1 2012-05-31 Whamcloud, Inc.
4 2.6.18-308.11.1.el5 (RHEL 5)
5 2.6.32-220.17.1.el6 (RHEL 6) - patchless client only
6 * Recommended e2fsprogs version: 1.41.90.wc4
8 See http://wiki.whamcloud.com/display/PUB/Changelog+1.8 for more information.
10 Severity : enhancement
12 Description: Kernel update [RHEL5.8 2.6.18-308.11.1.el5]
14 -------------------------------------------------------------------------------
16 2011-10-31 Whamcloud, Inc.
18 * Support for kernels:
19 2.6.18-274.3.1.el5 (RHEL 5)
20 2.6.32-131.12.1.el6 (RHEL 6) - patchless client only
21 * Recommended e2fsprogs version: 1.41.90.wc3
23 See http://wiki.whamcloud.com/display/PUB/Changelog+1.8 for more information.
25 -------------------------------------------------------------------------------
27 2011-06-30 Whamcloud, Inc.
29 * Support for kernels:
30 2.6.18-238.12.1.el5 (RHEL 5)
31 2.6.32-131.2.1.el6 (RHEL 6) - patchless client only
32 * Recommended e2fsprogs version: 1.41.90.wc3
33 * Use ext4-based ldiskfs as default for RHEL5
34 * Add support for 24TB LUN (RHEL5 server only)
36 Severity : enhancement
38 Description: Add support for RHEL5.6 (2.6.18-238.12.1.el5)
40 Severity : enhancement
41 Jira : LU-62, LU-73, LU-402, LU-130
42 Description: Add RHEL6 client support
46 Description: Fix client crash introduced in 1.8.5
47 Details : The patch from bugzilla ticket 18213 landed for 1.8.5 can cause
48 a crash in mdc_exit_request() when the process is interrupted.
52 Description: ls -l reports wrong file size
53 Details : Always update LVB from disk when a glimpse callback returns an
56 Severity : major (2.6.32-based kernels only)
58 Description: Add workaround for a race causing an assertion failure in
59 clear_inode() on lustre clients.
63 Description: Fix interoperability issue between 1.8 clients and 2.x servers
64 Details : The ldlm hash handling in Lustre 2.x is incompatible with the one
65 used in Lustre 1.8. Add support for 64-bit dir name hash to the
70 Description: slow IO with read-intense application
71 Details : Align the readahead extent by 1M after when it is trimed by
76 Description: Skip quotacheck on administratively disabled OSTs
80 Description: Some autogen improvements
82 Severity : enhancement
83 Jira : LU-123, LU-296, LU-230
84 Description: Add support for yaml data logging and auster script.
86 Severity : enhancement
88 Description: Allow OSTs to be created with no primary node
89 Details : Add a --servicenode parameter for mkfs.lustre to treat all
90 service nodes equally.
92 -------------------------------------------------------------------------------
94 2011-05-12 Oracle, Inc.
96 * Support for kernels:
97 2.6.16.60-0.42.8 (SLES 10),
98 2.6.27.39-0.3.1 (SLES 11),
99 2.6.18-194.3.1.el5 (RHEL 5)
100 2.6.18-194.3.1.0.1.el5 (OEL 5)
101 * Client support for unpatched kernels:
102 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
103 2.6.16 - 2.6.32 vanilla (kernel.org)
104 * Recommended e2fsprogs version: 1.41.10-sun2
105 * The async journal commit feature (bug 19128) and the cancel
106 lock before replay feature (bug 16774) are disabled by default.
110 Description: llap pages are moved to head of LRU list instead of tail
114 Description: disable statahead by default due to important races found in the
115 statahead implementation.
119 Description: release request ref through ptlrpc_server_drop_request()
120 Details : make sure the request is protected by rq_refcount while
121 being processed in ptlrpc_server_handle_req_in().
125 Description: correct check for transno value in filter_finish_transno()
129 Description: Set body->eadatasize in mdc_getattr_pack()
133 Description: Handle unsent requests with rq_net_err in ptlrpc_check_set()
137 Description: apps stuck in ptlrpc_check_set() during direct I/O
138 Details : start async bulk unregistering at the same time at reply unlink
142 Description: lfs setstripe --pool broken in 1.8.4
143 Details : fix llapi_search_fsname() to handle relative pathnames
147 Description: Only force the mode change if we're changing the size as well
148 Details : Fix regression introduced in 1.8.0. The offending code was added
149 by commit 77ba4b2141d04180211efa8a75c11ab0abf7fafb to remove
150 setgid/setuid bits when do_truncate() is called on the file.
151 We should only force the change when that occurs, similarly to
152 ll_setattr() in lustre/llite/llite_lib.c
154 Severity : enhancement
156 Description: add procfs tunable to enable/disable lockless direct I/O
157 Details : llite.lustre-*.lockless_direct_io=0 will disable default semantics
158 of direct I/O that forces it to be lockless. lockless_direct_io
159 value, however, will be ignored if per-file LL_FILE_LOCKED_DIRECTIO
164 Description: Lustre servers crashing with NULL pointer errors
165 Details : Make sure the request is protected by rq_refcount while being
166 processed in ptlrpc_server_handle_req_in().
170 Description: kernel BUG at drivers/scsi/sd.c
171 Details : Remove sd iostats patch from sles11 patch series.
175 Description: Modified value of at_min is not taken into account
179 Description: lustre grants flock exclusive locks to two fd's in the same process
183 Description: LBUG in the mfs_verify_child()
187 Description: ASSERTION(atomic_read(&client_stat->nid_exp_ref_count) == 0)
188 Details : Reconnecting from a different nid causes old per-nid stats not to be
193 Description: create proper macro check for bdi interface.
195 -------------------------------------------------------------------------------
197 2010-10-29 Oracle, Inc.
199 * Support for kernels:
200 2.6.16.60-0.69.1 (SLES 10),
201 2.6.32.19-0.2.1 (SLES11),
202 2.6.18-194.17.1.el5 (RHEL 5)
203 2.6.18-194.17.1.0.1.el5 (OEL 5)
204 * Client support for unpatched kernels:
205 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
206 2.6.16 - 2.6.32 vanilla (kernel.org)
207 * Recommended e2fsprogs version: 1.41.10-sun2
208 * The async journal commit feature (bug 19128) and the cancel
209 lock before replay feature (bug 16774) are disabled by default.
213 Description: atime is not properly updated on an MDS
215 Severity : enhancement
217 Description: Update to RHEL5.5 kernel 2.6.19-194.17.1.el5.
218 Update to OEL5.5 kernel 2.6.19-194.17.1.0.1.el5.
220 Severity : enhancement
222 Description: Update to SLES10 SP3 kernel 2.6.16.60-0.69.1.
225 Frequency : only with SLES10
227 Description: Use OFED "KMP" provided by Novell
228 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
229 Lustre is now built against this package. That means you need to
230 install the ofed-kmp package from Novell for the patchless client
231 and from our download site for the server. Note that the ofed-kmp
232 that Novell ships may not exactly match the kernel version but
233 should still be compatible.
235 Severity : enhancement
237 Description: Update SLES11 SP1 kernel to 2.6.32.19-0.2.1.
241 Description: Enabling quotas fails with non-consecutive OST numbering.
245 Description: Fix kernel warning due to lookup_on_len() called without i_mutex
250 Description: Account direct i/o inflight rpcs separately from non-direct i/o so
251 that direct i/o, which is limited by max_rpcs_in_flight, should not
252 block non-direct i/o, which is not limited by max_rpcs_in_flight.
256 Description: Fix per-NID reporting on outstanding writes
260 Description: Reduce stack pressure by uninlining some mds and ptlrpc functions.
264 Description: Remove LASSERT in lprocfs_rd_conn_uuid() since conn == NULL is a
269 Description: fix obdo leak issue in ll_setattr_raw()
273 Description: limit MMP interval
275 Severity : enhancement
277 Description: add several lfs ost enhancements
281 Description: Too many default ACLs break directory access on new directories
285 Description: Lustre inode size is not coherent across nodes.
286 Details : Update lvbo from disk when AST fails with EINVAL. Lvbo is updated
287 on EINVAL error in ldlm_handle_ast_error(). The updates in
288 filter_intent_policy() and ldlm_cb_interpret() have been removed as
293 Description: Oops at __percpu_counter_add+0x1b
294 Details : Use bdi_init()/bdi_destroy() to proper initialize backing_dev_info
299 Description: add mount option to generate 32bit ino, this can be used for 32bit
300 application compatibility.
304 Description: keep reference count for "lli_sai" to prevent it to be released
305 when "statahead_enter()"
309 Description: allow quotacheck over OSTs with sparse indices
313 Description: Objects not getting deleted for files which have been removed
314 Details : ll_have_md_lock() should differentiate between CR and CW OPEN
319 Description: pin object's inode in memory to avoid certain timeouts
323 Description: fix LBUG when obdfilter-survey is interrupted.
325 -------------------------------------------------------------------------------
327 2010-07-31 Oracle, Inc.
329 * Support for kernels:
330 2.6.16.60-0.42.8 (SLES 10),
331 2.6.27.39-0.3.1 (SLES11),
332 2.6.18-194.3.1.el5 (RHEL 5)
333 2.6.18-194.3.1.0.1.el5 (OEL 5)
334 * Client support for unpatched kernels:
335 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
336 2.6.16 - 2.6.30 vanilla (kernel.org)
337 * Recommended e2fsprogs version: 1.41.10-sun2
338 * The async journal commit feature (bug 19128) and the cancel
339 lock before replay feature (bug 16774) are disabled by default.
341 Severity : enhancement
343 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
344 2.6.18-194.8.1.0.1.el5.
346 Severity : enhancement
348 Description: using inkernel OFED stack for rhel5 & oel5.
350 Severity : enhancement
352 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
353 Details : lfs_migrate does a "poor man's" migration of files from their
354 current OST layout to a new OST layout as chosen by the MDS.
358 Description: mds_orphan_add_link()) error linking orphan to PENDING
359 Details : quota limits might disallow linking orphans to PENDING
360 when unlinking a file - temporary raise threads' privileges
361 when processing unlinks.
363 Severity : enhancement
365 Description: add conf-param -d option to remove permanent settings.
366 Details : Add the ability to remove permanent lctl conf_param settings.
367 (Previously conf_param settings could only be changed, not
368 removed.) This also provides a method to change failover
369 nid locations. Improve lctl man page.
371 Severity : enhancement
373 Description: add list_param to b1_8 and add "-R" option to list params
376 Severity : enhancement
378 Description: lfs quota output is not very convenient for awk/sed-parsing
379 Details : Some positions in lfs quota output table could be empty or
380 non-empty which made it hard to parse it with scripts, now a dash
381 is put instead of space where there is not supposed to be any data.
383 Severity : enhancement
385 Description: fix obdfilter-survey script to work properly with remote oss-s
387 Severity : enhancement
389 Description: add new OBDFILTER_SURVEY test suite
391 Severity : enhancement
393 Description: add new multiple mount protection (MMP) test suite
395 Severity : enhancement
397 Description: add support for async journal commit in echo client
399 Severity : enhancement
401 Description: allow userland programs to include <lustre/lustre_idl.h>
402 from stardard include directories
404 Severity : enhancement
406 Description: The prune-icache-use-trylock is no longer needed now that
407 the patch from bug 20008 is landed.
411 Description: The shrink grant feature is still active on the client although the
412 connect flag is not set.
416 Description: Don't leak grant space if the write failed with quota exceeded.
420 Description: Don't consume grant space twice on recoverable resent.
424 Description: a race condition could lead to SIGBUS being sent to an
425 application using mmap-ped files from Lustre
426 Details : truncate_complete_page implementation for the patchless
427 client could arbitrarily unset PG_Uptodate flag for a
428 page being kicked from the page cache, an uptodate check
429 right after a readpage call in filemap_fault could fail
430 because of that as though the page read had been unsuccessful.
434 Description: dlm lock slab shrinking is not efficient
435 Details : The dlm_locks slab can grow significantly and consumes a lot of
436 memory on the server. Set a hardlimit to grant_plan.
440 Description: Lustre does not do 1MB IOs to HW RAID
441 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
442 This is what we do already for SLES kernels.
446 Description: bump maximum number of phys/hw segments in the SLES11 kernel
447 until s/g chaining works properly.
452 Description: LSI Fusion MPT driver hacks to improve performance
453 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
455 Severity : enhancement
457 Description: increase default md stripe_cache_size to 16k
460 Bugzilla : 15587/21439
461 Description: don't handle security.capability xattr
462 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
463 This results in additional getxattr calls, causing VBR test
464 failures as well as a preformance drop when writing.
468 Description: obdfilter-survey is no longer working
469 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
470 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
474 Description: ll_shrink_cache does not handle __GFP_FS properly
478 Description: lfs getstripe shows wrong info for directories
479 Details : Set correct LOVEA default values for filesystem-wide.
483 Description: FSX checksum false positves due to mmap IO
484 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
485 checksum errors, if the flag is set on a request.
489 Description: file operations after eviction have successful return values
490 Details : use vfs ->flush callback to return any pending async errors
495 Description: mdsrate fails to write after 1.3+M files opened
496 Details : decrease memory usage on clients by recycling dentries and
501 Description: obdfilter-survey gives unreasonably high numbers
502 Details : Wait for all threads to complete when running test_brw.
506 Description: do not set lustre read_only device when server umount and keep
507 client records for recoverable ones
511 Description: move sync_on_lock_cancel tunable to the obdfilter layer
512 Details : move the tunable to trigger a journal flush on lock cancel from
513 the ost layer to the obdfilter layer. This tunable is useful
514 when using the async journal commit feature.
518 Description: exp->exp_nid_stats == NULL in filter_tally()
519 Details : fix race with per-nid stats by delaying procfs cleanup until
524 Description: extent lock cancellation on client can keep the cpu busy for too
529 Description: Do not fail OST activation when a llog is not found, just
530 issue an error message.
534 Description: Don't enable extents by default for MDT.
538 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
539 the AT code can access it concurrently while sending early
544 Description: Disable lockless truncate by default since it is sometimes flawed
545 and causes the write_disjoint test to fail.
549 Description: OSSs which don't have the patch from bug 20278 can trigger an
552 Severity : enhancement
554 Description: don't print message to the console when we have not managed to
559 Description: The MDS fails to synchronize OSTs which registered with the MGS
560 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
561 raised too early and thus discarded by the MDT stack.
562 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
567 Description: Fix race when the ping evictor and a service thread execute
568 target_recovery_check_and_stop() concurrently.
572 Description: quota broadcast can trigger a LBUG on the MDT if there are
577 Description: Resetting the lov_objid values to last_id reported by the OST
578 during orphan recovery is incorrect and can cause the same
579 objects to be allocated twice.
581 Severity : enhancement
583 Description: "weak-modules" support
584 Details : Implement "weak-modules" support which enables kernel modules
585 to be used with any kernel that implements the same kABI. In
586 order to achieve this modules are now installed in
587 /lib/modules/$(uname -r)/updates/kernel on all distributions.
589 Severity : enhancement
591 Description: add writeconf as mount option
593 Severity : enhancement
595 Description: produce debuginfo packages for SLES.
597 Severity : enhancement
599 Description: add failover nidlist to the import proc file.
601 Severity : enhancement
603 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
604 clients and 2.0 servers.
606 Severity : enhancement
608 Description: lfs find -s does not work correctly because of a bug in
613 Description: ll_read_ahead_page() must validate the dlm lock before using
618 Description: Prevent failover nids from registering with MGS first.
622 Description: fix lock inversion in ll_setattr_raw().
626 Description: object allocation is not balanced across OSTs.
627 Details : osc_precreate() should return 0, if there are enough objects left.
629 -------------------------------------------------------------------------------
631 2010-04-30 Oracle, Inc.
633 * Support for kernels:
634 2.6.16.60-0.42.8 (SLES 10),
635 2.6.27.39-0.3.1 (SLES11),
636 2.6.18-164.11.1.el5 (RHEL 5)
637 2.6.18-164.11.1.0.1.el5 (OEL 5)
638 * Client support for unpatched kernels:
639 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
640 2.6.16 - 2.6.30 vanilla (kernel.org)
641 * Recommended e2fsprogs version: 1.41.10-sun2
642 * The async journal commit feature (bug 19128) and the cancel
643 lock before replay feature (bug 16774) are disabled by default.
647 Description: open-unlinked directories trigger MDS LBUG
648 Details : Fix regression introduced by the patch from bug 19640.
649 ext3_inc_count() can reset nlink to 1 when the directory
650 is indexed and inode->i_nlink == 2. Work around the problem
651 by incrementing nlink by 2 instead of 1.
655 Description: Reconnects are not throttled
656 Details : Don't wake up pinger on reconnect failures and rely on regular
657 pings to trigger the next reconnection. Please note that the
658 pinger already uses a smaller interval if the import is
662 Frequency : only with NFS export
664 Description: Console flooded with error message from ll_inode_from_lock()
666 Details : in mds_open, initialize the child_res_id before enqueuing
667 the OPEN lock for the child inode, then to avoid senting
668 wrong ldlm_res_id to client.
672 Description: allow multiple instances of the same nid in NID hash
673 Details : Case of multiple separate clients from the same NID (as
674 with liblustre) is legitimate and so we should allow
675 multiple instances of the same NID in nid hash.
679 Description: the readahead code can sleep on a semaphore while holding a
681 Details : in ras_update, "lov_get_info" could be called during increasing
682 readahead windows, which tries to get the mutex lock "lov_lock"
683 while holding the spin_lock "ras_lock", then causes system
688 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
689 Details : fix assertion failure in the grant code.
693 Description: Use CNETERR (which is rate limited) in specific places in
694 the portal's LNET driver to avoid flooding the console.
698 Description: include last created object in precreate slow case
702 Description: don't do rep-ack if not created anything
703 Details : mds_open currently always put a lock into a rep-ack regardless
704 if something was created or not. This is pointless and only
705 creates needless contention. In fact the entire idea was to do
706 this for real creates as a recovery protection.
710 Description: Spurious error messages from smp_processor_id() on preemptible
712 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
713 first. The function fs_trace_get_tcd() was moved up.
717 Description: interval_erase() fix
718 Details : interval_erase() calls update_maxhigh() properly when child
723 Description: "lfs df" does not print stats for all mountpoints
724 Details : Print all mounted lustre filesystems with "lfs df"
728 Description: lfs setstripe -p not longer work with a relative pathname
729 Details : Use realpath() to provide absolute pathname.
733 Description: fix for truncated reply buffer
734 Details : reply buffer could be referred by reply_in_callback after released
739 Description: lustre.lov error when backing up symlinks with extended attributes
740 Details : Improved logic in ll_listxattr()
744 Description: properly handle null value for setattr -n lustre.lov
745 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
746 in ll_setxattr() due to no checking if "value" is NULL.
747 This command now resets to the default striping when executed
752 Description: stack overflow on lock cancellation due to fsync call
753 Details : sync_on_lock_cancel is needed for recovery when async journal
754 is enabled, but we actually just need to make sure that
755 metadata blocks have hit the journal, so doing a fs sync
756 should be enough and should consume less stack (just create an
757 empty handle and commmit it).
761 Description: using current->journal_info to store per-thread data leads
762 to problem under memory pressure
763 Details : disable the per-thread data (current->journal_info) containing
764 the lock info during I/O to work around the issue for short tem
768 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
769 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
770 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
775 Description: Cannot send after transport shutdown
776 Details : Clear imp_vbr_failed flag upon eviction
780 Description: soft lock in request set code during recovery
781 Details : during recovery, uses req->rq_set itself to replay the request
782 instead of ptlrpcd_recovery_pc
786 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
787 Details : During an ll_readahead under ll_readpage, we have seen the the
788 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
789 a page that is already locked by the readahead code.
794 Description: stop waitting for next replay transno if shutdown
795 Details : if the system is shutting down, wake up service thread blocked
796 to wait for next replay transno during recovery, then all the
797 references held by queued requests can be dropped and device
802 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
803 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
807 Description: per-nid stats should not access lustre-hash internal structures
812 Description: mount.lustre fails to pass some options to mount()
817 Description: ext4 extent allocation is slower than in ext3
818 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
819 enlarge ext4 preallocation table for 2048 4K blocks extents
824 Description: incorrect triggering of synchronous IO
825 Details : The OSC can mistakenly fall back to synchronous IO when the
826 max_dirty_mb limit is reached and no write requests have yet
827 been issued. This can occur when the dirty pages are spread
828 over many files all of which are below the optimal request size.
832 Description: Optimize quota_ctl operations by sending requests in parallel
833 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
834 Compiled from two attachments in the ticket.
838 Description: ordering issue between transaction start & i_mutex
839 Details : start the transaction earlier in llog_lvfs_destroy to get
840 transaction start and inode mutex lock nested properly.
844 Description: lru resize SLV can get stuck
845 Details : calculate SLV with a greater precision to not lose small
846 changes due to interger math truncation; round up SLV only
847 if the amount of granted locks less than the limit to not
848 get stuck with this SLV
852 Description: avoid divide-by-zero in lprocfs_rd_import()
856 Description: lfs quota failed when OSTs are down
857 Details : really return approximate block/inode usage when OSTs are down
861 Description: llobdstat fix and enhancement
862 Details : add a counter to set a limit to how many samples will be returned
863 fix a wildcard in the path to limit to obdfilter stats only
867 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
869 Details : abort lock enqueue processing sooner when we umount is in progress.
871 Severity : enhancement
873 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
874 2.6.18-164.11.1.0.1.el5.
876 Severity : enhancement
878 Description: error message improvements
879 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
880 Simplify MDT/OST service start message.
881 Suppress "changing the import ..." warning.
884 Bugzilla : 21961/17914
885 Description: ignore trailing -mdc when determining index number
888 Frequency : only with SLES11
890 Description: fix for a race condition in linux quotas implementation
891 Details : dq_flags(struct dquot) access is not properly locked which could
892 lead to certain inconsistencies when accessing it using non-atomic
893 bit operations like __set_bit in do_set_dqblk.
897 Description: return any pending async errors in close(2) using flush callback
899 -------------------------------------------------------------------------------
901 2010-01-29 Sun Microsystems, Inc.
903 * Support for kernels:
904 2.6.16.60-0.42.8 (SLES 10),
905 2.6.27.39-0.3.1 (SLES11),
906 2.6.18-164.11.1.el5 (RHEL 5)
907 2.6.18-164.6.1.0.1.el5 (OEL 5)
908 * Client support for unpatched kernels:
909 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
910 2.6.16 - 2.6.30 vanilla (kernel.org)
911 * Recommended e2fsprogs version: 1.41.6.sun1
912 * The async journal commit feature (bug 19128) and the cancel
913 lock before replay feature (bug 16774) are disabled by default.
915 Severity : enhancement
917 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
918 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
920 Severity : enhancement
921 Bugzilla : 21511/19848
922 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
924 Severity : enhancement
926 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
928 Severity : enhancement
930 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
931 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
933 Severity : enhancement
935 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
938 Severity : enhancement
940 Description: Vanilla kernel 2.6.30 patchless client support.
945 Description: bad entry in directory xxx: inode out of bounds
946 Details : fix locking issue in the rename path which could race with any
947 other operations updating the same directory.
949 Severity : enhancement
951 Description: Make watchdog timer messages to be more clear and descriptive.
955 Description: cp -p command does not preserve the dates and timestamp
956 Details : mtime could be spoiled by a write callback
960 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
964 Description: Allow non-root access for "lfs check".
965 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
967 Severity : enhancement
969 Description: quotacheck performance/scaling issues
970 Details : reduce quotacheck time on empty filesystem by skipping uninit
973 Severity : enhancement
975 Description: Enhancement for lfs(1) command to use numeric uid/gid.
977 Severity : enhancement
979 Description: Adjust locks' extents on their first enqueue, so that at the time
980 they get granted, there is no need for another pass through the
981 queues since they are already shaped into the proper forms.
985 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
986 arguments and prevent LBUG() in lustre_shrink_reply_v2().
990 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
991 exactly the specified mount options are used. Leaving off
992 any "mandatory" mount options is an error. Leaving off any
993 default mount options causes a warning, but is allowed.
994 Change errors=remount-ro from mandatory to default. Sanitize
995 the mount string before storing it. Update man pages accordingly.
999 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
1000 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
1003 Severity : enhancement
1005 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
1007 Severity : enhancement
1009 Description: Clients can replay thousands of unused locks during recovery
1010 Details : Don't replay unused locks (only read locks for now) during
1011 recovery. This feature is disabled by default and can be
1012 enabled by running the following command on the clients:
1013 lctl get_param ldlm.cancel_unused_locks_before_replay
1017 Description: can't stat file in some situation.
1018 Details : improve initialize osc date when target is added to mds and
1019 ability to resend too big getattr request is client isn't have info
1024 Description: Prevent inconsistences between linux and lustre mount structures.
1025 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1026 Make the sleep interruptible.
1028 Severity : enhancement
1030 Description: Communicate OST degraded/readonly state via statfs to MDS
1031 Details : Flags in the statfs returned from OSTs indicate whether the
1032 OST is in a degraded RAID state, or if the filesystem has
1033 turned read-only after a filesystem error is detected.
1038 Description: don't panic if EPROTO was hit when reading symlink
1039 Details : correctly handling request reference in error cases.
1044 Description: open sometimes returns ENOENT instead of EACCES
1045 Details : checking permission should be part of open part of mds_open, not
1046 lookup part. so server should be set DISP_OPEN_OPEN disposition
1047 before starting permission check.
1048 Also not need revalidate dentry if client already have LOOKUP lock.
1052 Description: enable client interface failover
1053 Frequency : on servers with multiple network interfaces
1054 Details : When a child reconnects from another NID, properly update export
1055 nid hash position and ldlm reverse import.
1057 Severity : enhancement
1059 Description: implemented direct I/O with arbitrary (nonaligned) memory
1060 addresses and file offsets.
1062 Severity : enhancement
1064 Description: added more recovery timeout options.
1066 Severity : enhancement
1068 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
1073 Description: Avoid deadlock for local client writes
1074 Frequency : only on systems with clients writing to an OST on the same node
1075 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1076 memory freeing context. This allows OST threads to set the
1077 PF_MEMALLOC flag on task structures in order to allocate memory
1078 from reserved pools and complete IO.
1079 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1080 so that the OST threads generate memory pressure and allow
1081 inactive pages to be reclaimed.
1086 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1087 Details : move ldlm namespace creation in setup phase to avoid grab
1088 _lprocfs_lock with cli_sem held.
1092 Frequency : only during format of test systems
1093 Description: Unable to run several mkfs.lustre on loop devices at the same time
1094 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1095 formatting. The solution is to proper handle the error.
1097 Severity : enhancement
1099 Description: implement async create (obd_async_create) method for osc, to avoid
1100 too long waiting new ost objects with holding ldlm lock.
1104 Frequency : occasionally during network problems
1105 Description: client not allowed to reconnect to OST because of active request
1106 Details : abort bulk requests received by the OST once the client has timed
1107 out since the client will resend the request anyway. The client
1108 also now retries to reconnect to the same server if a connect
1109 request failed with EBUSY or -EAGAIN.
1112 Frequency : rare, if used wide striped file and one ost in down.
1114 Descriptoin: don't return error if we created a subset of objects for file.
1115 Details : lov_update_create_set() uses set->set_success as index for created
1116 objects, so if some requests failed, they will have hole at end of
1117 array and we can use qos_shrink_lsm for allocate correct lsm.
1121 Description: Slow stale export processing during normal start up
1122 Details : The global mgc lock prevents OST setup to be run in parallel.
1123 Replace the global lock with a per-config_llog_data semaphore.
1127 Description: Out or order replies might be lost on replay
1128 Details : In ptlrpc_retain_replayable_request if we cannot find retained
1129 request with tid smaller then one currently being added, add it
1130 to the start, not end of the list.
1134 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
1135 Details : add cond_resched() calls to avoid hogging the cpu for too long
1136 in the hash code. Make also lustre_hash_for_each_empty() more
1139 Severity : enhancement
1141 Description: Performance improvements for debug messages with D_RPCTRACE,
1142 D_LDLM, D_QUOTA options.
1145 Frequency : only with NFS export
1147 Description: (lov_merge.c:74:lov_merge_lvb())
1148 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
1149 Details : Fix a race in the nfs export code by populating inode
1150 info while the new inode is still locked
1152 Severity : enhancement
1154 Description: add a new file in procfs called force_lbug. Writting to this
1155 file triggers a LBUG. Only for test purpose.
1159 Description: OOM killer causes node hang
1160 Details : really interrupt the sleep in osc_enter_cache on signals
1164 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1165 Details : fix race during quota release on the slave.
1167 Severity : enhancement
1169 Description: smaller hash bucket sizes, cleanups
1170 Details : increase hash table sizes and enabled rehashing for pools, quota,
1171 uuid, nid & per-nid stats.
1173 Severity : enhancement
1175 Description: Add ldiskfs maxdirsize mount option
1176 Details : add max_dir size mount option
1180 Description: panic in ll_statahead_thread
1181 Details : prevent parent thread to be killed before its child
1184 Frequency : only with 16TB device
1186 Description: unable to perform "mount -t lustre" of 16TB OST device
1187 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
1191 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
1192 Details : unregistering should be zero if no RPC inflight.
1196 Description: hyperion: Oops during metabench
1197 Details : Correct the refcount of lov_request_set
1199 Severity : enhancement
1201 Description: Add mptlinux and nxge drivers to Lustre builds
1203 Severity : enhancement
1205 Description: Fix watchdog timer message to be more clear
1206 Details : Make watchdog timer messages more clear and descriptive.
1210 Description: LNET soft lockups in socknal_cd thread
1211 Details : don't hog CPU for active-connecting if another connd is
1212 accepting connecting-requst from the same peer
1216 Description: recovery-small test_17 hang
1217 Details : Land several AT improvements & fixes.
1221 Description: MDS panic and hanging client processes
1222 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
1226 Description: OSS stuck in recovery.
1227 Details : fix race during recovery. class_unlink_export,
1228 class_set_export_delayed and target_queue_last_replay_reply
1229 may race while increasing/decreasing obd_recoverable_clients
1230 and obd_delayed_clients, causing recovery to wait forever.
1232 Severity : enhancement
1234 Description: add cascading_rw.c to lustre/tests
1238 Description: filter_last_id() NULL deref
1239 Details : lprocfs_filter_rd_last_id() should check for the fully
1240 setup obd device, before proceeding further.
1242 Severity : enhancement
1244 Description: Loadgen improvements
1245 Details : stacksize and locking fixes for loadgen
1249 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
1250 Details : The atomic_read() allowing the atomic_inc() are not covered
1251 by a lock. Thus they may safely race and trip this CERROR()
1252 unless we add in a small fudge factor (+1).
1254 Severity : enhancement
1256 Description: shrink_slab: nr=-9223362083340912175
1257 Details : fix spurious message from shrink_slab reporing negative nr
1261 Description: Quiet bogus previously committed transno error
1262 Details : suppress the "server went back in time" error message which
1263 is always printed even in the common case after a client eviction
1265 Severity : enhancement
1267 Description: Parallel statfs() calls result in client eviction
1268 Details : cache statfs data for 1s.
1272 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
1274 Details : fix serveral issues in pinger code causing clients not to ping
1275 servers for too long, resulting in evictions.
1279 Description: e2fsck should warn when MMP update interval is extended
1280 Details : print mmp_check_interval and make it possible to abort mount
1281 operation in case it takes too long.
1285 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
1286 Details : fix bug in the RHEL5's jbd2 callback patch.
1290 Description: drop number of active requests when queued for recovery
1291 Details : Now that we take a reference on the original request instead of
1292 making a copy of it for recovery. We need to drop the number of
1293 active requests or the queued requests will prevent all request
1294 processing when they exceed (srv->srv_threads_running - 1).
1296 Severity : enhancement
1298 Description: refuse to invalidate operational quota files when they are in use
1299 Details : an attempt to invalidate operational quota files on the quota
1300 master is not actually permitted by VFS (returning -EPERM), but we
1301 should not depend on that and should return the error earlier.
1305 Description: Applications stuck in jbd2_log_wait_commit during exit
1306 Details : fix deadlock between kjournald2 trying to acquire the page lock
1307 owned by an ost_io thread waiting for journal commit.
1309 -------------------------------------------------------------------------------
1311 2009-10-16 Sun Microsystems, Inc.
1313 * Support for kernels:
1314 2.6.16.60-0.42.4 (SLES 10),
1315 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1316 2.6.18-128.7.1.el5 (RHEL 5),
1317 2.6.18-128.7.1.el5 (OEL 5),
1318 * Client support for unpatched kernels:
1319 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1320 2.6.16 - 2.6.27 vanilla (kernel.org)
1321 * Recommended e2fsprogs version: 1.41.6.sun1
1322 * ext4 support for RHEL5 is experimental and thus should not be
1325 Severity : enhancement
1327 Description: Add OEL5 support.
1329 Severity : enhancement
1331 Description: Update kernel to SLES11 2.6.27.29-0.1.
1335 Description: File checksum failures with OST read cache on
1336 Details : Disable page poisoning when the bulk transfer has to be aborted
1337 because the client got evicted.
1341 Description: Don't allow make backward step on assiging osc next id.
1342 Details : race between allocation next id and ll_sync thread can be cause
1343 of set wrong osc next id and can be kill valid ost objects.
1345 Severity : enhancement
1347 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1349 Severity : enhancement
1351 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1355 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1359 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1363 Description: Error handling in osc_statfs_interpret() has been improved.
1364 Details : Check in osc_statfs_interpret() for EBADR.
1368 Description: Do not send statfs() requests to OSTs disabled by administrator.
1369 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1373 Description: Do not update ctime for the deleted inode.
1374 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1378 Description: Increase of the size of the LDLM resource hash.
1379 Details : Bump up RES_HASH_BITS=12.
1383 Description: correctly send lsm on open replay
1384 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1389 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1390 Details : filter_destroy() does not hold the DLM lock over the whole
1391 operation. If the DLM lock is dropped, filter_commitrw() can go
1392 through, causing the deadlock between page lock and i_mutex.
1393 The i_alloc_sem should also be hold in filter_destroy() while
1394 truncating the file.
1398 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1399 Details : ldiskfs_truncate calls grab_cache_page which may start page
1400 allocation under an open transaction. This may lead to
1401 calling prune_icache with consequent lustre reentrance.
1404 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1405 still up and when the OST pool feature is used
1407 Description: interop testing got LBUG when run dd with OST pool
1408 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1409 Details : down/upgrading the MDS to a version that doesn't/does support OST
1410 pool can cause clients to crash because the lsm has changed
1415 Description: missing tree_status on 1.8.1 RPM build
1416 Details : make rpms failed due because the tree_status file is missing.
1420 Description: continuing LustreError "mds adjust qunit failed!"
1421 Details : don't print message on the console when ->adjust_qunit fails.
1425 Description: don't increase ldlm timeout if previous client was evicted
1426 Details : if a client doesn't respond to a blocking callback within the
1427 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1428 when the lock is next granted.
1432 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1433 affects recovery negatively.
1434 Details : make sure all exports have been properly destroyed by the zombie
1435 thread processed before stopping the target.
1439 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1440 Details : disable write barrier for ext4/SLES11.
1444 Description: Kernel panic - not syncing: Out of memory and no killable
1445 processes... on OSS when iozone
1446 Details : fix memory leak in the journal checksum patch.
1450 Description: group quota "too many blocks" OSS crashes
1451 Details : we should keep the same uid/gid for lquota_chkquota() and
1452 lquota_pending_commit()
1456 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1457 Details : don't LBUG on release quota error. Just a workaround until the
1458 problem is understood.
1460 ------------------------------------------------------------------------------
1461 2009-07-31 Sun Microsystems, Inc.
1463 * Support for kernels:
1464 2.6.16.60-0.39.3 (SLES 10),
1465 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1466 2.6.18-128.1.14.el5 (RHEL 5),
1467 * Client support for unpatched kernels:
1468 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1469 2.6.16 - 2.6.27 vanilla (kernel.org)
1470 * Recommended e2fsprogs version: 1.41.6.sun1
1471 * File join has been disabled in this release, refer to Bugzilla 16929.
1472 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1473 of Lustre filesystem with 4K stack may cause a stack overflow. For
1474 more information, please refer to bugzilla 17630.
1475 * ext4 support for RHEL5 is experimental and thus should not be
1478 Severity : enhancement
1480 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1483 Frequency : with 1.8 server and 1.6 clients
1485 Description: correctly shrink reply for avoid send too big message to client.
1486 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1487 some problems with sending this reply to 1.6 client.
1491 Description: Repeated atomic allocation failures.
1492 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1493 to generate memory pressure and allow reclaiming of inactive pages.
1494 At the same time, do not allow to exhaust emergency pools.
1495 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1497 Severity : enhancement
1498 Bugzilla : 19846, 18289
1499 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1501 Severity : enhancement
1502 Bugzilla : 19625, 16893, 18668, 19848
1503 Description: Add support for SLES11 2.6.27.23-0.1.
1505 Severity : enhancement
1507 Description: Update client support to vanila kernels up to 2.6.27.
1509 Severity : enhancement
1511 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1513 Severity : enhancement
1515 Description: Compile with -Werror by default for i686 and x86_64.
1519 Description: resolve race between obd_disconnect and class_disconnect_exports
1520 Details : if obd_disconnect will be called to already disconnected export he
1521 forget release one reference and osc module can't unloaded.
1523 Severity : enhancement
1525 Description: move AT tunable parameters for more consistent usage
1526 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1530 Descriptoin: correctly skip time estimate if in recovery
1531 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1535 Descriptoin: OSS DeadLock
1536 Details : Use trylock to prevent deadlock when shrink icache.
1538 Severity : enhancement
1540 Description: Allow tuning service thread via /proc
1541 Details : For each service a new
1542 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1543 created that can be used to set min/max thread counts, and get the
1544 current number of running threads.
1546 Severity : enhancement
1548 Description: Add state history info file, enhance import info file
1549 Details : Track import connection state changes in a new osc/mdc proc file;
1550 add overview-type data to the osc/mdc import proc file.
1554 Description: Reduce small size read RPC
1555 Details : Set read-ahead limite for every file and only do read-ahead when
1556 available read-ahead pages are bigger than 1M to avoid small size
1561 Description: free_entry erroneously used groups_free instead of put_group_info
1563 Severity : enhancement
1565 Description: Make read-ahead stripe size aligned.
1567 Severity : enhancement
1569 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1572 Frequency : rare, connect and disconnect target at same time
1574 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1575 Details : don't call obd_disconnect under lov_lock. this long time
1576 operation and can block ptlrpcd which answer to connect request.
1579 Frequency : start MDS on uncleanly shutdowned MDS device
1581 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1582 Details : stay in waiting mds<>ost recovery finished produce random bugs
1583 due race between two ll_sync thread for one lov target. send
1584 ACTIVATE event only if connect realy finished and import have
1588 Frequency : start MDS on uncleanly shutdowned MDS device
1590 Descriptoin: aborting recovery hang on MDS
1591 Details : don't throttle destroy RPCs for the MDT.
1595 Description: Slow reads beyond 8Tb offsets.
1596 Details : Page index integer overflow in ll_read_ahead_page
1600 Description: Soft lockup on OSS after MDS failover
1601 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1602 As a conseqence, the patch from bug 18224 is not operational
1603 and the MDS export cannot be reused on the OSTs until it gets
1607 Frequency : rare, only if using MMP with Linux RAID
1609 Description: MMP doesn't work with Linux RAID
1610 Details : While using HA for Lustre servers with Linux RAID, it is possible
1611 that MMP will not detect multiple mounts. To make this work we
1612 need to unplug the device queue in RAID when the MMP block is being
1613 written. Also while reading the MMP block, we should read it from
1614 disk and not the cached one.
1617 Frequency : rare, during recovery
1619 Description: Assertion failure in ldlm_lock_put
1620 Details : Do not put cancelled locks into replay list, hold references on
1621 locks in replay list
1625 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1626 Details : create_count always drops to the min value (=32) because grow_count
1627 is being changed before the precreate RPC completes.
1630 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1633 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1634 Details : add proper locking for creation of jbd_4k slab cache
1638 Description: MMP check in ext3_remount() fails without displaying any error
1639 Details : When multiple mount protection fails during remount, proper error
1644 Description: Rare Client crash on resend if the file was deleted.
1645 Details : When file is opened, but open reply is lost and file is
1646 subsequently deleted before resend, resend processing logic
1647 breaks trying to open the file again, should not try to open.
1651 Description: add check for >8TB ldiskfs filesystems
1652 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1653 Don't allow >8TB ldiskfs filesystems to be mounted without
1654 force_over_8tb mount option
1658 Description: Client locked up when running multiple instances of an app. on
1659 multiple mount points
1660 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1661 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1665 Description: Cannot acces an NFS-mounted Lustre filesystem
1666 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1667 from a Lustre-client exporting the Lustre filesystem via NFS.
1671 Description: panic in ll_statahead_thread
1672 Details : grab dentry reference in parent process.
1675 -------------------------------------------------------------------------------
1677 tbd Sun Microsystems, Inc.
1679 * Support for kernels:
1680 2.6.16.60-0.31 (SLES 10),
1681 2.6.18-128.1.6.el5 (RHEL 5),
1682 2.6.22.14 vanilla (kernel.org)
1683 * Client support for unpatched kernels:
1684 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1685 2.6.16 - 2.6.22 vanilla (kernel.org)
1686 * Recommended e2fsprogs version: 1.40.11-sun1
1687 * File join has been disabled, refer to Bugzilla 16929.
1688 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1689 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1690 Lustre filesystem with 4K stack may cause a stack overflow. For more
1691 information, please refer to bugzilla 17630.
1693 Severity : enhancement
1695 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1697 Severity : enhancement
1699 Description: Update OFED release to 1.4.1 RC4
1701 Severity : major, only with big OST
1703 Description: Very poor metadata performance on Infiniband lustre configuration
1704 Details : OST object precreation becomes very slow on big OSTs. This is due
1705 to the ialloc patch spending too much time scanning groups.
1708 Frequency : during recovery
1710 Description: don't mix llog inodes with normal.
1711 Details : allocate inodes for log in last inode group
1716 Description: fix lqs' reference which won't be put in some situations
1717 Details : This patch fixes:
1718 1. In quota_check_common(), this function will check quota
1719 for user and group, but only send one return via "pending".
1720 In most cases, the pendings should be same. But that is not
1722 2. If quotaoff runs between lquota_chkquota() and
1723 lquota_pending_commit(), the same thing will happen too.
1724 That is why it comes:
1725 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1728 -------------------------------------------------------------------------------
1730 2008-12-31 Sun Microsystems, Inc.
1732 * Support for kernels:
1733 2.6.5-7.314 (SLES 9),
1734 2.6.9-67.0.22.EL (RHEL 4),
1735 2.6.16.60-0.31 (SLES 10),
1736 2.6.18-92.1.17.el5 (RHEL 5),
1737 2.6.22.14 vanilla (kernel.org)
1738 * Client support for unpatched kernels:
1739 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1740 2.6.16 - 2.6.22 vanilla (kernel.org)
1741 * Client support for unpatched kernels:
1742 we do not recommend using patchless RHEL4 clients with kernels
1743 prior to 2.6.9-55EL (RHEL4U5).
1744 * Recommended e2fsprogs version: 1.40.11-sun1
1745 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1746 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1747 removed cwd "./" (refer to Bugzilla 14399).
1748 * A new quota file format has been introduced in 1.6.5.
1749 The format conversion from prior releases is handled transparently,
1750 but releases older than 1.4.12/1.6.5 don't understand this new
1751 format. The automatic format conversion can be avoided by running
1752 the following command on the MDS:
1753 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1754 For more information, please refer to bugzilla 13904.
1755 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1756 The format conversion from prior releases is handled transparently,
1757 but releases older than 1.6.6/1.8.0 don't understand this new
1758 format. The automatic format conversion can be avoided by running
1759 the following commands on the MDS and OSS servers (for
1760 pre 1.4.12-1.6.5 quota files):
1761 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1762 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1763 or (for 1.4.12/1.6.5 quota files)
1764 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1765 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1766 For more information, please refer to bugzilla 13904.
1767 * Output of lfs quota has been made less detailed by default,
1768 old (verbose) output can be obtained by using -v option.
1769 * File join has been disabled in this release, refer to Bugzilla 16929.
1770 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1771 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1772 Lustre filesystem with 4K stack may cause a stack overflow. For more
1773 information, please refer to bugzilla 17630.
1775 Severity : enhancement
1777 Description: Caching OSS
1778 Details : introduce data caching on the OSS. The OSS now relies on the linux
1779 kernel page cache to keep recently accessed data in memory.
1780 It is worth noting that all write requests are still flushed
1781 synchronously as in lustre 1.6.
1783 Severity : enhancement
1785 Description: Version based recovery
1786 Details : introduce finer grained recovery able to detect transaction
1787 dependencies and can deal with transaction gaps caused by clients
1788 failing at the same time as the server.
1790 Severity : enhancement
1792 Description: Enable adaptive timeouts by default
1793 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1794 managed dynamically based on server load and should not need
1795 to be tuned manually based on cluster size. This allows Lustre
1796 to work under a wider variety of system sizes and loads, without
1797 unnecessarily causing lengthy recovery times.
1799 Severity : enhancement
1801 Description: Add OST Pools support
1802 Details : File striping can now be set to use an arbitrary pool of OSTs
1804 Severity : enhancement
1806 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1807 Details : allow skip disconnected ost for send statfs request and hide error
1811 Frequency : rare, on llog test 6
1813 Descriptoin: don't allow connect to already connected import
1814 Details : allowing connect to already connected import is hide connecting problem.
1817 Frequency : rare, on failed llog setup
1819 Descriptoin: don't leak obd reference on failed llog setup
1820 Details : for failed llog setup - mgc forget call class_destroy_import for
1821 client import, move destroy import to more generic place.
1826 Descriptoin: allow kill process which wait statahead result
1827 Details : for some reasons 'ls' can stick in waiting result from statahead,
1828 in this case need way for kill this process.
1831 Frequency : rare, at shutdown
1833 Descriptoin: panic at umount
1834 Details : llap_shrinker can be raced with killing super block from list and
1835 this produce panic with access to already freeded pointer
1840 Descriptoin: don't lose wakeup for imp_recovery_waitq
1841 Details : recover_import_no_retry or invalidate_import and import_close can
1842 both sleep on imp_recovery_waitq, but we was send only one wakeup
1848 Descriptoin: panic in mds_open
1849 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1854 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1856 Details : release lock refernce only after add page to pages list.
1859 Frequency : always with long access acl
1861 Descriptoin: mds can't pack reply with long acl.
1862 Details : mds don't control size of acl but they limited by reint/getattr
1867 Frequency : on remount
1869 Description: external journal device not working after the remount
1870 Details : clear dev_rdonly flag for external journal devices in
1876 Description: shutdown vs evict race
1877 Details : client_disconnect_export vs connect request race.
1878 if client will evicted at this time - we start invalidate
1879 thread without referece to import and import can be freed
1885 Description: shrink LOV EAs before replying
1886 Details : correctly adjust LOV EA buffer for reply.
1891 Description: don't skip ost target if they assigned to file
1892 Details : Drop slow OSCs if we can, but not for requested start idx.
1893 This means "if OSC is slow and it is not the requested
1894 start OST, then it can be skipped, otherwise skip it only
1895 if it is inactive/recovering/out-of-space.
1897 Severity : enhancement
1899 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1902 Frequency : rare, need acl's on inode.
1904 Description: client can't handle ost additional correctly
1905 Details : if ost was added after client connected to mds client can have
1906 hit lnet_try_match_md ... to big messages to wide striped files.
1907 in this case need teach client to handle config events about add
1908 lov target and update client max ea size at that event.
1910 Severity : enhancement
1912 Description: Update to sles9 kernel-2.6.5-7.314.
1914 Severity : enhancement
1916 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1919 Frequency : Create a symlink file with a very long name
1921 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1922 Details : If there is no extra space in the request for early cancels,
1923 ldlm_req_handles_avail() returns 0 instead of a negative value.
1928 Description: mds is deadlocked
1929 Details : in rare cases, inode in catalog can have i_no less than have parent
1930 i_no, this produce wrong order for locking during open, and
1931 parallel unlink can be lock open. this need teach mds_open to grab
1932 locks in resource id order, not at parent -> child order.
1934 Severity : enhancement
1936 Description: Add /proc entry for import status
1937 Details : The mdc, osc, and mgc import directories now have
1938 an import directory that contains useful import data for debugging
1939 connection problems.
1941 Severity : enhancement
1943 Description: Re-disable certain /proc logging
1944 Details : Enable and disable client's offset_stats, extents_stats and
1945 extents_stats_per_process stats logging on the fly.
1948 Frequency : Only on FC kernels 2.6.22+
1950 Description: oops in statahead
1951 Details : Do not drop reference count for the dentry from VFS when lookup,
1952 VFS will do that by itself.
1954 Severity : enhancement
1956 Description: Generic /proc file permissions
1957 Details : Set /Proc file permissions in a more generic way to enable non-
1958 root users operate on some /proc files.
1962 Description: Hitting mdc_commit_close() ASSERTION
1963 Details : Properly handle request reference release in
1964 ll_release_openhandle().
1968 Frequency : only patchless client
1969 Description: add workaround for race between add/remove dentry from hash
1971 Severity : enhancement
1973 Description: Allow OST glimpses to return PW locks
1977 Description: LBUG when llog conf file is full
1978 Details : When llog bitmap is full, ENOSPC should be returned for plain
1983 Description: Prevent import from entering FULL state when server in recovery
1987 Description: service mount cannot take device name with ":"
1988 Details : Only when device name contains ":/" will mount treat it as
1994 Description: replace ptlrpcd with the statahead thread to interpret the async
1995 statahead RPC callback
1999 Frequency : on recovery
2000 Description: I/O failures after umount during fail back
2001 Details : if client reconnected to restarted server we need join to recovery
2002 instead of find server handler is changed and process self eviction
2003 with cancel all locks.
2005 Severity : enhancement
2007 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
2011 Description: Kernel BUG tries to release flock
2012 Details : Lustre does not destroy flock lock before last reference goes
2013 away. So always drop flock locks when client is evicted and
2014 perform unlock regardless of successfulness of speaking to MDS.
2016 Severity : enhancement
2018 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
2020 Severity : enhancement
2022 Description: Upcall on Lustre log has been dumped
2023 Details : Allow for a user mode script to be called once a Lustre log has
2024 been dumped. It passes the filename of the dumped log to the
2025 script, the location of the script can be specified via
2026 /proc/sys/lnet/debug_log_upcall.
2031 Description: avoid messages about idr_remove called for id that is not allocated
2032 Details : Move assigment s_dev for clustered nfs to end of initialization,
2033 for avoid problem with error handling.
2038 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
2039 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
2040 CONN_USED_HASH and this prodice warning when put connection again
2046 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
2047 Details : release reference to stats when client disconnected, not
2048 when export destroyed for avoid races when client destroyed
2049 after main ost export.
2053 Description: more cleanup in mds_lov
2054 Details : add workaround for get valid ost count for avoid warnings about
2055 drop too big messages, not init llog cat under semphore which
2056 can be blocked on reconnect and break normal replay, fix access
2059 Severity : enhancement
2061 Description: Export bytes_read/bytes_write count on OSC/OST.
2065 Description: Early reply size mismatch, MGC loses connection
2066 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
2067 the connect flags are properly negotiated.
2071 Description: Properly propagate oinfo flags from lov to osc for statfs
2072 Details : restore missing copy oi_flags to lov requests.
2076 Description: exports in /proc are broken
2077 Details : recreate /proc entries for clients when they reconnect.
2079 Severity : enhancement
2081 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
2082 : l_getgroups(8), lst(8), routerstat(8)
2083 Details : included man pages for llobdstat(8), llstat(8),
2084 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
2086 Severity : enhancement
2088 Description: Implement lustre ll_show_options method.
2090 Severity : enhancement
2092 Description: Update to SLES9 kernel-2.6.5-7.312.
2094 Severity : enhancement
2096 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
2100 Description: exports in /proc are broken
2101 Details : recreate /proc entries for clients when they reconnect.
2105 Description: don't fail open with -ERANGE
2106 Details : if client connected until mds will be know about real ost count
2107 get LOV EA can be fail because mds not allocate enougth buffer
2112 Description: Resolve device initialization race
2113 Details : Prevent proc handler from accessing devices added to the
2114 obd_devs array but yet be intialized.
2116 Severity : enhancement
2118 Description: configure's --enable-quota should check the
2119 : kernel .config for CONFIG_QUOTA
2120 Details : configure is terminated if --enable-quota is passed but
2121 : no quota support is in kernel
2123 Severity : enhancement
2125 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
2127 Severity : enhancement
2129 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
2133 Frequency : rare, on PPC clients
2134 Description: don't swab ost objects in response about directory, because
2136 Details : bug similar bug 14856, but in different function.
2138 Severity : enhancement
2140 Description: lfs quota tool enhancement
2141 Details : added units specifiers support for setquota, default to
2142 current uid/gid for quota report, short quota stats by
2143 default, nonpositional parameters for setquota, added
2144 llapi_quotactl manual page.
2146 Severity : enhancement
2148 Description: *optional* service tags registration
2149 Details : if the "service tags" package is installed on a Lustre node
2150 When the filesystem is mounted, a local-node service tag will
2151 be created. See http://inventory.sun.com/ for more information
2152 about the Service Tags asset management system.
2156 Description: Client runs out of low memory
2157 Details : Consider only lowmem when counting initial number of llap pages
2160 Frequency : occasional
2162 Description: add refcount for osc callbacks, so avoid panic on shutdown
2164 Severity : enhancement
2166 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
2169 Frequency : testing only
2171 Description: sanity test 65a fails if stripecount of -1 is set
2172 Details : handle -1 striping on filesystem in ll_dirstripe_verify
2175 Frequency : only in unusual configurations
2177 Description: Kernel panic with find ost index.
2178 Details : lov_obd have panic if some OST's have sparse indexes.
2180 Severity : enhancement
2182 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
2185 Frequency : rarely, if filesystem is mounted with -o flock
2187 Description: do not process already freed flock
2188 Details : flock can possibly be freed by another thread before it reaches
2189 to ldlm_flock_completion_ast.
2192 Frequency : rarely, if filesystem is mounted with -o flock
2194 Description: LBUG during stress test
2195 Details : Need properly lock accesses the flock deadlock detection list.
2198 Frequency : rarely, if binaries are being run from Lustre
2200 Description: oops in page fault handler
2201 Details : kernel page fault handler can return two special 'pages' in
2202 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
2205 Frequency : rarely, during shutdown
2207 Description: timeout with invalidate import.
2208 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
2209 should be handled by ptlrpcd. This produce long age waiting and
2210 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
2216 Description: ASSERTION(CheckWriteback(page,cmd)) failed
2217 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
2221 Frequency : only with broken builds/installations
2223 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
2224 Details : just return an error to a user, put a console error message
2226 Severity : enhancement
2228 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
2230 Severity : enhancement
2232 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
2234 Severity : enhancement
2236 Description: enable MGS and MDT services start separately
2237 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
2238 a co-located MGS without starting the MGS, which is a complement
2239 to 'nosvc' mount option.
2242 Frequency : always, on big-endian systems
2244 Description: cleanup in ptlrpc code, related to PPC platform
2245 Details : store magic in native order avoid panic's in recovery on PPC
2246 node and forbid from this error in future. Also fix posibily
2247 of twice swab data. Fix get lov striping to userpace.
2250 Frequency : rarely, if replay get lost on server
2252 Description: server incorrectly drop resent replays lead to recovery failure.
2253 Details : do not drop replay according to msg flags, instead we check the
2254 per-export recovery request queue for duplication of transno.
2257 Frequency : after recovery
2259 Description: precreate to many object's after del orphan.
2260 Details : del orphan st in oscc last_id == next_id and this triger growing
2261 count of precreated objects. Set flag LOW to skip increase count
2262 of precreated objects.
2265 Frequency : rare, on clear nid stats
2267 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
2268 Details : when clean nid stats sometimes try destroy live entry,
2269 and this produce panic in free.
2272 Frequency : occasionally since 1.6.4
2274 Description: Stack overflow during MDS log replay
2275 Details : ease stack pressure by using a thread dealing llog_process.
2278 Frequency : very rare
2280 Description: MDT cannot be unmounted, reporting "Mount still busy"
2281 Details : Mountpoint references were being leaked during open reply
2282 reconstruction after an MDS restart. Drop mountpoint reference
2283 in reconstruct_open() and free dentry reference also.
2288 Description: wait until IO finished before start new when do lock cancel.
2289 Details : VM protocol want old IO finished before start new, in this case
2290 need wait until PG_writeback is cleared until check dirty flag
2291 and call writepages in lock cancel callback.
2296 Description: mds_mfd_close() ASSERTION(rc == 0)
2297 Details : In mds_mfd_close(), we need protect inode's writecount change
2298 within its orphan write semaphore to prevent possible races.
2301 Frequency : rare, on shutdown ost
2303 Description: don't hit live lock with umount ost.
2304 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2305 use shrink_dcache_sb instead.
2308 Frequency : only when echo_client is used
2310 Description: don't panic with use echo_client
2311 Details : echo client pass NULL as client nid pointer and this produce NULL
2312 pointer dereference.
2315 Frequency : Always on 32-bit PowerPC systems
2317 Description: fix build on PPC32
2318 Details : compile code with -m64 flag produce wrong object file for PPC32.
2323 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2324 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2325 by a client, and the data on disk are dangerous and incorrect.
2326 Change them with client eviction.
2328 Severity : enhancement
2330 Description: skiplist implementation simplification
2331 Details : skiplists are used to group compatible locks on granted list
2332 that was implemented as tracking first and last lock of each lock
2333 group the patch changes that to using doubly linked lists
2337 Description: delete compatibility for 32bit qdata
2338 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2339 qunit. That means servers of b1_4 and servers of b1_8 can't be
2340 used together if users want to use quota.
2343 Frequency : only with administrator action
2345 Description: mount failure if config log has invalid conf_param setting
2346 Details : If administrator specified an incorrect configuration parameter
2347 with "lctl conf_param" this would cause an error during future
2348 client mounts. Instead, ignore the bad configuration parameter.
2351 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2353 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2354 Details : If blocks per group is less than blocksize*8, set rest of the
2358 Frequency : Application do stride read on lustre
2360 Description: The read performance will drop a lot if the application does
2362 Details : Because the stride_start_offset are missing in stride read-ahead,
2363 it will cause clients read a lot of unused pages in read-ahead,
2364 then the read-performance drops.
2368 Description: more ldlm soft lockups
2369 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2370 starve other threads from the resource lock for a long time in
2371 case of long waiting queue, so change the debug level from
2372 D_OTHER to the less frequently used D_INFO.
2374 Severity : enhancement
2376 Description: add -gid, -group, -uid, -user options to lfs find
2378 Severity : enhancement
2380 Description: ll_recover_lost_found_objs - recover objects in lost+found
2381 Details : OST corruption and subsequent e2fsck can leave objects in the
2382 lost+found directory. Using the "ll_recover_lost_found_objs"
2383 tool, these objects can be retrieved and data can be salvaged
2384 by using the object ID saved in the fid EA on each object.
2389 Description: this bug _only_ happens when inode quota limitation is very low
2390 (less than 12), so that inode quota unit is 1 at initialization.
2391 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2392 is effective now. So least quota qunit should be 2.
2396 Description: Hung threads in invalidate_inode_pages2_range
2397 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2398 one is completed. As a result, some RPCs are stuck in the queue
2403 Description: Procfs and llog threads access destoryed import sometimes.
2404 Details : Sync the import destoryed process with procfs and llog threads by
2405 the import refcount and semaphore.
2409 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2410 Details : Sort source/child resource pair after updating child resource.
2415 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2416 Details : If insertion of an extent fails, then discard the inode
2417 preallocation and free data blocks else it can lead to duplicate
2422 Description: don't always update ctime in ext3_xattr_set_handle()
2423 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2424 In some cases the ctime should not be updated, for example for
2425 2.0->1.8 compatibility it is necessary to delete an xattr and it
2426 should not update the ctime.
2430 Description: add quota statistics
2431 Details : 1. sort out quota proc entries and proc code.
2432 2. add quota statistics
2437 Description: quotas are not honored with O_DIRECT
2438 Details : all writes with the flag O_DIRECT will use grants which leads to
2439 this problem. Now using OBD_BRW_SYNC to guard this.
2443 Bugzilla : 15713/16362
2444 Description: Assertion in iopen_connect_dentry in 1.6.3
2445 Details : looking up an inode via iopen with the wrong generation number can
2446 populate the dcache with a disconneced dentry while the inode
2447 number is in the process of being reallocated. This causes an
2448 assertion failure in iopen since the inode's dentry list contains
2449 both a connected and disconnected dentry.
2453 Description: assertion failure in ldlm_handle2lock()
2454 Details : fix a race between class_handle_unhash() and class_handle2object()
2455 introduced in lustre 1.6.5 by bug 13622.
2457 Severity : enhancement
2459 Description: superblock lock contention with many SMP cores on one client
2460 Details : several client filesystem locks were highly contended on SMP
2461 NUMA systems with 8 or more cores. Per-CPU datastructures
2462 and more efficient locking implemented to reduce contention.
2467 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2468 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2473 Description: oops in sd_iostats_seq_show()
2474 Details : unloading/reloading the scsi low level driver triggers a kernel
2475 bug when trying to access the sd iostat file.
2480 Description: Kernel panics during QLogic driver reload
2481 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2482 patch, causing memory corruption.
2487 Description: journal_dev option does not work in b1_6
2488 Details : pass mount option during pre-mount.
2490 Severity : enhancement
2492 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2493 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2494 extent information of a file. It can be used to map logical blocks
2495 in a file to physical blocks in the block device.
2498 Frequency : only with adaptive timeout enabled
2500 Description: DEBUG_REQ() bad paging request
2501 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2502 because it can be accessed by reply_in_callback() without the
2506 Frequency : only on Cray X2
2508 Description: X2 build failures
2509 Details : fix build failures on Cray X2.
2513 Description: xid & resent requests
2514 Details : Initialize RPC XID from clock at startup (randomly if clock is
2519 Description: quota recovery deadlock during mds failover
2520 Details : This patch includes att18982, att18236, att18237 in bz14840.
2522 1. fix osts hang when mds does failover with quotaon
2523 2. prevent watchdog storm when osts threads wait for the
2528 Description: kernel panic on racer
2529 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2531 Severity : enhancement
2533 Description: Add lustre_start utility to start or stop multiple Lustre servers
2538 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2539 Details : In case of memory pressure, list_del() can be called twice on
2540 req->rq_history_list, causing a kernel oops.
2544 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2545 Details : fix stack overflow in the distributed lock manager by defering
2546 export eviction after a failed ast to the elt thread instead of
2547 handling it in the dlm interpret routine.
2549 Severity : enhancement
2551 Description: More exported tunables for mballoc
2552 Details : Add support for tunable preallocation window and new tunables for
2553 large/small requests
2557 Description: Detect corruption of block bitmap and checking for preallocations
2558 Details : Checks validity of on-disk block bitmap. Also it does better
2559 checking of number of applied preallocations. When corruption is
2560 found, it turns filesystem readonly to prevent further corruptions.
2564 Frequency : only for big-endian servers
2565 Description: Check for big-endian system while mounting fs with extents feature
2566 Details : Mounting a filesystem with extents feature will fail on big-endian
2567 systems since ext3-based ldiskfs is not supported on big-endian
2568 systems. Can be over-ridden with "bigendian_extents" mount option.
2572 Description: Excessive recovery window
2573 Details : With AT enabled, the recovery window can be excessively long (6000+
2574 seconds). To address this problem, we no longer use
2575 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2576 timeout no longer depends on the service time, it is set to
2577 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2578 time via pb_service_time.
2581 Descriptoin: Don't sync journal after every i/o
2582 Details : Implement write RPC replay to allow server replies for write RPCs
2583 before data is on disk. However, this feature is disabled by
2584 default since some issues leading to data corruptions have been
2585 found during recovery (e.g. bug 19128). This feature can be enabled
2586 by running the following command on the OSSs:
2587 lctl set_param obdfilter.*.sync_journal=0
2591 Description: Watchdog triggered on MDS failover
2592 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2593 the OSTs know that the MDS "UUID" can be reused for the same export
2594 from a different NID, so we do not need to wait for the export to
2599 Descriptoin: Lustre detected file system corruption with inode out of bounds
2600 Details : don't update i_size on MDS_CLOSE for directories. This causes
2601 directory corruptions on the MDT.
2603 -------------------------------------------------------------------------------
2605 2008-05-26 Sun Microsystems, Inc.
2607 * Support for kernels:
2608 2.6.5-7.311 (SLES 9),
2609 2.6.9-67.0.7.EL (RHEL 4),
2610 2.6.16.54-0.2.5 (SLES 10),
2611 2.6.18-53.1.14.el5 (RHEL 5),
2612 2.6.22.14 vanilla (kernel.org)
2613 * Client support for unpatched kernels:
2614 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2615 2.6.16 - 2.6.22 vanilla (kernel.org)
2616 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2617 we do not recommend using patchless RHEL4 clients with kernels
2618 prior to 2.6.9-55EL (RHEL4U5).
2619 * Recommended e2fsprogs version: 1.40.7-sun1
2620 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2621 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2622 removed cwd "./" (refer to Bugzilla 14399).
2623 * A new quota file format has been introduced in 1.6.5.
2624 The format conversion from prior releases is handled transparently,
2625 but releases older than 1.4.12/1.6.5 will not understand this new
2626 format. The automatic format conversion can be avoided by running
2627 the following command on the MDS before upgrading:
2628 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2629 For more information, please refer to bugzilla 13904.
2633 Description: quota performance fix
2634 Details : quota data is written in journalled mode instead of ordered to
2635 increase performance
2639 Description: lfs support for human-readable quota grace time strings
2640 Details : lfs setquota -t and lfs quota -t represent quota grace times
2641 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2644 Frequency : always with o2ib 1.3 and sles10
2646 Description: fix build with SLES10 and o2ib v3.
2647 Details : sles10 uses diffrent name for Module.symver file but configure
2648 assume this file has same name on RHEL/SLES/vanila kernels.
2651 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2653 Description: MDS may lose file striping (and hence file data) in some cases
2654 Details : If there are additional extended attributes stored on the MDS,
2655 in particular ACLs, SELinux, or user attributes (if user_xattr
2656 is specified for the client mount options) then there is a risk
2657 of attribute loss. Additionally, the Lustre file striping
2658 needs to be larger than default (e.g. striped over all OSTs),
2659 and an additional attribute must be stored initially in the
2660 inode and then increase in size enough to be moved to the
2661 external attribute block (e.g. ACL growing in size) for file
2664 Severity : enhancement
2666 Description: add message levels for liblustreapi
2671 Description: MDT cannot be unmounted, reporting "Mount still busy"
2672 Details : Mountpoint references were being leaked during open reply
2673 reconstruction after an MDS restart. Drop mountpoint reference
2674 in reconstruct_open() and free dentry reference also.
2679 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2680 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2681 will result in an -EAGAIN and in the exit call path we call
2682 alloc_rr() which will with increasing aggressiveness attempt to
2683 aquire precreated objects on the minimum number of required OSCs.
2687 Description: Use old size assignment to avoid deadlock
2688 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2689 the scheduling while holding a spinlock. We do not need locking
2690 for size in ll_update_inode() because size is only updated from
2691 the MDS for directories or files without objects, so there is no
2692 other place to do the update, and concurrent access to such inodes
2693 are protected by the inode lock.
2697 Description: Use __u64 instead of int for valid bits
2701 Description: resolve "_IOWR redefined" build error on SLES10
2705 Description: dump the memory debugging after all modules are unloaded to
2706 suppress false negative in conf_sanity test 39
2710 Description: the recovery timer never expires
2711 Details : for new client connect request, the recovery timer should not be
2712 reset, otherwise recovery timer will never expired, if the old
2713 client never come. Only old client connect and first connection
2714 req should trigger recovery timer reset.
2718 Description: the min numbers of lproc stats are wrong
2719 Details : adding a new constant LC_MIN_INIT and use it for initialization
2723 Frequency : always with interactive lfs
2725 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2728 Frequency : with multiple concurrent readdir processes in same directory
2729 Bugzilla : 15406, 15169, 15175
2730 Description: misc fixes for directory readahead.
2731 Details : prevent previous statahead async RPC callback from processing the
2732 current "statahead_info", race condition between async RPC callback
2733 add dentry into dentry hash table and "ls" thread revalidate such
2734 dentry, statahead his/miss control for hidden items, and so on.
2736 Severity : enhancement
2738 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2742 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2743 Details : several u32 variables declarations are replaced with u64 declarations
2748 Description: fix SLES kernel versioning
2749 Details : the kernel version for our SLES 10 kernel did not include a "-"
2750 before the "smp" at the end. while this was not a problem in
2751 general, it did mean that software trying to use the kernel
2752 version to try to detect a vendor specific kernel would fail.
2753 this was most evident by the OFED build scripts.
2758 Description: Don't update lov_desc members until making sure they are valid
2759 Details : When updating lov_desc members via proc fs, need fix their
2760 validities before doing the real update.
2763 Frequency : very rare
2765 Description: don't put request into delay list while invalidate in flight.
2766 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2767 import in flight. this produce timeout for invalidate and sometimes
2768 can cause stale data.
2770 Severity : enhancement
2772 Description: Update kernel to SLES9 2.6.5-7.311.
2774 Severity : enhancement
2776 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2781 Frequency : on PPC only
2782 Description: not convert ost objects for directory because it's not exist.
2783 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2785 Severity : enhancement
2787 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2788 __u64 as usingied long long type.
2791 Frequency : rare, on shutdown
2793 Description: race process ast vs remove callback
2794 Details : removing callback before disconnect import open race with
2795 processing callback.
2797 Severity : enhancement
2799 Description: Update kernel to SLES9 2.6.5-7.311.
2801 Severity : enhancement
2803 Description: Files open for execute are not marked busy on SLES10
2804 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2806 Severity : enhancement
2808 Description: Add server support for vanilla-2.6.22.14.
2811 Frequency : occasional
2813 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2814 Details : If an OST is down the MDS will hang indefinitely in
2815 obd_statfs_rqset() waiting for the statfs data. While for
2816 MDS QOS usage of statfs, it should not stuck in waiting.
2818 Severity : enhancement
2820 Description: Disable adaptive timeouts by default
2823 Frequency : on network error
2825 Description: panic with double free request if network error
2826 Details : mdc_finish_enqueue is finish request if any network error occuring,
2827 but it's true only for synchronus enqueue, for async enqueue
2828 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2832 Frequency : rare, on recovery
2834 Description: read procfs can produce deadlock in some situation
2835 Details : Holding lprocfs lock with send rpc can produce block for destroy
2836 obd objects and this also block reconnect with -EALREADY.
2837 This isn't fix all lprocfs bugs - but make it rare.
2839 Severity : enhancement
2841 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2844 Frequency : frequent on X2 node
2846 Description: mdc_set_open_replay_data LBUG
2847 Details : Set replay data for requests that are eligible for replay.
2852 Description: lustre_mgs: operation 101 on unconnected MGS
2853 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2854 MGC, and late on MGC cannot successfully connect to MGS and a lot
2855 of the error messages complaining that MGS is not connected.
2858 Frequency : rare, depends on device drivers and load
2860 Description: MDS or OSS nodes crash due to stack overflow
2861 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2862 In some cases, in conjunction with device drivers that use a lot
2863 of stack, the MDS (or possibly OSS) service threads could overflow
2864 the stack. One change which was identified to consume additional
2865 stack has been reworked to avoid the extra stack usage.
2867 Severity : enhancement
2869 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2871 Severity : enhancement
2873 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2875 Severity : enhancement
2877 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2879 Serverity : enhancement
2881 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2884 Frequency : rare on shutdown OST
2886 Description: Don't allow skipping OSTs if index has been specified.
2887 Details : Don't allow skipping OSTs if index has been specified, make locking
2888 in internal create lots better.
2893 Description: ASSERTION(!PageDirty(page)) failed
2894 Details : Wrong check could lead to an assertion failure under specific
2900 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2901 Details : access to bitfield in structure is always rounded to long
2902 and this produce problem with not atomic change any bit.
2907 Description: Lustre make rpms failed.
2908 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2909 builds Lustre rpms from tarball.
2911 Severity : enhancement
2913 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2916 Frequency : rare on shutdown OST
2918 Description: If llog cancel was not send before clean_exports phase, this can
2919 produce deadlock in llog code.
2920 Details : If llog thread has last reference to obd and call class_import_put
2921 this produce deadlock because llog_cleanup_commit_master wait when
2922 last llog_commit_thread exited, but this never success because was
2923 called from llog_commit_thread.
2926 Frequency : only if OST index is skipped
2928 Description: NULL lov_tgts causing MDS oops
2929 Details : more safe checks for NULL lov_tgts for avoid oops.
2931 Severity : enhancement
2933 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2935 Severity : enhancement
2937 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2942 Description: make mgs_setparam() handle fsname containing dash
2943 Details : fsname containing a dash does not work with lctl conf_param
2945 Severity : enhancement
2947 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2950 Frequency : rare, in recovery and (or) destroy lovobjid file.
2952 Description: rewrite lov objid code.
2953 Details : Cleanup for lov objid code, remove scability problems and wrong
2954 locking. Fix sending last_id into ost.
2956 Severity : enhancement
2958 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2960 Severity : enhancement
2962 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2963 Details : Use d_move_locked instead of __d_move.
2966 Frequency : rare, at shutdown
2968 Description: access already free / zero obd_namespace.
2969 Details : if client_disconnect_export was called without force flag set,
2970 and exist connect request in flight, this can produce access to
2971 NULL pointer (or already free pointer) when connect_interpret
2972 store ocd flags in obd_namespace.
2975 Frequency : only at startup
2977 Description: not alloc memory with spinlock held.
2978 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2979 if any spinlock held.
2984 Description: lfs find does not continue on file error
2985 Details : Continue other files processing when a file/dir is absent.
2989 Description: Inconsistent usage of lustre_pack_reply()
2990 Details : Standardize the usage of lustre_pack_reply() such that it
2991 always generate a CERROR on failure.
2994 Frequency : very rare
2996 Description: Fix replay if there is an un-replied request and open
2997 Details : In some cases, older replay request will revert the
2998 mcd->mcd_last_xid on MDS which is used to record the client's
2999 latest sent request.
3001 Severity : enhancement
3003 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
3005 Severity : enhancement
3007 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
3009 Severity : enhancement
3011 Description: Update to SLES9 kernel-2.6.5-7.287.3.
3013 Severity : enhancement
3015 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
3017 Severity : enhancement
3019 Description: Build SLES10 patchless client fails
3020 Details : The configure was broken by run ./configure with
3021 --with-linux-obj=.... argument for patchless client. When the
3022 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
3023 can't search header adequately. Use absolute path such as
3024 -I($LINUX)/include instead.
3026 Severity : enhancement
3028 Description: Lustre Page Accounting
3029 Details : New macros for page alloc and free which enable accounting
3030 of page allocation of Lustre. Use percpu counters to store memory
3031 and page statistics.
3034 Frequency : only if debugging is disabled
3036 Description: LASSERT_{REQ,REP}SWAB macros are buggy
3037 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
3038 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
3039 these macros and replace them with their definitions instead.
3044 Description: interrupt oig_wait produce painc on resend.
3045 Details : brw_redo_request can be used for resend requests from ptlrpcd and
3046 private set, and this produce situation when rq_ptlrpcd_data not
3047 copyed to new allocated request and triggered LBUG on assert
3048 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
3049 ptlrpcd set if request is changed and can be safety changed to use
3052 Severity : enhancement
3054 Description: organize the server-side client stats on per-nid basis
3055 Details : Change the structure of stats under obdfilter and mds to
3063 The "uuid"s file would list the uuids of _active_ exports.
3064 And the clear entry is to clear all stats and stale nids.
3069 Description: Processes looping in ll_readdir() on Lustre clients finally causing
3070 a full node pseudo-hang
3071 Details : Concurrent access to the same directory from multiple clients with
3072 intensive file creation/removal can cause a client node to spin in
3073 ll_readdir(). i_version must be increased every time the lock is
3074 cancelled to ensure a revalidate is done.
3079 Description: touch file failed when fs is not full
3080 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
3081 otherwise we can get ENOSP while fs is not full.
3084 Frequency : only for Cray XT3
3085 Bugzilla : 12829/13455
3086 Description: Changing primary group doesn't change the group lustre assigns to
3088 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
3089 overridden with the primary group provided by the group upcall,
3090 whereas the supplied fsgid can be trusted if it is in the list of
3091 supplementary groups returned by the group upcall.
3093 Severity : enhancement
3095 Description: Root Squash Functionality
3096 Details : Implementation of NFS-like root squash capability. Specifically,
3097 don't allow someone with root access on a client node to be able
3098 to manipulate files owned by root on a server node.
3100 Severity : enhancement
3102 Description: Slow trucate/writes to huge files at high offsets.
3103 Details : Directly associate cached pages to lock that protect those pages,
3104 this allows us to quickly find what pages to write and remove
3105 once lock callback is received.
3110 Description: Too many locks accumulating on client during NFS usage
3111 Details : mds_open improperly used accmode to find out access mode to a
3112 file. Also mdc_intent_lock logic to find out if we already have
3113 lock similar to just received was flawed since introduction of
3114 skiplists - locks are now added to the front of the granted
3119 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
3121 Details : In obd setup failure handler, obd_stopping will not necessarily to
3122 be 1, and obd_set_up should also be checked to make sure whether
3123 obd is completely setup.
3125 Severity : enhancement
3127 Description: Allow masking D_WARNING, D_ERROR messages from console
3128 Details : Console messages can now be disabled via lnet.printk.
3133 Description: User code with malformed file open parameter crashes client node
3134 Details : Before packing join_file req, all the related reference should be
3135 checked carefully in case some malformed flags cause fake join_file
3141 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
3142 grain of quota doing harm to quota's accuracy
3143 Details : qunit size will be changed when quota limitation is too low/high;
3144 record the pending quota write in order to get more accureate
3145 quota; delete the patch for bug12588, which is unnecessary when
3146 this patch is landed. This bug also contains fixes for bug 14526,
3147 14299, 14601 and 13794.
3151 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
3152 Details : ldlm_completion_ast() assumes that a lock is granted when the req
3153 mode is equal to the granted mode. However, it should also check
3154 that LDLM_FL_CP_REQD is not set.
3158 Description: Heavy nfs access might result in deadlocks
3159 Details : After ELC code landed, it is now improper to enqueue any mds
3160 locks under och_sem, because enqueue might want to decide to
3161 cancel open locks for same inode we are holding och_sem for.
3165 Description: 35% write performance drop with ldiskfs2 when quotas are on
3166 Details : Enable ext3 journalled quota by default to improve performance
3167 when quotas are turned on.
3171 Description: Client eviction while running blogbench
3172 Details : A lot of unlink operations with concurrent I/O can lead to a
3173 deadlock causing evictions. To address the problem, the number of
3174 oustanding OST_DESTROY requests is now throttled to
3175 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
3176 callbacks are processed in priority.
3179 Frequency : RHEL4 only
3181 Description: mkfs is very slow on IA64/RHEL4
3182 Details : A performance regression has been discovered in the MPT Fusion
3183 driver between versions 3.02.73rh and 3.02.99.00rh. As a
3184 consequence, we have downgraded the MPT Fusion driver in the RHEL4
3185 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
3188 Frequency : PPC/PPC64 only
3190 Description: conflicts between asm-ppc64/types.h and lustre_types.h
3191 Details : fix duplicated definitions between asm-ppc64/types.h and
3192 lustre_types.h on PPC.
3195 Frequency : PPC/PPC64 only
3197 Description: asm-ppc/segment.h does not exist
3198 Details : fix compile issue on PPC.
3202 Description: data checksumming impacts single node performance
3203 Details : add support for several checksum algorithms. Currently, CRC32 and
3204 Adler-32 are supported. The checksum type can be changed on the fly
3205 through /proc/fs/lustre/osc/*/checksum_type.
3209 Description: use adler32 for page checksums
3210 Details : when available, use the Adler-32 algorithm instead of CRC32 for
3215 Description: better handle error messages in extents code
3217 Severity : enhancement
3219 Description: SNMP support enhancement
3220 Details : Adding total number of sampled request for an MDS node in snmp
3223 Severity : enhancement
3225 Description: Optimize ldlm waiting list processing for PR extent locks
3226 Details : When processing waiting list for read extent lock and meeting read
3227 lock that is same or wider to it that is not contended, skip
3228 processing rest of the list and immediatelly return current
3229 status of conflictness, since we are guaranteed there are no
3230 conflicting locks in the rest of the list.
3234 Description: Time out and refuse to reconnect
3235 Details : When the failover node is the primary node, it is possible
3236 to have two identical connections in imp_conn_list. We must
3237 compare not conn's pointers but NIDs, otherwise we
3238 can defeat connection throttling.
3242 Description: Client not clear own cache if answer to reconnect is lost.
3243 Details : Client gets evicted from server. Now client also thinks it is
3244 disconnected (or gets ENOTCONN on its operation) and decides to
3245 reconnect. Server receives reconnect message, but cannot find
3246 export. New export is created that is fully valid (new cookie!),
3247 but reply is lost and not reported to client. Client reconnects
3248 again and gets back a just-created connection, but it is not new
3249 so client thinks it was not evicted and does not do recovery.
3253 Description: Detect stride IO mode in read-ahead
3254 Details : When a client does stride read, read-ahead should detect that and
3255 read-ahead pages according to the detected stride pattern.
3259 Description: build for x2 fails
3260 Details : fix compile issue on Cray systems.
3262 Severity : enhancement
3264 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
3265 Details : This greatly improves speed of NFS writes on 2.6 kernels.
3268 Frequency : only on PPC/SLES10
3270 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
3271 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
3272 be defined. Add a hack in mkfs_lustre.c to work around this compile
3277 Description: LASSERT on MDS when client holding flock lock dies
3278 Details : ldlm pool logic depends on number of granted locks equal to
3279 number of released locks which is not true for flock locks, so
3280 just exclude such locks from consideration.
3284 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
3285 Details : Use fsfilt_sync() for both the whole filesystem sync and
3286 individual file sync to eliminate dangerous inode locking
3287 with I_LOCK that can lead to a deadlock.
3291 Description: Update an obsolete wirecheck.c generator
3292 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
3296 Description: Client can panic on open sometimes
3297 Details : It is possible that we try to free already freed request in
3298 ll_file_open in some error cases when we send request from
3303 Description: performance in 1.6.3
3304 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3305 sure that 1MB requests are not fragmented by the block layer.
3309 Description: LDLM soft lockups - improvement
3310 Details : It is be possible to send the lock handle along with each read
3311 or write request because the client is already doing a lock match
3312 itself so there isn't any reason the OST should have to re-do that
3318 Description: lfs quota fails with deactivated OSTS
3319 Details : With this patch, three improvements are included:
3320 1. detete the softlimit in mds and osts when use "lfs quota".
3321 2. display the inaccurate data in the output of "lfs quota".
3322 3. try to get quota info when "lfs quota" is executed.
3327 Description: Extent locks not granted with no conflicts sometimes.
3328 Details : When race occurs in glimpse handler and nothing is returned,
3329 we do not reprocess the queue after lock cancel, and that leads
3330 to a stall until next activity on a resource
3333 Frequency : failover with quotaon
3335 Description: during mds failovers with quota on, OSTs got into deadlock state
3336 and causing dumpstack.
3337 Details : for every quota slave, at any time, there is only one quota req
3338 is sent to quota master for every uid/gid. Before that quota req
3339 returns, all the thread relative to the same uid/gid will wait.
3340 So if the quota req is lost because mds failovers or any other
3341 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3342 quota reqs who time out.
3344 Severity : enhancement
3347 Description: when quota slave checks if quota is enough, there is an unnecessary
3349 Details : place this wait on necessary place instead of always waiting.
3351 --------------------------------------------------------------------------------
3353 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3355 * Support for kernels:
3356 2.6.5-7.286 (SLES 9),
3357 2.6.9-55.0.9.EL (RHEL 4),
3358 2.6.16.53-0.8 (SLES 10),
3359 2.6.18-8.1.14.el5 (RHEL 5),
3360 2.6.18.8 vanilla (kernel.org)
3361 * Client support for unpatched kernels:
3362 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3363 2.6.16 - 2.6.22 vanilla (kernel.org)
3364 * Due to recently discovered recovery problems, we do not recommend
3365 using patchless RHEL 4 clients with this or any earlier release.
3366 * Recommended e2fsprogs version: 1.40.2-cfs1
3367 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3370 Frequency : occasional
3372 Description: MDS hang or stay in waiting lock
3373 Details : If client receive lock with CBPENDING flag ldlm need send lock
3374 cancel as separate rpc, to avoid situation when cancel request
3375 can't processed due all i/o threads stay in wait lock.
3378 Frequency : occasional
3380 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3381 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3382 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3383 oscc_has_objects and see OSC has no objests and call
3384 oscc_internal_create to resend create request.
3386 Severity : enhancement
3388 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3389 Details : Update which_patch & target file for SLES10 latest kernel.
3391 Severity : enhancement
3393 Description: add --type and --size parameters to lfs find
3394 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3395 multiple OBDs can now be specified for the --obd option.
3397 Severity : enhancement
3399 Description: eliminate client locks in face of contention
3400 Details : file contention detection and lockless i/o implementation
3401 for contended files.
3403 Severity : enhancement
3405 Description: Remove client patches from SLES 10 kernel.
3406 Details : This causes SLES 10 clients to behave as patchless clients
3407 even on a Lustre-patched (server) kernel.
3409 Severity : enhancement
3411 Description: use i_size_read and i_size_write in 2.6 port
3412 Details : replace inode->i_size access with i_size_read/write()
3415 Frequency : when removing large files
3417 Description: scheduling issue during removal of large Lustre files
3418 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3419 It causes scheduling issues when removing large files (17TB in the
3425 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3426 Details : Impossible (EPROTO is returned) to access a directory that has a
3427 non-default striping and ACLs.
3430 Frequency : only on ppc
3432 Description: /proc/fs/lustre/devices broken on ppc
3433 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3434 We should make sure the type of 'index' is loff_t and then cast
3435 explicitly as needed below. Do not assign an explicitly cast
3439 Frequency : only for rhel5
3441 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3442 Details : Modify the target file & which_kernel.
3445 Frequency : if the uninit_groups feature is enabled on ldiskfs
3447 Description: e2fsck reports "invalid unused inodes count"
3448 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3449 feature and only a single inode is created in a group then the
3450 "bg_unused_inodes" count is incorrectly updated. Creating a
3451 second inode in that group would update it correctly.
3456 Description: buffer overruns could theoretically occur
3457 Details : llapi_semantic_traverse() modifies the "path" argument by
3458 appending values to the end of the origin string, and a buffer
3459 overrun may occur. Adding buffer overrun check in liblustreapi.
3461 Severity : enhancement
3463 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3466 Frequency : only if filesystem is inconsistent
3468 Description: handle "serious error: objid * already exists" more gracefully
3469 Details : If LAST_ID value on disk is smaller than the objects existing in
3470 the O/0/d* directories, it indicates disk corruption and causes an
3471 LBUG(). If the object is 0-length, then we should use the existing
3472 object. This will help to avoid a full fsck in most cases.
3474 Severity : enhancement
3476 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3477 Details : Modify vm-tunables-rhel4.patch.
3479 Severity : enhancement
3481 Description: Kernel config for 2.6.18-vanilla.
3482 Details : Modify targets/2.6-vanilla.target.in.
3483 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3484 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3485 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3486 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3489 Frequency : occasional
3491 Description: improve handling recoverable errors
3492 Details : If request processed with error which can be recoverable on server
3493 request should be resend, otherwise page released from cache and
3499 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3500 Details : Modify target file & which_patch.
3501 A flaw was found in the IA32 system call emulation provided
3502 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3503 value could be stored in the %RAX register, which could trigger an
3504 out-of-bounds system call table access. An untrusted local user
3505 could exploit this flaw to run code in the kernel
3506 (ie a root privilege escalation). (CVE-2007-4573).
3510 Description: change order of libsysio includes
3511 Details : '#include sysio.h' should always come before '#include xtio.h'
3513 Severity : enhancement
3515 Description: adapt the lustre_config script to support the upgrade case
3516 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3517 server targets to 1.6 in parallel.
3522 Description: To avoid grant space > avaible space when the disk is almost
3523 full. Without this patch you might see the error "grant XXXX >
3524 available" or some LBUG about grant, when the disk is almost
3526 Details : In filter_check_grant, for non_grant cache write, we should
3527 check the left space by if (*left > ungranted + bytes), instead
3528 of (*left > ungranted), because only we are sure the left space
3529 is enough for another "bytes", then the ungrant space should be
3530 increase. In client, we should update cl_avail_grant only there is
3531 OBD_MD_FLGRANT in the reply.
3536 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3537 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3538 described in CVE-2007-4573. This problem could allow untrusted
3539 local users to gain root access.
3542 Frequency : when using O_DIRECT and quotas
3544 Description: Incorrect file ownership on O_DIRECT output files
3545 Details : block usage reported by 'lfs quota' does not take into account
3546 files that have been written with O_DIRECT.
3550 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3551 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3552 window to 0. If it does hit cache miss, and read-pattern is still
3553 stride-io mode, does not reset the stride window, but also does not
3554 increase the stride window length in this case.
3556 --------------------------------------------------------------------------------
3558 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3560 * Support for kernels:
3561 2.6.5-7.286 (SLES 9),
3562 2.6.9-55.0.2.EL (RHEL 4),
3563 2.6.16.46-0.14 (SLES 10),
3564 2.6.18-8.1.8.el5 (RHEL 5),
3565 2.6.18.8 vanilla (kernel.org)
3566 * Client support for unpatched kernels:
3567 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3568 2.6.16 - 2.6.21 vanilla (kernel.org)
3569 * Due to recently discovered recovery problems, we do not recommend
3570 using patchless RHEL 4 clients with this or any earlier release.
3571 * Recommended e2fsprogs version: 1.40.2-cfs1
3572 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3576 Description: Fix errors in lfs documentation
3577 Details : Fixes man pages
3579 Severity : enhancement
3581 Description: Adaptive timeouts
3582 Details : RPC timeouts adapt to changing server load and network
3583 conditions to reduce resend attempts and improve recovery time.
3585 Severity : enhancement
3587 Description: llapi_file_create() does not allow some changes
3588 Details : add llapi_file_open() that allows specifying the file creation
3589 mode and open flags, and also returns an open file handle.
3591 Severity : enhancement
3593 Description: df doesn't work properly if diskfs blocksize != 4K
3594 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3597 Frequency : always on directories with default striping set
3599 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3600 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3602 Severity : enhancement
3604 Description: merge and cleanup kernel patches.
3605 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3607 Severity : enhancement
3609 Description: RedHat Update kernel for RHEL5
3610 Details : Kernel config file for RHEL5.
3612 Severity : enhancement
3614 Description: OSS needs mutliple precreate threads
3615 Details : Add ability to start more than one create thread per OSS.
3618 Frequency : only with quota on the root user
3620 Description: mds_obd_create error creating tmp object
3621 Details : When the user sets quota on root, llog will be affected and can't
3622 create files and write files.
3625 Frequency : Always on ia64 patchless client, and possibly others.
3627 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3628 Details : This allows the patchless client to be loaded on architectures
3629 without this export.
3634 Description: disorder of journal start and llog_add cause deadlock.
3635 Details : in llog_origin_connect, journal start should happen before
3636 llog_add keep the same order as other functions to avoid
3639 Severity : enhancement
3641 Description: RedHat Update kernel for RHEL5
3642 Details : Modify the kernel config file more closer RHEL5.
3644 Severity : enhancement
3646 Description: Build failure against Centos5 (RHEL5)
3647 Details : Define PAGE_SIZE when it isn't present.
3650 Frequency : occasionally when using NFS
3652 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3653 Details : with NFS, the anon dentry's parent was set to itself in
3654 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3655 corresponding dentry other than use rec->ur_name.
3657 Severity : enhancement
3659 Description: enable data checksumming by default
3660 Details : enable checksum by default, allow --disable-checksum
3661 configure option and "-o nochecksum" mount option. Checksums
3662 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3665 Frequency : Occasionally with failover
3667 Description: Client eviction due to failover config
3668 Details : after a connection loss, the lustre client should attempt to
3669 reconnect to the last active server first before trying the
3670 other potential connections.
3672 Severity : enhancement
3674 Description: client-side metadata stat-ahead during readdir(directory readahead)
3675 Details : perform client-side metadata stat-ahead when the client detects
3676 readdir and sequential stat of dir entries therein
3681 Description: when mds and osts use different quota unit(32bit and 64bit),
3682 quota will be released repeatly.
3683 Details : void sending multiple quota reqs to mds, which will keep the
3684 status between the reqs.
3687 Frequency : only with liblustre clients on XT3
3689 Description: evictions taking too long
3690 Details : allow llrd to evict clients directly on OSTs
3694 Description: osts not allocated evenly to files
3695 Details : change the condition to increase offset_idx
3697 Severity : enhancement
3699 Description: self-adjustable client's lru lists
3700 Details : use adaptive algorithm for managing client cached locks lru
3701 lists according to current server load, other client's work
3702 pattern, memory activities, etc. Both, server and client
3703 side namespaces provide number of proc tunables for controlling
3708 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3709 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3710 readdir code from ext3) from scratch preserving functionality.
3714 Description: Only those disconnect error should be returned by rq_status.
3715 Details : In open/enqueue processs, Some errors, which will cause client
3716 disconnected, should be returned by rq_status, while other
3717 errors should still be returned by intent, then mdc or llite will
3720 Severity : enhancement
3722 Description: Tune the kernel for good SCSI performance.
3723 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3724 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3728 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3729 Bugzilla : 13547, 13627
3730 Description: Data corruption for OSTs that are formatted larger than 2TB
3732 Details : When generating the bio request for lustre file writes the
3733 sector number would overflow a temporary variable before being
3734 used for the IO. The data reads correctly from Lustre (which
3735 will overflow in a similar manner) but other file data or
3736 filesystem metadata may be corrupted in some cases.
3740 Description: "lfs find -obd UUID" prints directories
3741 Details : "lfs find -obd UUID" will return all directory names instead
3742 of just file names. It is incorrect because the directories
3743 do not reside on the OSTs.
3746 Bugzilla : 13596, 13614
3747 Description: MDS hang after unclean shutdown of lots of clients
3748 Details : Never resend AST requests.
3752 Frequency : Always, for kernels after 2.6.16
3753 Description: Fix warning idr_remove called for id=.. which is not allocated.
3754 Details : Last kernels save old s_dev before kill super and not allow
3755 to restore from callback - restore it before call kill_anon_super.
3759 Description: Fix error on 'ls .' at the top of the Lustre mount.
3760 Details : Don't revalidate dentry if it is a root dentry.
3762 --------------------------------------------------------------------------------
3764 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3766 * Support for kernels:
3767 2.6.5-7.286 (SLES 9),
3768 2.6.9-55.0.2.EL (RHEL 4),
3769 2.6.16.46-0.14 (SLES 10),
3770 2.6.18-8.1.8.el5 (RHEL 5),
3771 2.6.18.8 vanilla (kernel.org)
3772 * Client support for unpatched kernels:
3773 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3774 2.6.16 - 2.6.22 vanilla (kernel.org)
3775 * Due to recently discovered recovery problems, we do not recommend
3776 using patchless RHEL 4 clients with this or any earlier release.
3777 * Recommended e2fsprogs version: 1.39.cfs8
3778 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3783 Description: block reactivating mgc import until all deactivates complete
3784 Details : Fix race when failing back MDT/MGS to itself (testing)
3786 Severity : enhancement
3788 Description: lfs setstripe enhancement
3789 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3792 Frequency : mds/oss recovery
3794 Description: llog ctxt is refrenced after it has been freed.
3795 Details : llog ctxt refcount was added to avoide the race between ctxt free
3796 and llog recovery process. Each llog user must hold ctxt refcount
3797 before it access the llog. And the llog ctxt can only be freed
3798 when its refcount is zero.
3800 Severity : enhancement
3802 Description: randomly memory allocation failure util
3803 Details : Make lustre randomly failed allocating memory for testing purpose.
3806 Frequency : only for SLES10
3808 Description: Update kernel patch for SLES10 SP1
3809 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3811 Severity : enhancement
3813 Description: omit set fsid for export NFS
3814 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3815 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3818 Frequency : after network failures
3820 Description: Add sync option to mount_lustre.c
3821 Details : Client loses data written to lustre after a network interruption.
3823 Severity : enhancement
3825 Description: Error message improvement.
3826 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3828 Severity : enhancement
3830 Description: don't use GFP_* in generic Lustre code.
3831 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3833 Severity : enhancement
3835 Description: obdclass is limited by single OBD_ALLOC(idarray)
3836 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3838 Severity : enhancement
3840 Description: updated patchess for new RHEL4 kernel
3841 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3843 Severity : enhancement
3845 Description: warnings with build patchless client with vanila 2.6.19 and up
3846 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3847 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3849 Severity : enhancement
3851 Description: O_DIRECT bypasses client statistics.
3852 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3853 read_ahead_stats were updated. Copied stats section from
3854 osc_send_oap_rpc() into async_internal().
3857 Frequency : only for Cray XT3
3859 Description: peer credits not enough on many OST per OSS systems.
3860 Details : Use new lnet way to add credits as we need those for pings and ASTs
3863 Frequency : only with liblustre
3865 Description: Liblustre is not releasing flock locks on file close.
3866 Details : Release flock locks on file close.
3869 Frequency : only for RHEL4
3871 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3872 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3873 Update target file and kernel config.
3877 Description: lustre support for RHEL5
3878 Details : Add support for RHEL5.
3883 Description: ASSERTION(export != NULL) failed in target_handle_connect
3884 Details : Assetion hit is result of rare race between disconnect and connet
3885 to same nid. target_handle_connect found old connect cockie and
3886 tried to reconnect, but can't find export for this cockie.
3891 Description: umount blocks forever on error
3892 Details : In result of wrong using obd_no_recov and obd_force flags client
3893 can hand if cancel or some other requests is lost.
3896 Frequency : Only for SLES
3898 Description: sanity_quota fail test_1
3899 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3900 file, which makes TSTID[2] inunique.
3902 Severity : enhancement
3904 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3905 Details : Update target/ChangeLog/which_patch .
3907 Severity : enhancement
3909 Description: jbd statistics
3910 Details : Port older jbd statistics patch for sles10
3915 Description: lvbo_init failed for resource with missing objects.
3916 Details : Fix returning error if we do stat for file with missing/corrupted
3917 objects and i_size set to all sum of size of all avaible objects.
3918 if we to truncate/write to missing object - it is recreated.
3923 Description: Oops in read and write path when failing to allocate lock.
3924 Details : Check if lock allocation failed and return error back.
3927 Frequency : When flocks are used.
3929 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3930 Details : Flock locks might destroy just granted lock if it could be merged
3931 with another existing flock, this is done in completion handler,
3932 so teach ldlm_cli_enquque_fini that this is a valid case for
3938 Description: reply_lock_interpret crash due to race with it and lock cancel.
3939 Details : Do not replay locks that are being cancelled. Do not reference
3940 locks by their address during replay, just by their handle.
3943 Frequency : only with deactivated OSTs
3945 Description: lstripe command fails for valid OST index
3946 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3947 instead of lov->desc.ld_active_tgt_count.
3949 Severity : enhancement
3951 Description: Build failure against Centos5 (RHEL5)
3952 Details : Use getpagesize() instead of PAGE_SIZE.
3954 --------------------------------------------------------------------------------
3956 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3958 * Support for kernels:
3959 2.6.5-7.283 (SLES 9),
3960 2.6.9-55.EL (RHEL 4),
3961 2.6.16.46-0.14 (SLES 10),
3962 2.6.18.8 vanilla (kernel.org)
3963 * Client support for unpatched kernels:
3964 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3965 2.6.16 - 2.6.22 vanilla (kernel.org)
3966 * Due to recently discovered recovery problems, we do not recommend
3967 using patchless RHEL 4 clients with this or any earlier release.
3968 * Recommended e2fsprogs version: 1.39.cfs8
3969 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3970 * Starting with this release, the ldiskfs backing filesystem required
3971 by Lustre is now in its own package, lustre-ldiskfs. This package
3972 should be installed. It is versioned separately from Lustre and
3973 may be released separately in future.
3978 Description: server LBUG when shutting down
3979 Details : Block umount forever until the mount refcount is zero rather
3980 than giving up after an arbitrary timeout.
3982 Severity : enhancement
3984 Description: add optional extra BUILD_VERSION info
3985 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3986 to override the lustre version.
3989 Frequency : 2.6.18 servers only
3991 Description: ll_kern_mount() doesn't release the module reference
3992 Details : The ldiskfs module reference count never drops down to 0
3993 because ll_kern_mount() doesn't release the module reference.
3998 Description: server LBUG when using old ost_num_threads parameter
3999 Details : Accept the old ost_num_threads parameter but warn that it
4000 is deprecated, and fix an off-by-one error that caused an LBUG.
4005 Description: Transient SCSI error results in persistent IO issue
4006 Details : iobuf->dr_error is not reinitialized to 0 between two
4010 Frequency : sometimes when underlying device returns I/O errors
4012 Description: OSTs not going read-only during write failures
4013 Details : OSTs are not remounted read-only when the journal commit threads
4014 get I/O errors because fsfilt_ext3 calls journal_start/stop()
4015 instead of the ext3 wrappers.
4018 Frequency : only with 10000 clients or more
4020 Description: poor connect scaling with increasing client count
4021 Details : Don't run filter_grant_sanity_check for more than 100 exports
4022 to improve scaling for large numbers of clients.
4025 Frequency : SLES10 only
4027 Description: sanity-quota.sh quotacheck failed: rc = -22
4028 Details : Quotas cannot be enabled on SLES10.
4031 Frequency : liblustre clients only
4033 Description: getdirentries does not give error when run on compute nodes
4034 Details : getdirentries does not fail when the size specified as an argument
4035 is too small to contain at least one entry
4037 Severity : enhancement
4039 Description: Add LNET router traceability for debug purposes
4040 Details : If a checksum failure occurs with a router as part of the
4041 IO path, the NID of the last router that forwarded the bulk data
4042 is printed so it can be identified.
4047 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
4048 Details : Due to a race condition, liblustre clients were occasionally
4049 evicted incorrectly.
4051 Severity : enhancement
4053 Description: lfs setstripe use optional parameters instead of postional
4056 Severity : enhancement
4058 Description: Nanosecond timestamp support for ldiskfs
4059 Details : The on-disk ldiskfs filesystem has added support for nanosecond
4060 resolution timestamps. There is not yet support for this at
4061 the Lustre filesystem level.
4064 Frequency : during server recovery
4066 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
4067 Details : request with rq_no_resend flag not awake l_wait_event if they get a
4071 Frequency : nfs export on patchless client
4073 Description: connectathon hang when test nfs export over patchless client
4074 Details : Disconnected dentry cannot be found with lookup, so we do not need
4075 to unhash it or make it invalid
4078 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
4079 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
4080 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
4085 Description: open req refcounting wrong on reconnect
4086 Details : If reconnect happened between getting open reply from server and
4087 call to mdc_set_replay_data in ll_file_open, we will schedule
4088 replay for unreferenced request that we are about to free.
4089 Subsequent close will crash in variety of ways.
4090 Check that request is still eligible for replay in
4091 mdc_set_replay_data().
4096 Description: disable writes to filesystem when reading health_check file
4097 Details : the default for reading the health_check proc file has changed
4098 to NOT do a journal transaction and write to disk, because this
4099 can cause reads of the /proc file to hang and block HA state
4100 checking on a healthy but otherwise heavily loaded system. It
4101 is possible to return to the previous behaviour during configure
4102 with --enable-health-write.
4104 Severity : enhancement
4106 Description: 64-bit inode version
4107 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
4108 to the inode. This will be required for version-based recovery.
4113 Description: MDS fails to start if a duplicate client export is detected
4114 Details : in some rare cases it was possible for a client to connect to
4115 an MDS multiple times. Upon recovery the MDS would detect this
4116 and fail during startup. Handle this more gracefully.
4118 Severity : enhancement
4120 Description: Add -o localflock option to simulate old noflock behaviour.
4121 Details : This will achieve local-only flock/fcntl locks coherentness.
4126 Description: log_commit_thread vs filter_destroy race leads to crash
4127 Details : Take import reference before releasing llog record semaphore
4132 Description: Wrong request locking in request set processing
4133 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
4134 imp_delayed_list, in this place should be used imp_lock.
4137 Frequency : when reconnecting
4139 Description: Grant leak when OSC reconnect to OST
4140 Details : When osc reconnect ost, OST(filter) should check whether it
4141 should grant more space to client by comparing fed_grant and
4142 cl_avail_grant, and return the granted space to client instead
4143 of "new granted" space, because client will call osc_init_grant
4144 to update the client grant space info.
4147 Frequency : when client reconnects to OST
4149 Description: Grant leak when OSC does a resend and replays bulk write
4150 Details : When osc reconnect to OST, OST(filter) should clear grant info of
4151 bulk write request, because the grant info while be sync between
4152 OSC and OST when reconnect, and we should ignore the grant info
4153 these of resend/replay write req.
4158 Description: Grant space more than avaiable space sometimes.
4159 Details : When then OST is about to be full, if two bulk writing from
4160 different clients came to OST. Accord the avaliable space of the
4161 OST, the first req should be permitted, and the second one
4162 should be denied by ENOSPC. But if the seconde arrived before
4163 the first one is commited. The OST might wrongly permit second
4164 writing, which will cause grant space > avaiable space.
4167 Frequency : when client is evicted
4169 Description: Grant might be wrongly erased when osc is evicted by OST
4170 Details : when the import is evicted by server, it will fork another
4171 thread ptlrpc_invalidate_import_thread to invalidate the
4172 import, where the grant will be set to 0. While the original
4173 thread will update the grant it got when connecting. So if
4174 the former happened latter, the grant will be wrongly errased
4175 because of this race.
4180 Description: Checking Stale with correct fid
4181 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
4182 to check whether it is stale, because sometimes, we want the
4183 enqueue happened anyway, and op_data.fid2 will not be initialized.
4185 Severity : enhancement
4187 Description: update patchless client
4188 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
4191 Frequency : only with 2.4 kernel
4193 Description: random memory corruption
4194 Details : size of struct ll_inode_info is to big for union inode.u and this
4195 can be cause of random memory corruption.
4200 Description: Memory leak in recovery
4201 Details : Lov_mds_md was not free in an error handler in mds_create_object.
4202 It should also check obd_fail before fsfilt_start, otherwise if
4203 fsfilt_start return -EROFS,(failover mds during mds recovery).
4204 then the req will return with repmsg->transno = 0 and rc = EROFS.
4205 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
4206 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
4207 be freed no matter whether fsfilt_commit success or not.
4210 Frequency : only with large numbers of cores on a single node
4212 Description: Prevents from taking the superblock lock in llap_from_page for
4214 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
4215 for llap_from_page call in ll_removepage() prevents from taking
4216 the superblock lock for a soon died page.
4221 Description: Not check open intent error before release open handle
4222 Details : in some rare cases, the open intent error is not checked before
4223 release open handle, which may cause
4224 ASSERTION(open_req->rq_transno != 0), because it tries to release
4225 the failed open handle.
4230 Description: Set cat log bitmap only after create log success.
4231 Details : in some rare cases, the cat log bitmap is set too early. and it
4232 should be set only after create log success.
4236 Description: Accessing a block bevice can re-enable I/O when Lustre is
4237 tearing down a device.
4238 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
4243 Description: service threads may hog cpus when there are a lot of requests
4244 Details : Insert cond_resched to give other threads a chance to use some CPU
4249 Description: the cat log was not initialized in recovery
4250 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
4251 unlink log on mds will not be initialized until mds post
4252 recovery. And also in mds post recovery, the unlink log will
4253 initialization will be done asynchronausly, so there will be race
4254 between add unlink log and unlink log initialization.
4258 Description: brw_stats were being printed incorrectly
4259 Details : brw_stats were being printed as log2 but all of them were not
4260 recorded as log2. Also remove some code duplication arising from
4261 filter_tally_{read,write}.
4265 Frequency : rare, only in recovery.
4266 Description: ASSERTION(req->rq_type != LI_POISON) failed
4267 Details : imp_lock should be held while iterating over imp_sending_list for
4268 prevent destroy request after get timeout in ptlrpc_queue_wait.
4272 Description: replay-single.sh test 52 fails
4273 Details : A lock's skiplist need to be cleanup when it being unlinked
4274 from its resource list.
4279 Description: Short directio read returns full requested size rather than
4281 Details : Direct I/O operations should return actual amount of bytes
4282 transferred rather than requested size.
4284 Severity : enhancement
4286 Description: metadata RPC reduction (e.g. for rm performance)
4287 Details : decrease the amount of synchronous RPC between clients and servers
4288 by canceling conflicing lock before the operation on the client
4289 and packing thier handles into the main operation RPC to server.
4291 Severity : enhancement
4293 Description: add #ifdef HAVE_KERNEL_CONFIG_H
4294 Details : kernels from 2.6.19 not need include linux/config.h, but add
4295 include linux/autoconf.h in commpiler command line.
4297 Severity : enhancement
4299 Description: patchless client support for 2.6.22 kernel
4300 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
4301 removed. In this case we need drop using os depended interface to
4302 kmem_cache and use cfs_mem_cache API.
4307 Description: fix mal-formatted messages
4308 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4312 Frequency : always in liblustre
4313 Description: wrong IS_ERR implementation in liblustre.h
4314 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4319 Description: Correct condition for output debug message.
4320 Details : inode i_nlink equal zero is not enough for output message about
4321 disk corruption, i_ctime and i_mode should be also checked.
4325 Frequency : always in patchless client
4326 Description: add configure check for truncate_complete_page
4327 Details : improve checks for exported symbols. This allow run check without
4328 sources, but with Module.symvers shipped with kernel distribution.
4329 add check for truncate_complete_page used by patchless client.
4333 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4334 Details : test_77h uses a file which was messed by other test case.
4338 Description: Not Check whether lov_tgts is NULL in some lov functions
4339 Details : Checking whether lov_tgts is NULL in some functions.
4343 Description: replace obdo_alloc() with OBDO_ALLOC macro
4344 Details : nothing special is done in obdo_alloc() function, and for
4345 debugging purpose, it needs to be replaced with macros.
4349 Description: bad return value and errno from fcntl call
4350 Details : In liblustre API, errno should be a negative value if error
4355 Description: ptlrpc_check_set() LBUG
4356 Details : In case of positive reply from server and failed client bulk
4357 callback after bulk transfer shouldn't LBUG, but process this
4358 request as erroneous.
4360 Severity : enhancement
4362 Description: VFS operations stats tool.
4363 Details : Tool which collects stats by tracking value written in pid,
4364 ppid, gid and uses llstat to generate output to plot graph using
4366 Updated lustre/utils/Makefile.am
4367 Added lustre/utils/ltrack_stats.c
4369 Severity : enhancement
4371 Description: 2.6.18 server support (lustre 1.6.1)
4372 Details : Support for 2.6.18 kernels on the server side.
4377 Description: ASSERTION(imp->imp_conn_current) failed
4378 Details : an assertion failure is hit if a client node boots and attempts to
4379 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4382 Frequency : only for i686
4384 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4385 Details : Fixed config variable for build.
4390 Description: Updated patchess for new RHEL4 kernel
4391 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4392 Updated patch jbd-stats-2.6.9.patch
4393 Updated patch qsnet-rhel4-2.6.patch
4394 Updated patch quota-deadlock-on-pagelock-core.patch
4395 Updated patch vfs_intent-2.6-rhel4.patch
4396 Updated patch vfs_races-2.6-rhel4.patch
4397 Updated series file 2.6-rhel4-titech.series
4398 Updated series file 2.6-rhel4.series
4399 Updated kernel config files
4404 Description: lquota slave complains LBUG when reconnecting with mds
4406 Details : quota slave depends qctxt->lqc_import to send its quota request.
4407 This pointer will be invalid if mds did failover or broke its
4408 connect to osts, which leads to LBUG.
4411 Frequency : when qunit size is too small(less than 20M)
4413 Description: write is stopped by improper -EDQUOT
4414 Details : If the master is busy and qunit size is small enough(let's say 1M),
4415 the slave can not get quota from master on time, which will lead
4416 slave to trigger a -EQUOTA to client.
4421 Description: Deadlock during metadata tests
4422 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4423 called with the per-dentry lock held.
4426 Frequency : SLES9 only
4428 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4429 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4431 Severity : enhancement
4433 Description: remove fs_prep_san_write operation and related patches
4434 Details : remove the ext3-san-jdike patches which are no longer useful.
4439 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4440 Details : a disk corruption can cause the mballoc code to assert on a
4441 double free or other extent corruptions. Handle these with
4442 ext3_error() instead of assertions.
4445 Frequency : only with mballoc3 code and deep extent trees
4447 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4448 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4449 can cause the filesystem to be remounted read-only.
4454 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4455 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4456 few external users of this ioctl number is lfsck's call to
4457 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4458 Add the old ioctl number to the handler so both old and new
4463 Description: parallel lock callbacks
4464 Details : Instead of sending blocking and completion callbacks as separated
4465 requests, adding them to a set and sending in parallel.
4469 Description: Disable most debugging by default
4470 Details : To improve performance, disable most logging (for debug purposes)
4471 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4472 default, and HA includes fewer messages.
4476 Frequency : only run on patchless client.
4477 Description: use do_facet on sanity.sh for test handling recoverables errors
4478 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4482 Description: hash tables for lists of nids, connections and uuids
4483 Details : Hash tables noticeably help when a lot of clients connect to a
4484 server, to faster identify duplicate connections or reconnects,
4485 also to faster find export to evict in manual eviction case.
4489 Description: Sometimes, when the server evict a client, and the client will
4490 not be evicted as soon as possible.
4491 Details : In enqueue req, the error was returned by intent, instead of
4492 rq_status which make ptlrpc layer not detect this error, and
4493 does not evict the client. So enqueue error should be returned
4497 Frequency : only at startup
4499 Description: Delay client connections to MDT until fist MDT->OST connect
4500 Details : If a client tried to create a new file before the MDT had
4501 connected to any OSTs, the create would return EIO. Now
4502 the client will simply block until the MDT connects to the
4503 first OST and the create can succeed.
4506 Frequency : only for SLES9
4508 Description: Routinely utilize latest Quadrics drivers in CFS releases
4509 Details : Update patch qsnet-suse-2.6.patch.
4512 Frequency : only for sles10
4514 Description: Update patches for SLES 10 SP1 kernel.
4515 Details : Update the patch vfs_intent-2.6-sles10.patch.
4520 Description: Routinely utilize latest Quadrics drivers in CFS releases
4521 Details : Update patch qsnet-rhel4-2.6.patch.
4524 Frequency : at statup only
4526 Description: mds_lov_synchronize race leads to various problems
4527 Details : simultaneous MDT->OST connections at startup can cause the
4528 sync to abort, leaving the OSC in a bad state.
4533 Description: Using wrong pointer in osc_brw_prep_request
4534 Details : Access to array[-1] can produce panic if kernel compiled with
4535 CONFIG_PAGE_ALLOC enabled
4537 Severity : enhancement
4539 Description: Async OSC create to avoid the blocking unnecessarily.
4540 Details : If a OST has no remain object, system will block on the creating
4541 when need to create a new object on this OST. Now, ways use
4542 pre-created objects when available, instead of blocking on an
4543 empty osc while others are not empty. If we must block, we block
4544 for the shortest possible period of time.
4548 Frequency : only in recovery
4549 Description: Mark OST as early accessible if his start SYNC.
4550 Details : osc_precreate return flag early accessible if oscc marked as
4551 OSCC_FLAG_SYNC_IN_PROGRESS.
4556 Description: Sometimes precreate code can triger create object on wrong ost
4557 Details : Wrong protected or not not restored variables aftre precreate loop
4558 can produce creation object on wrong ost.
4561 Frequency : oss recovery
4563 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4565 Details : llog_commit_thread_count should be synced between llog_commit
4566 start and cleanup, so new llog_commit thread should not be started
4567 when llog_commit threads being stopped to avoid accessing some
4570 Severity : enhancement
4572 Description: Add printing inode info into message about error in writepage.
4574 --------------------------------------------------------------------------------
4576 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4581 Frequency : on some architectures
4583 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4584 Details : Uninitialized flags sometimes cause configuration commands to
4588 Frequency : patchless clients only
4590 Description: missing __iget() symbol export
4591 Details : The __iget() symbol export is missing. To avoid the need for
4592 this on patchless clients the deathrow inode reaper is turned
4593 off, and we depend on the VM to clean up old inodes. This
4594 dependency was during via the fix for bug 12181.
4599 Description: sanity.sh fail: test_52b
4600 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4601 incorrect inode's flags to client.
4603 --------------------------------------------------------------------------------
4605 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4607 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4608 INTEROPERATE with older versions automatically. In many cases a
4609 special upgrade step is needed. Please read the
4610 user documentation before upgrading any part of a 1.4.x system.
4611 * WARNING: Lustre configuration and startup changes are required with
4612 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4614 * Support for kernels:
4615 2.4.21-47.0.1.EL (RHEL 3),
4616 2.6.5-7.283 (SLES 9),
4617 2.6.9-42.0.10.EL (RHEL 4),
4618 2.6.12.6 vanilla (kernel.org),
4619 2.6.16.27-0.9 (SLES10)
4620 * Client support for unpatched kernels:
4621 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4622 2.6.16 - 2.6.19 vanilla (kernel.org),
4623 2.6.9-42.0.8EL (RHEL 4)
4624 * Recommended e2fsprogs version: 1.39.cfs6
4625 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4628 Severity : enhancement
4630 Description: Async OSC create to avoid the blocking unnecessarily.
4631 Details : If a OST has no remain object, system will block on the creating
4632 when need to create a new object on this OST. Now, ways use
4633 pre-created objects when available, instead of blocking on an
4634 empty osc while others are not empty. If we must block, we block
4635 for the shortest possible period of time.
4637 Severity : enhancement
4639 Description: MountConf
4640 Details : Lustre configuration is now managed via mkfs and mount
4641 commands instead of lmc and lconf. New obd types (MGS, MGC)
4642 are added for dynamic configuration management. See
4643 https://mail.clusterfs.com/wikis/lustre/MountConf for
4646 Severity : enhancement
4648 Description: dynamic OST addition
4649 Details : OSTs can now be added to a live filesystem
4651 Severity : enhancement
4653 Description: startup order invariance
4654 Details : MDTs and OSTs can be started in any order. Clients only
4655 require the MDT to complete startup.
4657 Severity : enhancement
4659 Description: parallel, asynchronous orphan cleanup
4660 Details : orphan cleanup is now performed in separate threads for each
4661 OST, allowing parallel non-blocking operation.
4663 Severity : enhancement
4665 Description: optimized stripe assignment
4666 Details : stripe assignments are now made based on ost space available,
4667 ost previous usage, and OSS previous usage, in order to try
4668 to optimize storage space and networking resources.
4670 Severity : enhancement
4672 Description: Permanently set tunables
4673 Details : All writable /proc/fs/lustre tunables can now be permanently
4674 set on a per-server basis, at mkfs time or on a live
4677 Severity : enhancement
4679 Description: Lustre message v2
4680 Details : Add lustre message format v2.
4682 Severity : enhancement
4684 Description: client OST exclusion list
4685 Details : Clients can be started with a list of OSTs that should be
4686 declared "inactive" for known non-responsive OSTs.
4690 Description: ENOENT returned for valid filehandle during dbench.
4691 Details : Check if a directory has children when invalidating dentries
4692 associated with an inode during lock cancellation. This fixes
4693 an incorrect ENOENT sometimes seen for valid filehandles during
4694 testing with dbench.
4697 Frequency : SFS test only (otherwise harmless)
4699 Description: SPEC SFS validation failure on NFS v2 over lustre.
4700 Details : Changes the blocksize for regular files to be 2x RPC size,
4701 and not depend on stripe size.
4703 Severity : enhancement
4705 Description: fine-grained SMP locking inside DLM
4706 Details : Improve DLM performance on SMP systems by removing the single
4707 per-namespace lock and replace it with per-resource locks.
4709 Severity : enhancement
4711 Description: don't hold multiple extent locks at one time
4712 Details : To avoid client eviction during large writes, locks are not
4713 held on multiple stripes at one time or for very large writes.
4714 Otherwise, clients can block waiting for a lock on a failed OST
4715 while holding locks on other OSTs and be evicted.
4717 Severity : enhancement
4719 Description: Multiple MD RPCs in flight.
4720 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4721 To avoid overly-overloading MDT, introduce a limit on number of
4722 MDT RPCs in flight for a single client and add /proc controls
4723 to adjust this limit.
4725 Severity : enhancement
4727 Description: client read/write statistics
4728 Details : Add client read/write call usage stats for performance
4729 analysis of user processes.
4730 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4731 file access. extents_stats shows chunk size distribution.
4732 extents_stats_per_process show chunk size distribution per
4735 Severity : enhancement
4737 Description: per-client statistics on server
4738 Details : Add ldlm and operations statistics for each client in
4739 /proc/fs/lustre/mds|obdfilter/*/exports/
4741 Severity : enhancement
4743 Description: improved MDT statistics
4744 Details : Add detailed MDT operations statistics in
4745 /proc/fs/lustre/mds/*/stats
4747 Severity : enhancement
4749 Description: VFS operations stats
4750 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4751 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4756 Description: Fix client-side osc byte counters
4757 Details : The osc read/write byte counters in
4758 /proc/fs/lustre/osc/*/stats are now working
4761 Frequency : always as root on SLES
4763 Description: Failure of copying files with lustre special EAs.
4764 Details : Client side always return success for setxattr call for lustre
4765 special xattr (currently only "trusted.lov").
4770 Description: Refcount LNET uuids
4771 Details : The global LNET uuid list grew linearly with every startup;
4772 refcount repeated list entries instead of always adding to
4775 Severity : enhancement
4777 Description: Dynamic service threads
4778 Details : Within a small range, start extra service threads
4779 automatically when the request queue builds up.
4782 Frequency : mixed-endian client/server environments
4784 Description: mixed-endian crashes
4785 Details : The new msg_v2 system had some failures in mixed-endian
4788 Severity : enhancement
4790 Description: Easy OST removal
4791 Details : OSTs can be permanently deactivated with e.g. 'lctl
4792 conf_param lustre-OST0001.osc.active=0'
4794 Severity : enhancement
4796 Description: MGS proc entries
4797 Details : Added basic proc entries for the MGS showing what filesystems
4800 Severity : enhancement
4802 Description: provide MGS failover
4803 Details : Added config lock reacquisition after MGS server failover.
4805 Severity : enhancement
4807 Description: add Linux 2.4 support
4808 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4812 Description: a large application tries to do I/O to the same resource and dies
4813 in the middle of it.
4814 Details : Check the req->rq_arrival time after the call to
4815 ost_brw_lock_get(), but before we do anything about
4816 processing it & sending the BULK transfer request. This
4817 should help move old stale pending locks off the queue as
4818 quickly as obd_timeout.
4821 Frequency : when an incorrect nid is specified during startup
4823 Description: ptlrpc connect to non-existant node causes kernel crash
4824 Details : LNET can't be re-entered from an event callback, which
4825 happened when we expire a message after the export has been
4826 cleaned up. Instead, hand the zombie cleanup off to another
4829 Severity : enhancement
4831 Description: plain/inodebits lock performance improvement
4832 Details : Grouping plain/inodebits in granted list by their request modes
4833 and bits policy, thus improving the performance of search through
4837 Frequency : only if OST filesystem is corrupted
4839 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4840 Details : for a short time RPCs with bulk IO are in the replay list,
4841 but replay of bulk IOs is unimplemented. If the OST filesystem
4842 is corrupted due to disk cache incoherency and then replay is
4843 started it is possible to trip an assertion. Avoid putting
4844 committed RPCs into the replay list at all to avoid this issue.
4847 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4849 Description: System hang on startup
4850 Details : This bug allowed the liblustre (e.g. catamount) client to
4851 return to the app before handling all startup RPCs. This
4852 could leave the node unresponsive to lustre network traffic
4853 and manifested as a server ptllnd timeout.
4855 Severity : enhancement
4857 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4858 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4859 boolean to control whether to print peer debug info when a
4860 client's RPC times out.
4863 Frequency : only for kernels with patches from Lustre below 1.4.3
4865 Description: Remove old rdonly API
4866 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4869 Frequency : only for devices with external journals
4871 Description: Set external device read-only also
4872 Details : During a commanded failover stop, we set the disk device
4873 read-only while the server shuts down. We now also set any
4874 external journal device read-only at the same time.
4877 Frequency : when upgrading from 1.4 while trying to change parameters
4879 Description: The wrong (new) MDC name was used when setting parameters for
4880 upgraded MDT's. Also allows changing of OSC (and MDC)
4881 parameters if --writeconf is specified at tunefs upgrade time.
4884 Frequency : when setting specific ost indicies
4886 Description: QOS code breaks on skipped indicies
4887 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4888 created with --index need not be sequential.
4890 Severity : enhancement
4892 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4893 Details : The uninit_groups feature works in conjunction with the kernel
4894 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4895 up the pass1 processing of e2fsck. This is a read-only feature
4896 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4897 mount filesystems that have had this feature enabled.
4899 Severity : enhancement
4901 Description: Improve multi-block allocation algorithm to avoid fragmentation
4902 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4903 allocation locality and avoid filesystem fragmentation.
4905 ------------------------------------------------------------------------------
4907 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4909 * Support for kernels:
4910 2.4.21-47.0.1.EL (RHEL 3)
4911 2.6.5-7.283 (SLES 9)
4912 2.6.9-42.0.10.EL (RHEL 4)
4913 2.6.12.6 vanilla (kernel.org)
4914 2.6.16.27-0.9 (SLES 10)
4915 * Recommended e2fsprogs version: 1.39.cfs5
4917 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4921 Frequency : occasional, depends on client load and configuration
4922 Bugzilla : 12181, 12203
4923 Description: data loss for recently-modified files
4925 Details : In some cases it is possible that recently written or created
4926 files may not be written to disk in a timely manner (this should
4927 normally be within 30s unless client IO load is very high).
4928 The problem appears as zero-length files or files that are a
4929 multiple of 1MB in size after a client crash or client eviction
4930 that are missing data at the end of the file.
4932 This problem is more likely to be hit on clients where files are
4933 repeatedly created and unlinked in the same directory, clients
4934 have a large amount of RAM, have many CPUs, the filesystem has
4935 many OSTs, the clients are rebooted frequently, and/or the files
4936 are not accessed by other nodes after being written.
4938 The presence of the problem can be detected by looking at
4939 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4940 smaller than the second (nr_unused) then dirty files will not
4941 be flushed automatically to disk. "sync; sleep 10" should be
4942 run several times on the node before unmounting it to update
4943 Lustre (this is also safe to run on nodes without this problem).
4945 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4946 that can cause this same problem, so customers using that kernel
4947 also need to update the kernel in addition to Lustre. In order
4948 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4950 It is normal that files written just before a client crash (less
4951 than 30s) may not yet have been flushed to disk, even for local
4955 Frequency : frequent on thin XT3 nodes
4957 Description: UUID collision on thin XT3 Linux nodes
4958 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4959 randomly, since we relied on an insufficiently-seeded PRNG.
4964 Description: OSS hangs after "All ost request buffers busy"
4965 Details : A deadlock between quota and journal operations caused OSS
4966 hangs after printing "All ost request buffers busy."
4969 Frequency : always on liblustre builds
4971 Description: Cleanup compiler warnings on liblustre
4974 Frequency : always on liblustre builds on XT3
4976 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4980 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4981 > 32000 subdirectories
4982 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4983 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4984 subdirectory count crosses 32000. This will aid e2fsck to
4985 correctly handle more than 32000 subdirectories.
4988 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4990 Description: System hang on startup
4991 Details : This bug allowed the liblustre (e.g. catamount) client to
4992 return to the app before handling all startup RPCs. This
4993 could leave the node unresponsive to lustre network traffic
4994 and manifested as a server ptllnd timeout.
4996 Severity : enhancement
4998 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4999 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
5000 boolean to control whether to print peer debug info when a
5001 client's RPC times out.
5006 Description: make O_SYNC working on 2.6 kernels
5007 Details : 2.6 kernels use different method for mark pages for write,
5008 so need add a code to lustre for O_SYNC work.
5013 Description: Failure to close file and release space on NFS
5014 Details : Put inode details into lock acquired in ll_intent_file_open.
5015 Use mdc_intent_lock in ll_intent_open to properly
5016 detect all kind of errors unhandled by mdc_enqueue.
5021 Description: proc file read during shutdown sometimes raced obd removal,
5023 Details : Add lock to prevent obd access after proc file removal.
5026 Frequency : Only for files larger than 4GB on 32-bit clients.
5028 Description: improperly doing page alignment of locks
5029 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
5030 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
5035 Description: RPCs being resent when they shouldn't be
5036 Details : Some RPCs that should not be resent are being resent. This
5037 can cause inconsistencies in the RPC state machine. Do not
5038 resend such requests.
5041 Frequency : rare, only with NFS export
5043 Description: Crash on NFS re-export node
5044 Details : under very unusual load conditions an assertion is hit in
5045 ll_intent_file_open()
5048 Frequency : only if OST filesystem is corrupted
5050 Description: client incorrectly hits assertion in ptlrpc_replay_req()
5051 Details : for a short time RPCs with bulk IO are in the replay list,
5052 but replay of bulk IOs is unimplemented. If the OST filesystem
5053 is corrupted due to disk cache incoherency and then replay is
5054 started it is possible to trip an assertion. Avoid putting
5055 committed RPCs into the replay list at all to avoid this issue.
5060 Description: large O_DIRECT requests fail under memory pressure/fragmentation
5061 Details : Large single O_DIRECT read and write calls can fail to allocate
5062 a sufficiently large buffer to process the request. In case of
5063 allocation failure the allocation is retried with a smaller
5064 buffer and broken into smaller requests.
5066 Severity : enhancement
5068 Description: Add -o localflock option to simulate old noflock behaviour.
5069 Details : This will achieve local-only flock/fcntl locks coherentness.
5074 Description: versioning check is incomplete
5075 Details : Checking the version difference of client vs. server, report
5076 error if the gap is too big.
5081 Description: add support PG_writeback bit
5082 Details : add support for PG_writeback bit for Lustre, for more carefull
5083 work with page cache in 2.6 kernel. This also fix some deadlocks
5084 and remove hack for work O_SYNC with 2.6 kernel.
5086 Severity : enhancement
5088 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
5089 Details : The uninit_groups feature works in conjunction with the kernel
5090 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
5091 up the pass1 processing of e2fsck. This is a read-only feature
5092 in ldiskfs2 only, so older kernels and current ldiskfs cannot
5093 mount filesystems that have had this feature enabled.
5095 Severity : enhancement
5097 Description: Improve multi-block allocation algorithm to avoid fragmentation
5098 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
5099 allocation locality and avoid filesystem fragmentation.
5101 ------------------------------------------------------------------------------
5103 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
5105 * Support for kernels:
5106 2.6.9-42.0.3.EL (RHEL 4)
5107 2.6.5-7.276 (SLES 9)
5108 2.4.21-47.0.1.EL (RHEL 3)
5109 2.6.12.6 vanilla (kernel.org)
5110 2.6.16.21-0.8 (SLES10)
5111 * Recommended e2fsprogs version: 1.39.cfs2-0
5113 * The backwards-compatible /proc/sys/portals symlink has been removed
5114 in this release. Before upgrading, please ensure that you change
5115 any configuration scripts or /etc/sysctl.conf files that access
5116 /proc/sys/portals/* or sysctl portals.* to use the corresponding
5117 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
5118 in advance of the upgrade on any system running Lustre 1.4.6 or
5119 newer, since /proc/sys/lnet was added in that version.
5120 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
5124 Frequency : only when quota is used
5126 Description: avoid scanning export list for quota master
5127 Details : Change the algorithms to avoid scanning export list in order
5128 to improve the efficiency.
5131 Frequency : MDS failover only, very rarely
5133 Description: "went back in time" messages on mds failover
5134 Details : The greatest transno may be lost when the current operation
5135 finishes with an error (transno==0) and the client's last_rcvd
5136 record is over-written. Save the greatest transno in the
5137 mds_last_transno for this case.
5140 Frequency : always for specific kernels and striping counts
5142 Description: client may get "Matching packet too big" without ACL support
5143 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
5144 when trying to access files in certain configurations. The
5145 clients should in fact be denied when mounting because they do
5146 not understand ACLs.
5149 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
5151 Description: many clients connecting with IO in progress causes connect timeouts
5152 Details : Avoid synchronous journal commits to avoid delays caused by many
5153 clients connecting/disconnecting when bulk IO is in progress.
5154 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
5155 OST_IO_PORTAL to avoid delays behind potentially many pending
5159 Frequency : occasionally with multiple writers to a single file
5161 Description: shared writes to file may result in wrong size reported by stat()
5162 Details : Allow growing of kms when extent lock is cancelled
5165 Frequency : always with random mmap IO to multi-striped file
5167 Description: mmap write might be lost if we are writing to a 'hole' in stripe
5168 Details : Only if the hole is at the end of OST object so that kms is too
5169 small. Fix is to increase kms accordingly in ll_nopage.
5172 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
5174 Description: writes to a missing object would leak memory on the OST
5175 Details : If there is an inconsistency between the MDS and OST filesystems,
5176 such that the MDS references an object that doesn't exist, writes
5177 to that object will leak memory due to incorrect cleanup in the
5178 error handling path, eventually running out of memory on the OST.
5183 Description: Creating too long symlink causes lustre errors
5184 Details : Check symlink and name lengths before sending requests to MDS.
5187 Frequency : only if flock is enabled (not on by default)
5189 Description: posix locks not released on fd closure on 2.6.9+
5190 Details : We failed to add posix locks to list of inode locks on 2.6.9+
5191 kernels, this caused such locks not to be released on fd close and
5192 then assertions on fs unmount about still used locks.
5195 Frequency : MDS failover only, very rarely
5197 Description: clients may get ASSERTION(granted_lock != NULL)
5198 Details : When request was taking a long time, and a client was resending
5199 a getattr by name lock request. The were multiple lock requests
5200 with the same client lock handle and
5201 mds_getattr_name->fixup_handle_for_resent_request found one of the
5202 lock handles but later failed with ASSERTION(granted_lock != NULL).
5207 Description: handle->h_buffer_credits > 0, assertion failure
5208 Details : h_buffer_credits is zero after truncate, causing assertion
5209 failure. This patch extends the transaction or creates a new
5213 Frequency : NFS re-export or patchless client
5214 Bugzilla : 11179, 10796
5215 Description: Crash on NFS re-export node (__d_move)
5216 Details : We do not want to hash the dentry if we don't have a lock.
5217 But if this dentry is later used in d_move, we'd hit uninitialised
5218 list head d_hash, so we just do this to init d_hash field but
5219 leave dentry unhashed.
5222 Frequency : NFS re-export or patchless client
5224 Description: NFS exports has problem with symbolic link
5225 Details : lustre client didn't properly install dentry when re-exported
5226 to NFS or running patchless client.
5229 Frequency : NFS re-export or patchless client
5231 Description: Various nfs/patchless fixes.
5232 Details : fixes reuse disconected alias for lookup process - this fixes
5233 warning "find_exported_dentry: npd != pd",
5234 fix permission error with open files at nfs.
5235 fix apply umask when do revalidate.
5238 Frequency : occasional
5240 Description: Crash on NFS re-export node
5241 Details : calling clear_page() on the wrong pointer triggered oops in
5242 generic_mapping_read().
5245 Frequency : rarely, using O_DIRECT IO
5247 Description: unaligned directio crashes client with LASSERT
5248 Details : check for unaligned buffers before trying any requests.
5251 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
5253 Description: stale data returned from RAID cache
5254 Details : If only a small amount of IO is done to the RAID device before
5255 reading it again it is possible to get stale data from the RAID
5256 cache instead of reading it from disk.
5259 Frequency : always for sles10 kernel
5261 Description: sles10 support
5262 Details : ll_follow_link: compile fixes and using of nd_set_link
5263 under newer kernels.
5266 Frequency : depends on arch, kernel and compiler version, always on sles10
5269 Description: recursive or deep enough symlinks cause stack overflow
5270 Details : getting rid of large stack-allocated variable in
5274 Frequency : depends on hardware
5276 Description: lustre write performance loss in the SLES10 kernel
5277 Details : the performance loss is caused by using of write barriers in the
5278 ext3 code. The SLES10 kernel turns barrier support on by
5279 default. The fix is to undo that change for ldiskfs.
5281 ------------------------------------------------------------------------------
5283 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
5285 * Support for kernels:
5286 2.6.9-42.0.3EL (RHEL 4)
5287 2.6.5-7.276 (SLES 9)
5288 2.4.21-47.0.1.EL (RHEL 3)
5289 2.6.12.6 vanilla (kernel.org)
5293 Frequency : quota enabled and large files being deleted
5295 Description: releasing more than 4GB of quota at once hangs OST
5296 Details : If a user deletes more than 4GB of files on a single OST it
5297 will cause the OST to spin in an infinite loop. Release
5298 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5303 Description: statfs data retrieved from /proc may be stale or zero
5304 Details : When reading per-device statfs data from /proc, in the
5305 {kbytes,files}_{total,free,avail} files, it may appear
5306 as zero or be out of date.
5309 Frequency : systems with MD RAID1 external journal devices
5311 Description: lconf's call to blkid is confused by RAID1 journal devices
5312 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5313 of returning all block devices that match the journal UUID.
5316 Frequency : always, for aggregate stripe size over 4GB
5318 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5319 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5320 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5323 Frequency : always if "lfs find" used on a local file/directory
5325 Description: "lfs find" segfaults if used on a local file/directory
5326 Details : The case where a directory component was not specified wasn't
5327 handled correctly. Handle this properly.
5330 Frequency : always on ppc64
5332 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5333 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5334 perform endianness conversion.
5337 Frequency : rarely (truncate to non-zero file size after write under load)
5338 Bugzilla : 10730, 10687
5339 Description: Files padded with zeros to next 4K multiple
5340 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5341 it is possible that files that are truncated to a non-zero size
5342 immediately after being written are filled with zero bytes beyond
5343 the truncated size. No file data is lost.
5345 Severity : enhancement
5347 Description: Allow recovery/failover for liblustre clients.
5348 Details : liblustre clients were unaware of failover configurations until
5351 Severity : enhancement
5353 Description: user file locks should fail when not mounting with flock option
5354 Details : Set up an error-returning stub in ll_file_operations.lock field
5355 to prevent incorrect behaviour when client is mounted without
5356 flock option. Also, set up properly f_op->flock field for
5360 Frequency : always on ia64
5362 Description: "lfs df" loops on printing out MDS statfs information
5363 Details : The obd_ioctl_data was not initialized and in some systems
5364 this caused a failure during the ioctl that did not return
5365 an error. Initialize the struct and return an error on failure.
5368 Frequency : SLES 9 only
5370 Description: Error of copying files with lustre special EAs as root
5371 Details : Client side always return success for setxattr call for lustre
5372 special xattr (currently only "trusted.lov").
5375 Frequency : rarely on clusters with both ia64+i386 clients
5377 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5378 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5379 clients are doing concurrent writes to the same file it is
5380 possible that the ia64 clients may LASSERT because the OST
5381 extent locks are not PAGE_SIZE aligned. Ensure that grown
5382 locks are always aligned on the request boundary.
5385 Frequency : specific use, occasional
5387 Description: Overwriting in use executable truncates on-disk binary image
5388 Details : If one node attempts to overwrite an executable in use by
5389 another node, we now correctly return ETXTBSY instead of
5390 truncating the file.
5392 Severity : enhancement
5394 Description: Async OSC create to avoid the blocking unnecessarily.
5395 Details : If a OST has no remain object, system will block on the creating
5396 when need to create a new object on this OST. Now, ways use
5397 pre-created objects when available, instead of blocking on an
5398 empty osc while others are not empty. If we must block, we block
5399 for the shortest possible period of time.
5404 Description: chmod on Lustre root is propagated to other clients
5405 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5406 invalid by the follow_mount time.
5409 Frequency : liblustre clients only
5411 Description: Race in 'instant cancel' lock handling could lead to such locks
5412 never to be granted in case of SMP MDS
5413 Details : Do not destroy not yet granted but cbpending locks in
5417 Frequency : replay/resend of open
5419 Description: non null lock assetion failure in mds_intent_policy
5420 Details : Trying to replay/resend lockless open requests resulted in
5421 mds_open() returning 0 with no lock. Now it sets a flag if
5422 a lock is going to be returned.
5424 Severity : enhancement
5426 Description: Checksum enhancements
5427 Details : New checksum enhancements allow for resending RPCs that failed
5430 Severity : enhancement
5432 Description: Tunables on number of dirty pages in cacche
5433 Details : Allow to set limit on number of dirty pages cached.
5438 Description: client crash on unmount - lock still has references
5439 Details : In some error handling cases it was possible to leak a lock
5440 reference on a client while accessing a file. This was not
5441 harmful to the client during operation, but would cause the
5442 client to crash when the filesystem is unmounted.
5445 Frequency : specific case, rare
5447 Description: ETXTBSY on mds though file not in use by client
5448 Details : ETXTBSY is no longer incorrectly returned when attempting to
5449 chmod or chown a directory that the user previously tried to
5450 execute or a currently-executing binary.
5453 Frequency : extremely rare except on liblustre-based clients
5455 Description: Lustre space not freed when files are deleted
5456 Details : Clean up open-unlinked files after client eviction. Previously
5457 the unlink was skipped and the files remained as orphans.
5462 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5463 Details : In some cases it was possible to send a blocking callback to
5464 a client doing a glimpse, even though that client didn't get
5465 a lock granted. When the glimpse lock is cancelled on the OST
5466 the freed lock is left on the waiting list and corrupted the list.
5469 Frequency : all core dumps
5471 Description: Broke core dumps to lustre
5472 Details : Negative dentry may be unhashed if parent does not have UPDATE
5473 lock, but some callers, e.g. do_coredump, expect dentry to be
5474 hashed after successful create, hash it in ll_create_it.
5476 ------------------------------------------------------------------------------
5478 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5480 * Support for kernels:
5481 2.6.9-42.0.2.EL (RHEL 4)
5482 2.6.5-7.276 (SLES 9)
5483 2.4.21-40.EL (RHEL 3)
5484 2.6.12.6 vanilla (kernel.org)
5488 Frequency : always on RHEL 3
5490 Description: Number of open files grows over time
5491 Details : The number of open files grows over time, whether or not
5492 Lustre is started. This was due to a filp leak introduced
5493 by one of our kernel patches.
5495 ------------------------------------------------------------------------------
5497 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5499 * Support for kernels:
5500 2.6.9-42.EL (RHEL 4)
5501 2.6.5-7.267 (SLES 9)
5502 2.4.21-40.EL (RHEL 3)
5503 2.6.12.6 vanilla (kernel.org)
5508 Bugzilla : 5719, 9635, 9792, 9684
5509 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5510 Details : If a server is under heavy load and cannot reply to new
5511 connection requests before the client resends the (re)connect,
5512 the connection handling code can behave badly if two service
5513 threads are concurrently handing separate (re)connections from
5514 the same client. Add better locking to the connection handling
5515 code, and ensure that only a single connection will be processed
5516 for a given client UUID, even if the lock is dropped.
5518 Severity : enhancement
5520 Description: add TCP zero-copy support to kernel
5521 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5522 sends if the hardware supports scatter-gather and checksumming.
5523 This allows socklnd to do client-write and server-read more
5524 efficiently and reduce CPU utilization from skbuf copying.
5527 Frequency : only if NFS exporting from client
5529 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5530 Details : If setting attributes on a file created under NFS that had
5531 never been opened it would be possible to oops the client
5532 if the file had no objects.
5536 Bugzilla : 9326, 10402, 10897
5537 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5538 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5539 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5540 before ptlrpcd_wake() was called and the request was freed.
5543 Frequency : always for liblustre
5545 Description: liblustre client does MDS+OSTs setattr RPC for each write
5546 Details : When doing a write from a liblustre client, the client
5547 incorrectly issued an RPC to the MDS and each OST the file was
5548 striped over in order to update the timestamps. When writing
5549 with small chunks and many clients this could overwhelm the MDS
5550 with RPCs. In all cases it would slow down the write because
5551 these RPCs are unnecessary.
5553 Severity : enhancement
5555 Description: allow number of MDS service threads to be changed at module load
5556 Details : It is now possible to change the number of MDS service threads
5557 running. Adding "options mds mds_num_threads={N}" to the MDS's
5558 /etc/modprobe.conf will set the number of threads for the next
5559 time Lustre is restarted (assuming the "mds" module is also
5560 reloaded at that time). The default number of threads will
5561 stay the same, 32 for most systems.
5566 Description: OST crash if filesystem is unformatted or corrupt
5567 Details : If an OST is started on a device that has never been formatted
5568 or if the filesystem is corrupt and cannot even mount then the
5569 error handling cleanup routines would dereference a NULL pointer.
5574 Description: NULL pointer deref in llap_from_page.
5575 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5576 incorrect) mapping if the page was truncated/reclaimed while it was
5577 searched for. Check for this condition and skip such pages when
5578 doing readahead. Introduce extra check to llap_from_page() to
5579 verify page->mapping->host is non-NULL (so page is not anonymous).
5582 Frequency : Sometimes when using sys_sendfile
5584 Description: "page not covered by a lock" warnings from ll_readpage
5585 Details : sendfile called ll_readpage without right page locks present.
5586 Now we introduced ll_file_sendfile that does necessary locking
5587 around call to generic_file_sendfile() much like we do in
5591 Frequency : with certain MDS communication failures at client mount time
5593 Description: NULL pointer deref after failed client mount
5594 Details : a client connection request may delayed by the network layer
5595 and not be sent until after the PTLRPC layer has timed out the
5596 request. If the client fails the mount immediately it will try
5597 to clean up before the network times out the request. Add a
5598 reference from the request import to the obd device and delay
5599 the cleanup until the network drops the request.
5602 Frequency : occasionally during client (re)connect
5604 Description: assertion failure during client (re)connect
5605 Details : processing a client connection request may be delayed by the
5606 client or server longer than the client connect timeout. This
5607 causes the client to resend the connection request. If the
5608 original connection request is replied in this interval, the
5609 client may trip an assertion failure in ptlrpc_connect_interpret()
5610 which thought it would be the only running connect process.
5613 Frequency : only with obd_echo servers and clients that are rebooted
5615 Description: kernel BUG accessing uninitialized data structure
5616 Details : When running an obd_echo server it did not start the ping_evictor
5617 thread, and when a client was evicted an uninitialized data
5618 structure was accessed. Start the ping_evictor in the RPC
5619 service startup instead of the OBD startup.
5621 Severity : enhancement
5622 Bugzilla : 10193 (patchless)
5623 Description: Remove dependency on various unexported kernel interfaces.
5624 Details : No longer need reparent_to_init, exit_mm, exit_files,
5625 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5628 Frequency : rare (only users of deprecated and unsupported LDAP config)
5630 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5631 Details : LDAP apparently contains 'lustreName' attributes instead of
5632 'name'. A simple remapping of the name is sufficient.
5635 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5637 Description: waiting_locks_callback trips kernel BUG if client is evicted
5638 Details : Running with the dump_on_timeout debug flag turned on makes
5639 it possible that the waiting_locks_callback() can try to dump
5640 the Lustre kernel debug logs from an interrupt handler. Defer
5641 this log dumping to the expired_lock_main() thread.
5643 Severity : enhancement
5645 Description: Support NFS exporting on 2.6 kernels.
5646 Details : Implement non-rawops metadata methods for NFS server to use without
5647 changing NFS server code.
5650 Frequency : very rare (synthetic metadata workload only)
5652 Description: two racing renames might cause an MDS thread to deadlock
5653 Details : Running the "racer" program may cause one MDS thread to rename
5654 a file from being the source of a rename to being the target of
5655 a rename at exactly the same time that another thread is doing
5656 so, and the second thread has already enqueued these locks after
5657 doing a lookup of the target and is trying to relock them in
5658 order. Ensure that we don't try to re-lock the same resource.
5661 Frequency : only very large systems with liblustre clients
5663 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5664 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5665 all OSTs in parallel. This allows the eviction of stale liblustre
5666 clients to proceed much faster than if they were done in series,
5667 and also offers similar improvements for other set_info RPCs.
5672 Description: excessive CPU usage during initial read phase on client
5673 Details : During the initial read phase on a client, it would agressively
5674 retry readahead on the file, consuming too much CPU and impacting
5675 performance (since 1.4.5.8). Improve the readahead algorithm
5676 to avoid this, and also improve some other common cases (read
5677 of small files in particular, where "small" is files smaller than
5678 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5683 Description: MDS crash when receiving packet with unknown intent.
5684 Details : Do not LBUG in unknown intent case, just return -EFAULT
5686 Severity : enhancement
5687 Bugzilla : 9293, 9385
5688 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5689 Details : Do not serialize getattr (non-intent version) and statfs.
5692 Frequency : occasional, when OST network is overloaded/intermittent
5694 Description: client evicted by OST after bulk IO timeout
5695 Details : If a client sends a bulk IO request (read or write) the OST
5696 may evict the client if it is unresposive to its data GET/PUT
5697 request. This is incorrect if the network is overloaded (takes
5698 too long to transfer the RPC data) or dropped the OST GET/PUT
5699 request. There is no need to evict the client at all, since
5700 the pinger and/or lock callbacks will handle this, and the
5701 client can restart the bulk request.
5704 Frequency : Always when mmapping file with no objects
5706 Description: client crashes when mmapping file with no objects
5707 Details : Check that we actually have objects in a file before doing any
5708 operations on objects in ll_vm_open, ll_vm_close and
5714 Description: Request leak when working with deleted CWD
5715 Details : Introduce advanced request refcount tracking for requests
5716 referenced from lustre intent.
5718 Severity : Enhancement
5720 Description: Cache open file handles on client.
5721 Details : MDS now will return special lock along with openhandle, if
5722 requested and client is allowed to hold openhandle, even if unused,
5723 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5724 closing files for every read/write openration.
5726 Severity : Enhancement
5728 Description: Cache open negative dentries on client when possible.
5729 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5730 negative dentries on lock revocation.
5735 Description: Remounting a client read-only wasn't possible with a zconf mount
5736 Details : It wasn't possible to remount a client read-only with llmount.
5738 Severity : enhancement
5739 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5740 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5741 MPICH is included to add Lustre-specific ADIO interfaces.
5742 This is based closely on the UFS ADIO layer and only differs
5743 in file creation, in order to allow the OST striping to be set.
5744 This is user-contributed code and not supported by CFS.
5749 Description: extended inode attributes (immutable, append-only) work improperly
5750 when 2.4 and 2.6 kernels are used on client/server or vice versa
5751 Details : Introduce kernel-independent values for these flags.
5753 Severity : enhancement
5756 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5757 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5758 granularity. Now, Fractional MB granularity is be supported,
5759 this is very useful for low memory system.
5761 Severity : enhancement
5763 Description: Getattr by fid
5764 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5765 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5766 have locks and avoids getattr rpc requests for it on every stat.
5769 Frequency : Always, for filesystems larger than 2TB
5771 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5772 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5773 filesystems larger than 2TB to be mounted with mballoc enabled.
5776 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5778 Description: filesystem corruption for non-standard kernels and very large OSTs
5779 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5780 filesystems larger than 2TB is mounted then the kernel will
5781 silently corrupt the start of the filesystem. CONFIG_LBD is
5782 enabled for all CFS-supported kernels, but the possibility of
5783 this happening with a modified kernel config exists.
5785 Severity : enhancement
5787 Description: add client O_DIRECT support for 2.6 kernels
5788 Details : It is now possible to do O_DIRECT reads and writes to files
5789 in the Lustre client mountpoint on 2.6 kernel clients.
5791 Severity : enhancement
5793 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5794 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5795 parallel, not waiting for response from every OST before sending
5796 a rpc to the next OST.
5801 Description: setattr vs write race when updating file timestamps
5802 Details : Client processes that update a file timestamp into the past
5803 right after writing to the file (e.g. tar) it is possible that
5804 the updated file modification time can be reset to the current
5805 time due to a race between processing the setattr and write RPC.
5807 Severity : enhancement
5809 Description: Bring 'lfs find' closer in line with regular Linux find.
5810 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5811 -print0 options and obtains all the needed info through the lustre
5814 Severity : enhancement
5816 Description: support up to 1024 configured devices on one node
5817 Details : change obd_dev array from statically allocated to dynamically
5818 allocated structs as they are first used to reduce memory usage
5823 Description: Flush dirty partially truncated pages during truncate
5824 Details : Immediatelly flush partially truncated pages in filter_setattr,
5825 this way we completely avoid having any pages in page cache on OST
5826 and can retire ugly workarounds during writes to flush such pages.
5831 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5832 Details : protect inode from truncation within vfs_unlink() context
5833 just take a reference before calling vfs_unlink() and release it
5834 when parent's i_sem is free.
5837 Frequency : always, if extents are used on OSTs
5839 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5840 Details : OSTs using the extents format would not zero the high 16 bits of
5841 the index physical block number. This is not a problem for any
5842 OST filesystems smaller than 16TB, and no kernels support ext3
5843 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5844 new/modified files) and can be fixed for existing filesystems
5845 with e2fsprogs-1.39-cfs1.
5850 Description: import connection selection may be incorrect if timer wraps
5851 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5852 import connections to be ignored if the 32-bit jiffies counter
5853 wraps. Use a 64-bit jiffies counter.
5856 Frequency : during server recovery
5858 Description: crash after server is denying duplicate export
5859 Details : If clients are resending connect requests to the server, the
5860 server refuses to allow a client to connect multiple times.
5861 Fixed a bug in the handling of this case.
5864 Frequency : very large clusters immediately after boot
5866 Description: LNET request buffers exhausted under heavy short-term load
5867 Details : If a large number of client requests are generated on a service
5868 that has previously never seen so many requests it is possible
5869 that the request buffer growth cannot keep up with the spike in
5870 demand. Instead of dropping incoming requests, they are held in
5871 the LND until the RPC service can accept more requests.
5874 Frequency : Sometimes during replay
5876 Description: Assertion failure in ll_local_open after replay.
5877 Details : If replay happened on an open request reply before we were able
5878 to set replay handler, reply will become not swabbed tripping the
5879 assertion in ll_local_open. Now we set the handler right after
5880 recognising of open request
5883 Frequency : very rare
5885 Description: kernel reports "badness in vsnprintf"
5886 Details : Reading from the "recovery_status" /proc file in small chunks
5887 may cause a negative length in lprocfs_obd_rd_recovery_status()
5888 call to vsnprintf() (which is otherwise harmless). Exit early
5889 if there is no more space in the output buffer.
5891 Severity : enhancement
5893 Description: clear OBD RPC statistics by writing to them
5894 Details : It is now possible to clear the OBD RPC statistics by writing
5895 to the "stats" file.
5900 Description: Client mtime is not the same on different clients after utimes
5901 Details : In some cases, the client was using the utimes() syscall on
5902 a file cached on another node. The clients now validate the
5903 ctime from the MDS + OSTs to determine which one is right.
5908 Description: Inability to activate failout mode
5909 Details : lconf script incorrectly assumed that in python string's numeric
5910 value is used in comparisons.
5913 Frequency : always with multiple stripes per file
5915 Description: Inefficient object allocation for mutli-stripe files
5916 Details : When selecting which OSTs to stripe files over, for files with
5917 a stripe count that divides evenly into the number of OSTs,
5918 the MDS is always picking the same starting OST for each file.
5919 Return the OST selection heuristic to the original design.
5924 Description: mount failures may take full timeout to return an error
5925 Details : Under some heavy load conditions it is possible that a
5926 failed mount can wait for the full obd_timeout interval,
5927 possibly several minutes, before reporting an error.
5928 Instead return an error as soon as the status is known.
5930 ------------------------------------------------------------------------------
5932 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5934 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5935 INTEROPERATE with older versions automatically. Please read the
5936 user documentation before upgrading any part of a live system.
5937 * WARNING: Lustre networking configuration changes are required with
5938 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5941 * Support for kernels:
5942 2.6.9-22.0.2.EL (RHEL 4)
5943 2.6.5-7.244 (SLES 9)
5944 2.6.12.6 vanilla (kernel.org)
5947 Severity : enhancement
5948 Bugzilla : 7981/8208
5949 Description: Introduced Lustre Networking (LNET)
5950 Details : LNET is new networking infrastructure for Lustre, it includes
5951 a reorganized network configuration mode (see the user
5952 documentation for full details) as well as support for routing
5953 between different network fabrics. Lustre Networking Devices
5954 (LNDs) for the supported network fabrics have also been
5955 created for this new infrastructure.
5957 Severity : enhancement
5958 Description: Introduced Access control lists
5959 Details : clients can set ACLs on files and directories in order to have
5960 more fine-grained permissions than the standard Unix UGO+RWX.
5961 The MDS must be started with the "-o acl" mount option.
5963 Severity : enhancement
5964 Description: Introduced filesystem quotas
5965 Details : Administrators may now establish per-user quotas on the
5968 Severity : enhancement
5970 Description: Configuration change for the XT3
5971 The PTLLND is now used to run Lustre over Portals on the XT3
5972 The configure option(s) --with-cray-portals are no longer used.
5973 Rather --with-portals=<path-to-portals-includes> is used to
5974 enable building on the XT3. In addition to enable XT3 specific
5975 features the option --enable-cray-xt3 must be used.
5980 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5981 Details : A race between allocating a new llcd and re-getting the llcd_lock
5982 allowed another thread to grab newly-allocated llcd.
5984 Severity : enhancement
5986 Description: 2.6 OST async journal commit and locking fix to improve performance
5987 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5988 2.6 kernels are now async as they already were in 2.4 kernels so
5989 that they can commit concurrently with the network bulk transfer.
5990 For block-allocated files the filter allocation semaphore is held
5991 to avoid filesystem fragmentation during allocation. BKL lock
5992 removed for 2.6 xattr operations where it is no longer needed.
5997 Description: lconf incorrectly determined whether two IP networks could talk
5998 Details : In some more complicated routing and multiple-network
5999 configurations, lconf will avoid trying to make a network
6000 connection to a disjoint part of the IP space. It was doing the
6001 math incorrectly for one set of cases.
6006 Description: Fix for potential infinite loop processing records in an llog.
6007 Details : If an llog record is corrupted/zeroed, it is possible to loop
6008 forever in llog_process(). Validate the llog record length
6009 and skip the remainder of the block on error.
6012 Frequency : occasional (liblustre only)
6014 Description: liblustre could not open files whose last component is a symlink
6015 Details : sysio_path_walk() would incorrectly pass the open intent to
6016 intermediate path components.
6019 Frequency : rare (liblustre only with non-standard tuning)
6020 Bugzilla : 7201 (7350)
6021 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
6022 Details : llu_lookup_finish_locks() tries to set lock data on a lock
6023 after it has been released, only do this for referenced locks
6025 Severity : enhancement
6027 Description: specifying an (invalid) directory default stripe_size of -1
6028 would reset the directory default striping
6029 Details : stripe_size -1 was used internally to signal directory stripe
6030 removal, now use "all default" to signal dir stripe removal
6031 as a directory striping of "all default" is not useful
6034 Frequency : common for large clusters running liblustre clients
6036 Description: doing an ls when liblustre clients are running is slow
6037 Details : sending a glimpse AST to a liblustre client waits for every AST
6038 to time out, as liblustre clients will not respond. Since they
6039 cannot cache data we refresh the OST lock LVB from disk instead.
6041 Severity : enhancement
6043 Description: doing an ls at the same time as file IO can be slow
6044 Details : enqueue and other "small" requests can be blocked behind many
6045 large IO requests. Create a new OST IO portal for non-IO
6046 requests so they can be processed faster.
6049 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
6051 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
6052 created file with mode 000
6053 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
6054 allow an ftruncate() as a non-root user to a file with mode 000.
6055 Lustre now respects this flag to disable mode checks when
6056 truncating a file owned by the user
6059 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
6062 Description: Revoking locks from clients that went dead or catatonic might take
6064 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
6065 cancellation of such locks instant on servers without waiting for
6066 any reply from clients. Clients drops these locks when cancel
6067 notification from server is received without replying.
6070 Frequency : liblustre-only, when liblustre client dies or becomes busy
6072 Description: Doing ls on Linux clients can take a long time with active
6074 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
6075 granting such locks to it in the first place if possible. Locks
6076 are taken by proxy on the OST during the read or write and
6077 dropped immediately afterward. Add connect flags handling, do
6078 not grant locks to liblustre clients for glimpse ASTs.
6080 Severity : enhancement
6082 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
6083 Details : Existing read-ahead algorithm is tuned for the case of streamlined
6084 sequential reads and behaves badly with applications doing random
6085 reads. Improve it by reading ahead at least read region, and
6086 avoiding excessive large RPC for small reads.
6088 Severity : enhancement
6090 Description: Creating more than 1000 files for a single job may cause a load
6091 imbalance on the OSTs if there are also a large number of OSTs.
6092 Details : qos_prep_create() uses an OST index reseed value that is an
6093 even multiple of the number of available OSTs so that if the
6094 reseed happens in the middle of the object allocation it will
6095 still utilize the OSTs as uniformly as possible.
6100 Description: OST or MDS may oops in ping_evictor_main()
6101 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
6102 but doesn't restart at beginning of obd_exports_timed list
6105 Severity : enhancement
6107 Description: improve by-nid export eviction on the MDS and OST
6108 Details : allow multiple exports with the same NID to be evicted at one
6109 time without re-searching the exports list.
6112 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
6114 Description: MDS may oops in groups_free()
6115 Details : in rare race conditions a newly allocated group_info struct is
6116 freed again, and this can be NULL. The 2.4 compatibility code
6117 for groups_free() checked for a NULL pointer, but 2.6 did not.
6120 Frequency : common for liblustre clients doing little filesystem IO
6121 Bugzilla : 9352, 7313
6122 Description: server may evict liblustre clients accessing contended locks
6123 Details : if a client is granted a lock or receives a completion AST
6124 with a blocking AST already set it would not reply to the AST
6125 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
6128 Frequency : lfs setstripe, only systems with more than 160 OSTs
6130 Description: unable to set striping with a starting offset beyond OST 160
6131 Details : llapi_create_file() incorrectly limited the starting stripe
6132 index to the maximum single-file stripe count.
6135 Frequency : LDAP users only
6137 Description: lconf did not handle in-kernel recovery with LDAP properly
6138 Details : lconf/LustreDB get_refs() is searching the wrong namespace
6140 Severity : enhancement
6142 Description: bind OST threads to NUMA nodes to improve performance
6143 Details : all OST threads are uniformly bound to CPUs on a single NUMA
6144 node and do their allocations there to localize memory access
6146 Severity : enhancement
6148 Description: llmount can determine client NID directly from Myrinet (GM)
6149 Details : the client NID code from gmnalnid was moved directly into
6150 llmount, removing the need to use this or specifying the
6151 client NID explicitly when mounting GM clients with zeroconf
6154 Frequency : if client is started with down MDS
6156 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
6157 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
6158 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
6159 ptlrpc_import_delayed_req() because we didn't check if the
6160 request was interrupted, and we also didn't break out of the
6161 event loop if there was no timeout
6166 Description: data loss during non-page-aligned writes to a single file from
6167 both multiple nodes and multiple threads on one node at same time
6168 Details : updates to KMS and lsm weren't protected by common lock. Resulting
6169 inconsistency led to false short-reads, that were cached and later
6170 used by ->prepare_write() to fill in partially written page,
6171 leading to data loss.
6174 Frequency : always, if lconf --abort_recovery used
6176 Description: lconf --abort_recovery fails with 'Operation not supported'
6177 Details : lconf was attempting to abort recovery on the MDT device and not
6180 Severity : enhancement
6182 Description: remove cleanup logs
6183 Details : replace lconf-generated cleanup logs with lustre internal
6184 cleanup routines. Eliminates the need for client-cleanup and
6187 Severity : enhancement
6189 Description: add support for EAs (user and system) on lustre filesystems
6190 Details : it is now possible to store extended attributes in the Lustre
6191 client filesystem, and with the user_xattr mount option it
6192 is possible to allow users to store EAs on their files also
6194 Severity : enhancement
6196 Description: Add possibility (config option) to show minimal available OST free
6198 Details : When compiled with --enable-mindf configure option, statfs(2)
6199 (and so, df) will return least minimal free space available from
6200 all OSTs as amount of free space on FS, instead of summary of
6201 free spaces of all OSTs.
6203 Severity : enhancement
6205 Description: do not expand extent locks acquired on OST-side
6206 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
6207 by server: they are not cached anyway.
6210 Frequency : when mmap is used/binaries executed from Lustre
6212 Description: Unmmap pages before throwing them away from read cache.
6213 Details : llap_shrink cache now attempts to unmap pages before discarding
6214 them (if unmapping failed - do not discard). SLES9 kernel has
6215 extra checks that trigger if this unmapping is not done first.
6220 Description: lconf didn't resolve symlinks before checking to see whether a
6221 given mountpoint was already in use
6224 Frequency : when migrating failover services
6225 Bugzilla : 6395, 9514
6226 Description: When migrating a subset of services from a node (e.g. failback
6227 from a failover service node) the remaining services would
6228 time out and evict clients.
6229 Details : lconf --force (implied by --failover) sets the global obd_timeout
6230 to 5 seconds in order to quickly disconnect, but this caused
6231 other RPCs to time out too quickly. Do not change the global
6232 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
6234 Severity : enhancement
6235 Frequency : if MDS is started with down OST
6236 Bugzilla : 9439,5706
6237 Description: Allow startup/shutdown of an MDS without depending on the
6238 availability of the OSTs.
6239 Details : Asynchronously call mds_lov_synchronize during MDS startup.
6240 Add appropriate locking and lov-osc refcounts for safe
6241 cleaning. Add osc abort_inflight calls in case the
6242 synchronize never started.
6245 Frequency : occasional (Cray XT3 only)
6247 Description: root not authorized to access files in CRAY_PORTALS environment
6248 Details : The client process capabilities were not honoured on the MDS in
6249 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
6250 been accessed by an authorized user then root was able to access
6251 the file on the local client also. The root user capabilities
6252 are now allowed on the MDS, as this environment has secure UID.
6255 Frequency : occasional
6257 Description: ldiskfs "too long searching" message happens too often
6258 Details : A debugging message (otherwise harmless) prints too often on
6259 the OST console. This has been reduced to only happen when
6260 there are fragmentation problems on the filesystem.
6265 Description: Division by zero in statfs when all OSCs are inactive
6266 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
6267 lov_statfs divides by value returned by lov_get_stripecnt().
6271 Bugzilla : 9489, 3273
6272 Description: First write from each client to each OST was only 4kB in size,
6273 to initialize client writeback cache, which caused sub-optimal
6274 RPCs and poor layout on disk for the first writen file.
6275 Details : Clients now request an initial cache grant at (re)connect time
6276 and so that they can start streaming writes to the cache right
6277 away and always do full-sized RPCs if there is enough data.
6278 If the OST is rebooted the client also re-establishes its grant
6279 so that client cached writes will be honoured under the grant.
6284 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
6285 Details : Now I/O RPCs go to different portal number and (presumably) fast
6286 lock requests (and glimses) and other RPCs get their own service
6287 threads pool that should be able to service those RPCs
6290 Severity : enhancement
6292 Description: Ability to exchange lustre version between client and servers and
6293 issue warnings at client side if client is too old. Also for
6294 liblustre clients there is ability to refuse connection of too old
6296 Details : New 'version' field is added to connect data structure that is
6297 filled with version info. That info is later checked by server and
6301 Frequency : rare, liblustre only.
6302 Bugzilla : 9296, 9581
6303 Description: Two simultaneous writes from liblustre at offset within same page
6304 might proceed at the same time overwriting eachother with stale
6306 Details : I/O lock withing llu_file_prwv was released too early, before data
6307 actually was hitting the wire. Extended lock-holding time until
6308 server acknowledges receiving data.
6311 Frequency : extremely rare. Never observed in practice.
6313 Description: avoid generating lustre_handle cookie of 0.
6314 Details : class_handle_hash() generates handle cookies by incrementing
6315 global counter, and can hit 0 occasionaly (this is unlikely, but
6316 not impossible, because initial value of cookie counter is
6317 selected randonly). Value of 0 is used as a sentinel meaning
6318 "unassigned handle" --- avoid it. Also coalesce two critical
6319 sections in this function into one.
6321 Severity : enhancement
6323 Description: allow liblustre clients to delegate truncate locking to OST
6324 Details : To avoid overhead of locking, liblustre client instructs OST to
6325 take extent lock in ost_punch() on client's behalf. New connection
6326 flag is added to handle backward compatibility.
6328 Severity : enhancement
6329 Bugzilla : 4928, 7341, 9758
6330 Description: allow number of OST service threads to be specified
6331 Details : a module parameter allows the number of OST service threads
6332 to be specified via "options ost ost_num_threads={N}" in the
6333 OSS's /etc/modules.conf or /etc/modprobe.conf.
6337 Bugzilla : 6146, 9635, 9895
6338 Description: servers crash with bad pointer in target_handle_connect()
6339 Details : In rare cases when a client is reconnecting it was possible that
6340 the connection request was the last reference for that export.
6341 We would temporarily drop the export reference and get a new
6342 one, but this may have been the last reference and the export
6343 was just destroyed. Get new reference before dropping old one.
6345 Severity : enhancement
6346 Frequency : if client is started with failover MDS
6348 Description: Allow multiple MDS hostnames in the mount command
6349 Details : Try to read the configuration from all specified MDS
6350 hostnames during a client mount in case the "primary"
6353 Severity : enhancement
6355 Description: Stop sending data to evicted clients as soon as possible.
6356 Details : Check if the client we are about to send or are sending data to
6357 was evicted already. (Check is done every second of waiting,
6358 for which l_wait_event interface was extended to allow checking
6359 of exit condition at specified intervals).
6362 Frequency : rare, normally only when NFS exporting is done from client
6364 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6366 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6369 Frequency : always, if srand() is called before liblustre initialization
6371 Description: Liblustre uses system PRNG disturbing its usage by user application
6372 Details : Introduce internal to lustre fast and high-quality PRNG for
6373 lustre usage and make liblustre and some other places in generic
6374 lustre code to use it.
6376 Severity : enhancement
6377 Bugzilla : 9477, 9557, 9870
6378 Description: Verify that the MDS configuration logs are updated when xml is
6379 Details : Check if the .xml configuration logs are newer than the config
6380 logs stored on the MDS and report an error if this is the case.
6381 Request --write-conf, or allow starting with --old_conf.
6383 Severity : enhancement
6385 Description: Handle symlinks in the path when checking if Lustre is mounted.
6386 Details : Resolve intermediate symlinks when checking if a client has
6387 mounted a filesystem to avoid duplicate client mounts.
6392 Description: lconf can hit an error exception but still return success.
6393 Details : The lconf command catches the Command error exception at the top
6394 level script context and will exit with the associated exit
6395 status, but doesn't ensure that this exit status is non-zero.
6400 Description: failure of ptlrpc thread startup can cause oops
6401 Details : Starting a ptlrpc service thread can fail if there are a large
6402 number of threads or the server memory is very fragmented.
6403 Handle this without oopsing.
6406 Frequency : always, only if liblustre and non-default acceptor port was used
6408 Description: liblustre cannot connect to servers with non-default acceptor port
6409 Details : tcpnal_set_default_params() was not called and was therefore
6410 ignoring the environment varaible TCPNAL_PORT, as well as other
6411 TCPNAL_ environment variables
6416 Description: two objects could be created on the same OST for a single file
6417 Details : If an OST is down, in some cases it was possible to create two
6418 objects on a single OST for a single file. No problems other
6419 than potential performance impact and spurious error messages.
6423 Bugzilla : 5681, 9562
6424 Description: Client may oops in ll_unhash_aliases
6425 Details : Client dcache may become inconsistent in race condition.
6426 In some cases "getcwd" can fail if the current directory is
6432 Description: Inode refcounting problems in NFS export code
6433 Details : link_raw functions used to call d_instantiate without obtaining
6434 extra inode reference first.
6438 Bugzilla : 9942, 9903
6439 Description: Referencing freed requests leading to crash, memleaks with NFS.
6440 Details : We used to require that call to ll_revalidate_it was always
6441 followed by ll_lookup_it. Also with revalidate_special() it is
6442 possible to call ll_revalidate_it() twice for the same dentry
6443 even if first occurence returned success. This fix changes semantic
6444 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6445 reference on a request referred from the intent.
6446 ll_intent_release() then releases such a request.
6449 Frequency : rare, normally benchmark loads only
6451 Description: unlinked inodes were kept in memory on the client
6452 Details : If a client is repeatedly creating and unlinking files it
6453 can accumulate a lot of stale inodes in the inode slab cache.
6454 If there is no other client load running this can cause the
6455 client node to run out of memory. Instead flush old inodes
6456 from client cache that have the same inode number as a new inode.
6459 Frequency : SLES9 2.6.5 kernel and long filenames only
6460 Bugzilla : 9969, 10379
6461 Description: utime reports stale NFS file handle
6462 Details : SLES9 uses out-of-dentry names in some cases, which confused
6463 the lustre dentry revalidation. Change it to always use the
6467 Frequency : rare, unless heavy write-truncate concurrency is continuous
6468 Bugzilla : 4180, 6984, 7171, 9963, 9331
6469 Description: OST becomes very slow and/or deadlocked during object unlink
6470 Details : filter_destroy() was holding onto the parent directory lock
6471 while truncating+unlinking objects. For very large objects this
6472 may block other threads for a long time and slow overall OST
6473 responsiveness. It may also be possible to get a lock ordering
6474 deadlock in this case, or run out of journal credits because of
6475 the combined truncate+unlink. Solution is to do object truncate
6476 first in one transaction without parent lock, and then do the
6477 final unlink in a new transaction with the parent lock. This
6478 reduces the lock hold time dramatically.
6481 Frequency : rare, 2.4 kernels only
6483 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6484 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6485 before unmounting their filesystem in order to avoid deadlock.
6486 The kernel_locked() function in 2.4 kernels only checks whether
6487 the kernel lock is held, not whether it is this process that is
6488 holding it as 2.6 kernels do.
6493 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6494 Details : The client ptlrpc code may be trying to reconnect to a down
6495 server before a previous connection attempt has timed out.
6496 Increase the reconnect interval to be longer than the connection
6497 timeout interval to avoid sending duplicate connections to
6501 Frequency : echo_client brw_test command
6503 Description: fix echo_client to work with OST preallocated code
6504 Details : OST preallocation code (5137) didn't take echo_client IO path
6505 into account: echo_client calls filter methods outside of any
6506 OST thread and, hence, there is no per-thread preallocated
6507 pages and buffers to use. Solution: hijack pga pages for IO. As
6508 a byproduct, this avoids unnecessary data copying.
6512 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6513 Description: Client can oops in mdc_commit_close() after open replay
6514 Details : It was possible for the MDS to return an open request with no
6515 transaction number in mds_finish_transno() if the client was
6516 evicted, but without actually returning an error. Clients
6517 would later try to replay that open and may trip an assertion
6518 Simplify the client close codepath, and always return an error
6519 from the MDS in case the open is not successful.
6522 Frequency : rare, 2.6 OSTs only
6524 Description: OST may deadlock under high load on fragmented files
6525 Details : If there was a heavy load and highly-fragmented OST filesystems
6526 it was possible to have all the OST threads deadlock waiting on
6527 allocation of biovecs, because the biovecs were not released
6528 until the entire RPC IO was completed. Instead, release biovecs
6529 as soon as they are complete to ensure forward IO progress.
6531 Severity : enhancement
6533 Description: Support for specifying external journal device at mount
6534 Details : If an OST or MDS device is formatted with an external journal
6535 device, this device major/minor is stored in the ext3 superblock
6536 and may not be valid for failover. Allow detecting and
6537 specifying the external journal at mount time.
6542 Description: Mounting an MDS with pending unlinked files may cause oops
6543 Details : target_finish_recovery() calls mds_postrecov() which returned
6544 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6545 considers this an error and immediately begins cleaning up the
6546 lov, just after starting the mds_lov process
6548 Severity : enhancement
6550 Description: Implement 'lfs df' to report actual free space on per-OST basis
6551 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6552 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6553 usage of objects; '-h' to report in human readable format.
6555 ------------------------------------------------------------------------------
6557 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6564 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6565 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6566 at startup during orphan unlinking. Instead, skip buddy update
6567 before setup, as it will be regenerated after recovery is complete.
6572 Description: If an OST is inactive, its locks might reference stale inodes.
6573 Details : lov_change_cbdata() must iterate over all namespaces, even if
6574 they are inactive to clear inode references from the lock.
6576 Severity : enhancement
6577 Frequency : occasional, if non-standard max_dirty_mb used
6579 Description: Client will block write RPCs if not enough grant
6580 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6581 then the client will block writes while waiting for another RPC
6582 to complete instead of consuming its dirty limit. With change
6583 we get improved performance when max_dirty_mb is small.
6585 Severity : enhancement
6586 Bugzilla : 3389, 6253
6587 Description: Add support for supplementary groups on the MDS.
6588 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6589 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6590 lookups for user supplementary groups into a cache.
6594 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6595 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6596 for liblustre O_EXCL is enforced by MDS.
6600 Description: readdir never returns NULL in liblustre.
6601 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6602 dentry in struct dirent.
6606 Frequency : liblustre only, depends on application IO pattern
6607 Description: liblustre clients evicted if not contacting servers
6608 Details : Don't put liblustre clients into the ping_evictor list, so
6609 they will not be evicted by the pinger ever.
6611 Severity : enhancement
6613 Description: Add ability to evict clients by NID from MDS.
6614 Details : By echoing "nid:$NID" string into
6615 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6616 $NID would be instantly evicted from this MDS and from all active
6617 OSTs connected to it.
6621 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6622 Details : lookup_it_finish() used to query file size from OSTs that was not
6627 Description: service threads change working directory to that of init
6628 Details : Starting lustre service threads may pin the working directory
6629 of the parent thread, making that filesystem busy. Threads
6630 now change to the working directory of init to avoid this.
6634 Frequency : during shutdown only
6635 Description: shutdown with a failed MDS or OST can cause unmount to hang
6636 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6641 Frequency : 2.6 only
6642 Description: chmod/chown may include an extra supplementary group
6643 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6644 supplementary group and if none is specified this is used.
6647 Bugzilla : 5479 (6816)
6649 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6650 Details : If lookup is in progress on a file that is unlinked we might try
6651 to revalidate the inode and fail in revalidate after lookup is
6652 complete and ll_file_open() enqueues the open again but
6653 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6656 Frequency : always, if lconf --abort_recovery used
6658 Description: lconf --abort_recovery fails with 'Operation not supported'
6659 Details : lconf was attempting to abort recovery on the MDT device and not
6662 ------------------------------------------------------------------------------
6664 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6669 Frequency : rare (only unsupported configurations with a node running as an
6671 Bugzilla : 6514, 5137
6672 Description: Mounting a Lustre file system on a node running as an OST could
6674 Details : OSTs now preallocates memory needed to write out data at
6675 startup, instead of when needed, to avoid having to
6676 allocate memory in possibly low memory situations.
6677 Specifically, if the file system is mounted on on OST,
6678 memory pressure could force it to try to write out data,
6679 which it needed to allocate memory to do. Due to the low
6680 memory, it would be unable to do so and the node would
6681 become unresponsive.
6683 Severity : enhancement
6685 Description: Addition of lconf --service command line option
6686 Details : lconf now accepts a '--service <arg>' option, which is
6687 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6689 Severity : enhancement
6691 Description: Failover mode is now the default for OSTs.
6692 Details : By default, OSTs will now run in failover mode. To return to
6693 the old behaviour, add '--failout' to the lmc line for OSTs.
6695 Severity : enhancement
6697 Description: Health checks are now provided for MDS and OSTs
6698 Details : Additional detailed health check information on MSD and OSTs
6699 is now provided through the procfs health_check value.
6702 Frequency : occasional, depends on IO load
6704 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6705 after numerous create/delete cycles
6706 Details : The ext3 inode allocation policy would not allocate new inodes
6707 very well on the OSTs because there are no new directories
6708 being created. Instead we look for groups with free space if
6709 the parent directories are nearly full.
6714 Description: Network or server problems during mount may cause partially
6715 mounted clients instead of returning an error.
6716 Details : The config llog parsing code may overwrite the error return
6717 code during mount error handling, returning success instead
6723 Description: MDS can fail to allocate large reply buffers
6724 Details : After long uptimes the MDS can fail to allocate large reply
6725 buffers (e.g. zconf client mount config records) due to memory
6726 fragmentation or consumption by the buffer cache. Preallocate
6727 some large reply buffers so that these replies can be sent even
6728 under memory pressure.
6732 Frequency : rare (liblustre)
6733 Description: fsx running with liblustre complained that using truncate() to
6734 extend the file doesn't work. This patch corrects that issue.
6735 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6736 ATTR_SIZE and lsm use in llu_setattr_raw.
6740 Frequency : rare, only 2.6 kernels
6741 Description: Unusual file access patterns on the MDS may result in inode
6742 data being lost in very rare circumstances.
6743 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6744 code in the 2.6 kernel caused the inode and/or EA data not to
6745 be read from disk, causing single-file corruption.
6749 Frequency : rare, only 2.6 filesystems using extents
6750 Description: Heavy concurrent write and delete load may cause data corruption.
6751 Details : It was possible under high-load situations to have an extent
6752 metadata block in the block device cache from a just-unlinked
6753 file overwrite a newly-allocated data block. We now unmap any
6754 metadata buffers that alias just-allocated data blocks.
6758 Frequency : filesystems with default stripe_count larger than 77
6759 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6760 Details : lconf specifies an inode size of 4096 bytes when the default
6761 stripe_count is larger than 77. This conflicts with the default
6762 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6763 this case to avoid pinning too much memory for large EAs.
6765 ------------------------------------------------------------------------------
6767 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6772 Frequency : rare (extremely heavy IO load with hundreds of clients)
6774 Description: Client is evicted, gets IO error writing to file
6775 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6776 caused clients to be evicted for AST timeouts. The fixes in
6777 bug 5192 mean we no longer need to have such short AST timeouts
6778 so ldlm_timeout has been increased.
6781 Frequency : occasional during --force or --failover shutdown under load
6782 Bugzilla : 5949, 4834
6783 Description: Server oops/LBUG if stopped with --force or --failover under load
6784 Details : a collection of import/export refcount and cleanup ordering
6785 issues fixed for safer force cleanup
6788 Frequency : only filesystems larger than 120 OSTs
6789 Bugzilla : 5990, 6223
6790 Description: lfs getstripe would oops on a very large filesystem
6791 Details : lov_getconfig used kfree on vmalloc'd memory
6794 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6795 Bugzilla : 6242, 6243
6796 Description: reading from files that had been truncated to a non-zero size
6797 but never opened returned no data
6798 Details : ll_file_read() reads zeros from no-object files to EOF
6803 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6804 Details : There's little harm in aborting MDS/OSS recovery and letting it
6805 try again, so I removed the LASSERT and return an error instead.
6807 Severity : enhancement
6809 Description: New debugging infrastructure for tracking down data corruption
6810 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6811 (b) cover more of the client-side code path, and (c) try to narrow
6812 down where problems occurred
6816 Bugzilla : 3819, 4364, 4397, 6313
6817 Description: Racing close and eviction MDS could cause assertion in mds_close
6818 Details : It was possible to get multiple mfd references during close and
6819 client eviction, leading to one thread referencing a freed mfd.
6821 Severity: : enhancement
6822 Bugzilla : 3262, 6359
6823 Description: Attempts to reconnect to servers are now more aggressive.
6824 Details : This builds on the enhanced upcall-less recovery that was added
6825 in 1.4.2. When trying to reconnect to servers, clients will
6826 now try each server in the failover group every 10 seconds. By
6827 default, clients would previously try one server every 25 seconds.
6832 Description: After recovery, certain operations trigger a failed
6833 assertion on a client.
6834 Details : Failing over an mds, using lconf -d --failover, while a
6835 client was doing a readdir() call would cause the client to
6836 LBUG after recovery completed and the readdir() was resent.
6838 Severity : enhancement
6840 Description: Default groups are now added by lconf
6841 Details : You can now run lconf --group <servicename> without having to
6842 manually add groups with lmc.
6845 Frequency : occasional
6847 Description: Nodes with an elan id of 0 trigger a failed assertion
6850 Frequency : always when accessing e.g. tty/console device nodes
6852 Description: tty and some other devices nodes cannot be used on lustre
6853 Details : file's private_data field is used by device data and lustre
6854 values in there got lost. New field was added to struct file to
6855 store fs-specific private data.
6858 Frequency : when exporting Lustre via NFS
6860 Description: NFSD failed occasionally when looking up a path component
6861 Details : NFSD is looking up ".." which was broken in ext3 directories
6862 that had grown large enough to become hashed.
6865 Frequency : Clusters with multiple interfaces not on the same subnet
6867 Description: Nodes will repeatedly try to reconnect to an interface which it
6868 cannot reach and report an error to the log.
6869 Details : Extra peer list entries will be created by lconf with some peers
6870 unreachable. lconf now validates the peer before adding it.
6873 Frequency : Only if a default stripe is set on the filesystem root.
6875 Description: Setting a default stripe on the filesystem root prevented the
6876 filesystem from being remounted.
6877 Details : The client was sending extra request flags in the root getattr
6878 request and did not allocate a reply buffer for the dir EA.
6881 Frequency : occasional, higher if lots of files are accessed by one client
6882 Bugzilla : 6159, 6097
6883 Description: Client trips assertion regarding lsm mismatch/magic
6884 Details : While revalidating inodes the VFS looks up inodes with ifind()
6885 and in rare cases can find an inode that is being freed.
6886 The ll_test_inode() code will free the lsm during ifind()
6887 when it finds an existing inode and then the VFS later attaches
6888 this free lsm to a new inode.
6892 Bugzilla : 6422, 7030
6893 Description: MDS deadlock between mkdir and client eviction
6894 Details : Creating a new file via mkdir or mknod (starting a transaction
6895 and getting the ns lock) can deadlock with client eviction
6896 (gets ns lock and trying to finish a synchronous transaction).
6899 Frequency : occasional
6900 Description: While starting a server, the fsfilt_ext3 module could not be
6902 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6903 kernels. Previously, lconf would still use the ext3 name
6904 when trying to load modules. Now, it will correctly use
6905 ext3 on 2.4 and ldiskfs on 2.6.
6907 Severity : enhancement
6908 Description: The default stripe count has been changed to 1
6909 Details : The interpretation of the default stripe count (0, to lfs
6910 or lmc) has been changed to mean striping across a single
6911 OST, rather than all available. For general usage we have
6912 found a stripe count of 1 or 2 works best.
6914 Severity : enhancement
6915 Description: Add support for compiling against Cray portals.
6916 Details : Conditional compiling for some areas that are different
6920 Frequency : occasional
6921 Bugzilla : 6409, 6834
6922 Description: Creating files with an explicit stripe count may lead to
6923 a failed assertion on the MDS
6924 Details : If some OSTs are full or unavailable, creating files may
6925 trigger a failed assertion on the MDS. Now, Lustre will
6926 try to use other servers or return an error to the
6930 Frequency : occasional
6932 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6933 caused lock timeout during readahead (since 1.4.2).
6934 Details : Processes doing readahead might match a lock that hasn't been
6935 granted yet if there are overlapping and conflicting lock
6936 requests. The readahead process waits on ungranted lock
6937 (original lock is CBPENDING), while OST waits for that process
6938 to cancel CBPENDING read lock and eventually evicts client.
6940 Severity : enhancement
6942 Description: Initial enabling of flock support for clients
6943 Details : Implements fcntl advisory locking and file status functions.
6944 This feature is provided as an optional mount flag (default
6945 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6946 locking are implemented yet, and those that are are not guaranteed
6947 to be completely correct in production environments.
6948 mount -t lustre -o [flock|noflock] ...
6951 Frequency : occasional
6953 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6954 assertion in the ext3 JBD (journaling) layer.
6955 Details : The b_committed_data struct is protected by the big kernel lock
6956 in 2.4 kernels, serializing journal_commit_transaction() and
6957 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6958 access to this struct. In 2.6 kernels there is finer grained
6959 locking to improve SMP performance of the JBD layer.
6963 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6965 -----------------------------------------------------------------------------
6967 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6969 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6970 versions of Lustre. Please update all servers and clients to
6971 version 1.4.2 or later at the same time. You must also run
6972 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6973 to update the configuration logs.
6975 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6976 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6977 - swab llog records of type '0' so we get proper header size/idx (5861)
6978 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6979 - fix rename of one directory over another leaking an inode (5953)
6980 - avoid SetPageDirty on 2.6 (5981)
6981 - don't re-add just-being-destroyed locks to the waiting list (5653)
6982 - when creating new directories, inherit the parent's custom
6983 striping settings if present parent (3048)
6984 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6985 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6986 - don't hold client locks on temporary worklist from l_lru (5666)
6987 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6988 - automatically evict dead clients (5921)
6989 - Update file size properly in create+truncate+fstat case (6196)
6990 - Do not unhash mountpoint dentries, do not allow removal of
6992 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6993 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6994 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6995 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6996 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6997 - Correctly handle reads of files with no objects (6243)
6998 - lctl recover will also mark a device active if deactivate used (5933)
7000 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
7001 - allow --write-conf on an MDS with different nettype than client (5619)
7002 - don't write config llogs to MDS for mounts not from that MDS (5617)
7003 - lconf should create multiple TCP connections from a client (5201)
7004 - init scripts are now turned off by default; run chkconfig --on
7005 lustre and chkconfig --on lustrefs to use them
7006 - upcalls are no longer needed for clients to recover to failover
7008 - add --abort-recovery option to lconf to abort recovery on device
7010 - add support for an arbitrary number of OSTs (3026)
7011 - Quota support protocol changes.
7012 - forward compatibility changes to wire structs (6007)
7013 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
7014 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
7015 - improved "lustre status" script
7016 - initialize blocksize for non-regular files (6062)
7017 - added --disable-server and --disable-client configure options (5782)
7018 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
7019 - Vanilla 2.4.29 support
7020 - increase maximum number of obd devices to 520 (6242)
7021 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
7022 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
7023 SLES 9 2.6.5 kernels
7024 - limit stripes per file to 160 (the maximum EA size) (6093)
7026 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7029 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
7030 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
7031 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
7032 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
7033 - small ext3 extents cleanups and fixes (5733)
7034 - improved mballoc code, several small races and bugs fixed (5733, 5638)
7035 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
7036 - avoid needless client->OST connect, fix handle mismatch (5317)
7037 - fix DLM error path that led to out-of-sync client, long delays (5779)
7038 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
7039 - fix several locking issues related to i_size (5492,5624,5654,5672)
7040 - don't move pending lock onto export if it is already evicted (5683)
7041 - fix kernel oops when creating .foo in unlinked directory (5548)
7042 - fix deadlock in obdfilter statistics vs. object create (5811)
7043 - use time_{before,after} to avoid timer jiffies wrap (5882)
7044 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
7045 - Do not leak request if server was not able to process it (5154)
7046 - If mds_open unable to find parent dir, make that negative lookup(5154)
7047 - don't create new directories with extent-mapping (5909, 5936)
7049 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
7050 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
7051 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
7052 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
7053 - support for RHEL4 kernel on i686 (5773)
7054 - provide error messages when incompatible logs are encountered (5898)
7056 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
7057 * version 1.4.0.10 (1.4.1 release candidate 1)
7059 - don't keep a lock reference when lock is not granted (4238)
7060 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7061 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
7062 - drop import inflight refcount on signal_completed_replay error (5255)
7063 - unlock page after async write error during send (3677)
7064 - handle missing objects in filter_preprw_read properly (5265)
7065 - no transno return for symlink open, don't save no-trasno open (3440)
7066 - don't try to complete elan receive that already failed (4012)
7067 - free RPC server reply state on error (5406)
7068 - clean up thread from ptlrpc_start_thread() on error (5160)
7069 - readahead could read extra page into cache that wasn't ejected (5388)
7070 - prevent races in class_attach/setup/cleanup/detach (5260)
7071 - don't dereference de->d_inode after l_dput of de (5458)
7072 - use "int" for stripe value returned from lock_to_stripe (5544)
7073 - mballoc allocation and error-checking fixes in 2.6 (5504)
7074 - block device patches to fix I/O request sizes in 2.6 (5482)
7075 - look up hostnames for IB nals (5602)
7076 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
7077 - don't start multiple acceptors for the same port (5277)
7078 - fix incorrect LASSERT in mds_getattr_name (5635)
7079 - export a proc file for general "ping" checking (5628)
7080 - fix "lfs check" to not block when the MDS is down (5628)
7082 - service request history (4965)
7083 - put {ll,lov,osc}_async_page structs in a single slab (4699)
7084 - create an "evict_client" /proc entry on OSTs, like the MDS has
7085 - fix mount usage message, return errors per mount(8) (5168)
7086 - change grep [] to grep "[]" in tests so they work in more UMLs
7087 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
7088 - remove ancient LOV_MAGIC_V0 EA support (5047)
7089 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
7090 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
7091 - allow readahead allocations to fail when low on memory (5383)
7092 - mmap locking landed again, after considerable improvement (2828)
7093 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
7095 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
7098 - send OST transaction number in read/write reply to free req (4966)
7099 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7100 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7101 - the watchdog thread now runs as interruptible (5246)
7102 - flock/lockf fixes (but it's still disabled, pending 5135)
7103 - don't use EXT3 constants in llite code (5094)
7104 - memory shortage at startup could cause assertion (5176)
7106 - reorganization of lov code
7107 - single portals codebase
7109 - add extents/mballoc support (5025)
7110 - direct I/O reads in the obdfilter (4048)
7111 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
7113 tbd Cluster File Systems, Inc. <info@clusterfs.com>
7116 - send OST transaction number in read/write reply to free req (4966)
7117 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7118 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7119 - the watchdog thread now runs as interruptible (5246)
7120 - handle missing objects in filter_preprw_read properly (5265)
7121 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7122 - drop import inflight refcount on signal_completed_replay error (5255)
7123 - unlock page after async write error during send (3677)
7124 - return original error code on reconstructed replies (3761)
7125 - no transno return for symlink open, don't save no-trasno open (3440)
7127 - add pid to ldlm debugging output (4922)
7128 - bump the watchdog timeouts -- we can't handle 30sec yet
7129 - extra debugging for orphan dentry/inode bug (5259)
7131 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
7134 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
7135 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
7136 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
7137 - fix NULL dereference in /proc/sys/portals/routes (4827)
7138 - allow failed mdc_close() operations to be interrupted (4561)
7139 - stop precreate on OST before MDS would time out on it (4778)
7140 - don't send partial-page writes before EOF from client (4410)
7141 - discard client grant for sub-page writes on large-page clients (4520)
7142 - don't free dentries not owned by NFS code, check generation (4806)
7143 - fix lsm leak if mds_create_objects() fails (4801)
7144 - limit debug_daemon file size, always print CERROR messages (4789)
7145 - use transno after validating reply (3892)
7146 - process timed out requests if import state changes (3754)
7147 - update mtime on OST during writes, return in glimpse (4829)
7148 - add mkfsoptions to LDAP (4679)
7149 - use ->max_readahead method instead of zapping global ra (5039)
7150 - don't interrupt __l_wait_event() during strace
7152 - add software watchdogs to catch hung threads quickly (4941)
7153 - make lustrefs init script start after nfs is mounted
7154 - fix CWARN/ERROR duplication (4930)
7155 - return async write errors to application if possible (2248)
7156 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
7157 - print NAL number in %x format (4645)
7158 - update barely-supported suse-2.4.21-171 series (4842)
7159 - support for sles 9 %post scripts
7160 - support for building 2.6 kernel-source packages
7161 - support for sles km_* packages
7163 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
7166 - ignore -ENOENT errors in osc_destroy (3639)
7167 - notify osc create thread that OSC is being cleaned up (4600)
7168 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
7169 - reconstruct ost_handle() like mds_handle() (4657)
7170 - create a new thread to do import eviction to avoid deadlock (3969)
7171 - let lconf resolve symlinked-to devices (4629)
7172 - don't unlink "objects" from directory with default EA (4554)
7173 - hold socknal file ref over connect in case target is down (4394)
7174 - allow more than 32000 subdirectories in a single directory (3244)
7175 - fix blocks count for O_DIRECT writes (3751)
7176 - OST returns ENOSPC from object create when no space left (4539)
7177 - don't send truncate RPC if file size isn't changing (4410)
7178 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
7179 - bind to privileged port in socknal and tcpnal (3689)
7181 - rate limit CERROR/CWARN console message to avoid overload (4519)
7182 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
7183 - basic mmap support (3918)
7184 - kernel patch series update from b1_4 (4711)
7186 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
7189 - avoid crash during MDS cleanup with OST shut down (2775)
7190 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
7191 - don't use bad inodes on the MDS (3744)
7192 - dynamic object preallocation to improve recovery speed (4236)
7193 - don't hold spinlock over lock dumping or change debug flags (4401)
7194 - don't zero obd_dev when it is force cleaned (3651)
7195 - print grants to console if they go negative (4431)
7196 - "lctl deactivate" will stop automatic recovery attempts (3406)
7197 - look for existing locks in ldlm_handle_enqueue() (3764)
7198 - don't resolve lock handle twice in recovery avoiding race (4401)
7199 - revalidate should check working dir is a directory (4134)
7201 - don't always mark "slow" obdfilter messages as errors (4418)
7203 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
7206 - don't close LustreDB during write_conf until it is done (3860)
7207 - fix typo in lconf for_each_profile (3821)
7208 - allow dumping logs from multiple threads at one time (3820)
7209 - don't allow multiple threads in OSC recovery (3812)
7210 - fix debug_size parameters (3864)
7211 - fix mds_postrecov to initialize import for llog ctxt (3121)
7212 - replace config semaphore with spinlock (3306)
7213 - be sure to send a reply for a CANCEL rpc with bad export (3863)
7214 - don't allow enqueue to complete on a destroyed export (3822)
7215 - down write_lock before checking llog header bitmap (3825)
7216 - recover from lock replay timeout (3764)
7217 - up llog sem before sending rpc (3652)
7218 - reduce ns lock hold times when setting kms (3267)
7219 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
7220 - fix NULL deref and obd_dev leak on setup error (3312)
7221 - replace some LBUG about llog ops with error handling (3841)
7222 - don't match INVALID dentries from d_lookup and spin (3784)
7223 - hold dcache_lock while marking dentries INVALID and hashing (4255)
7224 - fix invalid assertion in ptlrpc_set_wait (3880)
7226 - add libwrap support for the TCP acceptor (3996)
7227 - add /proc/sys/portals/routes for non-root route listing (3994)
7228 - allow setting MDS UUID in .xml (2580)
7229 - print the stack of a process that LBUGs (4228)
7231 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
7234 - don't cleanup request in ll_file_open() on failed MDS open (3430)
7235 - make sure to unset replay flag from failed open requests (3440)
7236 - if default stripe count is 0, use OST count for inode size (3636)
7237 - update parent mtime/ctime on client for create/unlink (2611)
7238 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
7239 - free recovery state on server during a forced cleanup (3571)
7240 - unregister_reply for resent reqs (3063)
7241 - loop back devices mounting and status check on 2.6 (3563)
7242 - fix resource-creation race that can provoke i_size == 0 (3513)
7243 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
7244 - more debugging for page-accounting assertion (3746)
7245 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
7246 - avoid deadlock after precreation failure (3758)
7247 - fix race and lock order deadlock in orphan handling (3450, 3750)
7248 - add validity checks when grabbing inodes from l_ast_data (3599)
7250 - add /proc/.../recovery_status to obdfilter (3428)
7251 - lightweight CDEBUG infrastructure, debug daemon (3668)
7252 - change default OSC RPC parameters to be better on small clusters
7253 - turn off OST read cache for files smaller than 32MB
7254 - install man pages and include them in rpms (3100)
7255 - add new init script for (un)mounting lustre filesystems (2593)
7256 - run chkconfig in %post for init scripts (3701)
7257 - drop scimac NAL (unmaintained)
7259 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
7262 - clean kiobufs before and after use (3485)
7263 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
7264 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7265 - change init script to fail more clearly if not run as root (1528)
7266 - allow clients to reconnect during replay (1742)
7267 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
7268 - don't do DNS lookups on NIDs too small for IP addresses (3442)
7269 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
7270 - fix cond_resched (3554)
7271 - only evict unfinished clients after recovery (3515)
7272 - allow bulk resend, prevent data loss (3570)
7273 - dynamic ptlrpc request buffer allocation (2102)
7274 - don't allow unlinking open directory if it isn't empty (2904)
7275 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
7276 - remove extraneous obd dereference causing LASSERT failure (3334)
7277 - don't use get_cycles() when creating temp. files on the mds (3156)
7278 - hold i_sem when setting i_size in ll_extent_lock() (3564)
7279 - handle EEXIST for set-stripe, set proper directory name (3336)
7281 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
7282 - fix ksocknal_fmb_callback() error messages (2918)
7284 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
7287 - don't copy lvb into (possibly NULL) reply on error (2983)
7288 - don't deref dentry after dput, don't free lvb on error (2922)
7289 - use the kms to determine writeback rpc length (2947)
7290 - increment oti_logcookies when osc is inactive (2948)
7291 - update client's i_blocks count via lvb messages (2543)
7292 - handle intent open/close of special files properly (1557)
7293 - mount MDS with errors=remount-ro, like obdfilter (2009)
7294 - initialize lock handle to avoid ASSERT on error cleanup (3057)
7295 - don't use cancelling-locks' kms values (2947)
7296 - use highest lock extent for kms, not last one (2925)
7297 - don't dereference ERR_PTR() dentry in error handling path (3107)
7298 - fix thread race in portals_debug_dumplog() (3122)
7299 - create lprocfs device entries at setup instead of at attach (1519)
7300 - common AST error handler, don't evict client on completion race (3145)
7301 - zero nameidata in detach_mnt in 2.6 (3118)
7302 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7303 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7304 - initialize RPC timeout timer earlier for 2.6 (3219)
7305 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7306 - print nal/nid for unknown nid (3258)
7307 - additional checks for oscc recovery before doing precreate (3284)
7308 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7309 - don't crash in mdc_close for bad permissions on open (3285)
7310 - zero i_rdev for non-device files (3147)
7311 - clear page->private before handing to FS, better assertion (3119)
7312 - tune the read pipeline (3236)
7313 - fix incorrect decref of invalidated dentry (2350)
7314 - provide read-ahead stats and refine rpc in flight stats (3328)
7315 - don't hold journal transaction open across create RPC (3313)
7316 - update atime on MDS at close time (3265)
7317 - close LDAP connection when recovering to avoid server load (3315)
7318 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7319 - don't leak open file on MDS after open resend (3325)
7320 - serialize filter_precreate and filter_destroy_precreated (3329)
7321 - loop device shouldn't call sync_dev() for nul device (3092)
7322 - clear page cache after eviction (2766)
7323 - resynchronize MDS->OST in background (2824)
7324 - refuse to mount the same filesystem twice on same mountpoint (3394)
7325 - allow llmount to create routes for mounting behind routers (3320)
7326 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7327 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7328 - fix rare problem with rename on htree directories (3417)
7330 - allow default OST striping configuration per directory (1414)
7331 - fix compilation for qswnal for 2.6 kernels (3125)
7332 - increase maximum number of MDS request buffers for large systems
7333 - change liblustreapi to be useful for external progs like lfsck (3098)
7334 - increase local configuration timeout for slow disks (3353)
7335 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7337 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7340 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7341 - don't overwrite extent policy data in reply if lock was blocked (2901)
7342 - drop filter export grants atomically with removal from device (2663)
7343 - del obd_self_export from work_list in class_disconnect_exports (2908)
7344 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7345 - swab reply message in mdc_close, other PPC fixes (2464)
7346 - fix destroying of named logs (2325)
7347 - overwrite old logs when running lconf --write_conf (2264)
7348 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7349 - fix race in target_handle_connect (2898)
7350 - mds_reint_create() should take same inode create lock (2926)
7351 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7352 - don't close files for self_export to avoid uninitialized obd (2936)
7353 - allow MDS with the same name as client node (2939)
7354 - hold dentry reference for closed log files for unlink (2325)
7355 - reserve space for all logs during transactions (2059)
7356 - don't evict page beyond end of stripe extent (2925)
7357 - don't oops on a deleted current working directory (2399)
7358 - handle hard links to targets without a parent properly (2517)
7359 - don't dereference NULL lock when racing during eviction (2867)
7360 - don't grow lock extents when lots of conflicting locks (2919)
7362 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7365 - account for cache space usage on clients to avoid data loss (974)
7366 - lfsck support in lustre kernel code (2349)
7367 - reduce journal credits needed for BRW writes (2370)
7368 - orphan handling to avoid losing space on client/server crashes
7369 - ptlrpcd can be blocked, stopping ALL progress (2477)
7370 - use lock value blocks to assist in proper KMS, faster stat (1021)
7371 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7372 - recovery for initial connections (2355)
7373 - fixes for mds_cleanup_orphans (1934)
7374 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7375 - block all file creations until orphan recovery completes (1901)
7376 - client remove rq_connection from request struct (2423)
7377 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7378 - recovery timer race (2670)
7379 - mdc_close recovey bug (2532)
7380 - ptlrpc cleanup bug (2710)
7381 - mds timeout on local locks (2588)
7382 - namespace lock held during RPCs (2431)
7383 - handle interrupted sync write properly (2503)
7384 - don't try to handle a message that hasn't been replied to (2699)
7385 - client assert failure during cleanup after abort recovery (2701)
7386 - leak mdc device after failed mount (2712)
7387 - ptlrpc_check_set allows timedout requests to complete (2714)
7388 - wait for inflight reqs when ptlrpcd finishes (2710)
7389 - make sure unregistered services are removed from the srv_list
7390 - reset bulk XID's when resending them (caught by 1138 test)
7391 - unregister_bulk after timeout
7392 - fix lconf error (2694)
7393 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7394 - readahead locks pages, leaves pending causing memory pressure (2673)
7395 - increase OST request buffers to 4096 on large machines (2729)
7396 - fix up permission of existing directories in simple_mkdir (2661)
7397 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7398 - don't assign transno to errored transactions (2742)
7399 - don't delete objects on OST if given a bogus objid from MDS (2751)
7400 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7401 - if rq_no_resend, then timeout request after recovery (2432)
7402 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7403 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7404 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7405 - MDS recovery shouldn't delete objects that it has given out (2730)
7406 - if enqueue arrives after completion, don't clobber LVB (2819)
7407 - don't unlock pages twice when trigger_group_io returns error (2814)
7408 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7409 - don't write pages to disk if there was an error (1450)
7410 - don't ping imports that have recovery disabled (2676)
7411 - take buffered bytes into account when balancing socknal conn (2817)
7412 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7413 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7414 - remove little-endian swabbing of llog records (1987)
7415 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7416 - retry reposting request buffers if they fail (1191)
7417 - grow extent at grant time to avoid granting a revoked lock (2809)
7418 - lock revoke doesn't evict page if covered by a second lock (2765)
7419 - disable VM readahead to avoid reading outside lock extents (2805)
7421 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7422 - updated kernel patches for hp-2.4.20 kernel (2681)
7424 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7427 - fix truncated write corruption (2366)
7428 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7430 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7431 - ensure len(uuid) < 37 in lmc (1171)
7432 - fix ia64 OOPS in llog_test (2255)
7433 - zero end of page at obdfilter for partial page writes (2648)
7434 - don't leave stale dentries around after renames (bug 2428)
7435 - fix timeouts when evicting a client with a single lock held (2642)
7436 - set deadline for the initial HELLO message to drain (2634)
7437 - print out dotted-quad IP addresses in the socknal (2302)
7439 - additional debugging for MDS client eviction problem (2443)
7440 - fix mkfsoptions support for osts (2603, 2604)
7442 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7445 - add series for the vanilla 2.6.0 kernel
7446 - add series for the vanilla 2.4.24 kernel
7447 - add series for a cray x86/64 UL kernel drop
7448 - fix xattr patches for the vanilla 2.4.19 series
7450 - generate true UUIDs in lmc (1171)
7451 - have portals stack dumping break in UML (2466)
7452 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7453 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7454 - wake callback waiting threads on client eviction (2460)
7455 - Add --ptldebug and --subsystem to lmc (1719)
7456 - update assertion to allow safe interrupt allocation
7457 - set rq_no_resend for cancel requests (2432)
7458 - recalculate ptlrpcd timeout after resend (2494)
7459 - call vfs_rmdir when removing pending directories (2368)
7460 - fix renaming a file to itself (2429)
7461 - lmc creates a default one-stripe lov (2454)
7462 - expand procfs space to handle large clusters (2326)
7463 - increase UML stack to avoid overflow
7464 - update lconf's list of debug and subsystem masks
7465 - fix lfs find --obd (2510)
7466 - /proc tunable for disabling filter read caching (2591)
7467 - stop rpm packages from altering slapd.conf (2301)
7468 - disable nagle in the socknal under 0conf (2578)
7469 - choose mds inode size based on stripe count (2572)
7470 - fix kernel-source rpm problems (2516)
7472 - add --disable-doc to avoid pdf generation (2421)
7473 - update documentation, tests, type-os, comments
7474 - avoid format warnings on ia64
7475 - remove the TOE NAL
7476 - tiny code cleanups by removing unused fields
7478 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7481 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7482 - remove the most blatant lies from BUILDING file (2371)
7483 - change default debug level to reasonable production setting
7484 - reduce client side cache size to reduce cache flush time
7485 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7486 - make TCP zerocopy and pinger support enabled by default (2476)
7487 - sync writes completed after process exits caused crashes (2319)
7488 - maintain correct mount count on the MDS (2356)
7489 - backout 1557, because 2316 wasn't really fixed
7490 - better file I/O statistics gathering in /proc
7491 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7492 - another kernel patch to fix zero-copy TCP function export
7493 - don't take duplicate lock when processing re-sent getattr (2420)
7494 - lctl uses obd_self_export instead of creating new conn (2353)
7495 - MDS/OST recovery case which requires object creation asserted (2425)
7496 - move lfs from /usr/sbin to /usr/bin in packages
7497 - fix race between mds_client_add and mds_client_free (2417)
7498 - use kmalloc instead of slabs in portals (2430)
7499 - don't create duplicate records when a failover MDS is present (2442)
7500 - remove unnecessary mount age check (2332)
7501 - don't remove directory inodes from locks prematurely (2451)
7502 - don't break if MDS service name is the same as hostname (2103)
7503 - fix races in client write RPC generation when cache full (2482)
7505 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7508 - remove now-unused request->rq_obd (278)
7509 - if an allocation fails, print out how much memory we've used (1933)
7510 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7511 - add the "configurable stack size" patch to most series files (1256)
7512 - ability to write large log records, for 100+ OST configs (2306)
7513 - fix NULL deref when filter_prep fails (2314)
7514 - fix operator precedence error in filter_sync
7515 - dynamic allocation of socknal TX descriptors (2315)
7516 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7517 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7518 - ensure that CWARN messages in Portals always get to the syslog
7519 - __init/__exit are not for prototype decls (ldlm_init/exit)
7520 - x86-64 compile warning fixes
7521 - fix gateway LMC keyword conflict (2318)
7522 - fix MDS lock inversions in getattr/reint paths (1844)
7523 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7524 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7525 - DLM race condition prevented some lock evictions (2328)
7526 - ENOMEM detection and retry on socknal sends (2230)
7527 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7528 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7529 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7530 - better error messages when a client is rejected during recovery (1505)
7531 - avoid cancelling locks which were never granted, after failure (2330)
7532 - fix i_sem/journal inversion in mds_client_add (2333)
7533 - fix truncate/getattr lock cycle deadlock (2334)
7534 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7535 - fix two rare exit paths which could leak an l_lock() ref (2321)
7536 - fencepost error in MDS/OST orphan recovery (2226)
7537 - make log record alignment 8 bytes (1988)
7538 - lstripe now fails when requested offset > ost_count (2237)
7539 - ensure that all kernel series have a complete list.h (1607)
7540 - fix crashes in special-file operations (2316)
7541 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7543 - allow configurable automake binary, for testing new versions
7544 - small update to the lfs documentation
7546 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7548 * fix negative export reference count in fsfilt_sync (2312)
7550 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7551 * release candidate 0.9.1
7553 - orphans are moved into the PENDING directory for possible recovery
7554 - replayed opens now open by fid for orphan/rename safety (1042)
7555 - last close of an orphan inode generates a transno (683)
7556 - chdir() and mount() now pin the directory entry (1020)
7557 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7558 - discard very old requests without processing them (1502)
7559 - remove some common, well-understood CERRORs (1505)
7560 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7561 - clear "grant" flags in OST replies until OST grant code lands (1644)
7562 - fix read performance by not clobbering i_blksize on client (1598)
7563 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7564 - make LRU size a /proc tunable, clears locks when reduced (707)
7565 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7566 - server locks take references on exports now (1558)
7567 - build fixes for 2.4.20-rh trees (1663)
7568 - return an error from lov_create if all OSCs are inactive (1751)
7569 - fix import levels when a reconnect happens without a timeout (1597)
7570 - exit early from mds_open if we get a lookup error (1749)
7571 - partial page read at EOF wouldn't wait for disk before sending (1642)
7572 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7573 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7574 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7575 - write/truncate lock inversion (1639)
7576 - Don't auto-load obdclass, portals modules during cleanup (1495)
7577 - fix timestamps from jumping to "now" (1763)
7578 - extra journal assertions (1648)
7579 - add an extra multiunlink test (1771)
7580 - fix read_record/write_record API (1776)
7581 - fix leak of offset_extent, possible incorrect i_size later (1772)
7582 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7583 - Debugging for the kqswnal_get_idle_tx problems (1820)
7584 - Allow recovery to be attempted multiple times (1536)
7585 - Write out MDS last_rcvd file after it is first created (1600)
7586 - Fix tx_descriptor leak in failed transmit situations (1827)
7587 - ext3 journaling fixes for assertion failure after IO error (1871)
7588 - class_export_put() on freed export after completion AST error (1896)
7589 - Fix revalidate looping in VFS (1322)
7590 - Don't access a freed export during MDS_REINT timeout (1521)
7591 - Add open-unlink recovery support on the MDS (1673,1764)
7592 - Return an error if no MDS data was read from last_rcvd (1946)
7593 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7594 - Don't setup a disk device that doesn't match exported UUID (317)
7595 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7596 - avoid committing NULL handle in force close
7597 - local.sh is now a one-stripe LOV configuration
7598 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7599 - fix ext3 htree duplicate directory entry corruption (1516)
7600 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7601 - update to new LOV EA format (2097)
7602 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7604 2003-06-15 Phil Schwan <phil@clusterfs.com>
7607 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7608 - per-import/export recovery handling (958, 931, 959)
7609 - multiple last-rcvd slots, for serving multiple FSes (949)
7610 - connections are again shared between multiple imp/exports (963, 964)
7611 - "umount -f" would hang if any requests needed to be sent (393, 978)
7612 - avoid pinning large req buffer by copying for queued messages (989)
7613 - add "uuid" to "lctl device" command to help upcalls (991)
7614 - "open" RPCs with transnos would confuse recovery counters (1037)
7615 - do proper endian conversion of all wire messages (288, 340, 891)
7616 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7617 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7618 - fix ptlrpc_connection leak in target_handle_connect (1174)
7619 - fix import refcounting bug in OST and MDS cleanup (1134)
7620 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7621 - fix very unlikely obd_types race condition (501)
7622 - remove osc_open hack for echo_client (1187)
7623 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7624 - a failure in read_inode2 leads to deadlock (1139)
7625 - cancel ack-locks as soon as transaction is committed (1072)
7626 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7627 - make sure to commitrw after any preprw to avoid deadlock (1162)
7628 - failing to execute a file in a lustre FS would lock inode (1203)
7629 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7630 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7631 - obd_uuid2tgt would walk off the end of the list (1255)
7632 - on IA64 the osc would give portals incorrect bulk size (1258)
7633 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7634 - fix lock inversion caused by new llite matching code (1282)
7635 - limit the number of dirty pages on a client to 10MB (1286)
7636 - timed out locks were not being corrected cancelled (1289)
7637 - fix O_DIRECT above 4GB on IA-32 (1292)
7638 * major user-visible changes
7639 - fail out/fail over policy now controlled by the upcall (993)
7641 - add OBD_PING to check server availability and failure (954)
7642 - lustre messages are now sent in sending host order (288, 340, 891)
7643 - add eadatalen to MDS getattr reply (340)
7644 - OST read replies may contain second buffer, with per-page status (593)
7646 2003-03-11 Phil Schwan <phil@clusterfs.com>
7649 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7650 - failed open()s could cause deadlock; fixed (867, 869)
7651 - stop cancelling OST locks when files are closed (481)
7652 - overlapping XID spaces caused network corruption (851, 853)
7653 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7654 - setattr_raw added, to do single-RPC, server-side setattrs
7655 - lmc/lconf syntax change for OST UUIDs
7656 - fix crashy race condition between ptlrpc_free_req and osc_close
7657 - don't use request in mdc_enqueue if we hit a timeout (889)
7658 - don't set the inode i_size for regular files from the MDS (896)
7659 - handle out of order completion AST (842)
7660 - don't LBUG if a lock request times out after receiving AST (913)
7661 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7662 - if a bad lock AST arrives, send an error instead of dropping entirely
7663 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7664 - fix leak in bulk IO when only partially completed (899, 900, 926)
7665 - fix O_DIRECT for ia64 (55)
7666 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7667 - C-z after timeout could hang a process forever; fixed (977)
7669 - client-side I/O cache (678, 924, 929, 941, 970)
7671 - READPAGE and SETATTRs which don't take server-side locks get
7674 2003-02-11 Phil Schwan <phil@clusterfs.com>
7677 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7678 - Fix fsfilt_extN_readpage() to read a full page of directory
7679 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7680 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7681 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7682 - Most intent operations were removed, in favour of a new RPC mode
7683 that does a single RPC to the server and bypasses most of the VFS
7684 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7685 - Aggressively cancel local locks on DLM servers
7686 - mds_reint_unlink sends EA to the client if it's the last nlink.
7687 client uses that EA to unlink OST objects.
7688 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7689 - recursive symlinks were fixed (439)
7690 - fixed NULL deref in DEBUG_REQ
7691 - filter_update_lastobjid no longer calls sync, which annoyed extN
7692 - fixed multi-client small-writes to a single file problem (445)
7693 - fixed mtime updates during file writes (607)
7694 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7695 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7696 - fixed Linux OST size reporting problem (444, 656)
7697 - OST now updates object mtime with writes or setattr (607, 619)
7698 - client verifies file size before zeroing page past EOF (445)
7699 - OST now writes last allocated objid to disk with allocation (108)
7700 - LOV on echo now works (409)
7702 - mds_reint_unlink sends a new buffer, with the EA included. this
7703 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7704 set if a regular file was being unlinked, and it was the last link
7705 - use PtlGet from the target for bulk writes (315)
7706 - OST now updates object mtime with writes or setattr (607, 619)
7707 - LDLM now has a grant-time callback to revalidate locked items, if
7709 - Many MDS operations were reorganized to combat race conditions
7711 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7712 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7714 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7717 - Fully reactivate OST imports after reconnection (512, others)
7718 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7719 - More graceful error handling for truncating on dead OST (515)
7720 - Don't error out unless we're actually accessing dead stripes (474)
7721 - Fix garbage sizes when stripes are missing (410)
7722 - LRU counters were broken, causing constant lock purge (433, 432)
7723 - garbage on read from stripes with failed OSTs (441)
7724 - mark OSCs as active before reconnecting during recovery (438)
7725 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7726 - lfind did not preserve OST order in output (443)
7727 - symlinks cause hung clients, incorrect data (439)
7728 - stop dereferencing request after dropping refcount (457)
7729 - don't LASSERT(spin_is_locked) on non-SMP (455)
7730 - fixes for many rename() bugs
7731 - fstat didn't correctly synchronize attributes (399)
7732 - server must handle lock cancellation during blocking AST prep (487)
7733 - bulk descriptors were free()d too soon (511)
7734 - fix paths in lconf, which would load incorrect modules (451, 507)
7735 - fix confusing lconf 'host not found' error message (386)
7736 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7737 - fix race condition in mdc_blocking_ast() for inode access (526)
7738 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7739 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7740 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7741 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7742 - fix lstripe in several ways (kernel side) (527)
7743 - fix request leak in ldlm_cli_enqueue (262)
7744 - incorrect OSC was marked inactive after OST failure
7745 - call mds_fs_cleanup before unmounting filesystem (524)
7746 - fix races between taking ns_lock and ldlm_lock_change_resource
7747 - fix races updating LOV export open file list
7748 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7749 - fix recovery NULL deref in ldlm_cli_cancel_unused
7750 - fix some DLM races by using new hash table for lock handles (419)
7751 - permit the client to specify desired inodes, at replay
7752 - duplicate requests when we queue them for replay reintegration
7753 - fix last_rcvd offset calculation
7754 - sync after each recovered transaction, so we always make progress
7755 - never, not always, ERESTART requests without transnos
7756 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7758 - skip replay if the MDS says that the client is already connected
7759 - don't check for a recovery-enabled export to match lctl's UUID
7760 - don't INC_USE_COUNT for phantom exports
7761 - don't crash when cleaning up phantom exports (567)
7762 - don't double-finish or set replay data for errored mdc_open requests
7763 - abort requests when they time out, so we don't get old replies
7764 - send/receive replies for AST messages again
7765 - if the client says that it doesn't have the lock, cancel it on the
7767 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7768 mark it as destroyed, it will all work out when decref is called
7769 - fix module use counts (22, 581)
7771 - ASTs now expect a reply (server cancels lock on error reply)
7773 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7776 - fix many simultaneous client startup (392)
7777 - fix dentry->d_it clobbering
7778 - credentials weren't being shipped for readdir/getattr operations
7779 - remove invalid assertions triggered during some concurrent MD
7781 - proper Lustre versions added (336, 389)
7782 - fix memory leak for create error case (398)
7783 - fix LOV locking bug that would get cli/srv out of sync
7784 - fix echo client over LOV (409)
7785 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7786 - fix double-O_EXCL intent crash (424)
7787 - avoid sending multiple lock CANCELs (352)
7789 - MDS can do multi-client recovery (modulo bugs in new code)
7791 - many updates, edits, cleanups
7793 2002-11-18 Phil Schwan <phil@clusterfs.com>
7796 - fix null d_it dereference (346)
7797 - fix full OST/dbench hang (333)
7798 - fix permission problem with file removal (286)
7799 - fix removal of OSCs from LOV when they fail
7800 - fix NULL deref during bulk timeout (214)
7801 - fix problems related to multiple filesystems on one MDS (241)
7802 - fixed serious subtle metadata locking bugs
7803 - free locks on clients when inodes are removed due to memory
7805 - fix inode pointer in lock data (285)
7806 - partial support for multiple MDS on a single host (241)
7807 - data locks weren't cancelled at clear_inode time (290, 311)
7808 - intent locks could lead to unbounded lock growth (205)
7809 - added a maximum lock count, an LRU list, and a flusher
7810 - fix multiple rename (365)
7811 - properly abstracted the echo client
7812 - OSC locked 1 byte too many; fixed
7813 - rewrote brw callback code:
7814 - fixed recovery bugs related to LOVs (306)
7815 - fixed too-many-pages-in-one-write crash (191)
7816 - fixed (again) crash in sync_io_timeout (214)
7817 - probably fixed callback-related race (385)
7819 - Add capability to MDS protocol
7820 - LDLM cancellations and callbacks on different portals
7822 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7825 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7826 - defer open object destruction to close time (601981, 138)
7827 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7828 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7829 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7830 - don't call obd_setattr in ll_file_release for destroyed objects
7831 * protocol change to lustre_msg: move |version| and add |flags|
7832 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7833 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7834 * added replay of create, unlink, link and rename operations during
7835 MDS failover; recovery should be much more robust now
7836 * remove failed OSCs from LOVs (only lov_create uses this so far)
7837 * the lustre-HOWTO was brought (more) up to date (582544)
7839 2002-10-23 Phil Schwan <phil@clusterfs.com>
7842 - in-use dentries weren't being reused properly (617851)
7843 - prevent multiple LDLM setup (599178)
7844 - fix LOV size calculations for truncate (617853)
7845 - fix client handling of MDS intent errors (POSIX)
7846 - fix permission bug in lovstripe.c test (624321)
7847 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7848 - truncate past end of file could corrupt data
7849 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7850 - a race in recovery could return ETIMEDOUT to apps (623947)
7851 - building outside the source directory was fixed
7852 * the lustre-HOWTO was brought (more) up to date (582544)
7853 * major progress was made on recovery functionality
7855 2002-10-10 Phil Schwan <phil@clusterfs.com>
7858 - recovery deadlock fix
7859 - rm -rf causes LBUG fix (617817)
7860 - file open by multiple tasks fix (618962)
7861 - directory permissions bugs (602707 and 620007)
7862 - journal_stop fixed with locking (611313)
7863 - O_APPEND failures resolved (618273, perhaps 614459)
7864 - lconf PATH fix (619770)
7865 - IA64 build fix (621450)
7866 - RPC buffer sizes scale with amount of memory
7868 2002-10-01 Phil Schwan <phil@clusterfs.com>
7871 - locks would be cancelled without throwing away data pages,
7872 resulting in inconsistent data (605627)
7873 - inode attributes were not always being refreshed (605627, 612449)
7874 - lconf now continues to cleanup after lctl reports an error
7875 - MDS now enforces user permissions (602707)
7876 - lprocfs cleanup fixed, but not yet enabled (614157)
7877 - fixed infinite server hang, should a client not respond to an AST
7878 - avoid going into recovery if user calls readlink() with a buffer
7879 that's too small (613941)
7880 - AST RPCs no longer require replies (614867) -- this may be changed
7881 - don't crash server if client sends an IOV that's too big (611336)
7882 - fixed lock conversion deadlock (611892)
7883 - fixed the following of symlinks (614622)
7884 * recovery: the server can remove locks from a client that dies, other
7885 clients can make progress
7886 * more extN patch fixes
7887 * compile-time configurable ptlrpc buffer allocations
7889 - collaborative read cache document
7890 - Lustre Lite Performance CDR document-in-progress
7892 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7895 - fix typo in patch-2.4.18
7897 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7900 - clear ptlrpc request each time in handle_incoming_request()
7901 - unlink of files now destroys the object on the OST
7903 2002-09-19 Peter Braam <braam@clusterfs.com>
7905 * add hard link support
7906 * change obdfile creation method
7907 * kernel patch changed
7909 2002-09-19 Peter Braam <braam@clusterfs.com>
7912 - stack overflow bug in extN fixed
7914 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7916 * documentation updates
7917 - add man pages for config tools
7918 - update tests/README to describe testing with new config tools
7919 - finish metadata API descriptions
7920 * bug fixes and cleanups
7921 - statfs workaround for 16TB limit
7922 - LOV stripe allocation improved, can stripe on subset of OSTs
7923 - LOV file size/IO offset was wrong for files > 4GB in size
7924 - object EA data was being dropped, caused files to be unreadable
7925 - memory overflow with non-LOV OST caused memory corruption
7926 - fixed regression tests to work with new config tools, obdfilter
7927 - fixed bug when directory size became larger than 1 block
7928 - fixed bug (for single client case) when PWD was deleted
7929 - invalidate local directory pages when doing intent-based ops
7930 - avoid LDLM oops when lock callback contained bad data
7932 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7934 * documentation updates
7935 * bug fixes and cleanups
7936 - configuration tools
7939 - 64-bit compile warnings
7940 - 64-bit internal statfs data
7942 * test_brw on persistent OST devices
7944 * lprocfs (disabled)
7946 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7948 * documentation updates
7949 * bug fixes and cleanups
7950 * configuration tools
7952 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7955 * many small fixes to 0_5_4
7956 * io/network handling
7957 * thinkos in MDS operations
7959 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7962 * crucial basic fixes to 0.5.3
7963 * IOR, Iozone work over Elan
7966 2002-08-07 Phil Schwan <phil@clusterfs.com>
7967 * version 0_5_3, our first alpha
7968 * we use the new Portals iovs
7969 * documentation updates
7970 * bug fixes and cleanups
7971 * small changes in the DLM wire protocol
7973 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7974 * version 0_5_1 with some initial stability,
7975 * locking on MD and file I/O.
7976 * documentation updates
7977 * several bug fixes since 0.5.0
7978 * small changes in wire protocol
7980 2002-07-18 Phil Schwan <phil@clusterfs.com>
7982 * delivered as Lustre Light Alpha
7983 * fixed a crash after handling invalid MDS requests
7984 * fixed directory pages for architectures with non-4k pages sizes
7986 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7987 * release version v0_4_4
7988 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7990 * Fixes a number of interruption problems with OST operations.
7991 * Update documentation for portals header changes
7992 * Move all wire protocol structs/defines to lustre_idl.h
7993 * Fixes symlink length bug.
7994 * Add tcpdump to repository.
7996 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7997 * release version v0_4_3
7998 * Fixes statfs for inodes on extN.
7999 * Fixes bug in runtests which would delete /etc/hosts.
8000 * Use 64-bit object IDs wherever possible (not into VFS though)
8001 Remove ost_get_info, which is unused by lustre, and out of date.
8003 2002-07-03 Peter Braam <braam@clusterfs.com>
8004 * release version v0_4_2 Fixes a lookup error (type not passed)
8005 * move forward to head of Portals
8006 * move forward to latest Lustre kernel
8008 2002-06-25 Peter Braam <braam@clusterfs.com>
8009 * release version v0_4_1. Hopefully stable on single node use.