allow recovery to be attempted multiple times (1536)
7592 - Write out MDS last_rcvd file after it is first created (1600)
7593 - Fix tx_descriptor leak in failed transmit situations (1827)
7594 - ext3 journaling fixes for assertion failure after IO error (1871)
7595 - class_export_put() on freed export after completion AST error (1896)
7596 - Fix revalidate looping in VFS (1322)
7597 - Don't access a freed export during MDS_REINT timeout (1521)
7598 - Add open-unlink recovery support on the MDS (1673,1764)
7599 - Return an error if no MDS data was read from last_rcvd (1946)
7600 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7601 - Don't setup a disk device that doesn't match exported UUID (317)
7602 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7603 - avoid committing NULL handle in force close
7604 - local.sh is now a one-stripe LOV configuration
7605 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7606 - fix ext3 htree duplicate directory entry corruption (1516)
7607 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7608 - update to new LOV EA format (2097)
7609 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7611 2003-06-15 Phil Schwan <phil@clusterfs.com>
7614 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7615 - per-import/export recovery handling (958, 931, 959)
7616 - multiple last-rcvd slots, for serving multiple FSes (949)
7617 - connections are again shared between multiple imp/exports (963, 964)
7618 - "umount -f" would hang if any requests needed to be sent (393, 978)
7619 - avoid pinning large req buffer by copying for queued messages (989)
7620 - add "uuid" to "lctl device" command to help upcalls (991)
7621 - "open" RPCs with transnos would confuse recovery counters (1037)
7622 - do proper endian conversion of all wire messages (288, 340, 891)
7623 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7624 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7625 - fix ptlrpc_connection leak in target_handle_connect (1174)
7626 - fix import refcounting bug in OST and MDS cleanup (1134)
7627 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7628 - fix very unlikely obd_types race condition (501)
7629 - remove osc_open hack for echo_client (1187)
7630 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7631 - a failure in read_inode2 leads to deadlock (1139)
7632 - cancel ack-locks as soon as transaction is committed (1072)
7633 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7634 - make sure to commitrw after any preprw to avoid deadlock (1162)
7635 - failing to execute a file in a lustre FS would lock inode (1203)
7636 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7637 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7638 - obd_uuid2tgt would walk off the end of the list (1255)
7639 - on IA64 the osc would give portals incorrect bulk size (1258)
7640 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7641 - fix lock inversion caused by new llite matching code (1282)
7642 - limit the number of dirty pages on a client to 10MB (1286)
7643 - timed out locks were not being corrected cancelled (1289)
7644 - fix O_DIRECT above 4GB on IA-32 (1292)
7645 * major user-visible changes
7646 - fail out/fail over policy now controlled by the upcall (993)
7648 - add OBD_PING to check server availability and failure (954)
7649 - lustre messages are now sent in sending host order (288, 340, 891)
7650 - add eadatalen to MDS getattr reply (340)
7651 - OST read replies may contain second buffer, with per-page status (593)
7653 2003-03-11 Phil Schwan <phil@clusterfs.com>
7656 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7657 - failed open()s could cause deadlock; fixed (867, 869)
7658 - stop cancelling OST locks when files are closed (481)
7659 - overlapping XID spaces caused network corruption (851, 853)
7660 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7661 - setattr_raw added, to do single-RPC, server-side setattrs
7662 - lmc/lconf syntax change for OST UUIDs
7663 - fix crashy race condition between ptlrpc_free_req and osc_close
7664 - don't use request in mdc_enqueue if we hit a timeout (889)
7665 - don't set the inode i_size for regular files from the MDS (896)
7666 - handle out of order completion AST (842)
7667 - don't LBUG if a lock request times out after receiving AST (913)
7668 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7669 - if a bad lock AST arrives, send an error instead of dropping entirely
7670 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7671 - fix leak in bulk IO when only partially completed (899, 900, 926)
7672 - fix O_DIRECT for ia64 (55)
7673 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7674 - C-z after timeout could hang a process forever; fixed (977)
7676 - client-side I/O cache (678, 924, 929, 941, 970)
7678 - READPAGE and SETATTRs which don't take server-side locks get
7681 2003-02-11 Phil Schwan <phil@clusterfs.com>
7684 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7685 - Fix fsfilt_extN_readpage() to read a full page of directory
7686 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7687 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7688 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7689 - Most intent operations were removed, in favour of a new RPC mode
7690 that does a single RPC to the server and bypasses most of the VFS
7691 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7692 - Aggressively cancel local locks on DLM servers
7693 - mds_reint_unlink sends EA to the client if it's the last nlink.
7694 client uses that EA to unlink OST objects.
7695 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7696 - recursive symlinks were fixed (439)
7697 - fixed NULL deref in DEBUG_REQ
7698 - filter_update_lastobjid no longer calls sync, which annoyed extN
7699 - fixed multi-client small-writes to a single file problem (445)
7700 - fixed mtime updates during file writes (607)
7701 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7702 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7703 - fixed Linux OST size reporting problem (444, 656)
7704 - OST now updates object mtime with writes or setattr (607, 619)
7705 - client verifies file size before zeroing page past EOF (445)
7706 - OST now writes last allocated objid to disk with allocation (108)
7707 - LOV on echo now works (409)
7709 - mds_reint_unlink sends a new buffer, with the EA included. this
7710 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7711 set if a regular file was being unlinked, and it was the last link
7712 - use PtlGet from the target for bulk writes (315)
7713 - OST now updates object mtime with writes or setattr (607, 619)
7714 - LDLM now has a grant-time callback to revalidate locked items, if
7716 - Many MDS operations were reorganized to combat race conditions
7718 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7719 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7721 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7724 - Fully reactivate OST imports after reconnection (512, others)
7725 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7726 - More graceful error handling for truncating on dead OST (515)
7727 - Don't error out unless we're actually accessing dead stripes (474)
7728 - Fix garbage sizes when stripes are missing (410)
7729 - LRU counters were broken, causing constant lock purge (433, 432)
7730 - garbage on read from stripes with failed OSTs (441)
7731 - mark OSCs as active before reconnecting during recovery (438)
7732 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7733 - lfind did not preserve OST order in output (443)
7734 - symlinks cause hung clients, incorrect data (439)
7735 - stop dereferencing request after dropping refcount (457)
7736 - don't LASSERT(spin_is_locked) on non-SMP (455)
7737 - fixes for many rename() bugs
7738 - fstat didn't correctly synchronize attributes (399)
7739 - server must handle lock cancellation during blocking AST prep (487)
7740 - bulk descriptors were free()d too soon (511)
7741 - fix paths in lconf, which would load incorrect modules (451, 507)
7742 - fix confusing lconf 'host not found' error message (386)
7743 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7744 - fix race condition in mdc_blocking_ast() for inode access (526)
7745 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7746 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7747 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7748 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7749 - fix lstripe in several ways (kernel side) (527)
7750 - fix request leak in ldlm_cli_enqueue (262)
7751 - incorrect OSC was marked inactive after OST failure
7752 - call mds_fs_cleanup before unmounting filesystem (524)
7753 - fix races between taking ns_lock and ldlm_lock_change_resource
7754 - fix races updating LOV export open file list
7755 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7756 - fix recovery NULL deref in ldlm_cli_cancel_unused
7757 - fix some DLM races by using new hash table for lock handles (419)
7758 - permit the client to specify desired inodes, at replay
7759 - duplicate requests when we queue them for replay reintegration
7760 - fix last_rcvd offset calculation
7761 - sync after each recovered transaction, so we always make progress
7762 - never, not always, ERESTART requests without transnos
7763 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7765 - skip replay if the MDS says that the client is already connected
7766 - don't check for a recovery-enabled export to match lctl's UUID
7767 - don't INC_USE_COUNT for phantom exports
7768 - don't crash when cleaning up phantom exports (567)
7769 - don't double-finish or set replay data for errored mdc_open requests
7770 - abort requests when they time out, so we don't get old replies
7771 - send/receive replies for AST messages again
7772 - if the client says that it doesn't have the lock, cancel it on the
7774 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7775 mark it as destroyed, it will all work out when decref is called
7776 - fix module use counts (22, 581)
7778 - ASTs now expect a reply (server cancels lock on error reply)
7780 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7783 - fix many simultaneous client startup (392)
7784 - fix dentry->d_it clobbering
7785 - credentials weren't being shipped for readdir/getattr operations
7786 - remove invalid assertions triggered during some concurrent MD
7788 - proper Lustre versions added (336, 389)
7789 - fix memory leak for create error case (398)
7790 - fix LOV locking bug that would get cli/srv out of sync
7791 - fix echo client over LOV (409)
7792 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7793 - fix double-O_EXCL intent crash (424)
7794 - avoid sending multiple lock CANCELs (352)
7796 - MDS can do multi-client recovery (modulo bugs in new code)
7798 - many updates, edits, cleanups
7800 2002-11-18 Phil Schwan <phil@clusterfs.com>
7803 - fix null d_it dereference (346)
7804 - fix full OST/dbench hang (333)
7805 - fix permission problem with file removal (286)
7806 - fix removal of OSCs from LOV when they fail
7807 - fix NULL deref during bulk timeout (214)
7808 - fix problems related to multiple filesystems on one MDS (241)
7809 - fixed serious subtle metadata locking bugs
7810 - free locks on clients when inodes are removed due to memory
7812 - fix inode pointer in lock data (285)
7813 - partial support for multiple MDS on a single host (241)
7814 - data locks weren't cancelled at clear_inode time (290, 311)
7815 - intent locks could lead to unbounded lock growth (205)
7816 - added a maximum lock count, an LRU list, and a flusher
7817 - fix multiple rename (365)
7818 - properly abstracted the echo client
7819 - OSC locked 1 byte too many; fixed
7820 - rewrote brw callback code:
7821 - fixed recovery bugs related to LOVs (306)
7822 - fixed too-many-pages-in-one-write crash (191)
7823 - fixed (again) crash in sync_io_timeout (214)
7824 - probably fixed callback-related race (385)
7826 - Add capability to MDS protocol
7827 - LDLM cancellations and callbacks on different portals
7829 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7832 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7833 - defer open object destruction to close time (601981, 138)
7834 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7835 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7836 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7837 - don't call obd_setattr in ll_file_release for destroyed objects
7838 * protocol change to lustre_msg: move |version| and add |flags|
7839 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7840 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7841 * added replay of create, unlink, link and rename operations during
7842 MDS failover; recovery should be much more robust now
7843 * remove failed OSCs from LOVs (only lov_create uses this so far)
7844 * the lustre-HOWTO was brought (more) up to date (582544)
7846 2002-10-23 Phil Schwan <phil@clusterfs.com>
7849 - in-use dentries weren't being reused properly (617851)
7850 - prevent multiple LDLM setup (599178)
7851 - fix LOV size calculations for truncate (617853)
7852 - fix client handling of MDS intent errors (POSIX)
7853 - fix permission bug in lovstripe.c test (624321)
7854 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7855 - truncate past end of file could corrupt data
7856 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7857 - a race in recovery could return ETIMEDOUT to apps (623947)
7858 - building outside the source directory was fixed
7859 * the lustre-HOWTO was brought (more) up to date (582544)
7860 * major progress was made on recovery functionality
7862 2002-10-10 Phil Schwan <phil@clusterfs.com>
7865 - recovery deadlock fix
7866 - rm -rf causes LBUG fix (617817)
7867 - file open by multiple tasks fix (618962)
7868 - directory permissions bugs (602707 and 620007)
7869 - journal_stop fixed with locking (611313)
7870 - O_APPEND failures resolved (618273, perhaps 614459)
7871 - lconf PATH fix (619770)
7872 - IA64 build fix (621450)
7873 - RPC buffer sizes scale with amount of memory
7875 2002-10-01 Phil Schwan <phil@clusterfs.com>
7878 - locks would be cancelled without throwing away data pages,
7879 resulting in inconsistent data (605627)
7880 - inode attributes were not always being refreshed (605627, 612449)
7881 - lconf now continues to cleanup after lctl reports an error
7882 - MDS now enforces user permissions (602707)
7883 - lprocfs cleanup fixed, but not yet enabled (614157)
7884 - fixed infinite server hang, should a client not respond to an AST
7885 - avoid going into recovery if user calls readlink() with a buffer
7886 that's too small (613941)
7887 - AST RPCs no longer require replies (614867) -- this may be changed
7888 - don't crash server if client sends an IOV that's too big (611336)
7889 - fixed lock conversion deadlock (611892)
7890 - fixed the following of symlinks (614622)
7891 * recovery: the server can remove locks from a client that dies, other
7892 clients can make progress
7893 * more extN patch fixes
7894 * compile-time configurable ptlrpc buffer allocations
7896 - collaborative read cache document
7897 - Lustre Lite Performance CDR document-in-progress
7899 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7902 - fix typo in patch-2.4.18
7904 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7907 - clear ptlrpc request each time in handle_incoming_request()
7908 - unlink of files now destroys the object on the OST
7910 2002-09-19 Peter Braam <braam@clusterfs.com>
7912 * add hard link support
7913 * change obdfile creation method
7914 * kernel patch changed
7916 2002-09-19 Peter Braam <braam@clusterfs.com>
7919 - stack overflow bug in extN fixed
7921 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7923 * documentation updates
7924 - add man pages for config tools
7925 - update tests/README to describe testing with new config tools
7926 - finish metadata API descriptions
7927 * bug fixes and cleanups
7928 - statfs workaround for 16TB limit
7929 - LOV stripe allocation improved, can stripe on subset of OSTs
7930 - LOV file size/IO offset was wrong for files > 4GB in size
7931 - object EA data was being dropped, caused files to be unreadable
7932 - memory overflow with non-LOV OST caused memory corruption
7933 - fixed regression tests to work with new config tools, obdfilter
7934 - fixed bug when directory size became larger than 1 block
7935 - fixed bug (for single client case) when PWD was deleted
7936 - invalidate local directory pages when doing intent-based ops
7937 - avoid LDLM oops when lock callback contained bad data
7939 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7941 * documentation updates
7942 * bug fixes and cleanups
7943 - configuration tools
7946 - 64-bit compile warnings
7947 - 64-bit internal statfs data
7949 * test_brw on persistent OST devices
7951 * lprocfs (disabled)
7953 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7955 * documentation updates
7956 * bug fixes and cleanups
7957 * configuration tools
7959 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7962 * many small fixes to 0_5_4
7963 * io/network handling
7964 * thinkos in MDS operations
7966 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7969 * crucial basic fixes to 0.5.3
7970 * IOR, Iozone work over Elan
7973 2002-08-07 Phil Schwan <phil@clusterfs.com>
7974 * version 0_5_3, our first alpha
7975 * we use the new Portals iovs
7976 * documentation updates
7977 * bug fixes and cleanups
7978 * small changes in the DLM wire protocol
7980 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7981 * version 0_5_1 with some initial stability,
7982 * locking on MD and file I/O.
7983 * documentation updates
7984 * several bug fixes since 0.5.0
7985 * small changes in wire protocol
7987 2002-07-18 Phil Schwan <phil@clusterfs.com>
7989 * delivered as Lustre Light Alpha
7990 * fixed a crash after handling invalid MDS requests
7991 * fixed directory pages for architectures with non-4k pages sizes
7993 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7994 * release version v0_4_4
7995 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7997 * Fixes a number of interruption problems with OST operations.
7998 * Update documentation for portals header changes
7999 * Move all wire protocol structs/defines to lustre_idl.h
8000 * Fixes symlink length bug.
8001 * Add tcpdump to repository.
8003 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
8004 * release version v0_4_3
8005 * Fixes statfs for inodes on extN.
8006 * Fixes bug in runtests which would delete /etc/hosts.
8007 * Use 64-bit object IDs wherever possible (not into VFS though)
8008 Remove ost_get_info, which is unused by lustre, and out of date.
8010 2002-07-03 Peter Braam <braam@clusterfs.com>
8011 * release version v0_4_2 Fixes a lookup error (type not passed)
8012 * move forward to head of Portals
8013 * move forward to latest Lustre kernel
8015 2002-06-25 Peter Braam <braam@clusterfs.com>
8016 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob