1 2011-06-30 Whamcloud, Inc.
4 2.6.18-238.12.1.el5 (RHEL 5)
5 2.6.32-131.2.1.el6 (RHEL 6) - patchless client only
6 * Recommended e2fsprogs version: 1.41.90.wc3
7 * Use ext4-based ldiskfs as default for RHEL5
8 * Add support for 24TB LUN (RHEL5 server only)
10 Severity : enhancement
12 Description: Add support for RHEL5.6 (2.6.18-238.12.1.el5)
14 Severity : enhancement
15 Jira : LU-62, LU-73, LU-402
16 Description: Add RHEL6 client support
20 Description: Fix client crash introduced in 1.8.5
21 Details : The patch from bugzilla ticket 18213 landed for 1.8.5 can cause
22 a crash in mdc_exit_request() when the process is interrupted.
26 Description: ls -l reports wrong file size
27 Details : Always update LVB from disk when a glimpse callback returns an
30 Severity : major (2.6.32-based kernels only)
32 Description: Add workaround for a race causing an assertion failure in
33 clear_inode() on lustre clients.
37 Description: Fix interoperability issue between 1.8 clients and 2.x servers
38 Details : The ldlm hash handling in Lustre 2.x is incompatible with the one
39 used in Lustre 1.8. Add support for 64-bit dir name hash to the
44 Description: slow IO with read-intense application
45 Details : Align the readahead extent by 1M after when it is trimed by
50 Description: Skip quotacheck on administratively disabled OSTs
54 Description: Some autogen improvements
56 Severity : enhancement
57 Jira : LU-123, LU-296, LU-230
58 Description: Add support for yaml data logging and auster script.
60 Severity : enhancement
62 Description: Allow OSTs to be created with no primary node
63 Details : Add a --servicenode parameter for mkfs.lustre to treat all
64 service nodes equally.
66 -------------------------------------------------------------------------------
68 2011-05-12 Oracle, Inc.
70 * Support for kernels:
71 2.6.16.60-0.42.8 (SLES 10),
72 2.6.27.39-0.3.1 (SLES 11),
73 2.6.18-194.3.1.el5 (RHEL 5)
74 2.6.18-194.3.1.0.1.el5 (OEL 5)
75 * Client support for unpatched kernels:
76 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
77 2.6.16 - 2.6.32 vanilla (kernel.org)
78 * Recommended e2fsprogs version: 1.41.10-sun2
79 * The async journal commit feature (bug 19128) and the cancel
80 lock before replay feature (bug 16774) are disabled by default.
84 Description: llap pages are moved to head of LRU list instead of tail
88 Description: disable statahead by default due to important races found in the
89 statahead implementation.
93 Description: release request ref through ptlrpc_server_drop_request()
94 Details : make sure the request is protected by rq_refcount while
95 being processed in ptlrpc_server_handle_req_in().
99 Description: correct check for transno value in filter_finish_transno()
103 Description: Set body->eadatasize in mdc_getattr_pack()
107 Description: Handle unsent requests with rq_net_err in ptlrpc_check_set()
111 Description: apps stuck in ptlrpc_check_set() during direct I/O
112 Details : start async bulk unregistering at the same time at reply unlink
116 Description: lfs setstripe --pool broken in 1.8.4
117 Details : fix llapi_search_fsname() to handle relative pathnames
121 Description: Only force the mode change if we're changing the size as well
122 Details : Fix regression introduced in 1.8.0. The offending code was added
123 by commit 77ba4b2141d04180211efa8a75c11ab0abf7fafb to remove
124 setgid/setuid bits when do_truncate() is called on the file.
125 We should only force the change when that occurs, similarly to
126 ll_setattr() in lustre/llite/llite_lib.c
128 Severity : enhancement
130 Description: add procfs tunable to enable/disable lockless direct I/O
131 Details : llite.lustre-*.lockless_direct_io=0 will disable default semantics
132 of direct I/O that forces it to be lockless. lockless_direct_io
133 value, however, will be ignored if per-file LL_FILE_LOCKED_DIRECTIO
138 Description: Lustre servers crashing with NULL pointer errors
139 Details : Make sure the request is protected by rq_refcount while being
140 processed in ptlrpc_server_handle_req_in().
144 Description: kernel BUG at drivers/scsi/sd.c
145 Details : Remove sd iostats patch from sles11 patch series.
149 Description: Modified value of at_min is not taken into account
153 Description: lustre grants flock exclusive locks to two fd's in the same process
157 Description: LBUG in the mfs_verify_child()
161 Description: ASSERTION(atomic_read(&client_stat->nid_exp_ref_count) == 0)
162 Details : Reconnecting from a different nid causes old per-nid stats not to be
167 Description: create proper macro check for bdi interface.
169 -------------------------------------------------------------------------------
171 2010-10-29 Oracle, Inc.
173 * Support for kernels:
174 2.6.16.60-0.69.1 (SLES 10),
175 2.6.32.19-0.2.1 (SLES11),
176 2.6.18-194.17.1.el5 (RHEL 5)
177 2.6.18-194.17.1.0.1.el5 (OEL 5)
178 * Client support for unpatched kernels:
179 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
180 2.6.16 - 2.6.32 vanilla (kernel.org)
181 * Recommended e2fsprogs version: 1.41.10-sun2
182 * The async journal commit feature (bug 19128) and the cancel
183 lock before replay feature (bug 16774) are disabled by default.
187 Description: atime is not properly updated on an MDS
189 Severity : enhancement
191 Description: Update to RHEL5.5 kernel 2.6.19-194.17.1.el5.
192 Update to OEL5.5 kernel 2.6.19-194.17.1.0.1.el5.
194 Severity : enhancement
196 Description: Update to SLES10 SP3 kernel 2.6.16.60-0.69.1.
199 Frequency : only with SLES10
201 Description: Use OFED "KMP" provided by Novell
202 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
203 Lustre is now built against this package. That means you need to
204 install the ofed-kmp package from Novell for the patchless client
205 and from our download site for the server. Note that the ofed-kmp
206 that Novell ships may not exactly match the kernel version but
207 should still be compatible.
209 Severity : enhancement
211 Description: Update SLES11 SP1 kernel to 2.6.32.19-0.2.1.
215 Description: Enabling quotas fails with non-consecutive OST numbering.
219 Description: Fix kernel warning due to lookup_on_len() called without i_mutex
224 Description: Account direct i/o inflight rpcs separately from non-direct i/o so
225 that direct i/o, which is limited by max_rpcs_in_flight, should not
226 block non-direct i/o, which is not limited by max_rpcs_in_flight.
230 Description: Fix per-NID reporting on outstanding writes
234 Description: Reduce stack pressure by uninlining some mds and ptlrpc functions.
238 Description: Remove LASSERT in lprocfs_rd_conn_uuid() since conn == NULL is a
243 Description: fix obdo leak issue in ll_setattr_raw()
247 Description: limit MMP interval
249 Severity : enhancement
251 Description: add several lfs ost enhancements
255 Description: Too many default ACLs break directory access on new directories
259 Description: Lustre inode size is not coherent across nodes.
260 Details : Update lvbo from disk when AST fails with EINVAL. Lvbo is updated
261 on EINVAL error in ldlm_handle_ast_error(). The updates in
262 filter_intent_policy() and ldlm_cb_interpret() have been removed as
267 Description: Oops at __percpu_counter_add+0x1b
268 Details : Use bdi_init()/bdi_destroy() to proper initialize backing_dev_info
273 Description: add mount option to generate 32bit ino, this can be used for 32bit
274 application compatibility.
278 Description: keep reference count for "lli_sai" to prevent it to be released
279 when "statahead_enter()"
283 Description: allow quotacheck over OSTs with sparse indices
287 Description: Objects not getting deleted for files which have been removed
288 Details : ll_have_md_lock() should differentiate between CR and CW OPEN
293 Description: pin object's inode in memory to avoid certain timeouts
297 Description: fix LBUG when obdfilter-survey is interrupted.
299 -------------------------------------------------------------------------------
301 2010-07-31 Oracle, Inc.
303 * Support for kernels:
304 2.6.16.60-0.42.8 (SLES 10),
305 2.6.27.39-0.3.1 (SLES11),
306 2.6.18-194.3.1.el5 (RHEL 5)
307 2.6.18-194.3.1.0.1.el5 (OEL 5)
308 * Client support for unpatched kernels:
309 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
310 2.6.16 - 2.6.30 vanilla (kernel.org)
311 * Recommended e2fsprogs version: 1.41.10-sun2
312 * The async journal commit feature (bug 19128) and the cancel
313 lock before replay feature (bug 16774) are disabled by default.
315 Severity : enhancement
317 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
318 2.6.18-194.8.1.0.1.el5.
320 Severity : enhancement
322 Description: using inkernel OFED stack for rhel5 & oel5.
324 Severity : enhancement
326 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
327 Details : lfs_migrate does a "poor man's" migration of files from their
328 current OST layout to a new OST layout as chosen by the MDS.
332 Description: mds_orphan_add_link()) error linking orphan to PENDING
333 Details : quota limits might disallow linking orphans to PENDING
334 when unlinking a file - temporary raise threads' privileges
335 when processing unlinks.
337 Severity : enhancement
339 Description: add conf-param -d option to remove permanent settings.
340 Details : Add the ability to remove permanent lctl conf_param settings.
341 (Previously conf_param settings could only be changed, not
342 removed.) This also provides a method to change failover
343 nid locations. Improve lctl man page.
345 Severity : enhancement
347 Description: add list_param to b1_8 and add "-R" option to list params
350 Severity : enhancement
352 Description: lfs quota output is not very convenient for awk/sed-parsing
353 Details : Some positions in lfs quota output table could be empty or
354 non-empty which made it hard to parse it with scripts, now a dash
355 is put instead of space where there is not supposed to be any data.
357 Severity : enhancement
359 Description: fix obdfilter-survey script to work properly with remote oss-s
361 Severity : enhancement
363 Description: add new OBDFILTER_SURVEY test suite
365 Severity : enhancement
367 Description: add new multiple mount protection (MMP) test suite
369 Severity : enhancement
371 Description: add support for async journal commit in echo client
373 Severity : enhancement
375 Description: allow userland programs to include <lustre/lustre_idl.h>
376 from stardard include directories
378 Severity : enhancement
380 Description: The prune-icache-use-trylock is no longer needed now that
381 the patch from bug 20008 is landed.
385 Description: The shrink grant feature is still active on the client although the
386 connect flag is not set.
390 Description: Don't leak grant space if the write failed with quota exceeded.
394 Description: Don't consume grant space twice on recoverable resent.
398 Description: a race condition could lead to SIGBUS being sent to an
399 application using mmap-ped files from Lustre
400 Details : truncate_complete_page implementation for the patchless
401 client could arbitrarily unset PG_Uptodate flag for a
402 page being kicked from the page cache, an uptodate check
403 right after a readpage call in filemap_fault could fail
404 because of that as though the page read had been unsuccessful.
408 Description: dlm lock slab shrinking is not efficient
409 Details : The dlm_locks slab can grow significantly and consumes a lot of
410 memory on the server. Set a hardlimit to grant_plan.
414 Description: Lustre does not do 1MB IOs to HW RAID
415 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
416 This is what we do already for SLES kernels.
420 Description: bump maximum number of phys/hw segments in the SLES11 kernel
421 until s/g chaining works properly.
426 Description: LSI Fusion MPT driver hacks to improve performance
427 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
429 Severity : enhancement
431 Description: increase default md stripe_cache_size to 16k
434 Bugzilla : 15587/21439
435 Description: don't handle security.capability xattr
436 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
437 This results in additional getxattr calls, causing VBR test
438 failures as well as a preformance drop when writing.
442 Description: obdfilter-survey is no longer working
443 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
444 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
448 Description: ll_shrink_cache does not handle __GFP_FS properly
452 Description: lfs getstripe shows wrong info for directories
453 Details : Set correct LOVEA default values for filesystem-wide.
457 Description: FSX checksum false positves due to mmap IO
458 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
459 checksum errors, if the flag is set on a request.
463 Description: file operations after eviction have successful return values
464 Details : use vfs ->flush callback to return any pending async errors
469 Description: mdsrate fails to write after 1.3+M files opened
470 Details : decrease memory usage on clients by recycling dentries and
475 Description: obdfilter-survey gives unreasonably high numbers
476 Details : Wait for all threads to complete when running test_brw.
480 Description: do not set lustre read_only device when server umount and keep
481 client records for recoverable ones
485 Description: move sync_on_lock_cancel tunable to the obdfilter layer
486 Details : move the tunable to trigger a journal flush on lock cancel from
487 the ost layer to the obdfilter layer. This tunable is useful
488 when using the async journal commit feature.
492 Description: exp->exp_nid_stats == NULL in filter_tally()
493 Details : fix race with per-nid stats by delaying procfs cleanup until
498 Description: extent lock cancellation on client can keep the cpu busy for too
503 Description: Do not fail OST activation when a llog is not found, just
504 issue an error message.
508 Description: Don't enable extents by default for MDT.
512 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
513 the AT code can access it concurrently while sending early
518 Description: Disable lockless truncate by default since it is sometimes flawed
519 and causes the write_disjoint test to fail.
523 Description: OSSs which don't have the patch from bug 20278 can trigger an
526 Severity : enhancement
528 Description: don't print message to the console when we have not managed to
533 Description: The MDS fails to synchronize OSTs which registered with the MGS
534 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
535 raised too early and thus discarded by the MDT stack.
536 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
541 Description: Fix race when the ping evictor and a service thread execute
542 target_recovery_check_and_stop() concurrently.
546 Description: quota broadcast can trigger a LBUG on the MDT if there are
551 Description: Resetting the lov_objid values to last_id reported by the OST
552 during orphan recovery is incorrect and can cause the same
553 objects to be allocated twice.
555 Severity : enhancement
557 Description: "weak-modules" support
558 Details : Implement "weak-modules" support which enables kernel modules
559 to be used with any kernel that implements the same kABI. In
560 order to achieve this modules are now installed in
561 /lib/modules/$(uname -r)/updates/kernel on all distributions.
563 Severity : enhancement
565 Description: add writeconf as mount option
567 Severity : enhancement
569 Description: produce debuginfo packages for SLES.
571 Severity : enhancement
573 Description: add failover nidlist to the import proc file.
575 Severity : enhancement
577 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
578 clients and 2.0 servers.
580 Severity : enhancement
582 Description: lfs find -s does not work correctly because of a bug in
587 Description: ll_read_ahead_page() must validate the dlm lock before using
592 Description: Prevent failover nids from registering with MGS first.
596 Description: fix lock inversion in ll_setattr_raw().
600 Description: object allocation is not balanced across OSTs.
601 Details : osc_precreate() should return 0, if there are enough objects left.
603 -------------------------------------------------------------------------------
605 2010-04-30 Oracle, Inc.
607 * Support for kernels:
608 2.6.16.60-0.42.8 (SLES 10),
609 2.6.27.39-0.3.1 (SLES11),
610 2.6.18-164.11.1.el5 (RHEL 5)
611 2.6.18-164.11.1.0.1.el5 (OEL 5)
612 * Client support for unpatched kernels:
613 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
614 2.6.16 - 2.6.30 vanilla (kernel.org)
615 * Recommended e2fsprogs version: 1.41.10-sun2
616 * The async journal commit feature (bug 19128) and the cancel
617 lock before replay feature (bug 16774) are disabled by default.
621 Description: open-unlinked directories trigger MDS LBUG
622 Details : Fix regression introduced by the patch from bug 19640.
623 ext3_inc_count() can reset nlink to 1 when the directory
624 is indexed and inode->i_nlink == 2. Work around the problem
625 by incrementing nlink by 2 instead of 1.
629 Description: Reconnects are not throttled
630 Details : Don't wake up pinger on reconnect failures and rely on regular
631 pings to trigger the next reconnection. Please note that the
632 pinger already uses a smaller interval if the import is
636 Frequency : only with NFS export
638 Description: Console flooded with error message from ll_inode_from_lock()
640 Details : in mds_open, initialize the child_res_id before enqueuing
641 the OPEN lock for the child inode, then to avoid senting
642 wrong ldlm_res_id to client.
646 Description: allow multiple instances of the same nid in NID hash
647 Details : Case of multiple separate clients from the same NID (as
648 with liblustre) is legitimate and so we should allow
649 multiple instances of the same NID in nid hash.
653 Description: the readahead code can sleep on a semaphore while holding a
655 Details : in ras_update, "lov_get_info" could be called during increasing
656 readahead windows, which tries to get the mutex lock "lov_lock"
657 while holding the spin_lock "ras_lock", then causes system
662 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
663 Details : fix assertion failure in the grant code.
667 Description: Use CNETERR (which is rate limited) in specific places in
668 the portal's LNET driver to avoid flooding the console.
672 Description: include last created object in precreate slow case
676 Description: don't do rep-ack if not created anything
677 Details : mds_open currently always put a lock into a rep-ack regardless
678 if something was created or not. This is pointless and only
679 creates needless contention. In fact the entire idea was to do
680 this for real creates as a recovery protection.
684 Description: Spurious error messages from smp_processor_id() on preemptible
686 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
687 first. The function fs_trace_get_tcd() was moved up.
691 Description: interval_erase() fix
692 Details : interval_erase() calls update_maxhigh() properly when child
697 Description: "lfs df" does not print stats for all mountpoints
698 Details : Print all mounted lustre filesystems with "lfs df"
702 Description: lfs setstripe -p not longer work with a relative pathname
703 Details : Use realpath() to provide absolute pathname.
707 Description: fix for truncated reply buffer
708 Details : reply buffer could be referred by reply_in_callback after released
713 Description: lustre.lov error when backing up symlinks with extended attributes
714 Details : Improved logic in ll_listxattr()
718 Description: properly handle null value for setattr -n lustre.lov
719 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
720 in ll_setxattr() due to no checking if "value" is NULL.
721 This command now resets to the default striping when executed
726 Description: stack overflow on lock cancellation due to fsync call
727 Details : sync_on_lock_cancel is needed for recovery when async journal
728 is enabled, but we actually just need to make sure that
729 metadata blocks have hit the journal, so doing a fs sync
730 should be enough and should consume less stack (just create an
731 empty handle and commmit it).
735 Description: using current->journal_info to store per-thread data leads
736 to problem under memory pressure
737 Details : disable the per-thread data (current->journal_info) containing
738 the lock info during I/O to work around the issue for short tem
742 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
743 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
744 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
749 Description: Cannot send after transport shutdown
750 Details : Clear imp_vbr_failed flag upon eviction
754 Description: soft lock in request set code during recovery
755 Details : during recovery, uses req->rq_set itself to replay the request
756 instead of ptlrpcd_recovery_pc
760 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
761 Details : During an ll_readahead under ll_readpage, we have seen the the
762 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
763 a page that is already locked by the readahead code.
768 Description: stop waitting for next replay transno if shutdown
769 Details : if the system is shutting down, wake up service thread blocked
770 to wait for next replay transno during recovery, then all the
771 references held by queued requests can be dropped and device
776 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
777 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
781 Description: per-nid stats should not access lustre-hash internal structures
786 Description: mount.lustre fails to pass some options to mount()
791 Description: ext4 extent allocation is slower than in ext3
792 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
793 enlarge ext4 preallocation table for 2048 4K blocks extents
798 Description: incorrect triggering of synchronous IO
799 Details : The OSC can mistakenly fall back to synchronous IO when the
800 max_dirty_mb limit is reached and no write requests have yet
801 been issued. This can occur when the dirty pages are spread
802 over many files all of which are below the optimal request size.
806 Description: Optimize quota_ctl operations by sending requests in parallel
807 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
808 Compiled from two attachments in the ticket.
812 Description: ordering issue between transaction start & i_mutex
813 Details : start the transaction earlier in llog_lvfs_destroy to get
814 transaction start and inode mutex lock nested properly.
818 Description: lru resize SLV can get stuck
819 Details : calculate SLV with a greater precision to not lose small
820 changes due to interger math truncation; round up SLV only
821 if the amount of granted locks less than the limit to not
822 get stuck with this SLV
826 Description: avoid divide-by-zero in lprocfs_rd_import()
830 Description: lfs quota failed when OSTs are down
831 Details : really return approximate block/inode usage when OSTs are down
835 Description: llobdstat fix and enhancement
836 Details : add a counter to set a limit to how many samples will be returned
837 fix a wildcard in the path to limit to obdfilter stats only
841 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
843 Details : abort lock enqueue processing sooner when we umount is in progress.
845 Severity : enhancement
847 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
848 2.6.18-164.11.1.0.1.el5.
850 Severity : enhancement
852 Description: error message improvements
853 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
854 Simplify MDT/OST service start message.
855 Suppress "changing the import ..." warning.
858 Bugzilla : 21961/17914
859 Description: ignore trailing -mdc when determining index number
862 Frequency : only with SLES11
864 Description: fix for a race condition in linux quotas implementation
865 Details : dq_flags(struct dquot) access is not properly locked which could
866 lead to certain inconsistencies when accessing it using non-atomic
867 bit operations like __set_bit in do_set_dqblk.
871 Description: return any pending async errors in close(2) using flush callback
873 -------------------------------------------------------------------------------
875 2010-01-29 Sun Microsystems, Inc.
877 * Support for kernels:
878 2.6.16.60-0.42.8 (SLES 10),
879 2.6.27.39-0.3.1 (SLES11),
880 2.6.18-164.11.1.el5 (RHEL 5)
881 2.6.18-164.6.1.0.1.el5 (OEL 5)
882 * Client support for unpatched kernels:
883 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
884 2.6.16 - 2.6.30 vanilla (kernel.org)
885 * Recommended e2fsprogs version: 1.41.6.sun1
886 * The async journal commit feature (bug 19128) and the cancel
887 lock before replay feature (bug 16774) are disabled by default.
889 Severity : enhancement
891 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
892 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
894 Severity : enhancement
895 Bugzilla : 21511/19848
896 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
898 Severity : enhancement
900 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
902 Severity : enhancement
904 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
905 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
907 Severity : enhancement
909 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
912 Severity : enhancement
914 Description: Vanilla kernel 2.6.30 patchless client support.
919 Description: bad entry in directory xxx: inode out of bounds
920 Details : fix locking issue in the rename path which could race with any
921 other operations updating the same directory.
923 Severity : enhancement
925 Description: Make watchdog timer messages to be more clear and descriptive.
929 Description: cp -p command does not preserve the dates and timestamp
930 Details : mtime could be spoiled by a write callback
934 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
938 Description: Allow non-root access for "lfs check".
939 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
941 Severity : enhancement
943 Description: quotacheck performance/scaling issues
944 Details : reduce quotacheck time on empty filesystem by skipping uninit
947 Severity : enhancement
949 Description: Enhancement for lfs(1) command to use numeric uid/gid.
951 Severity : enhancement
953 Description: Adjust locks' extents on their first enqueue, so that at the time
954 they get granted, there is no need for another pass through the
955 queues since they are already shaped into the proper forms.
959 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
960 arguments and prevent LBUG() in lustre_shrink_reply_v2().
964 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
965 exactly the specified mount options are used. Leaving off
966 any "mandatory" mount options is an error. Leaving off any
967 default mount options causes a warning, but is allowed.
968 Change errors=remount-ro from mandatory to default. Sanitize
969 the mount string before storing it. Update man pages accordingly.
973 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
974 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
977 Severity : enhancement
979 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
981 Severity : enhancement
983 Description: Clients can replay thousands of unused locks during recovery
984 Details : Don't replay unused locks (only read locks for now) during
985 recovery. This feature is disabled by default and can be
986 enabled by running the following command on the clients:
987 lctl get_param ldlm.cancel_unused_locks_before_replay
991 Description: can't stat file in some situation.
992 Details : improve initialize osc date when target is added to mds and
993 ability to resend too big getattr request is client isn't have info
998 Description: Prevent inconsistences between linux and lustre mount structures.
999 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1000 Make the sleep interruptible.
1002 Severity : enhancement
1004 Description: Communicate OST degraded/readonly state via statfs to MDS
1005 Details : Flags in the statfs returned from OSTs indicate whether the
1006 OST is in a degraded RAID state, or if the filesystem has
1007 turned read-only after a filesystem error is detected.
1012 Description: don't panic if EPROTO was hit when reading symlink
1013 Details : correctly handling request reference in error cases.
1018 Description: open sometimes returns ENOENT instead of EACCES
1019 Details : checking permission should be part of open part of mds_open, not
1020 lookup part. so server should be set DISP_OPEN_OPEN disposition
1021 before starting permission check.
1022 Also not need revalidate dentry if client already have LOOKUP lock.
1026 Description: enable client interface failover
1027 Frequency : on servers with multiple network interfaces
1028 Details : When a child reconnects from another NID, properly update export
1029 nid hash position and ldlm reverse import.
1031 Severity : enhancement
1033 Description: implemented direct I/O with arbitrary (nonaligned) memory
1034 addresses and file offsets.
1036 Severity : enhancement
1038 Description: added more recovery timeout options.
1040 Severity : enhancement
1042 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
1047 Description: Avoid deadlock for local client writes
1048 Frequency : only on systems with clients writing to an OST on the same node
1049 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1050 memory freeing context. This allows OST threads to set the
1051 PF_MEMALLOC flag on task structures in order to allocate memory
1052 from reserved pools and complete IO.
1053 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1054 so that the OST threads generate memory pressure and allow
1055 inactive pages to be reclaimed.
1060 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1061 Details : move ldlm namespace creation in setup phase to avoid grab
1062 _lprocfs_lock with cli_sem held.
1066 Frequency : only during format of test systems
1067 Description: Unable to run several mkfs.lustre on loop devices at the same time
1068 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1069 formatting. The solution is to proper handle the error.
1071 Severity : enhancement
1073 Description: implement async create (obd_async_create) method for osc, to avoid
1074 too long waiting new ost objects with holding ldlm lock.
1078 Frequency : occasionally during network problems
1079 Description: client not allowed to reconnect to OST because of active request
1080 Details : abort bulk requests received by the OST once the client has timed
1081 out since the client will resend the request anyway. The client
1082 also now retries to reconnect to the same server if a connect
1083 request failed with EBUSY or -EAGAIN.
1086 Frequency : rare, if used wide striped file and one ost in down.
1088 Descriptoin: don't return error if we created a subset of objects for file.
1089 Details : lov_update_create_set() uses set->set_success as index for created
1090 objects, so if some requests failed, they will have hole at end of
1091 array and we can use qos_shrink_lsm for allocate correct lsm.
1095 Description: Slow stale export processing during normal start up
1096 Details : The global mgc lock prevents OST setup to be run in parallel.
1097 Replace the global lock with a per-config_llog_data semaphore.
1101 Description: Out or order replies might be lost on replay
1102 Details : In ptlrpc_retain_replayable_request if we cannot find retained
1103 request with tid smaller then one currently being added, add it
1104 to the start, not end of the list.
1108 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
1109 Details : add cond_resched() calls to avoid hogging the cpu for too long
1110 in the hash code. Make also lustre_hash_for_each_empty() more
1113 Severity : enhancement
1115 Description: Performance improvements for debug messages with D_RPCTRACE,
1116 D_LDLM, D_QUOTA options.
1119 Frequency : only with NFS export
1121 Description: (lov_merge.c:74:lov_merge_lvb())
1122 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
1123 Details : Fix a race in the nfs export code by populating inode
1124 info while the new inode is still locked
1126 Severity : enhancement
1128 Description: add a new file in procfs called force_lbug. Writting to this
1129 file triggers a LBUG. Only for test purpose.
1133 Description: OOM killer causes node hang
1134 Details : really interrupt the sleep in osc_enter_cache on signals
1138 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1139 Details : fix race during quota release on the slave.
1141 Severity : enhancement
1143 Description: smaller hash bucket sizes, cleanups
1144 Details : increase hash table sizes and enabled rehashing for pools, quota,
1145 uuid, nid & per-nid stats.
1147 Severity : enhancement
1149 Description: Add ldiskfs maxdirsize mount option
1150 Details : add max_dir size mount option
1154 Description: panic in ll_statahead_thread
1155 Details : prevent parent thread to be killed before its child
1158 Frequency : only with 16TB device
1160 Description: unable to perform "mount -t lustre" of 16TB OST device
1161 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
1165 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
1166 Details : unregistering should be zero if no RPC inflight.
1170 Description: hyperion: Oops during metabench
1171 Details : Correct the refcount of lov_request_set
1173 Severity : enhancement
1175 Description: Add mptlinux and nxge drivers to Lustre builds
1177 Severity : enhancement
1179 Description: Fix watchdog timer message to be more clear
1180 Details : Make watchdog timer messages more clear and descriptive.
1184 Description: LNET soft lockups in socknal_cd thread
1185 Details : don't hog CPU for active-connecting if another connd is
1186 accepting connecting-requst from the same peer
1190 Description: recovery-small test_17 hang
1191 Details : Land several AT improvements & fixes.
1195 Description: MDS panic and hanging client processes
1196 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
1200 Description: OSS stuck in recovery.
1201 Details : fix race during recovery. class_unlink_export,
1202 class_set_export_delayed and target_queue_last_replay_reply
1203 may race while increasing/decreasing obd_recoverable_clients
1204 and obd_delayed_clients, causing recovery to wait forever.
1206 Severity : enhancement
1208 Description: add cascading_rw.c to lustre/tests
1212 Description: filter_last_id() NULL deref
1213 Details : lprocfs_filter_rd_last_id() should check for the fully
1214 setup obd device, before proceeding further.
1216 Severity : enhancement
1218 Description: Loadgen improvements
1219 Details : stacksize and locking fixes for loadgen
1223 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
1224 Details : The atomic_read() allowing the atomic_inc() are not covered
1225 by a lock. Thus they may safely race and trip this CERROR()
1226 unless we add in a small fudge factor (+1).
1228 Severity : enhancement
1230 Description: shrink_slab: nr=-9223362083340912175
1231 Details : fix spurious message from shrink_slab reporing negative nr
1235 Description: Quiet bogus previously committed transno error
1236 Details : suppress the "server went back in time" error message which
1237 is always printed even in the common case after a client eviction
1239 Severity : enhancement
1241 Description: Parallel statfs() calls result in client eviction
1242 Details : cache statfs data for 1s.
1246 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
1248 Details : fix serveral issues in pinger code causing clients not to ping
1249 servers for too long, resulting in evictions.
1253 Description: e2fsck should warn when MMP update interval is extended
1254 Details : print mmp_check_interval and make it possible to abort mount
1255 operation in case it takes too long.
1259 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
1260 Details : fix bug in the RHEL5's jbd2 callback patch.
1264 Description: drop number of active requests when queued for recovery
1265 Details : Now that we take a reference on the original request instead of
1266 making a copy of it for recovery. We need to drop the number of
1267 active requests or the queued requests will prevent all request
1268 processing when they exceed (srv->srv_threads_running - 1).
1270 Severity : enhancement
1272 Description: refuse to invalidate operational quota files when they are in use
1273 Details : an attempt to invalidate operational quota files on the quota
1274 master is not actually permitted by VFS (returning -EPERM), but we
1275 should not depend on that and should return the error earlier.
1279 Description: Applications stuck in jbd2_log_wait_commit during exit
1280 Details : fix deadlock between kjournald2 trying to acquire the page lock
1281 owned by an ost_io thread waiting for journal commit.
1283 -------------------------------------------------------------------------------
1285 2009-10-16 Sun Microsystems, Inc.
1287 * Support for kernels:
1288 2.6.16.60-0.42.4 (SLES 10),
1289 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1290 2.6.18-128.7.1.el5 (RHEL 5),
1291 2.6.18-128.7.1.el5 (OEL 5),
1292 * Client support for unpatched kernels:
1293 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1294 2.6.16 - 2.6.27 vanilla (kernel.org)
1295 * Recommended e2fsprogs version: 1.41.6.sun1
1296 * ext4 support for RHEL5 is experimental and thus should not be
1299 Severity : enhancement
1301 Description: Add OEL5 support.
1303 Severity : enhancement
1305 Description: Update kernel to SLES11 2.6.27.29-0.1.
1309 Description: File checksum failures with OST read cache on
1310 Details : Disable page poisoning when the bulk transfer has to be aborted
1311 because the client got evicted.
1315 Description: Don't allow make backward step on assiging osc next id.
1316 Details : race between allocation next id and ll_sync thread can be cause
1317 of set wrong osc next id and can be kill valid ost objects.
1319 Severity : enhancement
1321 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1323 Severity : enhancement
1325 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1329 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1333 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1337 Description: Error handling in osc_statfs_interpret() has been improved.
1338 Details : Check in osc_statfs_interpret() for EBADR.
1342 Description: Do not send statfs() requests to OSTs disabled by administrator.
1343 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1347 Description: Do not update ctime for the deleted inode.
1348 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1352 Description: Increase of the size of the LDLM resource hash.
1353 Details : Bump up RES_HASH_BITS=12.
1357 Description: correctly send lsm on open replay
1358 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1363 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1364 Details : filter_destroy() does not hold the DLM lock over the whole
1365 operation. If the DLM lock is dropped, filter_commitrw() can go
1366 through, causing the deadlock between page lock and i_mutex.
1367 The i_alloc_sem should also be hold in filter_destroy() while
1368 truncating the file.
1372 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1373 Details : ldiskfs_truncate calls grab_cache_page which may start page
1374 allocation under an open transaction. This may lead to
1375 calling prune_icache with consequent lustre reentrance.
1378 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1379 still up and when the OST pool feature is used
1381 Description: interop testing got LBUG when run dd with OST pool
1382 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1383 Details : down/upgrading the MDS to a version that doesn't/does support OST
1384 pool can cause clients to crash because the lsm has changed
1389 Description: missing tree_status on 1.8.1 RPM build
1390 Details : make rpms failed due because the tree_status file is missing.
1394 Description: continuing LustreError "mds adjust qunit failed!"
1395 Details : don't print message on the console when ->adjust_qunit fails.
1399 Description: don't increase ldlm timeout if previous client was evicted
1400 Details : if a client doesn't respond to a blocking callback within the
1401 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1402 when the lock is next granted.
1406 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1407 affects recovery negatively.
1408 Details : make sure all exports have been properly destroyed by the zombie
1409 thread processed before stopping the target.
1413 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1414 Details : disable write barrier for ext4/SLES11.
1418 Description: Kernel panic - not syncing: Out of memory and no killable
1419 processes... on OSS when iozone
1420 Details : fix memory leak in the journal checksum patch.
1424 Description: group quota "too many blocks" OSS crashes
1425 Details : we should keep the same uid/gid for lquota_chkquota() and
1426 lquota_pending_commit()
1430 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1431 Details : don't LBUG on release quota error. Just a workaround until the
1432 problem is understood.
1434 ------------------------------------------------------------------------------
1435 2009-07-31 Sun Microsystems, Inc.
1437 * Support for kernels:
1438 2.6.16.60-0.39.3 (SLES 10),
1439 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1440 2.6.18-128.1.14.el5 (RHEL 5),
1441 * Client support for unpatched kernels:
1442 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1443 2.6.16 - 2.6.27 vanilla (kernel.org)
1444 * Recommended e2fsprogs version: 1.41.6.sun1
1445 * File join has been disabled in this release, refer to Bugzilla 16929.
1446 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1447 of Lustre filesystem with 4K stack may cause a stack overflow. For
1448 more information, please refer to bugzilla 17630.
1449 * ext4 support for RHEL5 is experimental and thus should not be
1452 Severity : enhancement
1454 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1457 Frequency : with 1.8 server and 1.6 clients
1459 Description: correctly shrink reply for avoid send too big message to client.
1460 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1461 some problems with sending this reply to 1.6 client.
1465 Description: Repeated atomic allocation failures.
1466 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1467 to generate memory pressure and allow reclaiming of inactive pages.
1468 At the same time, do not allow to exhaust emergency pools.
1469 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1471 Severity : enhancement
1472 Bugzilla : 19846, 18289
1473 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1475 Severity : enhancement
1476 Bugzilla : 19625, 16893, 18668, 19848
1477 Description: Add support for SLES11 2.6.27.23-0.1.
1479 Severity : enhancement
1481 Description: Update client support to vanila kernels up to 2.6.27.
1483 Severity : enhancement
1485 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1487 Severity : enhancement
1489 Description: Compile with -Werror by default for i686 and x86_64.
1493 Description: resolve race between obd_disconnect and class_disconnect_exports
1494 Details : if obd_disconnect will be called to already disconnected export he
1495 forget release one reference and osc module can't unloaded.
1497 Severity : enhancement
1499 Description: move AT tunable parameters for more consistent usage
1500 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1504 Descriptoin: correctly skip time estimate if in recovery
1505 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1509 Descriptoin: OSS DeadLock
1510 Details : Use trylock to prevent deadlock when shrink icache.
1512 Severity : enhancement
1514 Description: Allow tuning service thread via /proc
1515 Details : For each service a new
1516 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1517 created that can be used to set min/max thread counts, and get the
1518 current number of running threads.
1520 Severity : enhancement
1522 Description: Add state history info file, enhance import info file
1523 Details : Track import connection state changes in a new osc/mdc proc file;
1524 add overview-type data to the osc/mdc import proc file.
1528 Description: Reduce small size read RPC
1529 Details : Set read-ahead limite for every file and only do read-ahead when
1530 available read-ahead pages are bigger than 1M to avoid small size
1535 Description: free_entry erroneously used groups_free instead of put_group_info
1537 Severity : enhancement
1539 Description: Make read-ahead stripe size aligned.
1541 Severity : enhancement
1543 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1546 Frequency : rare, connect and disconnect target at same time
1548 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1549 Details : don't call obd_disconnect under lov_lock. this long time
1550 operation and can block ptlrpcd which answer to connect request.
1553 Frequency : start MDS on uncleanly shutdowned MDS device
1555 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1556 Details : stay in waiting mds<>ost recovery finished produce random bugs
1557 due race between two ll_sync thread for one lov target. send
1558 ACTIVATE event only if connect realy finished and import have
1562 Frequency : start MDS on uncleanly shutdowned MDS device
1564 Descriptoin: aborting recovery hang on MDS
1565 Details : don't throttle destroy RPCs for the MDT.
1569 Description: Slow reads beyond 8Tb offsets.
1570 Details : Page index integer overflow in ll_read_ahead_page
1574 Description: Soft lockup on OSS after MDS failover
1575 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1576 As a conseqence, the patch from bug 18224 is not operational
1577 and the MDS export cannot be reused on the OSTs until it gets
1581 Frequency : rare, only if using MMP with Linux RAID
1583 Description: MMP doesn't work with Linux RAID
1584 Details : While using HA for Lustre servers with Linux RAID, it is possible
1585 that MMP will not detect multiple mounts. To make this work we
1586 need to unplug the device queue in RAID when the MMP block is being
1587 written. Also while reading the MMP block, we should read it from
1588 disk and not the cached one.
1591 Frequency : rare, during recovery
1593 Description: Assertion failure in ldlm_lock_put
1594 Details : Do not put cancelled locks into replay list, hold references on
1595 locks in replay list
1599 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1600 Details : create_count always drops to the min value (=32) because grow_count
1601 is being changed before the precreate RPC completes.
1604 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1607 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1608 Details : add proper locking for creation of jbd_4k slab cache
1612 Description: MMP check in ext3_remount() fails without displaying any error
1613 Details : When multiple mount protection fails during remount, proper error
1618 Description: Rare Client crash on resend if the file was deleted.
1619 Details : When file is opened, but open reply is lost and file is
1620 subsequently deleted before resend, resend processing logic
1621 breaks trying to open the file again, should not try to open.
1625 Description: add check for >8TB ldiskfs filesystems
1626 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1627 Don't allow >8TB ldiskfs filesystems to be mounted without
1628 force_over_8tb mount option
1632 Description: Client locked up when running multiple instances of an app. on
1633 multiple mount points
1634 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1635 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1639 Description: Cannot acces an NFS-mounted Lustre filesystem
1640 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1641 from a Lustre-client exporting the Lustre filesystem via NFS.
1645 Description: panic in ll_statahead_thread
1646 Details : grab dentry reference in parent process.
1649 -------------------------------------------------------------------------------
1651 tbd Sun Microsystems, Inc.
1653 * Support for kernels:
1654 2.6.16.60-0.31 (SLES 10),
1655 2.6.18-128.1.6.el5 (RHEL 5),
1656 2.6.22.14 vanilla (kernel.org)
1657 * Client support for unpatched kernels:
1658 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1659 2.6.16 - 2.6.22 vanilla (kernel.org)
1660 * Recommended e2fsprogs version: 1.40.11-sun1
1661 * File join has been disabled, refer to Bugzilla 16929.
1662 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1663 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1664 Lustre filesystem with 4K stack may cause a stack overflow. For more
1665 information, please refer to bugzilla 17630.
1667 Severity : enhancement
1669 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1671 Severity : enhancement
1673 Description: Update OFED release to 1.4.1 RC4
1675 Severity : major, only with big OST
1677 Description: Very poor metadata performance on Infiniband lustre configuration
1678 Details : OST object precreation becomes very slow on big OSTs. This is due
1679 to the ialloc patch spending too much time scanning groups.
1682 Frequency : during recovery
1684 Description: don't mix llog inodes with normal.
1685 Details : allocate inodes for log in last inode group
1690 Description: fix lqs' reference which won't be put in some situations
1691 Details : This patch fixes:
1692 1. In quota_check_common(), this function will check quota
1693 for user and group, but only send one return via "pending".
1694 In most cases, the pendings should be same. But that is not
1696 2. If quotaoff runs between lquota_chkquota() and
1697 lquota_pending_commit(), the same thing will happen too.
1698 That is why it comes:
1699 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1702 -------------------------------------------------------------------------------
1704 2008-12-31 Sun Microsystems, Inc.
1706 * Support for kernels:
1707 2.6.5-7.314 (SLES 9),
1708 2.6.9-67.0.22.EL (RHEL 4),
1709 2.6.16.60-0.31 (SLES 10),
1710 2.6.18-92.1.17.el5 (RHEL 5),
1711 2.6.22.14 vanilla (kernel.org)
1712 * Client support for unpatched kernels:
1713 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1714 2.6.16 - 2.6.22 vanilla (kernel.org)
1715 * Client support for unpatched kernels:
1716 we do not recommend using patchless RHEL4 clients with kernels
1717 prior to 2.6.9-55EL (RHEL4U5).
1718 * Recommended e2fsprogs version: 1.40.11-sun1
1719 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1720 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1721 removed cwd "./" (refer to Bugzilla 14399).
1722 * A new quota file format has been introduced in 1.6.5.
1723 The format conversion from prior releases is handled transparently,
1724 but releases older than 1.4.12/1.6.5 don't understand this new
1725 format. The automatic format conversion can be avoided by running
1726 the following command on the MDS:
1727 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1728 For more information, please refer to bugzilla 13904.
1729 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1730 The format conversion from prior releases is handled transparently,
1731 but releases older than 1.6.6/1.8.0 don't understand this new
1732 format. The automatic format conversion can be avoided by running
1733 the following commands on the MDS and OSS servers (for
1734 pre 1.4.12-1.6.5 quota files):
1735 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1736 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1737 or (for 1.4.12/1.6.5 quota files)
1738 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1739 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1740 For more information, please refer to bugzilla 13904.
1741 * Output of lfs quota has been made less detailed by default,
1742 old (verbose) output can be obtained by using -v option.
1743 * File join has been disabled in this release, refer to Bugzilla 16929.
1744 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1745 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1746 Lustre filesystem with 4K stack may cause a stack overflow. For more
1747 information, please refer to bugzilla 17630.
1749 Severity : enhancement
1751 Description: Caching OSS
1752 Details : introduce data caching on the OSS. The OSS now relies on the linux
1753 kernel page cache to keep recently accessed data in memory.
1754 It is worth noting that all write requests are still flushed
1755 synchronously as in lustre 1.6.
1757 Severity : enhancement
1759 Description: Version based recovery
1760 Details : introduce finer grained recovery able to detect transaction
1761 dependencies and can deal with transaction gaps caused by clients
1762 failing at the same time as the server.
1764 Severity : enhancement
1766 Description: Enable adaptive timeouts by default
1767 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1768 managed dynamically based on server load and should not need
1769 to be tuned manually based on cluster size. This allows Lustre
1770 to work under a wider variety of system sizes and loads, without
1771 unnecessarily causing lengthy recovery times.
1773 Severity : enhancement
1775 Description: Add OST Pools support
1776 Details : File striping can now be set to use an arbitrary pool of OSTs
1778 Severity : enhancement
1780 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1781 Details : allow skip disconnected ost for send statfs request and hide error
1785 Frequency : rare, on llog test 6
1787 Descriptoin: don't allow connect to already connected import
1788 Details : allowing connect to already connected import is hide connecting problem.
1791 Frequency : rare, on failed llog setup
1793 Descriptoin: don't leak obd reference on failed llog setup
1794 Details : for failed llog setup - mgc forget call class_destroy_import for
1795 client import, move destroy import to more generic place.
1800 Descriptoin: allow kill process which wait statahead result
1801 Details : for some reasons 'ls' can stick in waiting result from statahead,
1802 in this case need way for kill this process.
1805 Frequency : rare, at shutdown
1807 Descriptoin: panic at umount
1808 Details : llap_shrinker can be raced with killing super block from list and
1809 this produce panic with access to already freeded pointer
1814 Descriptoin: don't lose wakeup for imp_recovery_waitq
1815 Details : recover_import_no_retry or invalidate_import and import_close can
1816 both sleep on imp_recovery_waitq, but we was send only one wakeup
1822 Descriptoin: panic in mds_open
1823 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1828 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1830 Details : release lock refernce only after add page to pages list.
1833 Frequency : always with long access acl
1835 Descriptoin: mds can't pack reply with long acl.
1836 Details : mds don't control size of acl but they limited by reint/getattr
1841 Frequency : on remount
1843 Description: external journal device not working after the remount
1844 Details : clear dev_rdonly flag for external journal devices in
1850 Description: shutdown vs evict race
1851 Details : client_disconnect_export vs connect request race.
1852 if client will evicted at this time - we start invalidate
1853 thread without referece to import and import can be freed
1859 Description: shrink LOV EAs before replying
1860 Details : correctly adjust LOV EA buffer for reply.
1865 Description: don't skip ost target if they assigned to file
1866 Details : Drop slow OSCs if we can, but not for requested start idx.
1867 This means "if OSC is slow and it is not the requested
1868 start OST, then it can be skipped, otherwise skip it only
1869 if it is inactive/recovering/out-of-space.
1871 Severity : enhancement
1873 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1876 Frequency : rare, need acl's on inode.
1878 Description: client can't handle ost additional correctly
1879 Details : if ost was added after client connected to mds client can have
1880 hit lnet_try_match_md ... to big messages to wide striped files.
1881 in this case need teach client to handle config events about add
1882 lov target and update client max ea size at that event.
1884 Severity : enhancement
1886 Description: Update to sles9 kernel-2.6.5-7.314.
1888 Severity : enhancement
1890 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1893 Frequency : Create a symlink file with a very long name
1895 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1896 Details : If there is no extra space in the request for early cancels,
1897 ldlm_req_handles_avail() returns 0 instead of a negative value.
1902 Description: mds is deadlocked
1903 Details : in rare cases, inode in catalog can have i_no less than have parent
1904 i_no, this produce wrong order for locking during open, and
1905 parallel unlink can be lock open. this need teach mds_open to grab
1906 locks in resource id order, not at parent -> child order.
1908 Severity : enhancement
1910 Description: Add /proc entry for import status
1911 Details : The mdc, osc, and mgc import directories now have
1912 an import directory that contains useful import data for debugging
1913 connection problems.
1915 Severity : enhancement
1917 Description: Re-disable certain /proc logging
1918 Details : Enable and disable client's offset_stats, extents_stats and
1919 extents_stats_per_process stats logging on the fly.
1922 Frequency : Only on FC kernels 2.6.22+
1924 Description: oops in statahead
1925 Details : Do not drop reference count for the dentry from VFS when lookup,
1926 VFS will do that by itself.
1928 Severity : enhancement
1930 Description: Generic /proc file permissions
1931 Details : Set /Proc file permissions in a more generic way to enable non-
1932 root users operate on some /proc files.
1936 Description: Hitting mdc_commit_close() ASSERTION
1937 Details : Properly handle request reference release in
1938 ll_release_openhandle().
1942 Frequency : only patchless client
1943 Description: add workaround for race between add/remove dentry from hash
1945 Severity : enhancement
1947 Description: Allow OST glimpses to return PW locks
1951 Description: LBUG when llog conf file is full
1952 Details : When llog bitmap is full, ENOSPC should be returned for plain
1957 Description: Prevent import from entering FULL state when server in recovery
1961 Description: service mount cannot take device name with ":"
1962 Details : Only when device name contains ":/" will mount treat it as
1968 Description: replace ptlrpcd with the statahead thread to interpret the async
1969 statahead RPC callback
1973 Frequency : on recovery
1974 Description: I/O failures after umount during fail back
1975 Details : if client reconnected to restarted server we need join to recovery
1976 instead of find server handler is changed and process self eviction
1977 with cancel all locks.
1979 Severity : enhancement
1981 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1985 Description: Kernel BUG tries to release flock
1986 Details : Lustre does not destroy flock lock before last reference goes
1987 away. So always drop flock locks when client is evicted and
1988 perform unlock regardless of successfulness of speaking to MDS.
1990 Severity : enhancement
1992 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1994 Severity : enhancement
1996 Description: Upcall on Lustre log has been dumped
1997 Details : Allow for a user mode script to be called once a Lustre log has
1998 been dumped. It passes the filename of the dumped log to the
1999 script, the location of the script can be specified via
2000 /proc/sys/lnet/debug_log_upcall.
2005 Description: avoid messages about idr_remove called for id that is not allocated
2006 Details : Move assigment s_dev for clustered nfs to end of initialization,
2007 for avoid problem with error handling.
2012 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
2013 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
2014 CONN_USED_HASH and this prodice warning when put connection again
2020 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
2021 Details : release reference to stats when client disconnected, not
2022 when export destroyed for avoid races when client destroyed
2023 after main ost export.
2027 Description: more cleanup in mds_lov
2028 Details : add workaround for get valid ost count for avoid warnings about
2029 drop too big messages, not init llog cat under semphore which
2030 can be blocked on reconnect and break normal replay, fix access
2033 Severity : enhancement
2035 Description: Export bytes_read/bytes_write count on OSC/OST.
2039 Description: Early reply size mismatch, MGC loses connection
2040 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
2041 the connect flags are properly negotiated.
2045 Description: Properly propagate oinfo flags from lov to osc for statfs
2046 Details : restore missing copy oi_flags to lov requests.
2050 Description: exports in /proc are broken
2051 Details : recreate /proc entries for clients when they reconnect.
2053 Severity : enhancement
2055 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
2056 : l_getgroups(8), lst(8), routerstat(8)
2057 Details : included man pages for llobdstat(8), llstat(8),
2058 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
2060 Severity : enhancement
2062 Description: Implement lustre ll_show_options method.
2064 Severity : enhancement
2066 Description: Update to SLES9 kernel-2.6.5-7.312.
2068 Severity : enhancement
2070 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
2074 Description: exports in /proc are broken
2075 Details : recreate /proc entries for clients when they reconnect.
2079 Description: don't fail open with -ERANGE
2080 Details : if client connected until mds will be know about real ost count
2081 get LOV EA can be fail because mds not allocate enougth buffer
2086 Description: Resolve device initialization race
2087 Details : Prevent proc handler from accessing devices added to the
2088 obd_devs array but yet be intialized.
2090 Severity : enhancement
2092 Description: configure's --enable-quota should check the
2093 : kernel .config for CONFIG_QUOTA
2094 Details : configure is terminated if --enable-quota is passed but
2095 : no quota support is in kernel
2097 Severity : enhancement
2099 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
2101 Severity : enhancement
2103 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
2107 Frequency : rare, on PPC clients
2108 Description: don't swab ost objects in response about directory, because
2110 Details : bug similar bug 14856, but in different function.
2112 Severity : enhancement
2114 Description: lfs quota tool enhancement
2115 Details : added units specifiers support for setquota, default to
2116 current uid/gid for quota report, short quota stats by
2117 default, nonpositional parameters for setquota, added
2118 llapi_quotactl manual page.
2120 Severity : enhancement
2122 Description: *optional* service tags registration
2123 Details : if the "service tags" package is installed on a Lustre node
2124 When the filesystem is mounted, a local-node service tag will
2125 be created. See http://inventory.sun.com/ for more information
2126 about the Service Tags asset management system.
2130 Description: Client runs out of low memory
2131 Details : Consider only lowmem when counting initial number of llap pages
2134 Frequency : occasional
2136 Description: add refcount for osc callbacks, so avoid panic on shutdown
2138 Severity : enhancement
2140 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
2143 Frequency : testing only
2145 Description: sanity test 65a fails if stripecount of -1 is set
2146 Details : handle -1 striping on filesystem in ll_dirstripe_verify
2149 Frequency : only in unusual configurations
2151 Description: Kernel panic with find ost index.
2152 Details : lov_obd have panic if some OST's have sparse indexes.
2154 Severity : enhancement
2156 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
2159 Frequency : rarely, if filesystem is mounted with -o flock
2161 Description: do not process already freed flock
2162 Details : flock can possibly be freed by another thread before it reaches
2163 to ldlm_flock_completion_ast.
2166 Frequency : rarely, if filesystem is mounted with -o flock
2168 Description: LBUG during stress test
2169 Details : Need properly lock accesses the flock deadlock detection list.
2172 Frequency : rarely, if binaries are being run from Lustre
2174 Description: oops in page fault handler
2175 Details : kernel page fault handler can return two special 'pages' in
2176 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
2179 Frequency : rarely, during shutdown
2181 Description: timeout with invalidate import.
2182 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
2183 should be handled by ptlrpcd. This produce long age waiting and
2184 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
2190 Description: ASSERTION(CheckWriteback(page,cmd)) failed
2191 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
2195 Frequency : only with broken builds/installations
2197 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
2198 Details : just return an error to a user, put a console error message
2200 Severity : enhancement
2202 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
2204 Severity : enhancement
2206 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
2208 Severity : enhancement
2210 Description: enable MGS and MDT services start separately
2211 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
2212 a co-located MGS without starting the MGS, which is a complement
2213 to 'nosvc' mount option.
2216 Frequency : always, on big-endian systems
2218 Description: cleanup in ptlrpc code, related to PPC platform
2219 Details : store magic in native order avoid panic's in recovery on PPC
2220 node and forbid from this error in future. Also fix posibily
2221 of twice swab data. Fix get lov striping to userpace.
2224 Frequency : rarely, if replay get lost on server
2226 Description: server incorrectly drop resent replays lead to recovery failure.
2227 Details : do not drop replay according to msg flags, instead we check the
2228 per-export recovery request queue for duplication of transno.
2231 Frequency : after recovery
2233 Description: precreate to many object's after del orphan.
2234 Details : del orphan st in oscc last_id == next_id and this triger growing
2235 count of precreated objects. Set flag LOW to skip increase count
2236 of precreated objects.
2239 Frequency : rare, on clear nid stats
2241 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
2242 Details : when clean nid stats sometimes try destroy live entry,
2243 and this produce panic in free.
2246 Frequency : occasionally since 1.6.4
2248 Description: Stack overflow during MDS log replay
2249 Details : ease stack pressure by using a thread dealing llog_process.
2252 Frequency : very rare
2254 Description: MDT cannot be unmounted, reporting "Mount still busy"
2255 Details : Mountpoint references were being leaked during open reply
2256 reconstruction after an MDS restart. Drop mountpoint reference
2257 in reconstruct_open() and free dentry reference also.
2262 Description: wait until IO finished before start new when do lock cancel.
2263 Details : VM protocol want old IO finished before start new, in this case
2264 need wait until PG_writeback is cleared until check dirty flag
2265 and call writepages in lock cancel callback.
2270 Description: mds_mfd_close() ASSERTION(rc == 0)
2271 Details : In mds_mfd_close(), we need protect inode's writecount change
2272 within its orphan write semaphore to prevent possible races.
2275 Frequency : rare, on shutdown ost
2277 Description: don't hit live lock with umount ost.
2278 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2279 use shrink_dcache_sb instead.
2282 Frequency : only when echo_client is used
2284 Description: don't panic with use echo_client
2285 Details : echo client pass NULL as client nid pointer and this produce NULL
2286 pointer dereference.
2289 Frequency : Always on 32-bit PowerPC systems
2291 Description: fix build on PPC32
2292 Details : compile code with -m64 flag produce wrong object file for PPC32.
2297 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2298 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2299 by a client, and the data on disk are dangerous and incorrect.
2300 Change them with client eviction.
2302 Severity : enhancement
2304 Description: skiplist implementation simplification
2305 Details : skiplists are used to group compatible locks on granted list
2306 that was implemented as tracking first and last lock of each lock
2307 group the patch changes that to using doubly linked lists
2311 Description: delete compatibility for 32bit qdata
2312 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2313 qunit. That means servers of b1_4 and servers of b1_8 can't be
2314 used together if users want to use quota.
2317 Frequency : only with administrator action
2319 Description: mount failure if config log has invalid conf_param setting
2320 Details : If administrator specified an incorrect configuration parameter
2321 with "lctl conf_param" this would cause an error during future
2322 client mounts. Instead, ignore the bad configuration parameter.
2325 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2327 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2328 Details : If blocks per group is less than blocksize*8, set rest of the
2332 Frequency : Application do stride read on lustre
2334 Description: The read performance will drop a lot if the application does
2336 Details : Because the stride_start_offset are missing in stride read-ahead,
2337 it will cause clients read a lot of unused pages in read-ahead,
2338 then the read-performance drops.
2342 Description: more ldlm soft lockups
2343 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2344 starve other threads from the resource lock for a long time in
2345 case of long waiting queue, so change the debug level from
2346 D_OTHER to the less frequently used D_INFO.
2348 Severity : enhancement
2350 Description: add -gid, -group, -uid, -user options to lfs find
2352 Severity : enhancement
2354 Description: ll_recover_lost_found_objs - recover objects in lost+found
2355 Details : OST corruption and subsequent e2fsck can leave objects in the
2356 lost+found directory. Using the "ll_recover_lost_found_objs"
2357 tool, these objects can be retrieved and data can be salvaged
2358 by using the object ID saved in the fid EA on each object.
2363 Description: this bug _only_ happens when inode quota limitation is very low
2364 (less than 12), so that inode quota unit is 1 at initialization.
2365 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2366 is effective now. So least quota qunit should be 2.
2370 Description: Hung threads in invalidate_inode_pages2_range
2371 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2372 one is completed. As a result, some RPCs are stuck in the queue
2377 Description: Procfs and llog threads access destoryed import sometimes.
2378 Details : Sync the import destoryed process with procfs and llog threads by
2379 the import refcount and semaphore.
2383 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2384 Details : Sort source/child resource pair after updating child resource.
2389 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2390 Details : If insertion of an extent fails, then discard the inode
2391 preallocation and free data blocks else it can lead to duplicate
2396 Description: don't always update ctime in ext3_xattr_set_handle()
2397 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2398 In some cases the ctime should not be updated, for example for
2399 2.0->1.8 compatibility it is necessary to delete an xattr and it
2400 should not update the ctime.
2404 Description: add quota statistics
2405 Details : 1. sort out quota proc entries and proc code.
2406 2. add quota statistics
2411 Description: quotas are not honored with O_DIRECT
2412 Details : all writes with the flag O_DIRECT will use grants which leads to
2413 this problem. Now using OBD_BRW_SYNC to guard this.
2417 Bugzilla : 15713/16362
2418 Description: Assertion in iopen_connect_dentry in 1.6.3
2419 Details : looking up an inode via iopen with the wrong generation number can
2420 populate the dcache with a disconneced dentry while the inode
2421 number is in the process of being reallocated. This causes an
2422 assertion failure in iopen since the inode's dentry list contains
2423 both a connected and disconnected dentry.
2427 Description: assertion failure in ldlm_handle2lock()
2428 Details : fix a race between class_handle_unhash() and class_handle2object()
2429 introduced in lustre 1.6.5 by bug 13622.
2431 Severity : enhancement
2433 Description: superblock lock contention with many SMP cores on one client
2434 Details : several client filesystem locks were highly contended on SMP
2435 NUMA systems with 8 or more cores. Per-CPU datastructures
2436 and more efficient locking implemented to reduce contention.
2441 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2442 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2447 Description: oops in sd_iostats_seq_show()
2448 Details : unloading/reloading the scsi low level driver triggers a kernel
2449 bug when trying to access the sd iostat file.
2454 Description: Kernel panics during QLogic driver reload
2455 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2456 patch, causing memory corruption.
2461 Description: journal_dev option does not work in b1_6
2462 Details : pass mount option during pre-mount.
2464 Severity : enhancement
2466 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2467 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2468 extent information of a file. It can be used to map logical blocks
2469 in a file to physical blocks in the block device.
2472 Frequency : only with adaptive timeout enabled
2474 Description: DEBUG_REQ() bad paging request
2475 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2476 because it can be accessed by reply_in_callback() without the
2480 Frequency : only on Cray X2
2482 Description: X2 build failures
2483 Details : fix build failures on Cray X2.
2487 Description: xid & resent requests
2488 Details : Initialize RPC XID from clock at startup (randomly if clock is
2493 Description: quota recovery deadlock during mds failover
2494 Details : This patch includes att18982, att18236, att18237 in bz14840.
2496 1. fix osts hang when mds does failover with quotaon
2497 2. prevent watchdog storm when osts threads wait for the
2502 Description: kernel panic on racer
2503 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2505 Severity : enhancement
2507 Description: Add lustre_start utility to start or stop multiple Lustre servers
2512 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2513 Details : In case of memory pressure, list_del() can be called twice on
2514 req->rq_history_list, causing a kernel oops.
2518 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2519 Details : fix stack overflow in the distributed lock manager by defering
2520 export eviction after a failed ast to the elt thread instead of
2521 handling it in the dlm interpret routine.
2523 Severity : enhancement
2525 Description: More exported tunables for mballoc
2526 Details : Add support for tunable preallocation window and new tunables for
2527 large/small requests
2531 Description: Detect corruption of block bitmap and checking for preallocations
2532 Details : Checks validity of on-disk block bitmap. Also it does better
2533 checking of number of applied preallocations. When corruption is
2534 found, it turns filesystem readonly to prevent further corruptions.
2538 Frequency : only for big-endian servers
2539 Description: Check for big-endian system while mounting fs with extents feature
2540 Details : Mounting a filesystem with extents feature will fail on big-endian
2541 systems since ext3-based ldiskfs is not supported on big-endian
2542 systems. Can be over-ridden with "bigendian_extents" mount option.
2546 Description: Excessive recovery window
2547 Details : With AT enabled, the recovery window can be excessively long (6000+
2548 seconds). To address this problem, we no longer use
2549 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2550 timeout no longer depends on the service time, it is set to
2551 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2552 time via pb_service_time.
2555 Descriptoin: Don't sync journal after every i/o
2556 Details : Implement write RPC replay to allow server replies for write RPCs
2557 before data is on disk. However, this feature is disabled by
2558 default since some issues leading to data corruptions have been
2559 found during recovery (e.g. bug 19128). This feature can be enabled
2560 by running the following command on the OSSs:
2561 lctl set_param obdfilter.*.sync_journal=0
2565 Description: Watchdog triggered on MDS failover
2566 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2567 the OSTs know that the MDS "UUID" can be reused for the same export
2568 from a different NID, so we do not need to wait for the export to
2573 Descriptoin: Lustre detected file system corruption with inode out of bounds
2574 Details : don't update i_size on MDS_CLOSE for directories. This causes
2575 directory corruptions on the MDT.
2577 -------------------------------------------------------------------------------
2579 2008-05-26 Sun Microsystems, Inc.
2581 * Support for kernels:
2582 2.6.5-7.311 (SLES 9),
2583 2.6.9-67.0.7.EL (RHEL 4),
2584 2.6.16.54-0.2.5 (SLES 10),
2585 2.6.18-53.1.14.el5 (RHEL 5),
2586 2.6.22.14 vanilla (kernel.org)
2587 * Client support for unpatched kernels:
2588 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2589 2.6.16 - 2.6.22 vanilla (kernel.org)
2590 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2591 we do not recommend using patchless RHEL4 clients with kernels
2592 prior to 2.6.9-55EL (RHEL4U5).
2593 * Recommended e2fsprogs version: 1.40.7-sun1
2594 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2595 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2596 removed cwd "./" (refer to Bugzilla 14399).
2597 * A new quota file format has been introduced in 1.6.5.
2598 The format conversion from prior releases is handled transparently,
2599 but releases older than 1.4.12/1.6.5 will not understand this new
2600 format. The automatic format conversion can be avoided by running
2601 the following command on the MDS before upgrading:
2602 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2603 For more information, please refer to bugzilla 13904.
2607 Description: quota performance fix
2608 Details : quota data is written in journalled mode instead of ordered to
2609 increase performance
2613 Description: lfs support for human-readable quota grace time strings
2614 Details : lfs setquota -t and lfs quota -t represent quota grace times
2615 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2618 Frequency : always with o2ib 1.3 and sles10
2620 Description: fix build with SLES10 and o2ib v3.
2621 Details : sles10 uses diffrent name for Module.symver file but configure
2622 assume this file has same name on RHEL/SLES/vanila kernels.
2625 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2627 Description: MDS may lose file striping (and hence file data) in some cases
2628 Details : If there are additional extended attributes stored on the MDS,
2629 in particular ACLs, SELinux, or user attributes (if user_xattr
2630 is specified for the client mount options) then there is a risk
2631 of attribute loss. Additionally, the Lustre file striping
2632 needs to be larger than default (e.g. striped over all OSTs),
2633 and an additional attribute must be stored initially in the
2634 inode and then increase in size enough to be moved to the
2635 external attribute block (e.g. ACL growing in size) for file
2638 Severity : enhancement
2640 Description: add message levels for liblustreapi
2645 Description: MDT cannot be unmounted, reporting "Mount still busy"
2646 Details : Mountpoint references were being leaked during open reply
2647 reconstruction after an MDS restart. Drop mountpoint reference
2648 in reconstruct_open() and free dentry reference also.
2653 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2654 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2655 will result in an -EAGAIN and in the exit call path we call
2656 alloc_rr() which will with increasing aggressiveness attempt to
2657 aquire precreated objects on the minimum number of required OSCs.
2661 Description: Use old size assignment to avoid deadlock
2662 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2663 the scheduling while holding a spinlock. We do not need locking
2664 for size in ll_update_inode() because size is only updated from
2665 the MDS for directories or files without objects, so there is no
2666 other place to do the update, and concurrent access to such inodes
2667 are protected by the inode lock.
2671 Description: Use __u64 instead of int for valid bits
2675 Description: resolve "_IOWR redefined" build error on SLES10
2679 Description: dump the memory debugging after all modules are unloaded to
2680 suppress false negative in conf_sanity test 39
2684 Description: the recovery timer never expires
2685 Details : for new client connect request, the recovery timer should not be
2686 reset, otherwise recovery timer will never expired, if the old
2687 client never come. Only old client connect and first connection
2688 req should trigger recovery timer reset.
2692 Description: the min numbers of lproc stats are wrong
2693 Details : adding a new constant LC_MIN_INIT and use it for initialization
2697 Frequency : always with interactive lfs
2699 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2702 Frequency : with multiple concurrent readdir processes in same directory
2703 Bugzilla : 15406, 15169, 15175
2704 Description: misc fixes for directory readahead.
2705 Details : prevent previous statahead async RPC callback from processing the
2706 current "statahead_info", race condition between async RPC callback
2707 add dentry into dentry hash table and "ls" thread revalidate such
2708 dentry, statahead his/miss control for hidden items, and so on.
2710 Severity : enhancement
2712 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2716 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2717 Details : several u32 variables declarations are replaced with u64 declarations
2722 Description: fix SLES kernel versioning
2723 Details : the kernel version for our SLES 10 kernel did not include a "-"
2724 before the "smp" at the end. while this was not a problem in
2725 general, it did mean that software trying to use the kernel
2726 version to try to detect a vendor specific kernel would fail.
2727 this was most evident by the OFED build scripts.
2732 Description: Don't update lov_desc members until making sure they are valid
2733 Details : When updating lov_desc members via proc fs, need fix their
2734 validities before doing the real update.
2737 Frequency : very rare
2739 Description: don't put request into delay list while invalidate in flight.
2740 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2741 import in flight. this produce timeout for invalidate and sometimes
2742 can cause stale data.
2744 Severity : enhancement
2746 Description: Update kernel to SLES9 2.6.5-7.311.
2748 Severity : enhancement
2750 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2755 Frequency : on PPC only
2756 Description: not convert ost objects for directory because it's not exist.
2757 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2759 Severity : enhancement
2761 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2762 __u64 as usingied long long type.
2765 Frequency : rare, on shutdown
2767 Description: race process ast vs remove callback
2768 Details : removing callback before disconnect import open race with
2769 processing callback.
2771 Severity : enhancement
2773 Description: Update kernel to SLES9 2.6.5-7.311.
2775 Severity : enhancement
2777 Description: Files open for execute are not marked busy on SLES10
2778 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2780 Severity : enhancement
2782 Description: Add server support for vanilla-2.6.22.14.
2785 Frequency : occasional
2787 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2788 Details : If an OST is down the MDS will hang indefinitely in
2789 obd_statfs_rqset() waiting for the statfs data. While for
2790 MDS QOS usage of statfs, it should not stuck in waiting.
2792 Severity : enhancement
2794 Description: Disable adaptive timeouts by default
2797 Frequency : on network error
2799 Description: panic with double free request if network error
2800 Details : mdc_finish_enqueue is finish request if any network error occuring,
2801 but it's true only for synchronus enqueue, for async enqueue
2802 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2806 Frequency : rare, on recovery
2808 Description: read procfs can produce deadlock in some situation
2809 Details : Holding lprocfs lock with send rpc can produce block for destroy
2810 obd objects and this also block reconnect with -EALREADY.
2811 This isn't fix all lprocfs bugs - but make it rare.
2813 Severity : enhancement
2815 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2818 Frequency : frequent on X2 node
2820 Description: mdc_set_open_replay_data LBUG
2821 Details : Set replay data for requests that are eligible for replay.
2826 Description: lustre_mgs: operation 101 on unconnected MGS
2827 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2828 MGC, and late on MGC cannot successfully connect to MGS and a lot
2829 of the error messages complaining that MGS is not connected.
2832 Frequency : rare, depends on device drivers and load
2834 Description: MDS or OSS nodes crash due to stack overflow
2835 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2836 In some cases, in conjunction with device drivers that use a lot
2837 of stack, the MDS (or possibly OSS) service threads could overflow
2838 the stack. One change which was identified to consume additional
2839 stack has been reworked to avoid the extra stack usage.
2841 Severity : enhancement
2843 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2845 Severity : enhancement
2847 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2849 Severity : enhancement
2851 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2853 Serverity : enhancement
2855 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2858 Frequency : rare on shutdown OST
2860 Description: Don't allow skipping OSTs if index has been specified.
2861 Details : Don't allow skipping OSTs if index has been specified, make locking
2862 in internal create lots better.
2867 Description: ASSERTION(!PageDirty(page)) failed
2868 Details : Wrong check could lead to an assertion failure under specific
2874 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2875 Details : access to bitfield in structure is always rounded to long
2876 and this produce problem with not atomic change any bit.
2881 Description: Lustre make rpms failed.
2882 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2883 builds Lustre rpms from tarball.
2885 Severity : enhancement
2887 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2890 Frequency : rare on shutdown OST
2892 Description: If llog cancel was not send before clean_exports phase, this can
2893 produce deadlock in llog code.
2894 Details : If llog thread has last reference to obd and call class_import_put
2895 this produce deadlock because llog_cleanup_commit_master wait when
2896 last llog_commit_thread exited, but this never success because was
2897 called from llog_commit_thread.
2900 Frequency : only if OST index is skipped
2902 Description: NULL lov_tgts causing MDS oops
2903 Details : more safe checks for NULL lov_tgts for avoid oops.
2905 Severity : enhancement
2907 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2909 Severity : enhancement
2911 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2916 Description: make mgs_setparam() handle fsname containing dash
2917 Details : fsname containing a dash does not work with lctl conf_param
2919 Severity : enhancement
2921 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2924 Frequency : rare, in recovery and (or) destroy lovobjid file.
2926 Description: rewrite lov objid code.
2927 Details : Cleanup for lov objid code, remove scability problems and wrong
2928 locking. Fix sending last_id into ost.
2930 Severity : enhancement
2932 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2934 Severity : enhancement
2936 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2937 Details : Use d_move_locked instead of __d_move.
2940 Frequency : rare, at shutdown
2942 Description: access already free / zero obd_namespace.
2943 Details : if client_disconnect_export was called without force flag set,
2944 and exist connect request in flight, this can produce access to
2945 NULL pointer (or already free pointer) when connect_interpret
2946 store ocd flags in obd_namespace.
2949 Frequency : only at startup
2951 Description: not alloc memory with spinlock held.
2952 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2953 if any spinlock held.
2958 Description: lfs find does not continue on file error
2959 Details : Continue other files processing when a file/dir is absent.
2963 Description: Inconsistent usage of lustre_pack_reply()
2964 Details : Standardize the usage of lustre_pack_reply() such that it
2965 always generate a CERROR on failure.
2968 Frequency : very rare
2970 Description: Fix replay if there is an un-replied request and open
2971 Details : In some cases, older replay request will revert the
2972 mcd->mcd_last_xid on MDS which is used to record the client's
2973 latest sent request.
2975 Severity : enhancement
2977 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2979 Severity : enhancement
2981 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2983 Severity : enhancement
2985 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2987 Severity : enhancement
2989 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2991 Severity : enhancement
2993 Description: Build SLES10 patchless client fails
2994 Details : The configure was broken by run ./configure with
2995 --with-linux-obj=.... argument for patchless client. When the
2996 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2997 can't search header adequately. Use absolute path such as
2998 -I($LINUX)/include instead.
3000 Severity : enhancement
3002 Description: Lustre Page Accounting
3003 Details : New macros for page alloc and free which enable accounting
3004 of page allocation of Lustre. Use percpu counters to store memory
3005 and page statistics.
3008 Frequency : only if debugging is disabled
3010 Description: LASSERT_{REQ,REP}SWAB macros are buggy
3011 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
3012 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
3013 these macros and replace them with their definitions instead.
3018 Description: interrupt oig_wait produce painc on resend.
3019 Details : brw_redo_request can be used for resend requests from ptlrpcd and
3020 private set, and this produce situation when rq_ptlrpcd_data not
3021 copyed to new allocated request and triggered LBUG on assert
3022 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
3023 ptlrpcd set if request is changed and can be safety changed to use
3026 Severity : enhancement
3028 Description: organize the server-side client stats on per-nid basis
3029 Details : Change the structure of stats under obdfilter and mds to
3037 The "uuid"s file would list the uuids of _active_ exports.
3038 And the clear entry is to clear all stats and stale nids.
3043 Description: Processes looping in ll_readdir() on Lustre clients finally causing
3044 a full node pseudo-hang
3045 Details : Concurrent access to the same directory from multiple clients with
3046 intensive file creation/removal can cause a client node to spin in
3047 ll_readdir(). i_version must be increased every time the lock is
3048 cancelled to ensure a revalidate is done.
3053 Description: touch file failed when fs is not full
3054 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
3055 otherwise we can get ENOSP while fs is not full.
3058 Frequency : only for Cray XT3
3059 Bugzilla : 12829/13455
3060 Description: Changing primary group doesn't change the group lustre assigns to
3062 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
3063 overridden with the primary group provided by the group upcall,
3064 whereas the supplied fsgid can be trusted if it is in the list of
3065 supplementary groups returned by the group upcall.
3067 Severity : enhancement
3069 Description: Root Squash Functionality
3070 Details : Implementation of NFS-like root squash capability. Specifically,
3071 don't allow someone with root access on a client node to be able
3072 to manipulate files owned by root on a server node.
3074 Severity : enhancement
3076 Description: Slow trucate/writes to huge files at high offsets.
3077 Details : Directly associate cached pages to lock that protect those pages,
3078 this allows us to quickly find what pages to write and remove
3079 once lock callback is received.
3084 Description: Too many locks accumulating on client during NFS usage
3085 Details : mds_open improperly used accmode to find out access mode to a
3086 file. Also mdc_intent_lock logic to find out if we already have
3087 lock similar to just received was flawed since introduction of
3088 skiplists - locks are now added to the front of the granted
3093 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
3095 Details : In obd setup failure handler, obd_stopping will not necessarily to
3096 be 1, and obd_set_up should also be checked to make sure whether
3097 obd is completely setup.
3099 Severity : enhancement
3101 Description: Allow masking D_WARNING, D_ERROR messages from console
3102 Details : Console messages can now be disabled via lnet.printk.
3107 Description: User code with malformed file open parameter crashes client node
3108 Details : Before packing join_file req, all the related reference should be
3109 checked carefully in case some malformed flags cause fake join_file
3115 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
3116 grain of quota doing harm to quota's accuracy
3117 Details : qunit size will be changed when quota limitation is too low/high;
3118 record the pending quota write in order to get more accureate
3119 quota; delete the patch for bug12588, which is unnecessary when
3120 this patch is landed. This bug also contains fixes for bug 14526,
3121 14299, 14601 and 13794.
3125 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
3126 Details : ldlm_completion_ast() assumes that a lock is granted when the req
3127 mode is equal to the granted mode. However, it should also check
3128 that LDLM_FL_CP_REQD is not set.
3132 Description: Heavy nfs access might result in deadlocks
3133 Details : After ELC code landed, it is now improper to enqueue any mds
3134 locks under och_sem, because enqueue might want to decide to
3135 cancel open locks for same inode we are holding och_sem for.
3139 Description: 35% write performance drop with ldiskfs2 when quotas are on
3140 Details : Enable ext3 journalled quota by default to improve performance
3141 when quotas are turned on.
3145 Description: Client eviction while running blogbench
3146 Details : A lot of unlink operations with concurrent I/O can lead to a
3147 deadlock causing evictions. To address the problem, the number of
3148 oustanding OST_DESTROY requests is now throttled to
3149 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
3150 callbacks are processed in priority.
3153 Frequency : RHEL4 only
3155 Description: mkfs is very slow on IA64/RHEL4
3156 Details : A performance regression has been discovered in the MPT Fusion
3157 driver between versions 3.02.73rh and 3.02.99.00rh. As a
3158 consequence, we have downgraded the MPT Fusion driver in the RHEL4
3159 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
3162 Frequency : PPC/PPC64 only
3164 Description: conflicts between asm-ppc64/types.h and lustre_types.h
3165 Details : fix duplicated definitions between asm-ppc64/types.h and
3166 lustre_types.h on PPC.
3169 Frequency : PPC/PPC64 only
3171 Description: asm-ppc/segment.h does not exist
3172 Details : fix compile issue on PPC.
3176 Description: data checksumming impacts single node performance
3177 Details : add support for several checksum algorithms. Currently, CRC32 and
3178 Adler-32 are supported. The checksum type can be changed on the fly
3179 through /proc/fs/lustre/osc/*/checksum_type.
3183 Description: use adler32 for page checksums
3184 Details : when available, use the Adler-32 algorithm instead of CRC32 for
3189 Description: better handle error messages in extents code
3191 Severity : enhancement
3193 Description: SNMP support enhancement
3194 Details : Adding total number of sampled request for an MDS node in snmp
3197 Severity : enhancement
3199 Description: Optimize ldlm waiting list processing for PR extent locks
3200 Details : When processing waiting list for read extent lock and meeting read
3201 lock that is same or wider to it that is not contended, skip
3202 processing rest of the list and immediatelly return current
3203 status of conflictness, since we are guaranteed there are no
3204 conflicting locks in the rest of the list.
3208 Description: Time out and refuse to reconnect
3209 Details : When the failover node is the primary node, it is possible
3210 to have two identical connections in imp_conn_list. We must
3211 compare not conn's pointers but NIDs, otherwise we
3212 can defeat connection throttling.
3216 Description: Client not clear own cache if answer to reconnect is lost.
3217 Details : Client gets evicted from server. Now client also thinks it is
3218 disconnected (or gets ENOTCONN on its operation) and decides to
3219 reconnect. Server receives reconnect message, but cannot find
3220 export. New export is created that is fully valid (new cookie!),
3221 but reply is lost and not reported to client. Client reconnects
3222 again and gets back a just-created connection, but it is not new
3223 so client thinks it was not evicted and does not do recovery.
3227 Description: Detect stride IO mode in read-ahead
3228 Details : When a client does stride read, read-ahead should detect that and
3229 read-ahead pages according to the detected stride pattern.
3233 Description: build for x2 fails
3234 Details : fix compile issue on Cray systems.
3236 Severity : enhancement
3238 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
3239 Details : This greatly improves speed of NFS writes on 2.6 kernels.
3242 Frequency : only on PPC/SLES10
3244 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
3245 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
3246 be defined. Add a hack in mkfs_lustre.c to work around this compile
3251 Description: LASSERT on MDS when client holding flock lock dies
3252 Details : ldlm pool logic depends on number of granted locks equal to
3253 number of released locks which is not true for flock locks, so
3254 just exclude such locks from consideration.
3258 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
3259 Details : Use fsfilt_sync() for both the whole filesystem sync and
3260 individual file sync to eliminate dangerous inode locking
3261 with I_LOCK that can lead to a deadlock.
3265 Description: Update an obsolete wirecheck.c generator
3266 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
3270 Description: Client can panic on open sometimes
3271 Details : It is possible that we try to free already freed request in
3272 ll_file_open in some error cases when we send request from
3277 Description: performance in 1.6.3
3278 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3279 sure that 1MB requests are not fragmented by the block layer.
3283 Description: LDLM soft lockups - improvement
3284 Details : It is be possible to send the lock handle along with each read
3285 or write request because the client is already doing a lock match
3286 itself so there isn't any reason the OST should have to re-do that
3292 Description: lfs quota fails with deactivated OSTS
3293 Details : With this patch, three improvements are included:
3294 1. detete the softlimit in mds and osts when use "lfs quota".
3295 2. display the inaccurate data in the output of "lfs quota".
3296 3. try to get quota info when "lfs quota" is executed.
3301 Description: Extent locks not granted with no conflicts sometimes.
3302 Details : When race occurs in glimpse handler and nothing is returned,
3303 we do not reprocess the queue after lock cancel, and that leads
3304 to a stall until next activity on a resource
3307 Frequency : failover with quotaon
3309 Description: during mds failovers with quota on, OSTs got into deadlock state
3310 and causing dumpstack.
3311 Details : for every quota slave, at any time, there is only one quota req
3312 is sent to quota master for every uid/gid. Before that quota req
3313 returns, all the thread relative to the same uid/gid will wait.
3314 So if the quota req is lost because mds failovers or any other
3315 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3316 quota reqs who time out.
3318 Severity : enhancement
3321 Description: when quota slave checks if quota is enough, there is an unnecessary
3323 Details : place this wait on necessary place instead of always waiting.
3325 --------------------------------------------------------------------------------
3327 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3329 * Support for kernels:
3330 2.6.5-7.286 (SLES 9),
3331 2.6.9-55.0.9.EL (RHEL 4),
3332 2.6.16.53-0.8 (SLES 10),
3333 2.6.18-8.1.14.el5 (RHEL 5),
3334 2.6.18.8 vanilla (kernel.org)
3335 * Client support for unpatched kernels:
3336 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3337 2.6.16 - 2.6.22 vanilla (kernel.org)
3338 * Due to recently discovered recovery problems, we do not recommend
3339 using patchless RHEL 4 clients with this or any earlier release.
3340 * Recommended e2fsprogs version: 1.40.2-cfs1
3341 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3344 Frequency : occasional
3346 Description: MDS hang or stay in waiting lock
3347 Details : If client receive lock with CBPENDING flag ldlm need send lock
3348 cancel as separate rpc, to avoid situation when cancel request
3349 can't processed due all i/o threads stay in wait lock.
3352 Frequency : occasional
3354 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3355 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3356 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3357 oscc_has_objects and see OSC has no objests and call
3358 oscc_internal_create to resend create request.
3360 Severity : enhancement
3362 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3363 Details : Update which_patch & target file for SLES10 latest kernel.
3365 Severity : enhancement
3367 Description: add --type and --size parameters to lfs find
3368 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3369 multiple OBDs can now be specified for the --obd option.
3371 Severity : enhancement
3373 Description: eliminate client locks in face of contention
3374 Details : file contention detection and lockless i/o implementation
3375 for contended files.
3377 Severity : enhancement
3379 Description: Remove client patches from SLES 10 kernel.
3380 Details : This causes SLES 10 clients to behave as patchless clients
3381 even on a Lustre-patched (server) kernel.
3383 Severity : enhancement
3385 Description: use i_size_read and i_size_write in 2.6 port
3386 Details : replace inode->i_size access with i_size_read/write()
3389 Frequency : when removing large files
3391 Description: scheduling issue during removal of large Lustre files
3392 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3393 It causes scheduling issues when removing large files (17TB in the
3399 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3400 Details : Impossible (EPROTO is returned) to access a directory that has a
3401 non-default striping and ACLs.
3404 Frequency : only on ppc
3406 Description: /proc/fs/lustre/devices broken on ppc
3407 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3408 We should make sure the type of 'index' is loff_t and then cast
3409 explicitly as needed below. Do not assign an explicitly cast
3413 Frequency : only for rhel5
3415 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3416 Details : Modify the target file & which_kernel.
3419 Frequency : if the uninit_groups feature is enabled on ldiskfs
3421 Description: e2fsck reports "invalid unused inodes count"
3422 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3423 feature and only a single inode is created in a group then the
3424 "bg_unused_inodes" count is incorrectly updated. Creating a
3425 second inode in that group would update it correctly.
3430 Description: buffer overruns could theoretically occur
3431 Details : llapi_semantic_traverse() modifies the "path" argument by
3432 appending values to the end of the origin string, and a buffer
3433 overrun may occur. Adding buffer overrun check in liblustreapi.
3435 Severity : enhancement
3437 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3440 Frequency : only if filesystem is inconsistent
3442 Description: handle "serious error: objid * already exists" more gracefully
3443 Details : If LAST_ID value on disk is smaller than the objects existing in
3444 the O/0/d* directories, it indicates disk corruption and causes an
3445 LBUG(). If the object is 0-length, then we should use the existing
3446 object. This will help to avoid a full fsck in most cases.
3448 Severity : enhancement
3450 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3451 Details : Modify vm-tunables-rhel4.patch.
3453 Severity : enhancement
3455 Description: Kernel config for 2.6.18-vanilla.
3456 Details : Modify targets/2.6-vanilla.target.in.
3457 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3458 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3459 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3460 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3463 Frequency : occasional
3465 Description: improve handling recoverable errors
3466 Details : If request processed with error which can be recoverable on server
3467 request should be resend, otherwise page released from cache and
3473 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3474 Details : Modify target file & which_patch.
3475 A flaw was found in the IA32 system call emulation provided
3476 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3477 value could be stored in the %RAX register, which could trigger an
3478 out-of-bounds system call table access. An untrusted local user
3479 could exploit this flaw to run code in the kernel
3480 (ie a root privilege escalation). (CVE-2007-4573).
3484 Description: change order of libsysio includes
3485 Details : '#include sysio.h' should always come before '#include xtio.h'
3487 Severity : enhancement
3489 Description: adapt the lustre_config script to support the upgrade case
3490 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3491 server targets to 1.6 in parallel.
3496 Description: To avoid grant space > avaible space when the disk is almost
3497 full. Without this patch you might see the error "grant XXXX >
3498 available" or some LBUG about grant, when the disk is almost
3500 Details : In filter_check_grant, for non_grant cache write, we should
3501 check the left space by if (*left > ungranted + bytes), instead
3502 of (*left > ungranted), because only we are sure the left space
3503 is enough for another "bytes", then the ungrant space should be
3504 increase. In client, we should update cl_avail_grant only there is
3505 OBD_MD_FLGRANT in the reply.
3510 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3511 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3512 described in CVE-2007-4573. This problem could allow untrusted
3513 local users to gain root access.
3516 Frequency : when using O_DIRECT and quotas
3518 Description: Incorrect file ownership on O_DIRECT output files
3519 Details : block usage reported by 'lfs quota' does not take into account
3520 files that have been written with O_DIRECT.
3524 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3525 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3526 window to 0. If it does hit cache miss, and read-pattern is still
3527 stride-io mode, does not reset the stride window, but also does not
3528 increase the stride window length in this case.
3530 --------------------------------------------------------------------------------
3532 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3534 * Support for kernels:
3535 2.6.5-7.286 (SLES 9),
3536 2.6.9-55.0.2.EL (RHEL 4),
3537 2.6.16.46-0.14 (SLES 10),
3538 2.6.18-8.1.8.el5 (RHEL 5),
3539 2.6.18.8 vanilla (kernel.org)
3540 * Client support for unpatched kernels:
3541 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3542 2.6.16 - 2.6.21 vanilla (kernel.org)
3543 * Due to recently discovered recovery problems, we do not recommend
3544 using patchless RHEL 4 clients with this or any earlier release.
3545 * Recommended e2fsprogs version: 1.40.2-cfs1
3546 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3550 Description: Fix errors in lfs documentation
3551 Details : Fixes man pages
3553 Severity : enhancement
3555 Description: Adaptive timeouts
3556 Details : RPC timeouts adapt to changing server load and network
3557 conditions to reduce resend attempts and improve recovery time.
3559 Severity : enhancement
3561 Description: llapi_file_create() does not allow some changes
3562 Details : add llapi_file_open() that allows specifying the file creation
3563 mode and open flags, and also returns an open file handle.
3565 Severity : enhancement
3567 Description: df doesn't work properly if diskfs blocksize != 4K
3568 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3571 Frequency : always on directories with default striping set
3573 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3574 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3576 Severity : enhancement
3578 Description: merge and cleanup kernel patches.
3579 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3581 Severity : enhancement
3583 Description: RedHat Update kernel for RHEL5
3584 Details : Kernel config file for RHEL5.
3586 Severity : enhancement
3588 Description: OSS needs mutliple precreate threads
3589 Details : Add ability to start more than one create thread per OSS.
3592 Frequency : only with quota on the root user
3594 Description: mds_obd_create error creating tmp object
3595 Details : When the user sets quota on root, llog will be affected and can't
3596 create files and write files.
3599 Frequency : Always on ia64 patchless client, and possibly others.
3601 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3602 Details : This allows the patchless client to be loaded on architectures
3603 without this export.
3608 Description: disorder of journal start and llog_add cause deadlock.
3609 Details : in llog_origin_connect, journal start should happen before
3610 llog_add keep the same order as other functions to avoid
3613 Severity : enhancement
3615 Description: RedHat Update kernel for RHEL5
3616 Details : Modify the kernel config file more closer RHEL5.
3618 Severity : enhancement
3620 Description: Build failure against Centos5 (RHEL5)
3621 Details : Define PAGE_SIZE when it isn't present.
3624 Frequency : occasionally when using NFS
3626 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3627 Details : with NFS, the anon dentry's parent was set to itself in
3628 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3629 corresponding dentry other than use rec->ur_name.
3631 Severity : enhancement
3633 Description: enable data checksumming by default
3634 Details : enable checksum by default, allow --disable-checksum
3635 configure option and "-o nochecksum" mount option. Checksums
3636 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3639 Frequency : Occasionally with failover
3641 Description: Client eviction due to failover config
3642 Details : after a connection loss, the lustre client should attempt to
3643 reconnect to the last active server first before trying the
3644 other potential connections.
3646 Severity : enhancement
3648 Description: client-side metadata stat-ahead during readdir(directory readahead)
3649 Details : perform client-side metadata stat-ahead when the client detects
3650 readdir and sequential stat of dir entries therein
3655 Description: when mds and osts use different quota unit(32bit and 64bit),
3656 quota will be released repeatly.
3657 Details : void sending multiple quota reqs to mds, which will keep the
3658 status between the reqs.
3661 Frequency : only with liblustre clients on XT3
3663 Description: evictions taking too long
3664 Details : allow llrd to evict clients directly on OSTs
3668 Description: osts not allocated evenly to files
3669 Details : change the condition to increase offset_idx
3671 Severity : enhancement
3673 Description: self-adjustable client's lru lists
3674 Details : use adaptive algorithm for managing client cached locks lru
3675 lists according to current server load, other client's work
3676 pattern, memory activities, etc. Both, server and client
3677 side namespaces provide number of proc tunables for controlling
3682 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3683 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3684 readdir code from ext3) from scratch preserving functionality.
3688 Description: Only those disconnect error should be returned by rq_status.
3689 Details : In open/enqueue processs, Some errors, which will cause client
3690 disconnected, should be returned by rq_status, while other
3691 errors should still be returned by intent, then mdc or llite will
3694 Severity : enhancement
3696 Description: Tune the kernel for good SCSI performance.
3697 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3698 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3702 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3703 Bugzilla : 13547, 13627
3704 Description: Data corruption for OSTs that are formatted larger than 2TB
3706 Details : When generating the bio request for lustre file writes the
3707 sector number would overflow a temporary variable before being
3708 used for the IO. The data reads correctly from Lustre (which
3709 will overflow in a similar manner) but other file data or
3710 filesystem metadata may be corrupted in some cases.
3714 Description: "lfs find -obd UUID" prints directories
3715 Details : "lfs find -obd UUID" will return all directory names instead
3716 of just file names. It is incorrect because the directories
3717 do not reside on the OSTs.
3720 Bugzilla : 13596, 13614
3721 Description: MDS hang after unclean shutdown of lots of clients
3722 Details : Never resend AST requests.
3726 Frequency : Always, for kernels after 2.6.16
3727 Description: Fix warning idr_remove called for id=.. which is not allocated.
3728 Details : Last kernels save old s_dev before kill super and not allow
3729 to restore from callback - restore it before call kill_anon_super.
3733 Description: Fix error on 'ls .' at the top of the Lustre mount.
3734 Details : Don't revalidate dentry if it is a root dentry.
3736 --------------------------------------------------------------------------------
3738 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3740 * Support for kernels:
3741 2.6.5-7.286 (SLES 9),
3742 2.6.9-55.0.2.EL (RHEL 4),
3743 2.6.16.46-0.14 (SLES 10),
3744 2.6.18-8.1.8.el5 (RHEL 5),
3745 2.6.18.8 vanilla (kernel.org)
3746 * Client support for unpatched kernels:
3747 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3748 2.6.16 - 2.6.22 vanilla (kernel.org)
3749 * Due to recently discovered recovery problems, we do not recommend
3750 using patchless RHEL 4 clients with this or any earlier release.
3751 * Recommended e2fsprogs version: 1.39.cfs8
3752 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3757 Description: block reactivating mgc import until all deactivates complete
3758 Details : Fix race when failing back MDT/MGS to itself (testing)
3760 Severity : enhancement
3762 Description: lfs setstripe enhancement
3763 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3766 Frequency : mds/oss recovery
3768 Description: llog ctxt is refrenced after it has been freed.
3769 Details : llog ctxt refcount was added to avoide the race between ctxt free
3770 and llog recovery process. Each llog user must hold ctxt refcount
3771 before it access the llog. And the llog ctxt can only be freed
3772 when its refcount is zero.
3774 Severity : enhancement
3776 Description: randomly memory allocation failure util
3777 Details : Make lustre randomly failed allocating memory for testing purpose.
3780 Frequency : only for SLES10
3782 Description: Update kernel patch for SLES10 SP1
3783 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3785 Severity : enhancement
3787 Description: omit set fsid for export NFS
3788 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3789 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3792 Frequency : after network failures
3794 Description: Add sync option to mount_lustre.c
3795 Details : Client loses data written to lustre after a network interruption.
3797 Severity : enhancement
3799 Description: Error message improvement.
3800 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3802 Severity : enhancement
3804 Description: don't use GFP_* in generic Lustre code.
3805 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3807 Severity : enhancement
3809 Description: obdclass is limited by single OBD_ALLOC(idarray)
3810 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3812 Severity : enhancement
3814 Description: updated patchess for new RHEL4 kernel
3815 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3817 Severity : enhancement
3819 Description: warnings with build patchless client with vanila 2.6.19 and up
3820 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3821 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3823 Severity : enhancement
3825 Description: O_DIRECT bypasses client statistics.
3826 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3827 read_ahead_stats were updated. Copied stats section from
3828 osc_send_oap_rpc() into async_internal().
3831 Frequency : only for Cray XT3
3833 Description: peer credits not enough on many OST per OSS systems.
3834 Details : Use new lnet way to add credits as we need those for pings and ASTs
3837 Frequency : only with liblustre
3839 Description: Liblustre is not releasing flock locks on file close.
3840 Details : Release flock locks on file close.
3843 Frequency : only for RHEL4
3845 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3846 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3847 Update target file and kernel config.
3851 Description: lustre support for RHEL5
3852 Details : Add support for RHEL5.
3857 Description: ASSERTION(export != NULL) failed in target_handle_connect
3858 Details : Assetion hit is result of rare race between disconnect and connet
3859 to same nid. target_handle_connect found old connect cockie and
3860 tried to reconnect, but can't find export for this cockie.
3865 Description: umount blocks forever on error
3866 Details : In result of wrong using obd_no_recov and obd_force flags client
3867 can hand if cancel or some other requests is lost.
3870 Frequency : Only for SLES
3872 Description: sanity_quota fail test_1
3873 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3874 file, which makes TSTID[2] inunique.
3876 Severity : enhancement
3878 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3879 Details : Update target/ChangeLog/which_patch .
3881 Severity : enhancement
3883 Description: jbd statistics
3884 Details : Port older jbd statistics patch for sles10
3889 Description: lvbo_init failed for resource with missing objects.
3890 Details : Fix returning error if we do stat for file with missing/corrupted
3891 objects and i_size set to all sum of size of all avaible objects.
3892 if we to truncate/write to missing object - it is recreated.
3897 Description: Oops in read and write path when failing to allocate lock.
3898 Details : Check if lock allocation failed and return error back.
3901 Frequency : When flocks are used.
3903 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3904 Details : Flock locks might destroy just granted lock if it could be merged
3905 with another existing flock, this is done in completion handler,
3906 so teach ldlm_cli_enquque_fini that this is a valid case for
3912 Description: reply_lock_interpret crash due to race with it and lock cancel.
3913 Details : Do not replay locks that are being cancelled. Do not reference
3914 locks by their address during replay, just by their handle.
3917 Frequency : only with deactivated OSTs
3919 Description: lstripe command fails for valid OST index
3920 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3921 instead of lov->desc.ld_active_tgt_count.
3923 Severity : enhancement
3925 Description: Build failure against Centos5 (RHEL5)
3926 Details : Use getpagesize() instead of PAGE_SIZE.
3928 --------------------------------------------------------------------------------
3930 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3932 * Support for kernels:
3933 2.6.5-7.283 (SLES 9),
3934 2.6.9-55.EL (RHEL 4),
3935 2.6.16.46-0.14 (SLES 10),
3936 2.6.18.8 vanilla (kernel.org)
3937 * Client support for unpatched kernels:
3938 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3939 2.6.16 - 2.6.22 vanilla (kernel.org)
3940 * Due to recently discovered recovery problems, we do not recommend
3941 using patchless RHEL 4 clients with this or any earlier release.
3942 * Recommended e2fsprogs version: 1.39.cfs8
3943 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3944 * Starting with this release, the ldiskfs backing filesystem required
3945 by Lustre is now in its own package, lustre-ldiskfs. This package
3946 should be installed. It is versioned separately from Lustre and
3947 may be released separately in future.
3952 Description: server LBUG when shutting down
3953 Details : Block umount forever until the mount refcount is zero rather
3954 than giving up after an arbitrary timeout.
3956 Severity : enhancement
3958 Description: add optional extra BUILD_VERSION info
3959 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3960 to override the lustre version.
3963 Frequency : 2.6.18 servers only
3965 Description: ll_kern_mount() doesn't release the module reference
3966 Details : The ldiskfs module reference count never drops down to 0
3967 because ll_kern_mount() doesn't release the module reference.
3972 Description: server LBUG when using old ost_num_threads parameter
3973 Details : Accept the old ost_num_threads parameter but warn that it
3974 is deprecated, and fix an off-by-one error that caused an LBUG.
3979 Description: Transient SCSI error results in persistent IO issue
3980 Details : iobuf->dr_error is not reinitialized to 0 between two
3984 Frequency : sometimes when underlying device returns I/O errors
3986 Description: OSTs not going read-only during write failures
3987 Details : OSTs are not remounted read-only when the journal commit threads
3988 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3989 instead of the ext3 wrappers.
3992 Frequency : only with 10000 clients or more
3994 Description: poor connect scaling with increasing client count
3995 Details : Don't run filter_grant_sanity_check for more than 100 exports
3996 to improve scaling for large numbers of clients.
3999 Frequency : SLES10 only
4001 Description: sanity-quota.sh quotacheck failed: rc = -22
4002 Details : Quotas cannot be enabled on SLES10.
4005 Frequency : liblustre clients only
4007 Description: getdirentries does not give error when run on compute nodes
4008 Details : getdirentries does not fail when the size specified as an argument
4009 is too small to contain at least one entry
4011 Severity : enhancement
4013 Description: Add LNET router traceability for debug purposes
4014 Details : If a checksum failure occurs with a router as part of the
4015 IO path, the NID of the last router that forwarded the bulk data
4016 is printed so it can be identified.
4021 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
4022 Details : Due to a race condition, liblustre clients were occasionally
4023 evicted incorrectly.
4025 Severity : enhancement
4027 Description: lfs setstripe use optional parameters instead of postional
4030 Severity : enhancement
4032 Description: Nanosecond timestamp support for ldiskfs
4033 Details : The on-disk ldiskfs filesystem has added support for nanosecond
4034 resolution timestamps. There is not yet support for this at
4035 the Lustre filesystem level.
4038 Frequency : during server recovery
4040 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
4041 Details : request with rq_no_resend flag not awake l_wait_event if they get a
4045 Frequency : nfs export on patchless client
4047 Description: connectathon hang when test nfs export over patchless client
4048 Details : Disconnected dentry cannot be found with lookup, so we do not need
4049 to unhash it or make it invalid
4052 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
4053 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
4054 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
4059 Description: open req refcounting wrong on reconnect
4060 Details : If reconnect happened between getting open reply from server and
4061 call to mdc_set_replay_data in ll_file_open, we will schedule
4062 replay for unreferenced request that we are about to free.
4063 Subsequent close will crash in variety of ways.
4064 Check that request is still eligible for replay in
4065 mdc_set_replay_data().
4070 Description: disable writes to filesystem when reading health_check file
4071 Details : the default for reading the health_check proc file has changed
4072 to NOT do a journal transaction and write to disk, because this
4073 can cause reads of the /proc file to hang and block HA state
4074 checking on a healthy but otherwise heavily loaded system. It
4075 is possible to return to the previous behaviour during configure
4076 with --enable-health-write.
4078 Severity : enhancement
4080 Description: 64-bit inode version
4081 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
4082 to the inode. This will be required for version-based recovery.
4087 Description: MDS fails to start if a duplicate client export is detected
4088 Details : in some rare cases it was possible for a client to connect to
4089 an MDS multiple times. Upon recovery the MDS would detect this
4090 and fail during startup. Handle this more gracefully.
4092 Severity : enhancement
4094 Description: Add -o localflock option to simulate old noflock behaviour.
4095 Details : This will achieve local-only flock/fcntl locks coherentness.
4100 Description: log_commit_thread vs filter_destroy race leads to crash
4101 Details : Take import reference before releasing llog record semaphore
4106 Description: Wrong request locking in request set processing
4107 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
4108 imp_delayed_list, in this place should be used imp_lock.
4111 Frequency : when reconnecting
4113 Description: Grant leak when OSC reconnect to OST
4114 Details : When osc reconnect ost, OST(filter) should check whether it
4115 should grant more space to client by comparing fed_grant and
4116 cl_avail_grant, and return the granted space to client instead
4117 of "new granted" space, because client will call osc_init_grant
4118 to update the client grant space info.
4121 Frequency : when client reconnects to OST
4123 Description: Grant leak when OSC does a resend and replays bulk write
4124 Details : When osc reconnect to OST, OST(filter) should clear grant info of
4125 bulk write request, because the grant info while be sync between
4126 OSC and OST when reconnect, and we should ignore the grant info
4127 these of resend/replay write req.
4132 Description: Grant space more than avaiable space sometimes.
4133 Details : When then OST is about to be full, if two bulk writing from
4134 different clients came to OST. Accord the avaliable space of the
4135 OST, the first req should be permitted, and the second one
4136 should be denied by ENOSPC. But if the seconde arrived before
4137 the first one is commited. The OST might wrongly permit second
4138 writing, which will cause grant space > avaiable space.
4141 Frequency : when client is evicted
4143 Description: Grant might be wrongly erased when osc is evicted by OST
4144 Details : when the import is evicted by server, it will fork another
4145 thread ptlrpc_invalidate_import_thread to invalidate the
4146 import, where the grant will be set to 0. While the original
4147 thread will update the grant it got when connecting. So if
4148 the former happened latter, the grant will be wrongly errased
4149 because of this race.
4154 Description: Checking Stale with correct fid
4155 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
4156 to check whether it is stale, because sometimes, we want the
4157 enqueue happened anyway, and op_data.fid2 will not be initialized.
4159 Severity : enhancement
4161 Description: update patchless client
4162 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
4165 Frequency : only with 2.4 kernel
4167 Description: random memory corruption
4168 Details : size of struct ll_inode_info is to big for union inode.u and this
4169 can be cause of random memory corruption.
4174 Description: Memory leak in recovery
4175 Details : Lov_mds_md was not free in an error handler in mds_create_object.
4176 It should also check obd_fail before fsfilt_start, otherwise if
4177 fsfilt_start return -EROFS,(failover mds during mds recovery).
4178 then the req will return with repmsg->transno = 0 and rc = EROFS.
4179 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
4180 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
4181 be freed no matter whether fsfilt_commit success or not.
4184 Frequency : only with large numbers of cores on a single node
4186 Description: Prevents from taking the superblock lock in llap_from_page for
4188 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
4189 for llap_from_page call in ll_removepage() prevents from taking
4190 the superblock lock for a soon died page.
4195 Description: Not check open intent error before release open handle
4196 Details : in some rare cases, the open intent error is not checked before
4197 release open handle, which may cause
4198 ASSERTION(open_req->rq_transno != 0), because it tries to release
4199 the failed open handle.
4204 Description: Set cat log bitmap only after create log success.
4205 Details : in some rare cases, the cat log bitmap is set too early. and it
4206 should be set only after create log success.
4210 Description: Accessing a block bevice can re-enable I/O when Lustre is
4211 tearing down a device.
4212 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
4217 Description: service threads may hog cpus when there are a lot of requests
4218 Details : Insert cond_resched to give other threads a chance to use some CPU
4223 Description: the cat log was not initialized in recovery
4224 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
4225 unlink log on mds will not be initialized until mds post
4226 recovery. And also in mds post recovery, the unlink log will
4227 initialization will be done asynchronausly, so there will be race
4228 between add unlink log and unlink log initialization.
4232 Description: brw_stats were being printed incorrectly
4233 Details : brw_stats were being printed as log2 but all of them were not
4234 recorded as log2. Also remove some code duplication arising from
4235 filter_tally_{read,write}.
4239 Frequency : rare, only in recovery.
4240 Description: ASSERTION(req->rq_type != LI_POISON) failed
4241 Details : imp_lock should be held while iterating over imp_sending_list for
4242 prevent destroy request after get timeout in ptlrpc_queue_wait.
4246 Description: replay-single.sh test 52 fails
4247 Details : A lock's skiplist need to be cleanup when it being unlinked
4248 from its resource list.
4253 Description: Short directio read returns full requested size rather than
4255 Details : Direct I/O operations should return actual amount of bytes
4256 transferred rather than requested size.
4258 Severity : enhancement
4260 Description: metadata RPC reduction (e.g. for rm performance)
4261 Details : decrease the amount of synchronous RPC between clients and servers
4262 by canceling conflicing lock before the operation on the client
4263 and packing thier handles into the main operation RPC to server.
4265 Severity : enhancement
4267 Description: add #ifdef HAVE_KERNEL_CONFIG_H
4268 Details : kernels from 2.6.19 not need include linux/config.h, but add
4269 include linux/autoconf.h in commpiler command line.
4271 Severity : enhancement
4273 Description: patchless client support for 2.6.22 kernel
4274 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
4275 removed. In this case we need drop using os depended interface to
4276 kmem_cache and use cfs_mem_cache API.
4281 Description: fix mal-formatted messages
4282 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4286 Frequency : always in liblustre
4287 Description: wrong IS_ERR implementation in liblustre.h
4288 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4293 Description: Correct condition for output debug message.
4294 Details : inode i_nlink equal zero is not enough for output message about
4295 disk corruption, i_ctime and i_mode should be also checked.
4299 Frequency : always in patchless client
4300 Description: add configure check for truncate_complete_page
4301 Details : improve checks for exported symbols. This allow run check without
4302 sources, but with Module.symvers shipped with kernel distribution.
4303 add check for truncate_complete_page used by patchless client.
4307 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4308 Details : test_77h uses a file which was messed by other test case.
4312 Description: Not Check whether lov_tgts is NULL in some lov functions
4313 Details : Checking whether lov_tgts is NULL in some functions.
4317 Description: replace obdo_alloc() with OBDO_ALLOC macro
4318 Details : nothing special is done in obdo_alloc() function, and for
4319 debugging purpose, it needs to be replaced with macros.
4323 Description: bad return value and errno from fcntl call
4324 Details : In liblustre API, errno should be a negative value if error
4329 Description: ptlrpc_check_set() LBUG
4330 Details : In case of positive reply from server and failed client bulk
4331 callback after bulk transfer shouldn't LBUG, but process this
4332 request as erroneous.
4334 Severity : enhancement
4336 Description: VFS operations stats tool.
4337 Details : Tool which collects stats by tracking value written in pid,
4338 ppid, gid and uses llstat to generate output to plot graph using
4340 Updated lustre/utils/Makefile.am
4341 Added lustre/utils/ltrack_stats.c
4343 Severity : enhancement
4345 Description: 2.6.18 server support (lustre 1.6.1)
4346 Details : Support for 2.6.18 kernels on the server side.
4351 Description: ASSERTION(imp->imp_conn_current) failed
4352 Details : an assertion failure is hit if a client node boots and attempts to
4353 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4356 Frequency : only for i686
4358 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4359 Details : Fixed config variable for build.
4364 Description: Updated patchess for new RHEL4 kernel
4365 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4366 Updated patch jbd-stats-2.6.9.patch
4367 Updated patch qsnet-rhel4-2.6.patch
4368 Updated patch quota-deadlock-on-pagelock-core.patch
4369 Updated patch vfs_intent-2.6-rhel4.patch
4370 Updated patch vfs_races-2.6-rhel4.patch
4371 Updated series file 2.6-rhel4-titech.series
4372 Updated series file 2.6-rhel4.series
4373 Updated kernel config files
4378 Description: lquota slave complains LBUG when reconnecting with mds
4380 Details : quota slave depends qctxt->lqc_import to send its quota request.
4381 This pointer will be invalid if mds did failover or broke its
4382 connect to osts, which leads to LBUG.
4385 Frequency : when qunit size is too small(less than 20M)
4387 Description: write is stopped by improper -EDQUOT
4388 Details : If the master is busy and qunit size is small enough(let's say 1M),
4389 the slave can not get quota from master on time, which will lead
4390 slave to trigger a -EQUOTA to client.
4395 Description: Deadlock during metadata tests
4396 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4397 called with the per-dentry lock held.
4400 Frequency : SLES9 only
4402 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4403 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4405 Severity : enhancement
4407 Description: remove fs_prep_san_write operation and related patches
4408 Details : remove the ext3-san-jdike patches which are no longer useful.
4413 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4414 Details : a disk corruption can cause the mballoc code to assert on a
4415 double free or other extent corruptions. Handle these with
4416 ext3_error() instead of assertions.
4419 Frequency : only with mballoc3 code and deep extent trees
4421 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4422 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4423 can cause the filesystem to be remounted read-only.
4428 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4429 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4430 few external users of this ioctl number is lfsck's call to
4431 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4432 Add the old ioctl number to the handler so both old and new
4437 Description: parallel lock callbacks
4438 Details : Instead of sending blocking and completion callbacks as separated
4439 requests, adding them to a set and sending in parallel.
4443 Description: Disable most debugging by default
4444 Details : To improve performance, disable most logging (for debug purposes)
4445 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4446 default, and HA includes fewer messages.
4450 Frequency : only run on patchless client.
4451 Description: use do_facet on sanity.sh for test handling recoverables errors
4452 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4456 Description: hash tables for lists of nids, connections and uuids
4457 Details : Hash tables noticeably help when a lot of clients connect to a
4458 server, to faster identify duplicate connections or reconnects,
4459 also to faster find export to evict in manual eviction case.
4463 Description: Sometimes, when the server evict a client, and the client will
4464 not be evicted as soon as possible.
4465 Details : In enqueue req, the error was returned by intent, instead of
4466 rq_status which make ptlrpc layer not detect this error, and
4467 does not evict the client. So enqueue error should be returned
4471 Frequency : only at startup
4473 Description: Delay client connections to MDT until fist MDT->OST connect
4474 Details : If a client tried to create a new file before the MDT had
4475 connected to any OSTs, the create would return EIO. Now
4476 the client will simply block until the MDT connects to the
4477 first OST and the create can succeed.
4480 Frequency : only for SLES9
4482 Description: Routinely utilize latest Quadrics drivers in CFS releases
4483 Details : Update patch qsnet-suse-2.6.patch.
4486 Frequency : only for sles10
4488 Description: Update patches for SLES 10 SP1 kernel.
4489 Details : Update the patch vfs_intent-2.6-sles10.patch.
4494 Description: Routinely utilize latest Quadrics drivers in CFS releases
4495 Details : Update patch qsnet-rhel4-2.6.patch.
4498 Frequency : at statup only
4500 Description: mds_lov_synchronize race leads to various problems
4501 Details : simultaneous MDT->OST connections at startup can cause the
4502 sync to abort, leaving the OSC in a bad state.
4507 Description: Using wrong pointer in osc_brw_prep_request
4508 Details : Access to array[-1] can produce panic if kernel compiled with
4509 CONFIG_PAGE_ALLOC enabled
4511 Severity : enhancement
4513 Description: Async OSC create to avoid the blocking unnecessarily.
4514 Details : If a OST has no remain object, system will block on the creating
4515 when need to create a new object on this OST. Now, ways use
4516 pre-created objects when available, instead of blocking on an
4517 empty osc while others are not empty. If we must block, we block
4518 for the shortest possible period of time.
4522 Frequency : only in recovery
4523 Description: Mark OST as early accessible if his start SYNC.
4524 Details : osc_precreate return flag early accessible if oscc marked as
4525 OSCC_FLAG_SYNC_IN_PROGRESS.
4530 Description: Sometimes precreate code can triger create object on wrong ost
4531 Details : Wrong protected or not not restored variables aftre precreate loop
4532 can produce creation object on wrong ost.
4535 Frequency : oss recovery
4537 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4539 Details : llog_commit_thread_count should be synced between llog_commit
4540 start and cleanup, so new llog_commit thread should not be started
4541 when llog_commit threads being stopped to avoid accessing some
4544 Severity : enhancement
4546 Description: Add printing inode info into message about error in writepage.
4548 --------------------------------------------------------------------------------
4550 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4555 Frequency : on some architectures
4557 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4558 Details : Uninitialized flags sometimes cause configuration commands to
4562 Frequency : patchless clients only
4564 Description: missing __iget() symbol export
4565 Details : The __iget() symbol export is missing. To avoid the need for
4566 this on patchless clients the deathrow inode reaper is turned
4567 off, and we depend on the VM to clean up old inodes. This
4568 dependency was during via the fix for bug 12181.
4573 Description: sanity.sh fail: test_52b
4574 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4575 incorrect inode's flags to client.
4577 --------------------------------------------------------------------------------
4579 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4581 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4582 INTEROPERATE with older versions automatically. In many cases a
4583 special upgrade step is needed. Please read the
4584 user documentation before upgrading any part of a 1.4.x system.
4585 * WARNING: Lustre configuration and startup changes are required with
4586 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4588 * Support for kernels:
4589 2.4.21-47.0.1.EL (RHEL 3),
4590 2.6.5-7.283 (SLES 9),
4591 2.6.9-42.0.10.EL (RHEL 4),
4592 2.6.12.6 vanilla (kernel.org),
4593 2.6.16.27-0.9 (SLES10)
4594 * Client support for unpatched kernels:
4595 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4596 2.6.16 - 2.6.19 vanilla (kernel.org),
4597 2.6.9-42.0.8EL (RHEL 4)
4598 * Recommended e2fsprogs version: 1.39.cfs6
4599 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4602 Severity : enhancement
4604 Description: Async OSC create to avoid the blocking unnecessarily.
4605 Details : If a OST has no remain object, system will block on the creating
4606 when need to create a new object on this OST. Now, ways use
4607 pre-created objects when available, instead of blocking on an
4608 empty osc while others are not empty. If we must block, we block
4609 for the shortest possible period of time.
4611 Severity : enhancement
4613 Description: MountConf
4614 Details : Lustre configuration is now managed via mkfs and mount
4615 commands instead of lmc and lconf. New obd types (MGS, MGC)
4616 are added for dynamic configuration management. See
4617 https://mail.clusterfs.com/wikis/lustre/MountConf for
4620 Severity : enhancement
4622 Description: dynamic OST addition
4623 Details : OSTs can now be added to a live filesystem
4625 Severity : enhancement
4627 Description: startup order invariance
4628 Details : MDTs and OSTs can be started in any order. Clients only
4629 require the MDT to complete startup.
4631 Severity : enhancement
4633 Description: parallel, asynchronous orphan cleanup
4634 Details : orphan cleanup is now performed in separate threads for each
4635 OST, allowing parallel non-blocking operation.
4637 Severity : enhancement
4639 Description: optimized stripe assignment
4640 Details : stripe assignments are now made based on ost space available,
4641 ost previous usage, and OSS previous usage, in order to try
4642 to optimize storage space and networking resources.
4644 Severity : enhancement
4646 Description: Permanently set tunables
4647 Details : All writable /proc/fs/lustre tunables can now be permanently
4648 set on a per-server basis, at mkfs time or on a live
4651 Severity : enhancement
4653 Description: Lustre message v2
4654 Details : Add lustre message format v2.
4656 Severity : enhancement
4658 Description: client OST exclusion list
4659 Details : Clients can be started with a list of OSTs that should be
4660 declared "inactive" for known non-responsive OSTs.
4664 Description: ENOENT returned for valid filehandle during dbench.
4665 Details : Check if a directory has children when invalidating dentries
4666 associated with an inode during lock cancellation. This fixes
4667 an incorrect ENOENT sometimes seen for valid filehandles during
4668 testing with dbench.
4671 Frequency : SFS test only (otherwise harmless)
4673 Description: SPEC SFS validation failure on NFS v2 over lustre.
4674 Details : Changes the blocksize for regular files to be 2x RPC size,
4675 and not depend on stripe size.
4677 Severity : enhancement
4679 Description: fine-grained SMP locking inside DLM
4680 Details : Improve DLM performance on SMP systems by removing the single
4681 per-namespace lock and replace it with per-resource locks.
4683 Severity : enhancement
4685 Description: don't hold multiple extent locks at one time
4686 Details : To avoid client eviction during large writes, locks are not
4687 held on multiple stripes at one time or for very large writes.
4688 Otherwise, clients can block waiting for a lock on a failed OST
4689 while holding locks on other OSTs and be evicted.
4691 Severity : enhancement
4693 Description: Multiple MD RPCs in flight.
4694 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4695 To avoid overly-overloading MDT, introduce a limit on number of
4696 MDT RPCs in flight for a single client and add /proc controls
4697 to adjust this limit.
4699 Severity : enhancement
4701 Description: client read/write statistics
4702 Details : Add client read/write call usage stats for performance
4703 analysis of user processes.
4704 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4705 file access. extents_stats shows chunk size distribution.
4706 extents_stats_per_process show chunk size distribution per
4709 Severity : enhancement
4711 Description: per-client statistics on server
4712 Details : Add ldlm and operations statistics for each client in
4713 /proc/fs/lustre/mds|obdfilter/*/exports/
4715 Severity : enhancement
4717 Description: improved MDT statistics
4718 Details : Add detailed MDT operations statistics in
4719 /proc/fs/lustre/mds/*/stats
4721 Severity : enhancement
4723 Description: VFS operations stats
4724 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4725 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4730 Description: Fix client-side osc byte counters
4731 Details : The osc read/write byte counters in
4732 /proc/fs/lustre/osc/*/stats are now working
4735 Frequency : always as root on SLES
4737 Description: Failure of copying files with lustre special EAs.
4738 Details : Client side always return success for setxattr call for lustre
4739 special xattr (currently only "trusted.lov").
4744 Description: Refcount LNET uuids
4745 Details : The global LNET uuid list grew linearly with every startup;
4746 refcount repeated list entries instead of always adding to
4749 Severity : enhancement
4751 Description: Dynamic service threads
4752 Details : Within a small range, start extra service threads
4753 automatically when the request queue builds up.
4756 Frequency : mixed-endian client/server environments
4758 Description: mixed-endian crashes
4759 Details : The new msg_v2 system had some failures in mixed-endian
4762 Severity : enhancement
4764 Description: Easy OST removal
4765 Details : OSTs can be permanently deactivated with e.g. 'lctl
4766 conf_param lustre-OST0001.osc.active=0'
4768 Severity : enhancement
4770 Description: MGS proc entries
4771 Details : Added basic proc entries for the MGS showing what filesystems
4774 Severity : enhancement
4776 Description: provide MGS failover
4777 Details : Added config lock reacquisition after MGS server failover.
4779 Severity : enhancement
4781 Description: add Linux 2.4 support
4782 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4786 Description: a large application tries to do I/O to the same resource and dies
4787 in the middle of it.
4788 Details : Check the req->rq_arrival time after the call to
4789 ost_brw_lock_get(), but before we do anything about
4790 processing it & sending the BULK transfer request. This
4791 should help move old stale pending locks off the queue as
4792 quickly as obd_timeout.
4795 Frequency : when an incorrect nid is specified during startup
4797 Description: ptlrpc connect to non-existant node causes kernel crash
4798 Details : LNET can't be re-entered from an event callback, which
4799 happened when we expire a message after the export has been
4800 cleaned up. Instead, hand the zombie cleanup off to another
4803 Severity : enhancement
4805 Description: plain/inodebits lock performance improvement
4806 Details : Grouping plain/inodebits in granted list by their request modes
4807 and bits policy, thus improving the performance of search through
4811 Frequency : only if OST filesystem is corrupted
4813 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4814 Details : for a short time RPCs with bulk IO are in the replay list,
4815 but replay of bulk IOs is unimplemented. If the OST filesystem
4816 is corrupted due to disk cache incoherency and then replay is
4817 started it is possible to trip an assertion. Avoid putting
4818 committed RPCs into the replay list at all to avoid this issue.
4821 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4823 Description: System hang on startup
4824 Details : This bug allowed the liblustre (e.g. catamount) client to
4825 return to the app before handling all startup RPCs. This
4826 could leave the node unresponsive to lustre network traffic
4827 and manifested as a server ptllnd timeout.
4829 Severity : enhancement
4831 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4832 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4833 boolean to control whether to print peer debug info when a
4834 client's RPC times out.
4837 Frequency : only for kernels with patches from Lustre below 1.4.3
4839 Description: Remove old rdonly API
4840 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4843 Frequency : only for devices with external journals
4845 Description: Set external device read-only also
4846 Details : During a commanded failover stop, we set the disk device
4847 read-only while the server shuts down. We now also set any
4848 external journal device read-only at the same time.
4851 Frequency : when upgrading from 1.4 while trying to change parameters
4853 Description: The wrong (new) MDC name was used when setting parameters for
4854 upgraded MDT's. Also allows changing of OSC (and MDC)
4855 parameters if --writeconf is specified at tunefs upgrade time.
4858 Frequency : when setting specific ost indicies
4860 Description: QOS code breaks on skipped indicies
4861 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4862 created with --index need not be sequential.
4864 Severity : enhancement
4866 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4867 Details : The uninit_groups feature works in conjunction with the kernel
4868 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4869 up the pass1 processing of e2fsck. This is a read-only feature
4870 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4871 mount filesystems that have had this feature enabled.
4873 Severity : enhancement
4875 Description: Improve multi-block allocation algorithm to avoid fragmentation
4876 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4877 allocation locality and avoid filesystem fragmentation.
4879 ------------------------------------------------------------------------------
4881 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4883 * Support for kernels:
4884 2.4.21-47.0.1.EL (RHEL 3)
4885 2.6.5-7.283 (SLES 9)
4886 2.6.9-42.0.10.EL (RHEL 4)
4887 2.6.12.6 vanilla (kernel.org)
4888 2.6.16.27-0.9 (SLES 10)
4889 * Recommended e2fsprogs version: 1.39.cfs5
4891 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4895 Frequency : occasional, depends on client load and configuration
4896 Bugzilla : 12181, 12203
4897 Description: data loss for recently-modified files
4899 Details : In some cases it is possible that recently written or created
4900 files may not be written to disk in a timely manner (this should
4901 normally be within 30s unless client IO load is very high).
4902 The problem appears as zero-length files or files that are a
4903 multiple of 1MB in size after a client crash or client eviction
4904 that are missing data at the end of the file.
4906 This problem is more likely to be hit on clients where files are
4907 repeatedly created and unlinked in the same directory, clients
4908 have a large amount of RAM, have many CPUs, the filesystem has
4909 many OSTs, the clients are rebooted frequently, and/or the files
4910 are not accessed by other nodes after being written.
4912 The presence of the problem can be detected by looking at
4913 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4914 smaller than the second (nr_unused) then dirty files will not
4915 be flushed automatically to disk. "sync; sleep 10" should be
4916 run several times on the node before unmounting it to update
4917 Lustre (this is also safe to run on nodes without this problem).
4919 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4920 that can cause this same problem, so customers using that kernel
4921 also need to update the kernel in addition to Lustre. In order
4922 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4924 It is normal that files written just before a client crash (less
4925 than 30s) may not yet have been flushed to disk, even for local
4929 Frequency : frequent on thin XT3 nodes
4931 Description: UUID collision on thin XT3 Linux nodes
4932 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4933 randomly, since we relied on an insufficiently-seeded PRNG.
4938 Description: OSS hangs after "All ost request buffers busy"
4939 Details : A deadlock between quota and journal operations caused OSS
4940 hangs after printing "All ost request buffers busy."
4943 Frequency : always on liblustre builds
4945 Description: Cleanup compiler warnings on liblustre
4948 Frequency : always on liblustre builds on XT3
4950 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4954 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4955 > 32000 subdirectories
4956 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4957 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4958 subdirectory count crosses 32000. This will aid e2fsck to
4959 correctly handle more than 32000 subdirectories.
4962 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4964 Description: System hang on startup
4965 Details : This bug allowed the liblustre (e.g. catamount) client to
4966 return to the app before handling all startup RPCs. This
4967 could leave the node unresponsive to lustre network traffic
4968 and manifested as a server ptllnd timeout.
4970 Severity : enhancement
4972 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4973 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4974 boolean to control whether to print peer debug info when a
4975 client's RPC times out.
4980 Description: make O_SYNC working on 2.6 kernels
4981 Details : 2.6 kernels use different method for mark pages for write,
4982 so need add a code to lustre for O_SYNC work.
4987 Description: Failure to close file and release space on NFS
4988 Details : Put inode details into lock acquired in ll_intent_file_open.
4989 Use mdc_intent_lock in ll_intent_open to properly
4990 detect all kind of errors unhandled by mdc_enqueue.
4995 Description: proc file read during shutdown sometimes raced obd removal,
4997 Details : Add lock to prevent obd access after proc file removal.
5000 Frequency : Only for files larger than 4GB on 32-bit clients.
5002 Description: improperly doing page alignment of locks
5003 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
5004 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
5009 Description: RPCs being resent when they shouldn't be
5010 Details : Some RPCs that should not be resent are being resent. This
5011 can cause inconsistencies in the RPC state machine. Do not
5012 resend such requests.
5015 Frequency : rare, only with NFS export
5017 Description: Crash on NFS re-export node
5018 Details : under very unusual load conditions an assertion is hit in
5019 ll_intent_file_open()
5022 Frequency : only if OST filesystem is corrupted
5024 Description: client incorrectly hits assertion in ptlrpc_replay_req()
5025 Details : for a short time RPCs with bulk IO are in the replay list,
5026 but replay of bulk IOs is unimplemented. If the OST filesystem
5027 is corrupted due to disk cache incoherency and then replay is
5028 started it is possible to trip an assertion. Avoid putting
5029 committed RPCs into the replay list at all to avoid this issue.
5034 Description: large O_DIRECT requests fail under memory pressure/fragmentation
5035 Details : Large single O_DIRECT read and write calls can fail to allocate
5036 a sufficiently large buffer to process the request. In case of
5037 allocation failure the allocation is retried with a smaller
5038 buffer and broken into smaller requests.
5040 Severity : enhancement
5042 Description: Add -o localflock option to simulate old noflock behaviour.
5043 Details : This will achieve local-only flock/fcntl locks coherentness.
5048 Description: versioning check is incomplete
5049 Details : Checking the version difference of client vs. server, report
5050 error if the gap is too big.
5055 Description: add support PG_writeback bit
5056 Details : add support for PG_writeback bit for Lustre, for more carefull
5057 work with page cache in 2.6 kernel. This also fix some deadlocks
5058 and remove hack for work O_SYNC with 2.6 kernel.
5060 Severity : enhancement
5062 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
5063 Details : The uninit_groups feature works in conjunction with the kernel
5064 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
5065 up the pass1 processing of e2fsck. This is a read-only feature
5066 in ldiskfs2 only, so older kernels and current ldiskfs cannot
5067 mount filesystems that have had this feature enabled.
5069 Severity : enhancement
5071 Description: Improve multi-block allocation algorithm to avoid fragmentation
5072 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
5073 allocation locality and avoid filesystem fragmentation.
5075 ------------------------------------------------------------------------------
5077 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
5079 * Support for kernels:
5080 2.6.9-42.0.3.EL (RHEL 4)
5081 2.6.5-7.276 (SLES 9)
5082 2.4.21-47.0.1.EL (RHEL 3)
5083 2.6.12.6 vanilla (kernel.org)
5084 2.6.16.21-0.8 (SLES10)
5085 * Recommended e2fsprogs version: 1.39.cfs2-0
5087 * The backwards-compatible /proc/sys/portals symlink has been removed
5088 in this release. Before upgrading, please ensure that you change
5089 any configuration scripts or /etc/sysctl.conf files that access
5090 /proc/sys/portals/* or sysctl portals.* to use the corresponding
5091 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
5092 in advance of the upgrade on any system running Lustre 1.4.6 or
5093 newer, since /proc/sys/lnet was added in that version.
5094 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
5098 Frequency : only when quota is used
5100 Description: avoid scanning export list for quota master
5101 Details : Change the algorithms to avoid scanning export list in order
5102 to improve the efficiency.
5105 Frequency : MDS failover only, very rarely
5107 Description: "went back in time" messages on mds failover
5108 Details : The greatest transno may be lost when the current operation
5109 finishes with an error (transno==0) and the client's last_rcvd
5110 record is over-written. Save the greatest transno in the
5111 mds_last_transno for this case.
5114 Frequency : always for specific kernels and striping counts
5116 Description: client may get "Matching packet too big" without ACL support
5117 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
5118 when trying to access files in certain configurations. The
5119 clients should in fact be denied when mounting because they do
5120 not understand ACLs.
5123 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
5125 Description: many clients connecting with IO in progress causes connect timeouts
5126 Details : Avoid synchronous journal commits to avoid delays caused by many
5127 clients connecting/disconnecting when bulk IO is in progress.
5128 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
5129 OST_IO_PORTAL to avoid delays behind potentially many pending
5133 Frequency : occasionally with multiple writers to a single file
5135 Description: shared writes to file may result in wrong size reported by stat()
5136 Details : Allow growing of kms when extent lock is cancelled
5139 Frequency : always with random mmap IO to multi-striped file
5141 Description: mmap write might be lost if we are writing to a 'hole' in stripe
5142 Details : Only if the hole is at the end of OST object so that kms is too
5143 small. Fix is to increase kms accordingly in ll_nopage.
5146 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
5148 Description: writes to a missing object would leak memory on the OST
5149 Details : If there is an inconsistency between the MDS and OST filesystems,
5150 such that the MDS references an object that doesn't exist, writes
5151 to that object will leak memory due to incorrect cleanup in the
5152 error handling path, eventually running out of memory on the OST.
5157 Description: Creating too long symlink causes lustre errors
5158 Details : Check symlink and name lengths before sending requests to MDS.
5161 Frequency : only if flock is enabled (not on by default)
5163 Description: posix locks not released on fd closure on 2.6.9+
5164 Details : We failed to add posix locks to list of inode locks on 2.6.9+
5165 kernels, this caused such locks not to be released on fd close and
5166 then assertions on fs unmount about still used locks.
5169 Frequency : MDS failover only, very rarely
5171 Description: clients may get ASSERTION(granted_lock != NULL)
5172 Details : When request was taking a long time, and a client was resending
5173 a getattr by name lock request. The were multiple lock requests
5174 with the same client lock handle and
5175 mds_getattr_name->fixup_handle_for_resent_request found one of the
5176 lock handles but later failed with ASSERTION(granted_lock != NULL).
5181 Description: handle->h_buffer_credits > 0, assertion failure
5182 Details : h_buffer_credits is zero after truncate, causing assertion
5183 failure. This patch extends the transaction or creates a new
5187 Frequency : NFS re-export or patchless client
5188 Bugzilla : 11179, 10796
5189 Description: Crash on NFS re-export node (__d_move)
5190 Details : We do not want to hash the dentry if we don't have a lock.
5191 But if this dentry is later used in d_move, we'd hit uninitialised
5192 list head d_hash, so we just do this to init d_hash field but
5193 leave dentry unhashed.
5196 Frequency : NFS re-export or patchless client
5198 Description: NFS exports has problem with symbolic link
5199 Details : lustre client didn't properly install dentry when re-exported
5200 to NFS or running patchless client.
5203 Frequency : NFS re-export or patchless client
5205 Description: Various nfs/patchless fixes.
5206 Details : fixes reuse disconected alias for lookup process - this fixes
5207 warning "find_exported_dentry: npd != pd",
5208 fix permission error with open files at nfs.
5209 fix apply umask when do revalidate.
5212 Frequency : occasional
5214 Description: Crash on NFS re-export node
5215 Details : calling clear_page() on the wrong pointer triggered oops in
5216 generic_mapping_read().
5219 Frequency : rarely, using O_DIRECT IO
5221 Description: unaligned directio crashes client with LASSERT
5222 Details : check for unaligned buffers before trying any requests.
5225 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
5227 Description: stale data returned from RAID cache
5228 Details : If only a small amount of IO is done to the RAID device before
5229 reading it again it is possible to get stale data from the RAID
5230 cache instead of reading it from disk.
5233 Frequency : always for sles10 kernel
5235 Description: sles10 support
5236 Details : ll_follow_link: compile fixes and using of nd_set_link
5237 under newer kernels.
5240 Frequency : depends on arch, kernel and compiler version, always on sles10
5243 Description: recursive or deep enough symlinks cause stack overflow
5244 Details : getting rid of large stack-allocated variable in
5248 Frequency : depends on hardware
5250 Description: lustre write performance loss in the SLES10 kernel
5251 Details : the performance loss is caused by using of write barriers in the
5252 ext3 code. The SLES10 kernel turns barrier support on by
5253 default. The fix is to undo that change for ldiskfs.
5255 ------------------------------------------------------------------------------
5257 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
5259 * Support for kernels:
5260 2.6.9-42.0.3EL (RHEL 4)
5261 2.6.5-7.276 (SLES 9)
5262 2.4.21-47.0.1.EL (RHEL 3)
5263 2.6.12.6 vanilla (kernel.org)
5267 Frequency : quota enabled and large files being deleted
5269 Description: releasing more than 4GB of quota at once hangs OST
5270 Details : If a user deletes more than 4GB of files on a single OST it
5271 will cause the OST to spin in an infinite loop. Release
5272 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5277 Description: statfs data retrieved from /proc may be stale or zero
5278 Details : When reading per-device statfs data from /proc, in the
5279 {kbytes,files}_{total,free,avail} files, it may appear
5280 as zero or be out of date.
5283 Frequency : systems with MD RAID1 external journal devices
5285 Description: lconf's call to blkid is confused by RAID1 journal devices
5286 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5287 of returning all block devices that match the journal UUID.
5290 Frequency : always, for aggregate stripe size over 4GB
5292 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5293 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5294 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5297 Frequency : always if "lfs find" used on a local file/directory
5299 Description: "lfs find" segfaults if used on a local file/directory
5300 Details : The case where a directory component was not specified wasn't
5301 handled correctly. Handle this properly.
5304 Frequency : always on ppc64
5306 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5307 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5308 perform endianness conversion.
5311 Frequency : rarely (truncate to non-zero file size after write under load)
5312 Bugzilla : 10730, 10687
5313 Description: Files padded with zeros to next 4K multiple
5314 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5315 it is possible that files that are truncated to a non-zero size
5316 immediately after being written are filled with zero bytes beyond
5317 the truncated size. No file data is lost.
5319 Severity : enhancement
5321 Description: Allow recovery/failover for liblustre clients.
5322 Details : liblustre clients were unaware of failover configurations until
5325 Severity : enhancement
5327 Description: user file locks should fail when not mounting with flock option
5328 Details : Set up an error-returning stub in ll_file_operations.lock field
5329 to prevent incorrect behaviour when client is mounted without
5330 flock option. Also, set up properly f_op->flock field for
5334 Frequency : always on ia64
5336 Description: "lfs df" loops on printing out MDS statfs information
5337 Details : The obd_ioctl_data was not initialized and in some systems
5338 this caused a failure during the ioctl that did not return
5339 an error. Initialize the struct and return an error on failure.
5342 Frequency : SLES 9 only
5344 Description: Error of copying files with lustre special EAs as root
5345 Details : Client side always return success for setxattr call for lustre
5346 special xattr (currently only "trusted.lov").
5349 Frequency : rarely on clusters with both ia64+i386 clients
5351 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5352 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5353 clients are doing concurrent writes to the same file it is
5354 possible that the ia64 clients may LASSERT because the OST
5355 extent locks are not PAGE_SIZE aligned. Ensure that grown
5356 locks are always aligned on the request boundary.
5359 Frequency : specific use, occasional
5361 Description: Overwriting in use executable truncates on-disk binary image
5362 Details : If one node attempts to overwrite an executable in use by
5363 another node, we now correctly return ETXTBSY instead of
5364 truncating the file.
5366 Severity : enhancement
5368 Description: Async OSC create to avoid the blocking unnecessarily.
5369 Details : If a OST has no remain object, system will block on the creating
5370 when need to create a new object on this OST. Now, ways use
5371 pre-created objects when available, instead of blocking on an
5372 empty osc while others are not empty. If we must block, we block
5373 for the shortest possible period of time.
5378 Description: chmod on Lustre root is propagated to other clients
5379 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5380 invalid by the follow_mount time.
5383 Frequency : liblustre clients only
5385 Description: Race in 'instant cancel' lock handling could lead to such locks
5386 never to be granted in case of SMP MDS
5387 Details : Do not destroy not yet granted but cbpending locks in
5391 Frequency : replay/resend of open
5393 Description: non null lock assetion failure in mds_intent_policy
5394 Details : Trying to replay/resend lockless open requests resulted in
5395 mds_open() returning 0 with no lock. Now it sets a flag if
5396 a lock is going to be returned.
5398 Severity : enhancement
5400 Description: Checksum enhancements
5401 Details : New checksum enhancements allow for resending RPCs that failed
5404 Severity : enhancement
5406 Description: Tunables on number of dirty pages in cacche
5407 Details : Allow to set limit on number of dirty pages cached.
5412 Description: client crash on unmount - lock still has references
5413 Details : In some error handling cases it was possible to leak a lock
5414 reference on a client while accessing a file. This was not
5415 harmful to the client during operation, but would cause the
5416 client to crash when the filesystem is unmounted.
5419 Frequency : specific case, rare
5421 Description: ETXTBSY on mds though file not in use by client
5422 Details : ETXTBSY is no longer incorrectly returned when attempting to
5423 chmod or chown a directory that the user previously tried to
5424 execute or a currently-executing binary.
5427 Frequency : extremely rare except on liblustre-based clients
5429 Description: Lustre space not freed when files are deleted
5430 Details : Clean up open-unlinked files after client eviction. Previously
5431 the unlink was skipped and the files remained as orphans.
5436 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5437 Details : In some cases it was possible to send a blocking callback to
5438 a client doing a glimpse, even though that client didn't get
5439 a lock granted. When the glimpse lock is cancelled on the OST
5440 the freed lock is left on the waiting list and corrupted the list.
5443 Frequency : all core dumps
5445 Description: Broke core dumps to lustre
5446 Details : Negative dentry may be unhashed if parent does not have UPDATE
5447 lock, but some callers, e.g. do_coredump, expect dentry to be
5448 hashed after successful create, hash it in ll_create_it.
5450 ------------------------------------------------------------------------------
5452 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5454 * Support for kernels:
5455 2.6.9-42.0.2.EL (RHEL 4)
5456 2.6.5-7.276 (SLES 9)
5457 2.4.21-40.EL (RHEL 3)
5458 2.6.12.6 vanilla (kernel.org)
5462 Frequency : always on RHEL 3
5464 Description: Number of open files grows over time
5465 Details : The number of open files grows over time, whether or not
5466 Lustre is started. This was due to a filp leak introduced
5467 by one of our kernel patches.
5469 ------------------------------------------------------------------------------
5471 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5473 * Support for kernels:
5474 2.6.9-42.EL (RHEL 4)
5475 2.6.5-7.267 (SLES 9)
5476 2.4.21-40.EL (RHEL 3)
5477 2.6.12.6 vanilla (kernel.org)
5482 Bugzilla : 5719, 9635, 9792, 9684
5483 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5484 Details : If a server is under heavy load and cannot reply to new
5485 connection requests before the client resends the (re)connect,
5486 the connection handling code can behave badly if two service
5487 threads are concurrently handing separate (re)connections from
5488 the same client. Add better locking to the connection handling
5489 code, and ensure that only a single connection will be processed
5490 for a given client UUID, even if the lock is dropped.
5492 Severity : enhancement
5494 Description: add TCP zero-copy support to kernel
5495 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5496 sends if the hardware supports scatter-gather and checksumming.
5497 This allows socklnd to do client-write and server-read more
5498 efficiently and reduce CPU utilization from skbuf copying.
5501 Frequency : only if NFS exporting from client
5503 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5504 Details : If setting attributes on a file created under NFS that had
5505 never been opened it would be possible to oops the client
5506 if the file had no objects.
5510 Bugzilla : 9326, 10402, 10897
5511 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5512 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5513 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5514 before ptlrpcd_wake() was called and the request was freed.
5517 Frequency : always for liblustre
5519 Description: liblustre client does MDS+OSTs setattr RPC for each write
5520 Details : When doing a write from a liblustre client, the client
5521 incorrectly issued an RPC to the MDS and each OST the file was
5522 striped over in order to update the timestamps. When writing
5523 with small chunks and many clients this could overwhelm the MDS
5524 with RPCs. In all cases it would slow down the write because
5525 these RPCs are unnecessary.
5527 Severity : enhancement
5529 Description: allow number of MDS service threads to be changed at module load
5530 Details : It is now possible to change the number of MDS service threads
5531 running. Adding "options mds mds_num_threads={N}" to the MDS's
5532 /etc/modprobe.conf will set the number of threads for the next
5533 time Lustre is restarted (assuming the "mds" module is also
5534 reloaded at that time). The default number of threads will
5535 stay the same, 32 for most systems.
5540 Description: OST crash if filesystem is unformatted or corrupt
5541 Details : If an OST is started on a device that has never been formatted
5542 or if the filesystem is corrupt and cannot even mount then the
5543 error handling cleanup routines would dereference a NULL pointer.
5548 Description: NULL pointer deref in llap_from_page.
5549 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5550 incorrect) mapping if the page was truncated/reclaimed while it was
5551 searched for. Check for this condition and skip such pages when
5552 doing readahead. Introduce extra check to llap_from_page() to
5553 verify page->mapping->host is non-NULL (so page is not anonymous).
5556 Frequency : Sometimes when using sys_sendfile
5558 Description: "page not covered by a lock" warnings from ll_readpage
5559 Details : sendfile called ll_readpage without right page locks present.
5560 Now we introduced ll_file_sendfile that does necessary locking
5561 around call to generic_file_sendfile() much like we do in
5565 Frequency : with certain MDS communication failures at client mount time
5567 Description: NULL pointer deref after failed client mount
5568 Details : a client connection request may delayed by the network layer
5569 and not be sent until after the PTLRPC layer has timed out the
5570 request. If the client fails the mount immediately it will try
5571 to clean up before the network times out the request. Add a
5572 reference from the request import to the obd device and delay
5573 the cleanup until the network drops the request.
5576 Frequency : occasionally during client (re)connect
5578 Description: assertion failure during client (re)connect
5579 Details : processing a client connection request may be delayed by the
5580 client or server longer than the client connect timeout. This
5581 causes the client to resend the connection request. If the
5582 original connection request is replied in this interval, the
5583 client may trip an assertion failure in ptlrpc_connect_interpret()
5584 which thought it would be the only running connect process.
5587 Frequency : only with obd_echo servers and clients that are rebooted
5589 Description: kernel BUG accessing uninitialized data structure
5590 Details : When running an obd_echo server it did not start the ping_evictor
5591 thread, and when a client was evicted an uninitialized data
5592 structure was accessed. Start the ping_evictor in the RPC
5593 service startup instead of the OBD startup.
5595 Severity : enhancement
5596 Bugzilla : 10193 (patchless)
5597 Description: Remove dependency on various unexported kernel interfaces.
5598 Details : No longer need reparent_to_init, exit_mm, exit_files,
5599 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5602 Frequency : rare (only users of deprecated and unsupported LDAP config)
5604 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5605 Details : LDAP apparently contains 'lustreName' attributes instead of
5606 'name'. A simple remapping of the name is sufficient.
5609 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5611 Description: waiting_locks_callback trips kernel BUG if client is evicted
5612 Details : Running with the dump_on_timeout debug flag turned on makes
5613 it possible that the waiting_locks_callback() can try to dump
5614 the Lustre kernel debug logs from an interrupt handler. Defer
5615 this log dumping to the expired_lock_main() thread.
5617 Severity : enhancement
5619 Description: Support NFS exporting on 2.6 kernels.
5620 Details : Implement non-rawops metadata methods for NFS server to use without
5621 changing NFS server code.
5624 Frequency : very rare (synthetic metadata workload only)
5626 Description: two racing renames might cause an MDS thread to deadlock
5627 Details : Running the "racer" program may cause one MDS thread to rename
5628 a file from being the source of a rename to being the target of
5629 a rename at exactly the same time that another thread is doing
5630 so, and the second thread has already enqueued these locks after
5631 doing a lookup of the target and is trying to relock them in
5632 order. Ensure that we don't try to re-lock the same resource.
5635 Frequency : only very large systems with liblustre clients
5637 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5638 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5639 all OSTs in parallel. This allows the eviction of stale liblustre
5640 clients to proceed much faster than if they were done in series,
5641 and also offers similar improvements for other set_info RPCs.
5646 Description: excessive CPU usage during initial read phase on client
5647 Details : During the initial read phase on a client, it would agressively
5648 retry readahead on the file, consuming too much CPU and impacting
5649 performance (since 1.4.5.8). Improve the readahead algorithm
5650 to avoid this, and also improve some other common cases (read
5651 of small files in particular, where "small" is files smaller than
5652 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5657 Description: MDS crash when receiving packet with unknown intent.
5658 Details : Do not LBUG in unknown intent case, just return -EFAULT
5660 Severity : enhancement
5661 Bugzilla : 9293, 9385
5662 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5663 Details : Do not serialize getattr (non-intent version) and statfs.
5666 Frequency : occasional, when OST network is overloaded/intermittent
5668 Description: client evicted by OST after bulk IO timeout
5669 Details : If a client sends a bulk IO request (read or write) the OST
5670 may evict the client if it is unresposive to its data GET/PUT
5671 request. This is incorrect if the network is overloaded (takes
5672 too long to transfer the RPC data) or dropped the OST GET/PUT
5673 request. There is no need to evict the client at all, since
5674 the pinger and/or lock callbacks will handle this, and the
5675 client can restart the bulk request.
5678 Frequency : Always when mmapping file with no objects
5680 Description: client crashes when mmapping file with no objects
5681 Details : Check that we actually have objects in a file before doing any
5682 operations on objects in ll_vm_open, ll_vm_close and
5688 Description: Request leak when working with deleted CWD
5689 Details : Introduce advanced request refcount tracking for requests
5690 referenced from lustre intent.
5692 Severity : Enhancement
5694 Description: Cache open file handles on client.
5695 Details : MDS now will return special lock along with openhandle, if
5696 requested and client is allowed to hold openhandle, even if unused,
5697 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5698 closing files for every read/write openration.
5700 Severity : Enhancement
5702 Description: Cache open negative dentries on client when possible.
5703 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5704 negative dentries on lock revocation.
5709 Description: Remounting a client read-only wasn't possible with a zconf mount
5710 Details : It wasn't possible to remount a client read-only with llmount.
5712 Severity : enhancement
5713 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5714 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5715 MPICH is included to add Lustre-specific ADIO interfaces.
5716 This is based closely on the UFS ADIO layer and only differs
5717 in file creation, in order to allow the OST striping to be set.
5718 This is user-contributed code and not supported by CFS.
5723 Description: extended inode attributes (immutable, append-only) work improperly
5724 when 2.4 and 2.6 kernels are used on client/server or vice versa
5725 Details : Introduce kernel-independent values for these flags.
5727 Severity : enhancement
5730 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5731 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5732 granularity. Now, Fractional MB granularity is be supported,
5733 this is very useful for low memory system.
5735 Severity : enhancement
5737 Description: Getattr by fid
5738 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5739 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5740 have locks and avoids getattr rpc requests for it on every stat.
5743 Frequency : Always, for filesystems larger than 2TB
5745 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5746 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5747 filesystems larger than 2TB to be mounted with mballoc enabled.
5750 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5752 Description: filesystem corruption for non-standard kernels and very large OSTs
5753 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5754 filesystems larger than 2TB is mounted then the kernel will
5755 silently corrupt the start of the filesystem. CONFIG_LBD is
5756 enabled for all CFS-supported kernels, but the possibility of
5757 this happening with a modified kernel config exists.
5759 Severity : enhancement
5761 Description: add client O_DIRECT support for 2.6 kernels
5762 Details : It is now possible to do O_DIRECT reads and writes to files
5763 in the Lustre client mountpoint on 2.6 kernel clients.
5765 Severity : enhancement
5767 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5768 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5769 parallel, not waiting for response from every OST before sending
5770 a rpc to the next OST.
5775 Description: setattr vs write race when updating file timestamps
5776 Details : Client processes that update a file timestamp into the past
5777 right after writing to the file (e.g. tar) it is possible that
5778 the updated file modification time can be reset to the current
5779 time due to a race between processing the setattr and write RPC.
5781 Severity : enhancement
5783 Description: Bring 'lfs find' closer in line with regular Linux find.
5784 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5785 -print0 options and obtains all the needed info through the lustre
5788 Severity : enhancement
5790 Description: support up to 1024 configured devices on one node
5791 Details : change obd_dev array from statically allocated to dynamically
5792 allocated structs as they are first used to reduce memory usage
5797 Description: Flush dirty partially truncated pages during truncate
5798 Details : Immediatelly flush partially truncated pages in filter_setattr,
5799 this way we completely avoid having any pages in page cache on OST
5800 and can retire ugly workarounds during writes to flush such pages.
5805 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5806 Details : protect inode from truncation within vfs_unlink() context
5807 just take a reference before calling vfs_unlink() and release it
5808 when parent's i_sem is free.
5811 Frequency : always, if extents are used on OSTs
5813 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5814 Details : OSTs using the extents format would not zero the high 16 bits of
5815 the index physical block number. This is not a problem for any
5816 OST filesystems smaller than 16TB, and no kernels support ext3
5817 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5818 new/modified files) and can be fixed for existing filesystems
5819 with e2fsprogs-1.39-cfs1.
5824 Description: import connection selection may be incorrect if timer wraps
5825 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5826 import connections to be ignored if the 32-bit jiffies counter
5827 wraps. Use a 64-bit jiffies counter.
5830 Frequency : during server recovery
5832 Description: crash after server is denying duplicate export
5833 Details : If clients are resending connect requests to the server, the
5834 server refuses to allow a client to connect multiple times.
5835 Fixed a bug in the handling of this case.
5838 Frequency : very large clusters immediately after boot
5840 Description: LNET request buffers exhausted under heavy short-term load
5841 Details : If a large number of client requests are generated on a service
5842 that has previously never seen so many requests it is possible
5843 that the request buffer growth cannot keep up with the spike in
5844 demand. Instead of dropping incoming requests, they are held in
5845 the LND until the RPC service can accept more requests.
5848 Frequency : Sometimes during replay
5850 Description: Assertion failure in ll_local_open after replay.
5851 Details : If replay happened on an open request reply before we were able
5852 to set replay handler, reply will become not swabbed tripping the
5853 assertion in ll_local_open. Now we set the handler right after
5854 recognising of open request
5857 Frequency : very rare
5859 Description: kernel reports "badness in vsnprintf"
5860 Details : Reading from the "recovery_status" /proc file in small chunks
5861 may cause a negative length in lprocfs_obd_rd_recovery_status()
5862 call to vsnprintf() (which is otherwise harmless). Exit early
5863 if there is no more space in the output buffer.
5865 Severity : enhancement
5867 Description: clear OBD RPC statistics by writing to them
5868 Details : It is now possible to clear the OBD RPC statistics by writing
5869 to the "stats" file.
5874 Description: Client mtime is not the same on different clients after utimes
5875 Details : In some cases, the client was using the utimes() syscall on
5876 a file cached on another node. The clients now validate the
5877 ctime from the MDS + OSTs to determine which one is right.
5882 Description: Inability to activate failout mode
5883 Details : lconf script incorrectly assumed that in python string's numeric
5884 value is used in comparisons.
5887 Frequency : always with multiple stripes per file
5889 Description: Inefficient object allocation for mutli-stripe files
5890 Details : When selecting which OSTs to stripe files over, for files with
5891 a stripe count that divides evenly into the number of OSTs,
5892 the MDS is always picking the same starting OST for each file.
5893 Return the OST selection heuristic to the original design.
5898 Description: mount failures may take full timeout to return an error
5899 Details : Under some heavy load conditions it is possible that a
5900 failed mount can wait for the full obd_timeout interval,
5901 possibly several minutes, before reporting an error.
5902 Instead return an error as soon as the status is known.
5904 ------------------------------------------------------------------------------
5906 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5908 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5909 INTEROPERATE with older versions automatically. Please read the
5910 user documentation before upgrading any part of a live system.
5911 * WARNING: Lustre networking configuration changes are required with
5912 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5915 * Support for kernels:
5916 2.6.9-22.0.2.EL (RHEL 4)
5917 2.6.5-7.244 (SLES 9)
5918 2.6.12.6 vanilla (kernel.org)
5921 Severity : enhancement
5922 Bugzilla : 7981/8208
5923 Description: Introduced Lustre Networking (LNET)
5924 Details : LNET is new networking infrastructure for Lustre, it includes
5925 a reorganized network configuration mode (see the user
5926 documentation for full details) as well as support for routing
5927 between different network fabrics. Lustre Networking Devices
5928 (LNDs) for the supported network fabrics have also been
5929 created for this new infrastructure.
5931 Severity : enhancement
5932 Description: Introduced Access control lists
5933 Details : clients can set ACLs on files and directories in order to have
5934 more fine-grained permissions than the standard Unix UGO+RWX.
5935 The MDS must be started with the "-o acl" mount option.
5937 Severity : enhancement
5938 Description: Introduced filesystem quotas
5939 Details : Administrators may now establish per-user quotas on the
5942 Severity : enhancement
5944 Description: Configuration change for the XT3
5945 The PTLLND is now used to run Lustre over Portals on the XT3
5946 The configure option(s) --with-cray-portals are no longer used.
5947 Rather --with-portals=<path-to-portals-includes> is used to
5948 enable building on the XT3. In addition to enable XT3 specific
5949 features the option --enable-cray-xt3 must be used.
5954 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5955 Details : A race between allocating a new llcd and re-getting the llcd_lock
5956 allowed another thread to grab newly-allocated llcd.
5958 Severity : enhancement
5960 Description: 2.6 OST async journal commit and locking fix to improve performance
5961 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5962 2.6 kernels are now async as they already were in 2.4 kernels so
5963 that they can commit concurrently with the network bulk transfer.
5964 For block-allocated files the filter allocation semaphore is held
5965 to avoid filesystem fragmentation during allocation. BKL lock
5966 removed for 2.6 xattr operations where it is no longer needed.
5971 Description: lconf incorrectly determined whether two IP networks could talk
5972 Details : In some more complicated routing and multiple-network
5973 configurations, lconf will avoid trying to make a network
5974 connection to a disjoint part of the IP space. It was doing the
5975 math incorrectly for one set of cases.
5980 Description: Fix for potential infinite loop processing records in an llog.
5981 Details : If an llog record is corrupted/zeroed, it is possible to loop
5982 forever in llog_process(). Validate the llog record length
5983 and skip the remainder of the block on error.
5986 Frequency : occasional (liblustre only)
5988 Description: liblustre could not open files whose last component is a symlink
5989 Details : sysio_path_walk() would incorrectly pass the open intent to
5990 intermediate path components.
5993 Frequency : rare (liblustre only with non-standard tuning)
5994 Bugzilla : 7201 (7350)
5995 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5996 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5997 after it has been released, only do this for referenced locks
5999 Severity : enhancement
6001 Description: specifying an (invalid) directory default stripe_size of -1
6002 would reset the directory default striping
6003 Details : stripe_size -1 was used internally to signal directory stripe
6004 removal, now use "all default" to signal dir stripe removal
6005 as a directory striping of "all default" is not useful
6008 Frequency : common for large clusters running liblustre clients
6010 Description: doing an ls when liblustre clients are running is slow
6011 Details : sending a glimpse AST to a liblustre client waits for every AST
6012 to time out, as liblustre clients will not respond. Since they
6013 cannot cache data we refresh the OST lock LVB from disk instead.
6015 Severity : enhancement
6017 Description: doing an ls at the same time as file IO can be slow
6018 Details : enqueue and other "small" requests can be blocked behind many
6019 large IO requests. Create a new OST IO portal for non-IO
6020 requests so they can be processed faster.
6023 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
6025 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
6026 created file with mode 000
6027 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
6028 allow an ftruncate() as a non-root user to a file with mode 000.
6029 Lustre now respects this flag to disable mode checks when
6030 truncating a file owned by the user
6033 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
6036 Description: Revoking locks from clients that went dead or catatonic might take
6038 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
6039 cancellation of such locks instant on servers without waiting for
6040 any reply from clients. Clients drops these locks when cancel
6041 notification from server is received without replying.
6044 Frequency : liblustre-only, when liblustre client dies or becomes busy
6046 Description: Doing ls on Linux clients can take a long time with active
6048 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
6049 granting such locks to it in the first place if possible. Locks
6050 are taken by proxy on the OST during the read or write and
6051 dropped immediately afterward. Add connect flags handling, do
6052 not grant locks to liblustre clients for glimpse ASTs.
6054 Severity : enhancement
6056 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
6057 Details : Existing read-ahead algorithm is tuned for the case of streamlined
6058 sequential reads and behaves badly with applications doing random
6059 reads. Improve it by reading ahead at least read region, and
6060 avoiding excessive large RPC for small reads.
6062 Severity : enhancement
6064 Description: Creating more than 1000 files for a single job may cause a load
6065 imbalance on the OSTs if there are also a large number of OSTs.
6066 Details : qos_prep_create() uses an OST index reseed value that is an
6067 even multiple of the number of available OSTs so that if the
6068 reseed happens in the middle of the object allocation it will
6069 still utilize the OSTs as uniformly as possible.
6074 Description: OST or MDS may oops in ping_evictor_main()
6075 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
6076 but doesn't restart at beginning of obd_exports_timed list
6079 Severity : enhancement
6081 Description: improve by-nid export eviction on the MDS and OST
6082 Details : allow multiple exports with the same NID to be evicted at one
6083 time without re-searching the exports list.
6086 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
6088 Description: MDS may oops in groups_free()
6089 Details : in rare race conditions a newly allocated group_info struct is
6090 freed again, and this can be NULL. The 2.4 compatibility code
6091 for groups_free() checked for a NULL pointer, but 2.6 did not.
6094 Frequency : common for liblustre clients doing little filesystem IO
6095 Bugzilla : 9352, 7313
6096 Description: server may evict liblustre clients accessing contended locks
6097 Details : if a client is granted a lock or receives a completion AST
6098 with a blocking AST already set it would not reply to the AST
6099 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
6102 Frequency : lfs setstripe, only systems with more than 160 OSTs
6104 Description: unable to set striping with a starting offset beyond OST 160
6105 Details : llapi_create_file() incorrectly limited the starting stripe
6106 index to the maximum single-file stripe count.
6109 Frequency : LDAP users only
6111 Description: lconf did not handle in-kernel recovery with LDAP properly
6112 Details : lconf/LustreDB get_refs() is searching the wrong namespace
6114 Severity : enhancement
6116 Description: bind OST threads to NUMA nodes to improve performance
6117 Details : all OST threads are uniformly bound to CPUs on a single NUMA
6118 node and do their allocations there to localize memory access
6120 Severity : enhancement
6122 Description: llmount can determine client NID directly from Myrinet (GM)
6123 Details : the client NID code from gmnalnid was moved directly into
6124 llmount, removing the need to use this or specifying the
6125 client NID explicitly when mounting GM clients with zeroconf
6128 Frequency : if client is started with down MDS
6130 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
6131 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
6132 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
6133 ptlrpc_import_delayed_req() because we didn't check if the
6134 request was interrupted, and we also didn't break out of the
6135 event loop if there was no timeout
6140 Description: data loss during non-page-aligned writes to a single file from
6141 both multiple nodes and multiple threads on one node at same time
6142 Details : updates to KMS and lsm weren't protected by common lock. Resulting
6143 inconsistency led to false short-reads, that were cached and later
6144 used by ->prepare_write() to fill in partially written page,
6145 leading to data loss.
6148 Frequency : always, if lconf --abort_recovery used
6150 Description: lconf --abort_recovery fails with 'Operation not supported'
6151 Details : lconf was attempting to abort recovery on the MDT device and not
6154 Severity : enhancement
6156 Description: remove cleanup logs
6157 Details : replace lconf-generated cleanup logs with lustre internal
6158 cleanup routines. Eliminates the need for client-cleanup and
6161 Severity : enhancement
6163 Description: add support for EAs (user and system) on lustre filesystems
6164 Details : it is now possible to store extended attributes in the Lustre
6165 client filesystem, and with the user_xattr mount option it
6166 is possible to allow users to store EAs on their files also
6168 Severity : enhancement
6170 Description: Add possibility (config option) to show minimal available OST free
6172 Details : When compiled with --enable-mindf configure option, statfs(2)
6173 (and so, df) will return least minimal free space available from
6174 all OSTs as amount of free space on FS, instead of summary of
6175 free spaces of all OSTs.
6177 Severity : enhancement
6179 Description: do not expand extent locks acquired on OST-side
6180 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
6181 by server: they are not cached anyway.
6184 Frequency : when mmap is used/binaries executed from Lustre
6186 Description: Unmmap pages before throwing them away from read cache.
6187 Details : llap_shrink cache now attempts to unmap pages before discarding
6188 them (if unmapping failed - do not discard). SLES9 kernel has
6189 extra checks that trigger if this unmapping is not done first.
6194 Description: lconf didn't resolve symlinks before checking to see whether a
6195 given mountpoint was already in use
6198 Frequency : when migrating failover services
6199 Bugzilla : 6395, 9514
6200 Description: When migrating a subset of services from a node (e.g. failback
6201 from a failover service node) the remaining services would
6202 time out and evict clients.
6203 Details : lconf --force (implied by --failover) sets the global obd_timeout
6204 to 5 seconds in order to quickly disconnect, but this caused
6205 other RPCs to time out too quickly. Do not change the global
6206 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
6208 Severity : enhancement
6209 Frequency : if MDS is started with down OST
6210 Bugzilla : 9439,5706
6211 Description: Allow startup/shutdown of an MDS without depending on the
6212 availability of the OSTs.
6213 Details : Asynchronously call mds_lov_synchronize during MDS startup.
6214 Add appropriate locking and lov-osc refcounts for safe
6215 cleaning. Add osc abort_inflight calls in case the
6216 synchronize never started.
6219 Frequency : occasional (Cray XT3 only)
6221 Description: root not authorized to access files in CRAY_PORTALS environment
6222 Details : The client process capabilities were not honoured on the MDS in
6223 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
6224 been accessed by an authorized user then root was able to access
6225 the file on the local client also. The root user capabilities
6226 are now allowed on the MDS, as this environment has secure UID.
6229 Frequency : occasional
6231 Description: ldiskfs "too long searching" message happens too often
6232 Details : A debugging message (otherwise harmless) prints too often on
6233 the OST console. This has been reduced to only happen when
6234 there are fragmentation problems on the filesystem.
6239 Description: Division by zero in statfs when all OSCs are inactive
6240 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
6241 lov_statfs divides by value returned by lov_get_stripecnt().
6245 Bugzilla : 9489, 3273
6246 Description: First write from each client to each OST was only 4kB in size,
6247 to initialize client writeback cache, which caused sub-optimal
6248 RPCs and poor layout on disk for the first writen file.
6249 Details : Clients now request an initial cache grant at (re)connect time
6250 and so that they can start streaming writes to the cache right
6251 away and always do full-sized RPCs if there is enough data.
6252 If the OST is rebooted the client also re-establishes its grant
6253 so that client cached writes will be honoured under the grant.
6258 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
6259 Details : Now I/O RPCs go to different portal number and (presumably) fast
6260 lock requests (and glimses) and other RPCs get their own service
6261 threads pool that should be able to service those RPCs
6264 Severity : enhancement
6266 Description: Ability to exchange lustre version between client and servers and
6267 issue warnings at client side if client is too old. Also for
6268 liblustre clients there is ability to refuse connection of too old
6270 Details : New 'version' field is added to connect data structure that is
6271 filled with version info. That info is later checked by server and
6275 Frequency : rare, liblustre only.
6276 Bugzilla : 9296, 9581
6277 Description: Two simultaneous writes from liblustre at offset within same page
6278 might proceed at the same time overwriting eachother with stale
6280 Details : I/O lock withing llu_file_prwv was released too early, before data
6281 actually was hitting the wire. Extended lock-holding time until
6282 server acknowledges receiving data.
6285 Frequency : extremely rare. Never observed in practice.
6287 Description: avoid generating lustre_handle cookie of 0.
6288 Details : class_handle_hash() generates handle cookies by incrementing
6289 global counter, and can hit 0 occasionaly (this is unlikely, but
6290 not impossible, because initial value of cookie counter is
6291 selected randonly). Value of 0 is used as a sentinel meaning
6292 "unassigned handle" --- avoid it. Also coalesce two critical
6293 sections in this function into one.
6295 Severity : enhancement
6297 Description: allow liblustre clients to delegate truncate locking to OST
6298 Details : To avoid overhead of locking, liblustre client instructs OST to
6299 take extent lock in ost_punch() on client's behalf. New connection
6300 flag is added to handle backward compatibility.
6302 Severity : enhancement
6303 Bugzilla : 4928, 7341, 9758
6304 Description: allow number of OST service threads to be specified
6305 Details : a module parameter allows the number of OST service threads
6306 to be specified via "options ost ost_num_threads={N}" in the
6307 OSS's /etc/modules.conf or /etc/modprobe.conf.
6311 Bugzilla : 6146, 9635, 9895
6312 Description: servers crash with bad pointer in target_handle_connect()
6313 Details : In rare cases when a client is reconnecting it was possible that
6314 the connection request was the last reference for that export.
6315 We would temporarily drop the export reference and get a new
6316 one, but this may have been the last reference and the export
6317 was just destroyed. Get new reference before dropping old one.
6319 Severity : enhancement
6320 Frequency : if client is started with failover MDS
6322 Description: Allow multiple MDS hostnames in the mount command
6323 Details : Try to read the configuration from all specified MDS
6324 hostnames during a client mount in case the "primary"
6327 Severity : enhancement
6329 Description: Stop sending data to evicted clients as soon as possible.
6330 Details : Check if the client we are about to send or are sending data to
6331 was evicted already. (Check is done every second of waiting,
6332 for which l_wait_event interface was extended to allow checking
6333 of exit condition at specified intervals).
6336 Frequency : rare, normally only when NFS exporting is done from client
6338 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6340 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6343 Frequency : always, if srand() is called before liblustre initialization
6345 Description: Liblustre uses system PRNG disturbing its usage by user application
6346 Details : Introduce internal to lustre fast and high-quality PRNG for
6347 lustre usage and make liblustre and some other places in generic
6348 lustre code to use it.
6350 Severity : enhancement
6351 Bugzilla : 9477, 9557, 9870
6352 Description: Verify that the MDS configuration logs are updated when xml is
6353 Details : Check if the .xml configuration logs are newer than the config
6354 logs stored on the MDS and report an error if this is the case.
6355 Request --write-conf, or allow starting with --old_conf.
6357 Severity : enhancement
6359 Description: Handle symlinks in the path when checking if Lustre is mounted.
6360 Details : Resolve intermediate symlinks when checking if a client has
6361 mounted a filesystem to avoid duplicate client mounts.
6366 Description: lconf can hit an error exception but still return success.
6367 Details : The lconf command catches the Command error exception at the top
6368 level script context and will exit with the associated exit
6369 status, but doesn't ensure that this exit status is non-zero.
6374 Description: failure of ptlrpc thread startup can cause oops
6375 Details : Starting a ptlrpc service thread can fail if there are a large
6376 number of threads or the server memory is very fragmented.
6377 Handle this without oopsing.
6380 Frequency : always, only if liblustre and non-default acceptor port was used
6382 Description: liblustre cannot connect to servers with non-default acceptor port
6383 Details : tcpnal_set_default_params() was not called and was therefore
6384 ignoring the environment varaible TCPNAL_PORT, as well as other
6385 TCPNAL_ environment variables
6390 Description: two objects could be created on the same OST for a single file
6391 Details : If an OST is down, in some cases it was possible to create two
6392 objects on a single OST for a single file. No problems other
6393 than potential performance impact and spurious error messages.
6397 Bugzilla : 5681, 9562
6398 Description: Client may oops in ll_unhash_aliases
6399 Details : Client dcache may become inconsistent in race condition.
6400 In some cases "getcwd" can fail if the current directory is
6406 Description: Inode refcounting problems in NFS export code
6407 Details : link_raw functions used to call d_instantiate without obtaining
6408 extra inode reference first.
6412 Bugzilla : 9942, 9903
6413 Description: Referencing freed requests leading to crash, memleaks with NFS.
6414 Details : We used to require that call to ll_revalidate_it was always
6415 followed by ll_lookup_it. Also with revalidate_special() it is
6416 possible to call ll_revalidate_it() twice for the same dentry
6417 even if first occurence returned success. This fix changes semantic
6418 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6419 reference on a request referred from the intent.
6420 ll_intent_release() then releases such a request.
6423 Frequency : rare, normally benchmark loads only
6425 Description: unlinked inodes were kept in memory on the client
6426 Details : If a client is repeatedly creating and unlinking files it
6427 can accumulate a lot of stale inodes in the inode slab cache.
6428 If there is no other client load running this can cause the
6429 client node to run out of memory. Instead flush old inodes
6430 from client cache that have the same inode number as a new inode.
6433 Frequency : SLES9 2.6.5 kernel and long filenames only
6434 Bugzilla : 9969, 10379
6435 Description: utime reports stale NFS file handle
6436 Details : SLES9 uses out-of-dentry names in some cases, which confused
6437 the lustre dentry revalidation. Change it to always use the
6441 Frequency : rare, unless heavy write-truncate concurrency is continuous
6442 Bugzilla : 4180, 6984, 7171, 9963, 9331
6443 Description: OST becomes very slow and/or deadlocked during object unlink
6444 Details : filter_destroy() was holding onto the parent directory lock
6445 while truncating+unlinking objects. For very large objects this
6446 may block other threads for a long time and slow overall OST
6447 responsiveness. It may also be possible to get a lock ordering
6448 deadlock in this case, or run out of journal credits because of
6449 the combined truncate+unlink. Solution is to do object truncate
6450 first in one transaction without parent lock, and then do the
6451 final unlink in a new transaction with the parent lock. This
6452 reduces the lock hold time dramatically.
6455 Frequency : rare, 2.4 kernels only
6457 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6458 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6459 before unmounting their filesystem in order to avoid deadlock.
6460 The kernel_locked() function in 2.4 kernels only checks whether
6461 the kernel lock is held, not whether it is this process that is
6462 holding it as 2.6 kernels do.
6467 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6468 Details : The client ptlrpc code may be trying to reconnect to a down
6469 server before a previous connection attempt has timed out.
6470 Increase the reconnect interval to be longer than the connection
6471 timeout interval to avoid sending duplicate connections to
6475 Frequency : echo_client brw_test command
6477 Description: fix echo_client to work with OST preallocated code
6478 Details : OST preallocation code (5137) didn't take echo_client IO path
6479 into account: echo_client calls filter methods outside of any
6480 OST thread and, hence, there is no per-thread preallocated
6481 pages and buffers to use. Solution: hijack pga pages for IO. As
6482 a byproduct, this avoids unnecessary data copying.
6486 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6487 Description: Client can oops in mdc_commit_close() after open replay
6488 Details : It was possible for the MDS to return an open request with no
6489 transaction number in mds_finish_transno() if the client was
6490 evicted, but without actually returning an error. Clients
6491 would later try to replay that open and may trip an assertion
6492 Simplify the client close codepath, and always return an error
6493 from the MDS in case the open is not successful.
6496 Frequency : rare, 2.6 OSTs only
6498 Description: OST may deadlock under high load on fragmented files
6499 Details : If there was a heavy load and highly-fragmented OST filesystems
6500 it was possible to have all the OST threads deadlock waiting on
6501 allocation of biovecs, because the biovecs were not released
6502 until the entire RPC IO was completed. Instead, release biovecs
6503 as soon as they are complete to ensure forward IO progress.
6505 Severity : enhancement
6507 Description: Support for specifying external journal device at mount
6508 Details : If an OST or MDS device is formatted with an external journal
6509 device, this device major/minor is stored in the ext3 superblock
6510 and may not be valid for failover. Allow detecting and
6511 specifying the external journal at mount time.
6516 Description: Mounting an MDS with pending unlinked files may cause oops
6517 Details : target_finish_recovery() calls mds_postrecov() which returned
6518 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6519 considers this an error and immediately begins cleaning up the
6520 lov, just after starting the mds_lov process
6522 Severity : enhancement
6524 Description: Implement 'lfs df' to report actual free space on per-OST basis
6525 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6526 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6527 usage of objects; '-h' to report in human readable format.
6529 ------------------------------------------------------------------------------
6531 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6538 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6539 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6540 at startup during orphan unlinking. Instead, skip buddy update
6541 before setup, as it will be regenerated after recovery is complete.
6546 Description: If an OST is inactive, its locks might reference stale inodes.
6547 Details : lov_change_cbdata() must iterate over all namespaces, even if
6548 they are inactive to clear inode references from the lock.
6550 Severity : enhancement
6551 Frequency : occasional, if non-standard max_dirty_mb used
6553 Description: Client will block write RPCs if not enough grant
6554 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6555 then the client will block writes while waiting for another RPC
6556 to complete instead of consuming its dirty limit. With change
6557 we get improved performance when max_dirty_mb is small.
6559 Severity : enhancement
6560 Bugzilla : 3389, 6253
6561 Description: Add support for supplementary groups on the MDS.
6562 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6563 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6564 lookups for user supplementary groups into a cache.
6568 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6569 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6570 for liblustre O_EXCL is enforced by MDS.
6574 Description: readdir never returns NULL in liblustre.
6575 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6576 dentry in struct dirent.
6580 Frequency : liblustre only, depends on application IO pattern
6581 Description: liblustre clients evicted if not contacting servers
6582 Details : Don't put liblustre clients into the ping_evictor list, so
6583 they will not be evicted by the pinger ever.
6585 Severity : enhancement
6587 Description: Add ability to evict clients by NID from MDS.
6588 Details : By echoing "nid:$NID" string into
6589 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6590 $NID would be instantly evicted from this MDS and from all active
6591 OSTs connected to it.
6595 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6596 Details : lookup_it_finish() used to query file size from OSTs that was not
6601 Description: service threads change working directory to that of init
6602 Details : Starting lustre service threads may pin the working directory
6603 of the parent thread, making that filesystem busy. Threads
6604 now change to the working directory of init to avoid this.
6608 Frequency : during shutdown only
6609 Description: shutdown with a failed MDS or OST can cause unmount to hang
6610 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6615 Frequency : 2.6 only
6616 Description: chmod/chown may include an extra supplementary group
6617 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6618 supplementary group and if none is specified this is used.
6621 Bugzilla : 5479 (6816)
6623 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6624 Details : If lookup is in progress on a file that is unlinked we might try
6625 to revalidate the inode and fail in revalidate after lookup is
6626 complete and ll_file_open() enqueues the open again but
6627 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6630 Frequency : always, if lconf --abort_recovery used
6632 Description: lconf --abort_recovery fails with 'Operation not supported'
6633 Details : lconf was attempting to abort recovery on the MDT device and not
6636 ------------------------------------------------------------------------------
6638 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6643 Frequency : rare (only unsupported configurations with a node running as an
6645 Bugzilla : 6514, 5137
6646 Description: Mounting a Lustre file system on a node running as an OST could
6648 Details : OSTs now preallocates memory needed to write out data at
6649 startup, instead of when needed, to avoid having to
6650 allocate memory in possibly low memory situations.
6651 Specifically, if the file system is mounted on on OST,
6652 memory pressure could force it to try to write out data,
6653 which it needed to allocate memory to do. Due to the low
6654 memory, it would be unable to do so and the node would
6655 become unresponsive.
6657 Severity : enhancement
6659 Description: Addition of lconf --service command line option
6660 Details : lconf now accepts a '--service <arg>' option, which is
6661 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6663 Severity : enhancement
6665 Description: Failover mode is now the default for OSTs.
6666 Details : By default, OSTs will now run in failover mode. To return to
6667 the old behaviour, add '--failout' to the lmc line for OSTs.
6669 Severity : enhancement
6671 Description: Health checks are now provided for MDS and OSTs
6672 Details : Additional detailed health check information on MSD and OSTs
6673 is now provided through the procfs health_check value.
6676 Frequency : occasional, depends on IO load
6678 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6679 after numerous create/delete cycles
6680 Details : The ext3 inode allocation policy would not allocate new inodes
6681 very well on the OSTs because there are no new directories
6682 being created. Instead we look for groups with free space if
6683 the parent directories are nearly full.
6688 Description: Network or server problems during mount may cause partially
6689 mounted clients instead of returning an error.
6690 Details : The config llog parsing code may overwrite the error return
6691 code during mount error handling, returning success instead
6697 Description: MDS can fail to allocate large reply buffers
6698 Details : After long uptimes the MDS can fail to allocate large reply
6699 buffers (e.g. zconf client mount config records) due to memory
6700 fragmentation or consumption by the buffer cache. Preallocate
6701 some large reply buffers so that these replies can be sent even
6702 under memory pressure.
6706 Frequency : rare (liblustre)
6707 Description: fsx running with liblustre complained that using truncate() to
6708 extend the file doesn't work. This patch corrects that issue.
6709 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6710 ATTR_SIZE and lsm use in llu_setattr_raw.
6714 Frequency : rare, only 2.6 kernels
6715 Description: Unusual file access patterns on the MDS may result in inode
6716 data being lost in very rare circumstances.
6717 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6718 code in the 2.6 kernel caused the inode and/or EA data not to
6719 be read from disk, causing single-file corruption.
6723 Frequency : rare, only 2.6 filesystems using extents
6724 Description: Heavy concurrent write and delete load may cause data corruption.
6725 Details : It was possible under high-load situations to have an extent
6726 metadata block in the block device cache from a just-unlinked
6727 file overwrite a newly-allocated data block. We now unmap any
6728 metadata buffers that alias just-allocated data blocks.
6732 Frequency : filesystems with default stripe_count larger than 77
6733 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6734 Details : lconf specifies an inode size of 4096 bytes when the default
6735 stripe_count is larger than 77. This conflicts with the default
6736 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6737 this case to avoid pinning too much memory for large EAs.
6739 ------------------------------------------------------------------------------
6741 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6746 Frequency : rare (extremely heavy IO load with hundreds of clients)
6748 Description: Client is evicted, gets IO error writing to file
6749 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6750 caused clients to be evicted for AST timeouts. The fixes in
6751 bug 5192 mean we no longer need to have such short AST timeouts
6752 so ldlm_timeout has been increased.
6755 Frequency : occasional during --force or --failover shutdown under load
6756 Bugzilla : 5949, 4834
6757 Description: Server oops/LBUG if stopped with --force or --failover under load
6758 Details : a collection of import/export refcount and cleanup ordering
6759 issues fixed for safer force cleanup
6762 Frequency : only filesystems larger than 120 OSTs
6763 Bugzilla : 5990, 6223
6764 Description: lfs getstripe would oops on a very large filesystem
6765 Details : lov_getconfig used kfree on vmalloc'd memory
6768 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6769 Bugzilla : 6242, 6243
6770 Description: reading from files that had been truncated to a non-zero size
6771 but never opened returned no data
6772 Details : ll_file_read() reads zeros from no-object files to EOF
6777 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6778 Details : There's little harm in aborting MDS/OSS recovery and letting it
6779 try again, so I removed the LASSERT and return an error instead.
6781 Severity : enhancement
6783 Description: New debugging infrastructure for tracking down data corruption
6784 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6785 (b) cover more of the client-side code path, and (c) try to narrow
6786 down where problems occurred
6790 Bugzilla : 3819, 4364, 4397, 6313
6791 Description: Racing close and eviction MDS could cause assertion in mds_close
6792 Details : It was possible to get multiple mfd references during close and
6793 client eviction, leading to one thread referencing a freed mfd.
6795 Severity: : enhancement
6796 Bugzilla : 3262, 6359
6797 Description: Attempts to reconnect to servers are now more aggressive.
6798 Details : This builds on the enhanced upcall-less recovery that was added
6799 in 1.4.2. When trying to reconnect to servers, clients will
6800 now try each server in the failover group every 10 seconds. By
6801 default, clients would previously try one server every 25 seconds.
6806 Description: After recovery, certain operations trigger a failed
6807 assertion on a client.
6808 Details : Failing over an mds, using lconf -d --failover, while a
6809 client was doing a readdir() call would cause the client to
6810 LBUG after recovery completed and the readdir() was resent.
6812 Severity : enhancement
6814 Description: Default groups are now added by lconf
6815 Details : You can now run lconf --group <servicename> without having to
6816 manually add groups with lmc.
6819 Frequency : occasional
6821 Description: Nodes with an elan id of 0 trigger a failed assertion
6824 Frequency : always when accessing e.g. tty/console device nodes
6826 Description: tty and some other devices nodes cannot be used on lustre
6827 Details : file's private_data field is used by device data and lustre
6828 values in there got lost. New field was added to struct file to
6829 store fs-specific private data.
6832 Frequency : when exporting Lustre via NFS
6834 Description: NFSD failed occasionally when looking up a path component
6835 Details : NFSD is looking up ".." which was broken in ext3 directories
6836 that had grown large enough to become hashed.
6839 Frequency : Clusters with multiple interfaces not on the same subnet
6841 Description: Nodes will repeatedly try to reconnect to an interface which it
6842 cannot reach and report an error to the log.
6843 Details : Extra peer list entries will be created by lconf with some peers
6844 unreachable. lconf now validates the peer before adding it.
6847 Frequency : Only if a default stripe is set on the filesystem root.
6849 Description: Setting a default stripe on the filesystem root prevented the
6850 filesystem from being remounted.
6851 Details : The client was sending extra request flags in the root getattr
6852 request and did not allocate a reply buffer for the dir EA.
6855 Frequency : occasional, higher if lots of files are accessed by one client
6856 Bugzilla : 6159, 6097
6857 Description: Client trips assertion regarding lsm mismatch/magic
6858 Details : While revalidating inodes the VFS looks up inodes with ifind()
6859 and in rare cases can find an inode that is being freed.
6860 The ll_test_inode() code will free the lsm during ifind()
6861 when it finds an existing inode and then the VFS later attaches
6862 this free lsm to a new inode.
6866 Bugzilla : 6422, 7030
6867 Description: MDS deadlock between mkdir and client eviction
6868 Details : Creating a new file via mkdir or mknod (starting a transaction
6869 and getting the ns lock) can deadlock with client eviction
6870 (gets ns lock and trying to finish a synchronous transaction).
6873 Frequency : occasional
6874 Description: While starting a server, the fsfilt_ext3 module could not be
6876 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6877 kernels. Previously, lconf would still use the ext3 name
6878 when trying to load modules. Now, it will correctly use
6879 ext3 on 2.4 and ldiskfs on 2.6.
6881 Severity : enhancement
6882 Description: The default stripe count has been changed to 1
6883 Details : The interpretation of the default stripe count (0, to lfs
6884 or lmc) has been changed to mean striping across a single
6885 OST, rather than all available. For general usage we have
6886 found a stripe count of 1 or 2 works best.
6888 Severity : enhancement
6889 Description: Add support for compiling against Cray portals.
6890 Details : Conditional compiling for some areas that are different
6894 Frequency : occasional
6895 Bugzilla : 6409, 6834
6896 Description: Creating files with an explicit stripe count may lead to
6897 a failed assertion on the MDS
6898 Details : If some OSTs are full or unavailable, creating files may
6899 trigger a failed assertion on the MDS. Now, Lustre will
6900 try to use other servers or return an error to the
6904 Frequency : occasional
6906 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6907 caused lock timeout during readahead (since 1.4.2).
6908 Details : Processes doing readahead might match a lock that hasn't been
6909 granted yet if there are overlapping and conflicting lock
6910 requests. The readahead process waits on ungranted lock
6911 (original lock is CBPENDING), while OST waits for that process
6912 to cancel CBPENDING read lock and eventually evicts client.
6914 Severity : enhancement
6916 Description: Initial enabling of flock support for clients
6917 Details : Implements fcntl advisory locking and file status functions.
6918 This feature is provided as an optional mount flag (default
6919 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6920 locking are implemented yet, and those that are are not guaranteed
6921 to be completely correct in production environments.
6922 mount -t lustre -o [flock|noflock] ...
6925 Frequency : occasional
6927 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6928 assertion in the ext3 JBD (journaling) layer.
6929 Details : The b_committed_data struct is protected by the big kernel lock
6930 in 2.4 kernels, serializing journal_commit_transaction() and
6931 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6932 access to this struct. In 2.6 kernels there is finer grained
6933 locking to improve SMP performance of the JBD layer.
6937 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6939 -----------------------------------------------------------------------------
6941 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6943 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6944 versions of Lustre. Please update all servers and clients to
6945 version 1.4.2 or later at the same time. You must also run
6946 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6947 to update the configuration logs.
6949 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6950 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6951 - swab llog records of type '0' so we get proper header size/idx (5861)
6952 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6953 - fix rename of one directory over another leaking an inode (5953)
6954 - avoid SetPageDirty on 2.6 (5981)
6955 - don't re-add just-being-destroyed locks to the waiting list (5653)
6956 - when creating new directories, inherit the parent's custom
6957 striping settings if present parent (3048)
6958 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6959 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6960 - don't hold client locks on temporary worklist from l_lru (5666)
6961 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6962 - automatically evict dead clients (5921)
6963 - Update file size properly in create+truncate+fstat case (6196)
6964 - Do not unhash mountpoint dentries, do not allow removal of
6966 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6967 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6968 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6969 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6970 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6971 - Correctly handle reads of files with no objects (6243)
6972 - lctl recover will also mark a device active if deactivate used (5933)
6974 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6975 - allow --write-conf on an MDS with different nettype than client (5619)
6976 - don't write config llogs to MDS for mounts not from that MDS (5617)
6977 - lconf should create multiple TCP connections from a client (5201)
6978 - init scripts are now turned off by default; run chkconfig --on
6979 lustre and chkconfig --on lustrefs to use them
6980 - upcalls are no longer needed for clients to recover to failover
6982 - add --abort-recovery option to lconf to abort recovery on device
6984 - add support for an arbitrary number of OSTs (3026)
6985 - Quota support protocol changes.
6986 - forward compatibility changes to wire structs (6007)
6987 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6988 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6989 - improved "lustre status" script
6990 - initialize blocksize for non-regular files (6062)
6991 - added --disable-server and --disable-client configure options (5782)
6992 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6993 - Vanilla 2.4.29 support
6994 - increase maximum number of obd devices to 520 (6242)
6995 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6996 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6997 SLES 9 2.6.5 kernels
6998 - limit stripes per file to 160 (the maximum EA size) (6093)
7000 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7003 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
7004 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
7005 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
7006 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
7007 - small ext3 extents cleanups and fixes (5733)
7008 - improved mballoc code, several small races and bugs fixed (5733, 5638)
7009 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
7010 - avoid needless client->OST connect, fix handle mismatch (5317)
7011 - fix DLM error path that led to out-of-sync client, long delays (5779)
7012 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
7013 - fix several locking issues related to i_size (5492,5624,5654,5672)
7014 - don't move pending lock onto export if it is already evicted (5683)
7015 - fix kernel oops when creating .foo in unlinked directory (5548)
7016 - fix deadlock in obdfilter statistics vs. object create (5811)
7017 - use time_{before,after} to avoid timer jiffies wrap (5882)
7018 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
7019 - Do not leak request if server was not able to process it (5154)
7020 - If mds_open unable to find parent dir, make that negative lookup(5154)
7021 - don't create new directories with extent-mapping (5909, 5936)
7023 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
7024 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
7025 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
7026 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
7027 - support for RHEL4 kernel on i686 (5773)
7028 - provide error messages when incompatible logs are encountered (5898)
7030 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
7031 * version 1.4.0.10 (1.4.1 release candidate 1)
7033 - don't keep a lock reference when lock is not granted (4238)
7034 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7035 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
7036 - drop import inflight refcount on signal_completed_replay error (5255)
7037 - unlock page after async write error during send (3677)
7038 - handle missing objects in filter_preprw_read properly (5265)
7039 - no transno return for symlink open, don't save no-trasno open (3440)
7040 - don't try to complete elan receive that already failed (4012)
7041 - free RPC server reply state on error (5406)
7042 - clean up thread from ptlrpc_start_thread() on error (5160)
7043 - readahead could read extra page into cache that wasn't ejected (5388)
7044 - prevent races in class_attach/setup/cleanup/detach (5260)
7045 - don't dereference de->d_inode after l_dput of de (5458)
7046 - use "int" for stripe value returned from lock_to_stripe (5544)
7047 - mballoc allocation and error-checking fixes in 2.6 (5504)
7048 - block device patches to fix I/O request sizes in 2.6 (5482)
7049 - look up hostnames for IB nals (5602)
7050 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
7051 - don't start multiple acceptors for the same port (5277)
7052 - fix incorrect LASSERT in mds_getattr_name (5635)
7053 - export a proc file for general "ping" checking (5628)
7054 - fix "lfs check" to not block when the MDS is down (5628)
7056 - service request history (4965)
7057 - put {ll,lov,osc}_async_page structs in a single slab (4699)
7058 - create an "evict_client" /proc entry on OSTs, like the MDS has
7059 - fix mount usage message, return errors per mount(8) (5168)
7060 - change grep [] to grep "[]" in tests so they work in more UMLs
7061 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
7062 - remove ancient LOV_MAGIC_V0 EA support (5047)
7063 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
7064 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
7065 - allow readahead allocations to fail when low on memory (5383)
7066 - mmap locking landed again, after considerable improvement (2828)
7067 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
7069 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
7072 - send OST transaction number in read/write reply to free req (4966)
7073 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7074 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7075 - the watchdog thread now runs as interruptible (5246)
7076 - flock/lockf fixes (but it's still disabled, pending 5135)
7077 - don't use EXT3 constants in llite code (5094)
7078 - memory shortage at startup could cause assertion (5176)
7080 - reorganization of lov code
7081 - single portals codebase
7083 - add extents/mballoc support (5025)
7084 - direct I/O reads in the obdfilter (4048)
7085 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
7087 tbd Cluster File Systems, Inc. <info@clusterfs.com>
7090 - send OST transaction number in read/write reply to free req (4966)
7091 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7092 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7093 - the watchdog thread now runs as interruptible (5246)
7094 - handle missing objects in filter_preprw_read properly (5265)
7095 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7096 - drop import inflight refcount on signal_completed_replay error (5255)
7097 - unlock page after async write error during send (3677)
7098 - return original error code on reconstructed replies (3761)
7099 - no transno return for symlink open, don't save no-trasno open (3440)
7101 - add pid to ldlm debugging output (4922)
7102 - bump the watchdog timeouts -- we can't handle 30sec yet
7103 - extra debugging for orphan dentry/inode bug (5259)
7105 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
7108 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
7109 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
7110 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
7111 - fix NULL dereference in /proc/sys/portals/routes (4827)
7112 - allow failed mdc_close() operations to be interrupted (4561)
7113 - stop precreate on OST before MDS would time out on it (4778)
7114 - don't send partial-page writes before EOF from client (4410)
7115 - discard client grant for sub-page writes on large-page clients (4520)
7116 - don't free dentries not owned by NFS code, check generation (4806)
7117 - fix lsm leak if mds_create_objects() fails (4801)
7118 - limit debug_daemon file size, always print CERROR messages (4789)
7119 - use transno after validating reply (3892)
7120 - process timed out requests if import state changes (3754)
7121 - update mtime on OST during writes, return in glimpse (4829)
7122 - add mkfsoptions to LDAP (4679)
7123 - use ->max_readahead method instead of zapping global ra (5039)
7124 - don't interrupt __l_wait_event() during strace
7126 - add software watchdogs to catch hung threads quickly (4941)
7127 - make lustrefs init script start after nfs is mounted
7128 - fix CWARN/ERROR duplication (4930)
7129 - return async write errors to application if possible (2248)
7130 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
7131 - print NAL number in %x format (4645)
7132 - update barely-supported suse-2.4.21-171 series (4842)
7133 - support for sles 9 %post scripts
7134 - support for building 2.6 kernel-source packages
7135 - support for sles km_* packages
7137 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
7140 - ignore -ENOENT errors in osc_destroy (3639)
7141 - notify osc create thread that OSC is being cleaned up (4600)
7142 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
7143 - reconstruct ost_handle() like mds_handle() (4657)
7144 - create a new thread to do import eviction to avoid deadlock (3969)
7145 - let lconf resolve symlinked-to devices (4629)
7146 - don't unlink "objects" from directory with default EA (4554)
7147 - hold socknal file ref over connect in case target is down (4394)
7148 - allow more than 32000 subdirectories in a single directory (3244)
7149 - fix blocks count for O_DIRECT writes (3751)
7150 - OST returns ENOSPC from object create when no space left (4539)
7151 - don't send truncate RPC if file size isn't changing (4410)
7152 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
7153 - bind to privileged port in socknal and tcpnal (3689)
7155 - rate limit CERROR/CWARN console message to avoid overload (4519)
7156 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
7157 - basic mmap support (3918)
7158 - kernel patch series update from b1_4 (4711)
7160 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
7163 - avoid crash during MDS cleanup with OST shut down (2775)
7164 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
7165 - don't use bad inodes on the MDS (3744)
7166 - dynamic object preallocation to improve recovery speed (4236)
7167 - don't hold spinlock over lock dumping or change debug flags (4401)
7168 - don't zero obd_dev when it is force cleaned (3651)
7169 - print grants to console if they go negative (4431)
7170 - "lctl deactivate" will stop automatic recovery attempts (3406)
7171 - look for existing locks in ldlm_handle_enqueue() (3764)
7172 - don't resolve lock handle twice in recovery avoiding race (4401)
7173 - revalidate should check working dir is a directory (4134)
7175 - don't always mark "slow" obdfilter messages as errors (4418)
7177 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
7180 - don't close LustreDB during write_conf until it is done (3860)
7181 - fix typo in lconf for_each_profile (3821)
7182 - allow dumping logs from multiple threads at one time (3820)
7183 - don't allow multiple threads in OSC recovery (3812)
7184 - fix debug_size parameters (3864)
7185 - fix mds_postrecov to initialize import for llog ctxt (3121)
7186 - replace config semaphore with spinlock (3306)
7187 - be sure to send a reply for a CANCEL rpc with bad export (3863)
7188 - don't allow enqueue to complete on a destroyed export (3822)
7189 - down write_lock before checking llog header bitmap (3825)
7190 - recover from lock replay timeout (3764)
7191 - up llog sem before sending rpc (3652)
7192 - reduce ns lock hold times when setting kms (3267)
7193 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
7194 - fix NULL deref and obd_dev leak on setup error (3312)
7195 - replace some LBUG about llog ops with error handling (3841)
7196 - don't match INVALID dentries from d_lookup and spin (3784)
7197 - hold dcache_lock while marking dentries INVALID and hashing (4255)
7198 - fix invalid assertion in ptlrpc_set_wait (3880)
7200 - add libwrap support for the TCP acceptor (3996)
7201 - add /proc/sys/portals/routes for non-root route listing (3994)
7202 - allow setting MDS UUID in .xml (2580)
7203 - print the stack of a process that LBUGs (4228)
7205 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
7208 - don't cleanup request in ll_file_open() on failed MDS open (3430)
7209 - make sure to unset replay flag from failed open requests (3440)
7210 - if default stripe count is 0, use OST count for inode size (3636)
7211 - update parent mtime/ctime on client for create/unlink (2611)
7212 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
7213 - free recovery state on server during a forced cleanup (3571)
7214 - unregister_reply for resent reqs (3063)
7215 - loop back devices mounting and status check on 2.6 (3563)
7216 - fix resource-creation race that can provoke i_size == 0 (3513)
7217 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
7218 - more debugging for page-accounting assertion (3746)
7219 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
7220 - avoid deadlock after precreation failure (3758)
7221 - fix race and lock order deadlock in orphan handling (3450, 3750)
7222 - add validity checks when grabbing inodes from l_ast_data (3599)
7224 - add /proc/.../recovery_status to obdfilter (3428)
7225 - lightweight CDEBUG infrastructure, debug daemon (3668)
7226 - change default OSC RPC parameters to be better on small clusters
7227 - turn off OST read cache for files smaller than 32MB
7228 - install man pages and include them in rpms (3100)
7229 - add new init script for (un)mounting lustre filesystems (2593)
7230 - run chkconfig in %post for init scripts (3701)
7231 - drop scimac NAL (unmaintained)
7233 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
7236 - clean kiobufs before and after use (3485)
7237 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
7238 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7239 - change init script to fail more clearly if not run as root (1528)
7240 - allow clients to reconnect during replay (1742)
7241 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
7242 - don't do DNS lookups on NIDs too small for IP addresses (3442)
7243 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
7244 - fix cond_resched (3554)
7245 - only evict unfinished clients after recovery (3515)
7246 - allow bulk resend, prevent data loss (3570)
7247 - dynamic ptlrpc request buffer allocation (2102)
7248 - don't allow unlinking open directory if it isn't empty (2904)
7249 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
7250 - remove extraneous obd dereference causing LASSERT failure (3334)
7251 - don't use get_cycles() when creating temp. files on the mds (3156)
7252 - hold i_sem when setting i_size in ll_extent_lock() (3564)
7253 - handle EEXIST for set-stripe, set proper directory name (3336)
7255 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
7256 - fix ksocknal_fmb_callback() error messages (2918)
7258 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
7261 - don't copy lvb into (possibly NULL) reply on error (2983)
7262 - don't deref dentry after dput, don't free lvb on error (2922)
7263 - use the kms to determine writeback rpc length (2947)
7264 - increment oti_logcookies when osc is inactive (2948)
7265 - update client's i_blocks count via lvb messages (2543)
7266 - handle intent open/close of special files properly (1557)
7267 - mount MDS with errors=remount-ro, like obdfilter (2009)
7268 - initialize lock handle to avoid ASSERT on error cleanup (3057)
7269 - don't use cancelling-locks' kms values (2947)
7270 - use highest lock extent for kms, not last one (2925)
7271 - don't dereference ERR_PTR() dentry in error handling path (3107)
7272 - fix thread race in portals_debug_dumplog() (3122)
7273 - create lprocfs device entries at setup instead of at attach (1519)
7274 - common AST error handler, don't evict client on completion race (3145)
7275 - zero nameidata in detach_mnt in 2.6 (3118)
7276 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7277 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7278 - initialize RPC timeout timer earlier for 2.6 (3219)
7279 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7280 - print nal/nid for unknown nid (3258)
7281 - additional checks for oscc recovery before doing precreate (3284)
7282 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7283 - don't crash in mdc_close for bad permissions on open (3285)
7284 - zero i_rdev for non-device files (3147)
7285 - clear page->private before handing to FS, better assertion (3119)
7286 - tune the read pipeline (3236)
7287 - fix incorrect decref of invalidated dentry (2350)
7288 - provide read-ahead stats and refine rpc in flight stats (3328)
7289 - don't hold journal transaction open across create RPC (3313)
7290 - update atime on MDS at close time (3265)
7291 - close LDAP connection when recovering to avoid server load (3315)
7292 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7293 - don't leak open file on MDS after open resend (3325)
7294 - serialize filter_precreate and filter_destroy_precreated (3329)
7295 - loop device shouldn't call sync_dev() for nul device (3092)
7296 - clear page cache after eviction (2766)
7297 - resynchronize MDS->OST in background (2824)
7298 - refuse to mount the same filesystem twice on same mountpoint (3394)
7299 - allow llmount to create routes for mounting behind routers (3320)
7300 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7301 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7302 - fix rare problem with rename on htree directories (3417)
7304 - allow default OST striping configuration per directory (1414)
7305 - fix compilation for qswnal for 2.6 kernels (3125)
7306 - increase maximum number of MDS request buffers for large systems
7307 - change liblustreapi to be useful for external progs like lfsck (3098)
7308 - increase local configuration timeout for slow disks (3353)
7309 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7311 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7314 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7315 - don't overwrite extent policy data in reply if lock was blocked (2901)
7316 - drop filter export grants atomically with removal from device (2663)
7317 - del obd_self_export from work_list in class_disconnect_exports (2908)
7318 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7319 - swab reply message in mdc_close, other PPC fixes (2464)
7320 - fix destroying of named logs (2325)
7321 - overwrite old logs when running lconf --write_conf (2264)
7322 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7323 - fix race in target_handle_connect (2898)
7324 - mds_reint_create() should take same inode create lock (2926)
7325 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7326 - don't close files for self_export to avoid uninitialized obd (2936)
7327 - allow MDS with the same name as client node (2939)
7328 - hold dentry reference for closed log files for unlink (2325)
7329 - reserve space for all logs during transactions (2059)
7330 - don't evict page beyond end of stripe extent (2925)
7331 - don't oops on a deleted current working directory (2399)
7332 - handle hard links to targets without a parent properly (2517)
7333 - don't dereference NULL lock when racing during eviction (2867)
7334 - don't grow lock extents when lots of conflicting locks (2919)
7336 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7339 - account for cache space usage on clients to avoid data loss (974)
7340 - lfsck support in lustre kernel code (2349)
7341 - reduce journal credits needed for BRW writes (2370)
7342 - orphan handling to avoid losing space on client/server crashes
7343 - ptlrpcd can be blocked, stopping ALL progress (2477)
7344 - use lock value blocks to assist in proper KMS, faster stat (1021)
7345 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7346 - recovery for initial connections (2355)
7347 - fixes for mds_cleanup_orphans (1934)
7348 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7349 - block all file creations until orphan recovery completes (1901)
7350 - client remove rq_connection from request struct (2423)
7351 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7352 - recovery timer race (2670)
7353 - mdc_close recovey bug (2532)
7354 - ptlrpc cleanup bug (2710)
7355 - mds timeout on local locks (2588)
7356 - namespace lock held during RPCs (2431)
7357 - handle interrupted sync write properly (2503)
7358 - don't try to handle a message that hasn't been replied to (2699)
7359 - client assert failure during cleanup after abort recovery (2701)
7360 - leak mdc device after failed mount (2712)
7361 - ptlrpc_check_set allows timedout requests to complete (2714)
7362 - wait for inflight reqs when ptlrpcd finishes (2710)
7363 - make sure unregistered services are removed from the srv_list
7364 - reset bulk XID's when resending them (caught by 1138 test)
7365 - unregister_bulk after timeout
7366 - fix lconf error (2694)
7367 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7368 - readahead locks pages, leaves pending causing memory pressure (2673)
7369 - increase OST request buffers to 4096 on large machines (2729)
7370 - fix up permission of existing directories in simple_mkdir (2661)
7371 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7372 - don't assign transno to errored transactions (2742)
7373 - don't delete objects on OST if given a bogus objid from MDS (2751)
7374 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7375 - if rq_no_resend, then timeout request after recovery (2432)
7376 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7377 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7378 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7379 - MDS recovery shouldn't delete objects that it has given out (2730)
7380 - if enqueue arrives after completion, don't clobber LVB (2819)
7381 - don't unlock pages twice when trigger_group_io returns error (2814)
7382 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7383 - don't write pages to disk if there was an error (1450)
7384 - don't ping imports that have recovery disabled (2676)
7385 - take buffered bytes into account when balancing socknal conn (2817)
7386 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7387 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7388 - remove little-endian swabbing of llog records (1987)
7389 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7390 - retry reposting request buffers if they fail (1191)
7391 - grow extent at grant time to avoid granting a revoked lock (2809)
7392 - lock revoke doesn't evict page if covered by a second lock (2765)
7393 - disable VM readahead to avoid reading outside lock extents (2805)
7395 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7396 - updated kernel patches for hp-2.4.20 kernel (2681)
7398 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7401 - fix truncated write corruption (2366)
7402 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7404 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7405 - ensure len(uuid) < 37 in lmc (1171)
7406 - fix ia64 OOPS in llog_test (2255)
7407 - zero end of page at obdfilter for partial page writes (2648)
7408 - don't leave stale dentries around after renames (bug 2428)
7409 - fix timeouts when evicting a client with a single lock held (2642)
7410 - set deadline for the initial HELLO message to drain (2634)
7411 - print out dotted-quad IP addresses in the socknal (2302)
7413 - additional debugging for MDS client eviction problem (2443)
7414 - fix mkfsoptions support for osts (2603, 2604)
7416 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7419 - add series for the vanilla 2.6.0 kernel
7420 - add series for the vanilla 2.4.24 kernel
7421 - add series for a cray x86/64 UL kernel drop
7422 - fix xattr patches for the vanilla 2.4.19 series
7424 - generate true UUIDs in lmc (1171)
7425 - have portals stack dumping break in UML (2466)
7426 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7427 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7428 - wake callback waiting threads on client eviction (2460)
7429 - Add --ptldebug and --subsystem to lmc (1719)
7430 - update assertion to allow safe interrupt allocation
7431 - set rq_no_resend for cancel requests (2432)
7432 - recalculate ptlrpcd timeout after resend (2494)
7433 - call vfs_rmdir when removing pending directories (2368)
7434 - fix renaming a file to itself (2429)
7435 - lmc creates a default one-stripe lov (2454)
7436 - expand procfs space to handle large clusters (2326)
7437 - increase UML stack to avoid overflow
7438 - update lconf's list of debug and subsystem masks
7439 - fix lfs find --obd (2510)
7440 - /proc tunable for disabling filter read caching (2591)
7441 - stop rpm packages from altering slapd.conf (2301)
7442 - disable nagle in the socknal under 0conf (2578)
7443 - choose mds inode size based on stripe count (2572)
7444 - fix kernel-source rpm problems (2516)
7446 - add --disable-doc to avoid pdf generation (2421)
7447 - update documentation, tests, type-os, comments
7448 - avoid format warnings on ia64
7449 - remove the TOE NAL
7450 - tiny code cleanups by removing unused fields
7452 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7455 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7456 - remove the most blatant lies from BUILDING file (2371)
7457 - change default debug level to reasonable production setting
7458 - reduce client side cache size to reduce cache flush time
7459 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7460 - make TCP zerocopy and pinger support enabled by default (2476)
7461 - sync writes completed after process exits caused crashes (2319)
7462 - maintain correct mount count on the MDS (2356)
7463 - backout 1557, because 2316 wasn't really fixed
7464 - better file I/O statistics gathering in /proc
7465 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7466 - another kernel patch to fix zero-copy TCP function export
7467 - don't take duplicate lock when processing re-sent getattr (2420)
7468 - lctl uses obd_self_export instead of creating new conn (2353)
7469 - MDS/OST recovery case which requires object creation asserted (2425)
7470 - move lfs from /usr/sbin to /usr/bin in packages
7471 - fix race between mds_client_add and mds_client_free (2417)
7472 - use kmalloc instead of slabs in portals (2430)
7473 - don't create duplicate records when a failover MDS is present (2442)
7474 - remove unnecessary mount age check (2332)
7475 - don't remove directory inodes from locks prematurely (2451)
7476 - don't break if MDS service name is the same as hostname (2103)
7477 - fix races in client write RPC generation when cache full (2482)
7479 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7482 - remove now-unused request->rq_obd (278)
7483 - if an allocation fails, print out how much memory we've used (1933)
7484 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7485 - add the "configurable stack size" patch to most series files (1256)
7486 - ability to write large log records, for 100+ OST configs (2306)
7487 - fix NULL deref when filter_prep fails (2314)
7488 - fix operator precedence error in filter_sync
7489 - dynamic allocation of socknal TX descriptors (2315)
7490 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7491 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7492 - ensure that CWARN messages in Portals always get to the syslog
7493 - __init/__exit are not for prototype decls (ldlm_init/exit)
7494 - x86-64 compile warning fixes
7495 - fix gateway LMC keyword conflict (2318)
7496 - fix MDS lock inversions in getattr/reint paths (1844)
7497 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7498 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7499 - DLM race condition prevented some lock evictions (2328)
7500 - ENOMEM detection and retry on socknal sends (2230)
7501 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7502 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7503 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7504 - better error messages when a client is rejected during recovery (1505)
7505 - avoid cancelling locks which were never granted, after failure (2330)
7506 - fix i_sem/journal inversion in mds_client_add (2333)
7507 - fix truncate/getattr lock cycle deadlock (2334)
7508 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7509 - fix two rare exit paths which could leak an l_lock() ref (2321)
7510 - fencepost error in MDS/OST orphan recovery (2226)
7511 - make log record alignment 8 bytes (1988)
7512 - lstripe now fails when requested offset > ost_count (2237)
7513 - ensure that all kernel series have a complete list.h (1607)
7514 - fix crashes in special-file operations (2316)
7515 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7517 - allow configurable automake binary, for testing new versions
7518 - small update to the lfs documentation
7520 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7522 * fix negative export reference count in fsfilt_sync (2312)
7524 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7525 * release candidate 0.9.1
7527 - orphans are moved into the PENDING directory for possible recovery
7528 - replayed opens now open by fid for orphan/rename safety (1042)
7529 - last close of an orphan inode generates a transno (683)
7530 - chdir() and mount() now pin the directory entry (1020)
7531 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7532 - discard very old requests without processing them (1502)
7533 - remove some common, well-understood CERRORs (1505)
7534 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7535 - clear "grant" flags in OST replies until OST grant code lands (1644)
7536 - fix read performance by not clobbering i_blksize on client (1598)
7537 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7538 - make LRU size a /proc tunable, clears locks when reduced (707)
7539 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7540 - server locks take references on exports now (1558)
7541 - build fixes for 2.4.20-rh trees (1663)
7542 - return an error from lov_create if all OSCs are inactive (1751)
7543 - fix import levels when a reconnect happens without a timeout (1597)
7544 - exit early from mds_open if we get a lookup error (1749)
7545 - partial page read at EOF wouldn't wait for disk before sending (1642)
7546 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7547 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7548 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7549 - write/truncate lock inversion (1639)
7550 - Don't auto-load obdclass, portals modules during cleanup (1495)
7551 - fix timestamps from jumping to "now" (1763)
7552 - extra journal assertions (1648)
7553 - add an extra multiunlink test (1771)
7554 - fix read_record/write_record API (1776)
7555 - fix leak of offset_extent, possible incorrect i_size later (1772)
7556 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7557 - Debugging for the kqswnal_get_idle_tx problems (1820)
7558 - Allow recovery to be attempted multiple times (1536)
7559 - Write out MDS last_rcvd file after it is first created (1600)
7560 - Fix tx_descriptor leak in failed transmit situations (1827)
7561 - ext3 journaling fixes for assertion failure after IO error (1871)
7562 - class_export_put() on freed export after completion AST error (1896)
7563 - Fix revalidate looping in VFS (1322)
7564 - Don't access a freed export during MDS_REINT timeout (1521)
7565 - Add open-unlink recovery support on the MDS (1673,1764)
7566 - Return an error if no MDS data was read from last_rcvd (1946)
7567 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7568 - Don't setup a disk device that doesn't match exported UUID (317)
7569 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7570 - avoid committing NULL handle in force close
7571 - local.sh is now a one-stripe LOV configuration
7572 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7573 - fix ext3 htree duplicate directory entry corruption (1516)
7574 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7575 - update to new LOV EA format (2097)
7576 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7578 2003-06-15 Phil Schwan <phil@clusterfs.com>
7581 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7582 - per-import/export recovery handling (958, 931, 959)
7583 - multiple last-rcvd slots, for serving multiple FSes (949)
7584 - connections are again shared between multiple imp/exports (963, 964)
7585 - "umount -f" would hang if any requests needed to be sent (393, 978)
7586 - avoid pinning large req buffer by copying for queued messages (989)
7587 - add "uuid" to "lctl device" command to help upcalls (991)
7588 - "open" RPCs with transnos would confuse recovery counters (1037)
7589 - do proper endian conversion of all wire messages (288, 340, 891)
7590 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7591 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7592 - fix ptlrpc_connection leak in target_handle_connect (1174)
7593 - fix import refcounting bug in OST and MDS cleanup (1134)
7594 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7595 - fix very unlikely obd_types race condition (501)
7596 - remove osc_open hack for echo_client (1187)
7597 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7598 - a failure in read_inode2 leads to deadlock (1139)
7599 - cancel ack-locks as soon as transaction is committed (1072)
7600 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7601 - make sure to commitrw after any preprw to avoid deadlock (1162)
7602 - failing to execute a file in a lustre FS would lock inode (1203)
7603 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7604 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7605 - obd_uuid2tgt would walk off the end of the list (1255)
7606 - on IA64 the osc would give portals incorrect bulk size (1258)
7607 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7608 - fix lock inversion caused by new llite matching code (1282)
7609 - limit the number of dirty pages on a client to 10MB (1286)
7610 - timed out locks were not being corrected cancelled (1289)
7611 - fix O_DIRECT above 4GB on IA-32 (1292)
7612 * major user-visible changes
7613 - fail out/fail over policy now controlled by the upcall (993)
7615 - add OBD_PING to check server availability and failure (954)
7616 - lustre messages are now sent in sending host order (288, 340, 891)
7617 - add eadatalen to MDS getattr reply (340)
7618 - OST read replies may contain second buffer, with per-page status (593)
7620 2003-03-11 Phil Schwan <phil@clusterfs.com>
7623 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7624 - failed open()s could cause deadlock; fixed (867, 869)
7625 - stop cancelling OST locks when files are closed (481)
7626 - overlapping XID spaces caused network corruption (851, 853)
7627 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7628 - setattr_raw added, to do single-RPC, server-side setattrs
7629 - lmc/lconf syntax change for OST UUIDs
7630 - fix crashy race condition between ptlrpc_free_req and osc_close
7631 - don't use request in mdc_enqueue if we hit a timeout (889)
7632 - don't set the inode i_size for regular files from the MDS (896)
7633 - handle out of order completion AST (842)
7634 - don't LBUG if a lock request times out after receiving AST (913)
7635 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7636 - if a bad lock AST arrives, send an error instead of dropping entirely
7637 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7638 - fix leak in bulk IO when only partially completed (899, 900, 926)
7639 - fix O_DIRECT for ia64 (55)
7640 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7641 - C-z after timeout could hang a process forever; fixed (977)
7643 - client-side I/O cache (678, 924, 929, 941, 970)
7645 - READPAGE and SETATTRs which don't take server-side locks get
7648 2003-02-11 Phil Schwan <phil@clusterfs.com>
7651 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7652 - Fix fsfilt_extN_readpage() to read a full page of directory
7653 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7654 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7655 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7656 - Most intent operations were removed, in favour of a new RPC mode
7657 that does a single RPC to the server and bypasses most of the VFS
7658 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7659 - Aggressively cancel local locks on DLM servers
7660 - mds_reint_unlink sends EA to the client if it's the last nlink.
7661 client uses that EA to unlink OST objects.
7662 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7663 - recursive symlinks were fixed (439)
7664 - fixed NULL deref in DEBUG_REQ
7665 - filter_update_lastobjid no longer calls sync, which annoyed extN
7666 - fixed multi-client small-writes to a single file problem (445)
7667 - fixed mtime updates during file writes (607)
7668 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7669 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7670 - fixed Linux OST size reporting problem (444, 656)
7671 - OST now updates object mtime with writes or setattr (607, 619)
7672 - client verifies file size before zeroing page past EOF (445)
7673 - OST now writes last allocated objid to disk with allocation (108)
7674 - LOV on echo now works (409)
7676 - mds_reint_unlink sends a new buffer, with the EA included. this
7677 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7678 set if a regular file was being unlinked, and it was the last link
7679 - use PtlGet from the target for bulk writes (315)
7680 - OST now updates object mtime with writes or setattr (607, 619)
7681 - LDLM now has a grant-time callback to revalidate locked items, if
7683 - Many MDS operations were reorganized to combat race conditions
7685 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7686 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7688 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7691 - Fully reactivate OST imports after reconnection (512, others)
7692 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7693 - More graceful error handling for truncating on dead OST (515)
7694 - Don't error out unless we're actually accessing dead stripes (474)
7695 - Fix garbage sizes when stripes are missing (410)
7696 - LRU counters were broken, causing constant lock purge (433, 432)
7697 - garbage on read from stripes with failed OSTs (441)
7698 - mark OSCs as active before reconnecting during recovery (438)
7699 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7700 - lfind did not preserve OST order in output (443)
7701 - symlinks cause hung clients, incorrect data (439)
7702 - stop dereferencing request after dropping refcount (457)
7703 - don't LASSERT(spin_is_locked) on non-SMP (455)
7704 - fixes for many rename() bugs
7705 - fstat didn't correctly synchronize attributes (399)
7706 - server must handle lock cancellation during blocking AST prep (487)
7707 - bulk descriptors were free()d too soon (511)
7708 - fix paths in lconf, which would load incorrect modules (451, 507)
7709 - fix confusing lconf 'host not found' error message (386)
7710 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7711 - fix race condition in mdc_blocking_ast() for inode access (526)
7712 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7713 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7714 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7715 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7716 - fix lstripe in several ways (kernel side) (527)
7717 - fix request leak in ldlm_cli_enqueue (262)
7718 - incorrect OSC was marked inactive after OST failure
7719 - call mds_fs_cleanup before unmounting filesystem (524)
7720 - fix races between taking ns_lock and ldlm_lock_change_resource
7721 - fix races updating LOV export open file list
7722 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7723 - fix recovery NULL deref in ldlm_cli_cancel_unused
7724 - fix some DLM races by using new hash table for lock handles (419)
7725 - permit the client to specify desired inodes, at replay
7726 - duplicate requests when we queue them for replay reintegration
7727 - fix last_rcvd offset calculation
7728 - sync after each recovered transaction, so we always make progress
7729 - never, not always, ERESTART requests without transnos
7730 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7732 - skip replay if the MDS says that the client is already connected
7733 - don't check for a recovery-enabled export to match lctl's UUID
7734 - don't INC_USE_COUNT for phantom exports
7735 - don't crash when cleaning up phantom exports (567)
7736 - don't double-finish or set replay data for errored mdc_open requests
7737 - abort requests when they time out, so we don't get old replies
7738 - send/receive replies for AST messages again
7739 - if the client says that it doesn't have the lock, cancel it on the
7741 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7742 mark it as destroyed, it will all work out when decref is called
7743 - fix module use counts (22, 581)
7745 - ASTs now expect a reply (server cancels lock on error reply)
7747 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7750 - fix many simultaneous client startup (392)
7751 - fix dentry->d_it clobbering
7752 - credentials weren't being shipped for readdir/getattr operations
7753 - remove invalid assertions triggered during some concurrent MD
7755 - proper Lustre versions added (336, 389)
7756 - fix memory leak for create error case (398)
7757 - fix LOV locking bug that would get cli/srv out of sync
7758 - fix echo client over LOV (409)
7759 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7760 - fix double-O_EXCL intent crash (424)
7761 - avoid sending multiple lock CANCELs (352)
7763 - MDS can do multi-client recovery (modulo bugs in new code)
7765 - many updates, edits, cleanups
7767 2002-11-18 Phil Schwan <phil@clusterfs.com>
7770 - fix null d_it dereference (346)
7771 - fix full OST/dbench hang (333)
7772 - fix permission problem with file removal (286)
7773 - fix removal of OSCs from LOV when they fail
7774 - fix NULL deref during bulk timeout (214)
7775 - fix problems related to multiple filesystems on one MDS (241)
7776 - fixed serious subtle metadata locking bugs
7777 - free locks on clients when inodes are removed due to memory
7779 - fix inode pointer in lock data (285)
7780 - partial support for multiple MDS on a single host (241)
7781 - data locks weren't cancelled at clear_inode time (290, 311)
7782 - intent locks could lead to unbounded lock growth (205)
7783 - added a maximum lock count, an LRU list, and a flusher
7784 - fix multiple rename (365)
7785 - properly abstracted the echo client
7786 - OSC locked 1 byte too many; fixed
7787 - rewrote brw callback code:
7788 - fixed recovery bugs related to LOVs (306)
7789 - fixed too-many-pages-in-one-write crash (191)
7790 - fixed (again) crash in sync_io_timeout (214)
7791 - probably fixed callback-related race (385)
7793 - Add capability to MDS protocol
7794 - LDLM cancellations and callbacks on different portals
7796 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7799 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7800 - defer open object destruction to close time (601981, 138)
7801 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7802 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7803 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7804 - don't call obd_setattr in ll_file_release for destroyed objects
7805 * protocol change to lustre_msg: move |version| and add |flags|
7806 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7807 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7808 * added replay of create, unlink, link and rename operations during
7809 MDS failover; recovery should be much more robust now
7810 * remove failed OSCs from LOVs (only lov_create uses this so far)
7811 * the lustre-HOWTO was brought (more) up to date (582544)
7813 2002-10-23 Phil Schwan <phil@clusterfs.com>
7816 - in-use dentries weren't being reused properly (617851)
7817 - prevent multiple LDLM setup (599178)
7818 - fix LOV size calculations for truncate (617853)
7819 - fix client handling of MDS intent errors (POSIX)
7820 - fix permission bug in lovstripe.c test (624321)
7821 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7822 - truncate past end of file could corrupt data
7823 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7824 - a race in recovery could return ETIMEDOUT to apps (623947)
7825 - building outside the source directory was fixed
7826 * the lustre-HOWTO was brought (more) up to date (582544)
7827 * major progress was made on recovery functionality
7829 2002-10-10 Phil Schwan <phil@clusterfs.com>
7832 - recovery deadlock fix
7833 - rm -rf causes LBUG fix (617817)
7834 - file open by multiple tasks fix (618962)
7835 - directory permissions bugs (602707 and 620007)
7836 - journal_stop fixed with locking (611313)
7837 - O_APPEND failures resolved (618273, perhaps 614459)
7838 - lconf PATH fix (619770)
7839 - IA64 build fix (621450)
7840 - RPC buffer sizes scale with amount of memory
7842 2002-10-01 Phil Schwan <phil@clusterfs.com>
7845 - locks would be cancelled without throwing away data pages,
7846 resulting in inconsistent data (605627)
7847 - inode attributes were not always being refreshed (605627, 612449)
7848 - lconf now continues to cleanup after lctl reports an error
7849 - MDS now enforces user permissions (602707)
7850 - lprocfs cleanup fixed, but not yet enabled (614157)
7851 - fixed infinite server hang, should a client not respond to an AST
7852 - avoid going into recovery if user calls readlink() with a buffer
7853 that's too small (613941)
7854 - AST RPCs no longer require replies (614867) -- this may be changed
7855 - don't crash server if client sends an IOV that's too big (611336)
7856 - fixed lock conversion deadlock (611892)
7857 - fixed the following of symlinks (614622)
7858 * recovery: the server can remove locks from a client that dies, other
7859 clients can make progress
7860 * more extN patch fixes
7861 * compile-time configurable ptlrpc buffer allocations
7863 - collaborative read cache document
7864 - Lustre Lite Performance CDR document-in-progress
7866 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7869 - fix typo in patch-2.4.18
7871 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7874 - clear ptlrpc request each time in handle_incoming_request()
7875 - unlink of files now destroys the object on the OST
7877 2002-09-19 Peter Braam <braam@clusterfs.com>
7879 * add hard link support
7880 * change obdfile creation method
7881 * kernel patch changed
7883 2002-09-19 Peter Braam <braam@clusterfs.com>
7886 - stack overflow bug in extN fixed
7888 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7890 * documentation updates
7891 - add man pages for config tools
7892 - update tests/README to describe testing with new config tools
7893 - finish metadata API descriptions
7894 * bug fixes and cleanups
7895 - statfs workaround for 16TB limit
7896 - LOV stripe allocation improved, can stripe on subset of OSTs
7897 - LOV file size/IO offset was wrong for files > 4GB in size
7898 - object EA data was being dropped, caused files to be unreadable
7899 - memory overflow with non-LOV OST caused memory corruption
7900 - fixed regression tests to work with new config tools, obdfilter
7901 - fixed bug when directory size became larger than 1 block
7902 - fixed bug (for single client case) when PWD was deleted
7903 - invalidate local directory pages when doing intent-based ops
7904 - avoid LDLM oops when lock callback contained bad data
7906 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7908 * documentation updates
7909 * bug fixes and cleanups
7910 - configuration tools
7913 - 64-bit compile warnings
7914 - 64-bit internal statfs data
7916 * test_brw on persistent OST devices
7918 * lprocfs (disabled)
7920 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7922 * documentation updates
7923 * bug fixes and cleanups
7924 * configuration tools
7926 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7929 * many small fixes to 0_5_4
7930 * io/network handling
7931 * thinkos in MDS operations
7933 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7936 * crucial basic fixes to 0.5.3
7937 * IOR, Iozone work over Elan
7940 2002-08-07 Phil Schwan <phil@clusterfs.com>
7941 * version 0_5_3, our first alpha
7942 * we use the new Portals iovs
7943 * documentation updates
7944 * bug fixes and cleanups
7945 * small changes in the DLM wire protocol
7947 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7948 * version 0_5_1 with some initial stability,
7949 * locking on MD and file I/O.
7950 * documentation updates
7951 * several bug fixes since 0.5.0
7952 * small changes in wire protocol
7954 2002-07-18 Phil Schwan <phil@clusterfs.com>
7956 * delivered as Lustre Light Alpha
7957 * fixed a crash after handling invalid MDS requests
7958 * fixed directory pages for architectures with non-4k pages sizes
7960 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7961 * release version v0_4_4
7962 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7964 * Fixes a number of interruption problems with OST operations.
7965 * Update documentation for portals header changes
7966 * Move all wire protocol structs/defines to lustre_idl.h
7967 * Fixes symlink length bug.
7968 * Add tcpdump to repository.
7970 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7971 * release version v0_4_3
7972 * Fixes statfs for inodes on extN.
7973 * Fixes bug in runtests which would delete /etc/hosts.
7974 * Use 64-bit object IDs wherever possible (not into VFS though)
7975 Remove ost_get_info, which is unused by lustre, and out of date.
7977 2002-07-03 Peter Braam <braam@clusterfs.com>
7978 * release version v0_4_2 Fixes a lookup error (type not passed)
7979 * move forward to head of Portals
7980 * move forward to latest Lustre kernel
7982 2002-06-25 Peter Braam <braam@clusterfs.com>
7983 * release version v0_4_1. Hopefully stable on single node use.