1 xxxx-xx-xx 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 -------------------------------------------------------------------------------
12 2011-06-30 Whamcloud, Inc.
14 * Support for kernels:
15 2.6.18-238.12.1.el5 (RHEL 5)
16 2.6.32-131.2.1.el6 (RHEL 6) - patchless client only
17 * Recommended e2fsprogs version: 1.41.90.wc3
18 * Use ext4-based ldiskfs as default for RHEL5
19 * Add support for 24TB LUN (RHEL5 server only)
21 Severity : enhancement
23 Description: Add support for RHEL5.6 (2.6.18-238.12.1.el5)
25 Severity : enhancement
26 Jira : LU-62, LU-73, LU-402, LU-130
27 Description: Add RHEL6 client support
31 Description: Fix client crash introduced in 1.8.5
32 Details : The patch from bugzilla ticket 18213 landed for 1.8.5 can cause
33 a crash in mdc_exit_request() when the process is interrupted.
37 Description: ls -l reports wrong file size
38 Details : Always update LVB from disk when a glimpse callback returns an
41 Severity : major (2.6.32-based kernels only)
43 Description: Add workaround for a race causing an assertion failure in
44 clear_inode() on lustre clients.
48 Description: Fix interoperability issue between 1.8 clients and 2.x servers
49 Details : The ldlm hash handling in Lustre 2.x is incompatible with the one
50 used in Lustre 1.8. Add support for 64-bit dir name hash to the
55 Description: slow IO with read-intense application
56 Details : Align the readahead extent by 1M after when it is trimed by
61 Description: Skip quotacheck on administratively disabled OSTs
65 Description: Some autogen improvements
67 Severity : enhancement
68 Jira : LU-123, LU-296, LU-230
69 Description: Add support for yaml data logging and auster script.
71 Severity : enhancement
73 Description: Allow OSTs to be created with no primary node
74 Details : Add a --servicenode parameter for mkfs.lustre to treat all
75 service nodes equally.
77 -------------------------------------------------------------------------------
79 2011-05-12 Oracle, Inc.
81 * Support for kernels:
82 2.6.16.60-0.42.8 (SLES 10),
83 2.6.27.39-0.3.1 (SLES 11),
84 2.6.18-194.3.1.el5 (RHEL 5)
85 2.6.18-194.3.1.0.1.el5 (OEL 5)
86 * Client support for unpatched kernels:
87 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
88 2.6.16 - 2.6.32 vanilla (kernel.org)
89 * Recommended e2fsprogs version: 1.41.10-sun2
90 * The async journal commit feature (bug 19128) and the cancel
91 lock before replay feature (bug 16774) are disabled by default.
95 Description: llap pages are moved to head of LRU list instead of tail
99 Description: disable statahead by default due to important races found in the
100 statahead implementation.
104 Description: release request ref through ptlrpc_server_drop_request()
105 Details : make sure the request is protected by rq_refcount while
106 being processed in ptlrpc_server_handle_req_in().
110 Description: correct check for transno value in filter_finish_transno()
114 Description: Set body->eadatasize in mdc_getattr_pack()
118 Description: Handle unsent requests with rq_net_err in ptlrpc_check_set()
122 Description: apps stuck in ptlrpc_check_set() during direct I/O
123 Details : start async bulk unregistering at the same time at reply unlink
127 Description: lfs setstripe --pool broken in 1.8.4
128 Details : fix llapi_search_fsname() to handle relative pathnames
132 Description: Only force the mode change if we're changing the size as well
133 Details : Fix regression introduced in 1.8.0. The offending code was added
134 by commit 77ba4b2141d04180211efa8a75c11ab0abf7fafb to remove
135 setgid/setuid bits when do_truncate() is called on the file.
136 We should only force the change when that occurs, similarly to
137 ll_setattr() in lustre/llite/llite_lib.c
139 Severity : enhancement
141 Description: add procfs tunable to enable/disable lockless direct I/O
142 Details : llite.lustre-*.lockless_direct_io=0 will disable default semantics
143 of direct I/O that forces it to be lockless. lockless_direct_io
144 value, however, will be ignored if per-file LL_FILE_LOCKED_DIRECTIO
149 Description: Lustre servers crashing with NULL pointer errors
150 Details : Make sure the request is protected by rq_refcount while being
151 processed in ptlrpc_server_handle_req_in().
155 Description: kernel BUG at drivers/scsi/sd.c
156 Details : Remove sd iostats patch from sles11 patch series.
160 Description: Modified value of at_min is not taken into account
164 Description: lustre grants flock exclusive locks to two fd's in the same process
168 Description: LBUG in the mfs_verify_child()
172 Description: ASSERTION(atomic_read(&client_stat->nid_exp_ref_count) == 0)
173 Details : Reconnecting from a different nid causes old per-nid stats not to be
178 Description: create proper macro check for bdi interface.
180 -------------------------------------------------------------------------------
182 2010-10-29 Oracle, Inc.
184 * Support for kernels:
185 2.6.16.60-0.69.1 (SLES 10),
186 2.6.32.19-0.2.1 (SLES11),
187 2.6.18-194.17.1.el5 (RHEL 5)
188 2.6.18-194.17.1.0.1.el5 (OEL 5)
189 * Client support for unpatched kernels:
190 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
191 2.6.16 - 2.6.32 vanilla (kernel.org)
192 * Recommended e2fsprogs version: 1.41.10-sun2
193 * The async journal commit feature (bug 19128) and the cancel
194 lock before replay feature (bug 16774) are disabled by default.
198 Description: atime is not properly updated on an MDS
200 Severity : enhancement
202 Description: Update to RHEL5.5 kernel 2.6.19-194.17.1.el5.
203 Update to OEL5.5 kernel 2.6.19-194.17.1.0.1.el5.
205 Severity : enhancement
207 Description: Update to SLES10 SP3 kernel 2.6.16.60-0.69.1.
210 Frequency : only with SLES10
212 Description: Use OFED "KMP" provided by Novell
213 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
214 Lustre is now built against this package. That means you need to
215 install the ofed-kmp package from Novell for the patchless client
216 and from our download site for the server. Note that the ofed-kmp
217 that Novell ships may not exactly match the kernel version but
218 should still be compatible.
220 Severity : enhancement
222 Description: Update SLES11 SP1 kernel to 2.6.32.19-0.2.1.
226 Description: Enabling quotas fails with non-consecutive OST numbering.
230 Description: Fix kernel warning due to lookup_on_len() called without i_mutex
235 Description: Account direct i/o inflight rpcs separately from non-direct i/o so
236 that direct i/o, which is limited by max_rpcs_in_flight, should not
237 block non-direct i/o, which is not limited by max_rpcs_in_flight.
241 Description: Fix per-NID reporting on outstanding writes
245 Description: Reduce stack pressure by uninlining some mds and ptlrpc functions.
249 Description: Remove LASSERT in lprocfs_rd_conn_uuid() since conn == NULL is a
254 Description: fix obdo leak issue in ll_setattr_raw()
258 Description: limit MMP interval
260 Severity : enhancement
262 Description: add several lfs ost enhancements
266 Description: Too many default ACLs break directory access on new directories
270 Description: Lustre inode size is not coherent across nodes.
271 Details : Update lvbo from disk when AST fails with EINVAL. Lvbo is updated
272 on EINVAL error in ldlm_handle_ast_error(). The updates in
273 filter_intent_policy() and ldlm_cb_interpret() have been removed as
278 Description: Oops at __percpu_counter_add+0x1b
279 Details : Use bdi_init()/bdi_destroy() to proper initialize backing_dev_info
284 Description: add mount option to generate 32bit ino, this can be used for 32bit
285 application compatibility.
289 Description: keep reference count for "lli_sai" to prevent it to be released
290 when "statahead_enter()"
294 Description: allow quotacheck over OSTs with sparse indices
298 Description: Objects not getting deleted for files which have been removed
299 Details : ll_have_md_lock() should differentiate between CR and CW OPEN
304 Description: pin object's inode in memory to avoid certain timeouts
308 Description: fix LBUG when obdfilter-survey is interrupted.
310 -------------------------------------------------------------------------------
312 2010-07-31 Oracle, Inc.
314 * Support for kernels:
315 2.6.16.60-0.42.8 (SLES 10),
316 2.6.27.39-0.3.1 (SLES11),
317 2.6.18-194.3.1.el5 (RHEL 5)
318 2.6.18-194.3.1.0.1.el5 (OEL 5)
319 * Client support for unpatched kernels:
320 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
321 2.6.16 - 2.6.30 vanilla (kernel.org)
322 * Recommended e2fsprogs version: 1.41.10-sun2
323 * The async journal commit feature (bug 19128) and the cancel
324 lock before replay feature (bug 16774) are disabled by default.
326 Severity : enhancement
328 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
329 2.6.18-194.8.1.0.1.el5.
331 Severity : enhancement
333 Description: using inkernel OFED stack for rhel5 & oel5.
335 Severity : enhancement
337 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
338 Details : lfs_migrate does a "poor man's" migration of files from their
339 current OST layout to a new OST layout as chosen by the MDS.
343 Description: mds_orphan_add_link()) error linking orphan to PENDING
344 Details : quota limits might disallow linking orphans to PENDING
345 when unlinking a file - temporary raise threads' privileges
346 when processing unlinks.
348 Severity : enhancement
350 Description: add conf-param -d option to remove permanent settings.
351 Details : Add the ability to remove permanent lctl conf_param settings.
352 (Previously conf_param settings could only be changed, not
353 removed.) This also provides a method to change failover
354 nid locations. Improve lctl man page.
356 Severity : enhancement
358 Description: add list_param to b1_8 and add "-R" option to list params
361 Severity : enhancement
363 Description: lfs quota output is not very convenient for awk/sed-parsing
364 Details : Some positions in lfs quota output table could be empty or
365 non-empty which made it hard to parse it with scripts, now a dash
366 is put instead of space where there is not supposed to be any data.
368 Severity : enhancement
370 Description: fix obdfilter-survey script to work properly with remote oss-s
372 Severity : enhancement
374 Description: add new OBDFILTER_SURVEY test suite
376 Severity : enhancement
378 Description: add new multiple mount protection (MMP) test suite
380 Severity : enhancement
382 Description: add support for async journal commit in echo client
384 Severity : enhancement
386 Description: allow userland programs to include <lustre/lustre_idl.h>
387 from stardard include directories
389 Severity : enhancement
391 Description: The prune-icache-use-trylock is no longer needed now that
392 the patch from bug 20008 is landed.
396 Description: The shrink grant feature is still active on the client although the
397 connect flag is not set.
401 Description: Don't leak grant space if the write failed with quota exceeded.
405 Description: Don't consume grant space twice on recoverable resent.
409 Description: a race condition could lead to SIGBUS being sent to an
410 application using mmap-ped files from Lustre
411 Details : truncate_complete_page implementation for the patchless
412 client could arbitrarily unset PG_Uptodate flag for a
413 page being kicked from the page cache, an uptodate check
414 right after a readpage call in filemap_fault could fail
415 because of that as though the page read had been unsuccessful.
419 Description: dlm lock slab shrinking is not efficient
420 Details : The dlm_locks slab can grow significantly and consumes a lot of
421 memory on the server. Set a hardlimit to grant_plan.
425 Description: Lustre does not do 1MB IOs to HW RAID
426 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
427 This is what we do already for SLES kernels.
431 Description: bump maximum number of phys/hw segments in the SLES11 kernel
432 until s/g chaining works properly.
437 Description: LSI Fusion MPT driver hacks to improve performance
438 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
440 Severity : enhancement
442 Description: increase default md stripe_cache_size to 16k
445 Bugzilla : 15587/21439
446 Description: don't handle security.capability xattr
447 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
448 This results in additional getxattr calls, causing VBR test
449 failures as well as a preformance drop when writing.
453 Description: obdfilter-survey is no longer working
454 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
455 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
459 Description: ll_shrink_cache does not handle __GFP_FS properly
463 Description: lfs getstripe shows wrong info for directories
464 Details : Set correct LOVEA default values for filesystem-wide.
468 Description: FSX checksum false positves due to mmap IO
469 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
470 checksum errors, if the flag is set on a request.
474 Description: file operations after eviction have successful return values
475 Details : use vfs ->flush callback to return any pending async errors
480 Description: mdsrate fails to write after 1.3+M files opened
481 Details : decrease memory usage on clients by recycling dentries and
486 Description: obdfilter-survey gives unreasonably high numbers
487 Details : Wait for all threads to complete when running test_brw.
491 Description: do not set lustre read_only device when server umount and keep
492 client records for recoverable ones
496 Description: move sync_on_lock_cancel tunable to the obdfilter layer
497 Details : move the tunable to trigger a journal flush on lock cancel from
498 the ost layer to the obdfilter layer. This tunable is useful
499 when using the async journal commit feature.
503 Description: exp->exp_nid_stats == NULL in filter_tally()
504 Details : fix race with per-nid stats by delaying procfs cleanup until
509 Description: extent lock cancellation on client can keep the cpu busy for too
514 Description: Do not fail OST activation when a llog is not found, just
515 issue an error message.
519 Description: Don't enable extents by default for MDT.
523 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
524 the AT code can access it concurrently while sending early
529 Description: Disable lockless truncate by default since it is sometimes flawed
530 and causes the write_disjoint test to fail.
534 Description: OSSs which don't have the patch from bug 20278 can trigger an
537 Severity : enhancement
539 Description: don't print message to the console when we have not managed to
544 Description: The MDS fails to synchronize OSTs which registered with the MGS
545 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
546 raised too early and thus discarded by the MDT stack.
547 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
552 Description: Fix race when the ping evictor and a service thread execute
553 target_recovery_check_and_stop() concurrently.
557 Description: quota broadcast can trigger a LBUG on the MDT if there are
562 Description: Resetting the lov_objid values to last_id reported by the OST
563 during orphan recovery is incorrect and can cause the same
564 objects to be allocated twice.
566 Severity : enhancement
568 Description: "weak-modules" support
569 Details : Implement "weak-modules" support which enables kernel modules
570 to be used with any kernel that implements the same kABI. In
571 order to achieve this modules are now installed in
572 /lib/modules/$(uname -r)/updates/kernel on all distributions.
574 Severity : enhancement
576 Description: add writeconf as mount option
578 Severity : enhancement
580 Description: produce debuginfo packages for SLES.
582 Severity : enhancement
584 Description: add failover nidlist to the import proc file.
586 Severity : enhancement
588 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
589 clients and 2.0 servers.
591 Severity : enhancement
593 Description: lfs find -s does not work correctly because of a bug in
598 Description: ll_read_ahead_page() must validate the dlm lock before using
603 Description: Prevent failover nids from registering with MGS first.
607 Description: fix lock inversion in ll_setattr_raw().
611 Description: object allocation is not balanced across OSTs.
612 Details : osc_precreate() should return 0, if there are enough objects left.
614 -------------------------------------------------------------------------------
616 2010-04-30 Oracle, Inc.
618 * Support for kernels:
619 2.6.16.60-0.42.8 (SLES 10),
620 2.6.27.39-0.3.1 (SLES11),
621 2.6.18-164.11.1.el5 (RHEL 5)
622 2.6.18-164.11.1.0.1.el5 (OEL 5)
623 * Client support for unpatched kernels:
624 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
625 2.6.16 - 2.6.30 vanilla (kernel.org)
626 * Recommended e2fsprogs version: 1.41.10-sun2
627 * The async journal commit feature (bug 19128) and the cancel
628 lock before replay feature (bug 16774) are disabled by default.
632 Description: open-unlinked directories trigger MDS LBUG
633 Details : Fix regression introduced by the patch from bug 19640.
634 ext3_inc_count() can reset nlink to 1 when the directory
635 is indexed and inode->i_nlink == 2. Work around the problem
636 by incrementing nlink by 2 instead of 1.
640 Description: Reconnects are not throttled
641 Details : Don't wake up pinger on reconnect failures and rely on regular
642 pings to trigger the next reconnection. Please note that the
643 pinger already uses a smaller interval if the import is
647 Frequency : only with NFS export
649 Description: Console flooded with error message from ll_inode_from_lock()
651 Details : in mds_open, initialize the child_res_id before enqueuing
652 the OPEN lock for the child inode, then to avoid senting
653 wrong ldlm_res_id to client.
657 Description: allow multiple instances of the same nid in NID hash
658 Details : Case of multiple separate clients from the same NID (as
659 with liblustre) is legitimate and so we should allow
660 multiple instances of the same NID in nid hash.
664 Description: the readahead code can sleep on a semaphore while holding a
666 Details : in ras_update, "lov_get_info" could be called during increasing
667 readahead windows, which tries to get the mutex lock "lov_lock"
668 while holding the spin_lock "ras_lock", then causes system
673 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
674 Details : fix assertion failure in the grant code.
678 Description: Use CNETERR (which is rate limited) in specific places in
679 the portal's LNET driver to avoid flooding the console.
683 Description: include last created object in precreate slow case
687 Description: don't do rep-ack if not created anything
688 Details : mds_open currently always put a lock into a rep-ack regardless
689 if something was created or not. This is pointless and only
690 creates needless contention. In fact the entire idea was to do
691 this for real creates as a recovery protection.
695 Description: Spurious error messages from smp_processor_id() on preemptible
697 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
698 first. The function fs_trace_get_tcd() was moved up.
702 Description: interval_erase() fix
703 Details : interval_erase() calls update_maxhigh() properly when child
708 Description: "lfs df" does not print stats for all mountpoints
709 Details : Print all mounted lustre filesystems with "lfs df"
713 Description: lfs setstripe -p not longer work with a relative pathname
714 Details : Use realpath() to provide absolute pathname.
718 Description: fix for truncated reply buffer
719 Details : reply buffer could be referred by reply_in_callback after released
724 Description: lustre.lov error when backing up symlinks with extended attributes
725 Details : Improved logic in ll_listxattr()
729 Description: properly handle null value for setattr -n lustre.lov
730 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
731 in ll_setxattr() due to no checking if "value" is NULL.
732 This command now resets to the default striping when executed
737 Description: stack overflow on lock cancellation due to fsync call
738 Details : sync_on_lock_cancel is needed for recovery when async journal
739 is enabled, but we actually just need to make sure that
740 metadata blocks have hit the journal, so doing a fs sync
741 should be enough and should consume less stack (just create an
742 empty handle and commmit it).
746 Description: using current->journal_info to store per-thread data leads
747 to problem under memory pressure
748 Details : disable the per-thread data (current->journal_info) containing
749 the lock info during I/O to work around the issue for short tem
753 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
754 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
755 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
760 Description: Cannot send after transport shutdown
761 Details : Clear imp_vbr_failed flag upon eviction
765 Description: soft lock in request set code during recovery
766 Details : during recovery, uses req->rq_set itself to replay the request
767 instead of ptlrpcd_recovery_pc
771 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
772 Details : During an ll_readahead under ll_readpage, we have seen the the
773 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
774 a page that is already locked by the readahead code.
779 Description: stop waitting for next replay transno if shutdown
780 Details : if the system is shutting down, wake up service thread blocked
781 to wait for next replay transno during recovery, then all the
782 references held by queued requests can be dropped and device
787 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
788 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
792 Description: per-nid stats should not access lustre-hash internal structures
797 Description: mount.lustre fails to pass some options to mount()
802 Description: ext4 extent allocation is slower than in ext3
803 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
804 enlarge ext4 preallocation table for 2048 4K blocks extents
809 Description: incorrect triggering of synchronous IO
810 Details : The OSC can mistakenly fall back to synchronous IO when the
811 max_dirty_mb limit is reached and no write requests have yet
812 been issued. This can occur when the dirty pages are spread
813 over many files all of which are below the optimal request size.
817 Description: Optimize quota_ctl operations by sending requests in parallel
818 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
819 Compiled from two attachments in the ticket.
823 Description: ordering issue between transaction start & i_mutex
824 Details : start the transaction earlier in llog_lvfs_destroy to get
825 transaction start and inode mutex lock nested properly.
829 Description: lru resize SLV can get stuck
830 Details : calculate SLV with a greater precision to not lose small
831 changes due to interger math truncation; round up SLV only
832 if the amount of granted locks less than the limit to not
833 get stuck with this SLV
837 Description: avoid divide-by-zero in lprocfs_rd_import()
841 Description: lfs quota failed when OSTs are down
842 Details : really return approximate block/inode usage when OSTs are down
846 Description: llobdstat fix and enhancement
847 Details : add a counter to set a limit to how many samples will be returned
848 fix a wildcard in the path to limit to obdfilter stats only
852 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
854 Details : abort lock enqueue processing sooner when we umount is in progress.
856 Severity : enhancement
858 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
859 2.6.18-164.11.1.0.1.el5.
861 Severity : enhancement
863 Description: error message improvements
864 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
865 Simplify MDT/OST service start message.
866 Suppress "changing the import ..." warning.
869 Bugzilla : 21961/17914
870 Description: ignore trailing -mdc when determining index number
873 Frequency : only with SLES11
875 Description: fix for a race condition in linux quotas implementation
876 Details : dq_flags(struct dquot) access is not properly locked which could
877 lead to certain inconsistencies when accessing it using non-atomic
878 bit operations like __set_bit in do_set_dqblk.
882 Description: return any pending async errors in close(2) using flush callback
884 -------------------------------------------------------------------------------
886 2010-01-29 Sun Microsystems, Inc.
888 * Support for kernels:
889 2.6.16.60-0.42.8 (SLES 10),
890 2.6.27.39-0.3.1 (SLES11),
891 2.6.18-164.11.1.el5 (RHEL 5)
892 2.6.18-164.6.1.0.1.el5 (OEL 5)
893 * Client support for unpatched kernels:
894 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
895 2.6.16 - 2.6.30 vanilla (kernel.org)
896 * Recommended e2fsprogs version: 1.41.6.sun1
897 * The async journal commit feature (bug 19128) and the cancel
898 lock before replay feature (bug 16774) are disabled by default.
900 Severity : enhancement
902 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
903 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
905 Severity : enhancement
906 Bugzilla : 21511/19848
907 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
909 Severity : enhancement
911 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
913 Severity : enhancement
915 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
916 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
918 Severity : enhancement
920 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
923 Severity : enhancement
925 Description: Vanilla kernel 2.6.30 patchless client support.
930 Description: bad entry in directory xxx: inode out of bounds
931 Details : fix locking issue in the rename path which could race with any
932 other operations updating the same directory.
934 Severity : enhancement
936 Description: Make watchdog timer messages to be more clear and descriptive.
940 Description: cp -p command does not preserve the dates and timestamp
941 Details : mtime could be spoiled by a write callback
945 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
949 Description: Allow non-root access for "lfs check".
950 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
952 Severity : enhancement
954 Description: quotacheck performance/scaling issues
955 Details : reduce quotacheck time on empty filesystem by skipping uninit
958 Severity : enhancement
960 Description: Enhancement for lfs(1) command to use numeric uid/gid.
962 Severity : enhancement
964 Description: Adjust locks' extents on their first enqueue, so that at the time
965 they get granted, there is no need for another pass through the
966 queues since they are already shaped into the proper forms.
970 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
971 arguments and prevent LBUG() in lustre_shrink_reply_v2().
975 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
976 exactly the specified mount options are used. Leaving off
977 any "mandatory" mount options is an error. Leaving off any
978 default mount options causes a warning, but is allowed.
979 Change errors=remount-ro from mandatory to default. Sanitize
980 the mount string before storing it. Update man pages accordingly.
984 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
985 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
988 Severity : enhancement
990 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
992 Severity : enhancement
994 Description: Clients can replay thousands of unused locks during recovery
995 Details : Don't replay unused locks (only read locks for now) during
996 recovery. This feature is disabled by default and can be
997 enabled by running the following command on the clients:
998 lctl get_param ldlm.cancel_unused_locks_before_replay
1002 Description: can't stat file in some situation.
1003 Details : improve initialize osc date when target is added to mds and
1004 ability to resend too big getattr request is client isn't have info
1009 Description: Prevent inconsistences between linux and lustre mount structures.
1010 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1011 Make the sleep interruptible.
1013 Severity : enhancement
1015 Description: Communicate OST degraded/readonly state via statfs to MDS
1016 Details : Flags in the statfs returned from OSTs indicate whether the
1017 OST is in a degraded RAID state, or if the filesystem has
1018 turned read-only after a filesystem error is detected.
1023 Description: don't panic if EPROTO was hit when reading symlink
1024 Details : correctly handling request reference in error cases.
1029 Description: open sometimes returns ENOENT instead of EACCES
1030 Details : checking permission should be part of open part of mds_open, not
1031 lookup part. so server should be set DISP_OPEN_OPEN disposition
1032 before starting permission check.
1033 Also not need revalidate dentry if client already have LOOKUP lock.
1037 Description: enable client interface failover
1038 Frequency : on servers with multiple network interfaces
1039 Details : When a child reconnects from another NID, properly update export
1040 nid hash position and ldlm reverse import.
1042 Severity : enhancement
1044 Description: implemented direct I/O with arbitrary (nonaligned) memory
1045 addresses and file offsets.
1047 Severity : enhancement
1049 Description: added more recovery timeout options.
1051 Severity : enhancement
1053 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
1058 Description: Avoid deadlock for local client writes
1059 Frequency : only on systems with clients writing to an OST on the same node
1060 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1061 memory freeing context. This allows OST threads to set the
1062 PF_MEMALLOC flag on task structures in order to allocate memory
1063 from reserved pools and complete IO.
1064 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1065 so that the OST threads generate memory pressure and allow
1066 inactive pages to be reclaimed.
1071 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1072 Details : move ldlm namespace creation in setup phase to avoid grab
1073 _lprocfs_lock with cli_sem held.
1077 Frequency : only during format of test systems
1078 Description: Unable to run several mkfs.lustre on loop devices at the same time
1079 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1080 formatting. The solution is to proper handle the error.
1082 Severity : enhancement
1084 Description: implement async create (obd_async_create) method for osc, to avoid
1085 too long waiting new ost objects with holding ldlm lock.
1089 Frequency : occasionally during network problems
1090 Description: client not allowed to reconnect to OST because of active request
1091 Details : abort bulk requests received by the OST once the client has timed
1092 out since the client will resend the request anyway. The client
1093 also now retries to reconnect to the same server if a connect
1094 request failed with EBUSY or -EAGAIN.
1097 Frequency : rare, if used wide striped file and one ost in down.
1099 Descriptoin: don't return error if we created a subset of objects for file.
1100 Details : lov_update_create_set() uses set->set_success as index for created
1101 objects, so if some requests failed, they will have hole at end of
1102 array and we can use qos_shrink_lsm for allocate correct lsm.
1106 Description: Slow stale export processing during normal start up
1107 Details : The global mgc lock prevents OST setup to be run in parallel.
1108 Replace the global lock with a per-config_llog_data semaphore.
1112 Description: Out or order replies might be lost on replay
1113 Details : In ptlrpc_retain_replayable_request if we cannot find retained
1114 request with tid smaller then one currently being added, add it
1115 to the start, not end of the list.
1119 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
1120 Details : add cond_resched() calls to avoid hogging the cpu for too long
1121 in the hash code. Make also lustre_hash_for_each_empty() more
1124 Severity : enhancement
1126 Description: Performance improvements for debug messages with D_RPCTRACE,
1127 D_LDLM, D_QUOTA options.
1130 Frequency : only with NFS export
1132 Description: (lov_merge.c:74:lov_merge_lvb())
1133 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
1134 Details : Fix a race in the nfs export code by populating inode
1135 info while the new inode is still locked
1137 Severity : enhancement
1139 Description: add a new file in procfs called force_lbug. Writting to this
1140 file triggers a LBUG. Only for test purpose.
1144 Description: OOM killer causes node hang
1145 Details : really interrupt the sleep in osc_enter_cache on signals
1149 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1150 Details : fix race during quota release on the slave.
1152 Severity : enhancement
1154 Description: smaller hash bucket sizes, cleanups
1155 Details : increase hash table sizes and enabled rehashing for pools, quota,
1156 uuid, nid & per-nid stats.
1158 Severity : enhancement
1160 Description: Add ldiskfs maxdirsize mount option
1161 Details : add max_dir size mount option
1165 Description: panic in ll_statahead_thread
1166 Details : prevent parent thread to be killed before its child
1169 Frequency : only with 16TB device
1171 Description: unable to perform "mount -t lustre" of 16TB OST device
1172 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
1176 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
1177 Details : unregistering should be zero if no RPC inflight.
1181 Description: hyperion: Oops during metabench
1182 Details : Correct the refcount of lov_request_set
1184 Severity : enhancement
1186 Description: Add mptlinux and nxge drivers to Lustre builds
1188 Severity : enhancement
1190 Description: Fix watchdog timer message to be more clear
1191 Details : Make watchdog timer messages more clear and descriptive.
1195 Description: LNET soft lockups in socknal_cd thread
1196 Details : don't hog CPU for active-connecting if another connd is
1197 accepting connecting-requst from the same peer
1201 Description: recovery-small test_17 hang
1202 Details : Land several AT improvements & fixes.
1206 Description: MDS panic and hanging client processes
1207 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
1211 Description: OSS stuck in recovery.
1212 Details : fix race during recovery. class_unlink_export,
1213 class_set_export_delayed and target_queue_last_replay_reply
1214 may race while increasing/decreasing obd_recoverable_clients
1215 and obd_delayed_clients, causing recovery to wait forever.
1217 Severity : enhancement
1219 Description: add cascading_rw.c to lustre/tests
1223 Description: filter_last_id() NULL deref
1224 Details : lprocfs_filter_rd_last_id() should check for the fully
1225 setup obd device, before proceeding further.
1227 Severity : enhancement
1229 Description: Loadgen improvements
1230 Details : stacksize and locking fixes for loadgen
1234 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
1235 Details : The atomic_read() allowing the atomic_inc() are not covered
1236 by a lock. Thus they may safely race and trip this CERROR()
1237 unless we add in a small fudge factor (+1).
1239 Severity : enhancement
1241 Description: shrink_slab: nr=-9223362083340912175
1242 Details : fix spurious message from shrink_slab reporing negative nr
1246 Description: Quiet bogus previously committed transno error
1247 Details : suppress the "server went back in time" error message which
1248 is always printed even in the common case after a client eviction
1250 Severity : enhancement
1252 Description: Parallel statfs() calls result in client eviction
1253 Details : cache statfs data for 1s.
1257 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
1259 Details : fix serveral issues in pinger code causing clients not to ping
1260 servers for too long, resulting in evictions.
1264 Description: e2fsck should warn when MMP update interval is extended
1265 Details : print mmp_check_interval and make it possible to abort mount
1266 operation in case it takes too long.
1270 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
1271 Details : fix bug in the RHEL5's jbd2 callback patch.
1275 Description: drop number of active requests when queued for recovery
1276 Details : Now that we take a reference on the original request instead of
1277 making a copy of it for recovery. We need to drop the number of
1278 active requests or the queued requests will prevent all request
1279 processing when they exceed (srv->srv_threads_running - 1).
1281 Severity : enhancement
1283 Description: refuse to invalidate operational quota files when they are in use
1284 Details : an attempt to invalidate operational quota files on the quota
1285 master is not actually permitted by VFS (returning -EPERM), but we
1286 should not depend on that and should return the error earlier.
1290 Description: Applications stuck in jbd2_log_wait_commit during exit
1291 Details : fix deadlock between kjournald2 trying to acquire the page lock
1292 owned by an ost_io thread waiting for journal commit.
1294 -------------------------------------------------------------------------------
1296 2009-10-16 Sun Microsystems, Inc.
1298 * Support for kernels:
1299 2.6.16.60-0.42.4 (SLES 10),
1300 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1301 2.6.18-128.7.1.el5 (RHEL 5),
1302 2.6.18-128.7.1.el5 (OEL 5),
1303 * Client support for unpatched kernels:
1304 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1305 2.6.16 - 2.6.27 vanilla (kernel.org)
1306 * Recommended e2fsprogs version: 1.41.6.sun1
1307 * ext4 support for RHEL5 is experimental and thus should not be
1310 Severity : enhancement
1312 Description: Add OEL5 support.
1314 Severity : enhancement
1316 Description: Update kernel to SLES11 2.6.27.29-0.1.
1320 Description: File checksum failures with OST read cache on
1321 Details : Disable page poisoning when the bulk transfer has to be aborted
1322 because the client got evicted.
1326 Description: Don't allow make backward step on assiging osc next id.
1327 Details : race between allocation next id and ll_sync thread can be cause
1328 of set wrong osc next id and can be kill valid ost objects.
1330 Severity : enhancement
1332 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1334 Severity : enhancement
1336 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1340 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1344 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1348 Description: Error handling in osc_statfs_interpret() has been improved.
1349 Details : Check in osc_statfs_interpret() for EBADR.
1353 Description: Do not send statfs() requests to OSTs disabled by administrator.
1354 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1358 Description: Do not update ctime for the deleted inode.
1359 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1363 Description: Increase of the size of the LDLM resource hash.
1364 Details : Bump up RES_HASH_BITS=12.
1368 Description: correctly send lsm on open replay
1369 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1374 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1375 Details : filter_destroy() does not hold the DLM lock over the whole
1376 operation. If the DLM lock is dropped, filter_commitrw() can go
1377 through, causing the deadlock between page lock and i_mutex.
1378 The i_alloc_sem should also be hold in filter_destroy() while
1379 truncating the file.
1383 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1384 Details : ldiskfs_truncate calls grab_cache_page which may start page
1385 allocation under an open transaction. This may lead to
1386 calling prune_icache with consequent lustre reentrance.
1389 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1390 still up and when the OST pool feature is used
1392 Description: interop testing got LBUG when run dd with OST pool
1393 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1394 Details : down/upgrading the MDS to a version that doesn't/does support OST
1395 pool can cause clients to crash because the lsm has changed
1400 Description: missing tree_status on 1.8.1 RPM build
1401 Details : make rpms failed due because the tree_status file is missing.
1405 Description: continuing LustreError "mds adjust qunit failed!"
1406 Details : don't print message on the console when ->adjust_qunit fails.
1410 Description: don't increase ldlm timeout if previous client was evicted
1411 Details : if a client doesn't respond to a blocking callback within the
1412 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1413 when the lock is next granted.
1417 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1418 affects recovery negatively.
1419 Details : make sure all exports have been properly destroyed by the zombie
1420 thread processed before stopping the target.
1424 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1425 Details : disable write barrier for ext4/SLES11.
1429 Description: Kernel panic - not syncing: Out of memory and no killable
1430 processes... on OSS when iozone
1431 Details : fix memory leak in the journal checksum patch.
1435 Description: group quota "too many blocks" OSS crashes
1436 Details : we should keep the same uid/gid for lquota_chkquota() and
1437 lquota_pending_commit()
1441 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1442 Details : don't LBUG on release quota error. Just a workaround until the
1443 problem is understood.
1445 ------------------------------------------------------------------------------
1446 2009-07-31 Sun Microsystems, Inc.
1448 * Support for kernels:
1449 2.6.16.60-0.39.3 (SLES 10),
1450 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1451 2.6.18-128.1.14.el5 (RHEL 5),
1452 * Client support for unpatched kernels:
1453 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1454 2.6.16 - 2.6.27 vanilla (kernel.org)
1455 * Recommended e2fsprogs version: 1.41.6.sun1
1456 * File join has been disabled in this release, refer to Bugzilla 16929.
1457 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1458 of Lustre filesystem with 4K stack may cause a stack overflow. For
1459 more information, please refer to bugzilla 17630.
1460 * ext4 support for RHEL5 is experimental and thus should not be
1463 Severity : enhancement
1465 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1468 Frequency : with 1.8 server and 1.6 clients
1470 Description: correctly shrink reply for avoid send too big message to client.
1471 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1472 some problems with sending this reply to 1.6 client.
1476 Description: Repeated atomic allocation failures.
1477 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1478 to generate memory pressure and allow reclaiming of inactive pages.
1479 At the same time, do not allow to exhaust emergency pools.
1480 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1482 Severity : enhancement
1483 Bugzilla : 19846, 18289
1484 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1486 Severity : enhancement
1487 Bugzilla : 19625, 16893, 18668, 19848
1488 Description: Add support for SLES11 2.6.27.23-0.1.
1490 Severity : enhancement
1492 Description: Update client support to vanila kernels up to 2.6.27.
1494 Severity : enhancement
1496 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1498 Severity : enhancement
1500 Description: Compile with -Werror by default for i686 and x86_64.
1504 Description: resolve race between obd_disconnect and class_disconnect_exports
1505 Details : if obd_disconnect will be called to already disconnected export he
1506 forget release one reference and osc module can't unloaded.
1508 Severity : enhancement
1510 Description: move AT tunable parameters for more consistent usage
1511 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1515 Descriptoin: correctly skip time estimate if in recovery
1516 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1520 Descriptoin: OSS DeadLock
1521 Details : Use trylock to prevent deadlock when shrink icache.
1523 Severity : enhancement
1525 Description: Allow tuning service thread via /proc
1526 Details : For each service a new
1527 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1528 created that can be used to set min/max thread counts, and get the
1529 current number of running threads.
1531 Severity : enhancement
1533 Description: Add state history info file, enhance import info file
1534 Details : Track import connection state changes in a new osc/mdc proc file;
1535 add overview-type data to the osc/mdc import proc file.
1539 Description: Reduce small size read RPC
1540 Details : Set read-ahead limite for every file and only do read-ahead when
1541 available read-ahead pages are bigger than 1M to avoid small size
1546 Description: free_entry erroneously used groups_free instead of put_group_info
1548 Severity : enhancement
1550 Description: Make read-ahead stripe size aligned.
1552 Severity : enhancement
1554 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1557 Frequency : rare, connect and disconnect target at same time
1559 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1560 Details : don't call obd_disconnect under lov_lock. this long time
1561 operation and can block ptlrpcd which answer to connect request.
1564 Frequency : start MDS on uncleanly shutdowned MDS device
1566 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1567 Details : stay in waiting mds<>ost recovery finished produce random bugs
1568 due race between two ll_sync thread for one lov target. send
1569 ACTIVATE event only if connect realy finished and import have
1573 Frequency : start MDS on uncleanly shutdowned MDS device
1575 Descriptoin: aborting recovery hang on MDS
1576 Details : don't throttle destroy RPCs for the MDT.
1580 Description: Slow reads beyond 8Tb offsets.
1581 Details : Page index integer overflow in ll_read_ahead_page
1585 Description: Soft lockup on OSS after MDS failover
1586 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1587 As a conseqence, the patch from bug 18224 is not operational
1588 and the MDS export cannot be reused on the OSTs until it gets
1592 Frequency : rare, only if using MMP with Linux RAID
1594 Description: MMP doesn't work with Linux RAID
1595 Details : While using HA for Lustre servers with Linux RAID, it is possible
1596 that MMP will not detect multiple mounts. To make this work we
1597 need to unplug the device queue in RAID when the MMP block is being
1598 written. Also while reading the MMP block, we should read it from
1599 disk and not the cached one.
1602 Frequency : rare, during recovery
1604 Description: Assertion failure in ldlm_lock_put
1605 Details : Do not put cancelled locks into replay list, hold references on
1606 locks in replay list
1610 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1611 Details : create_count always drops to the min value (=32) because grow_count
1612 is being changed before the precreate RPC completes.
1615 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1618 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1619 Details : add proper locking for creation of jbd_4k slab cache
1623 Description: MMP check in ext3_remount() fails without displaying any error
1624 Details : When multiple mount protection fails during remount, proper error
1629 Description: Rare Client crash on resend if the file was deleted.
1630 Details : When file is opened, but open reply is lost and file is
1631 subsequently deleted before resend, resend processing logic
1632 breaks trying to open the file again, should not try to open.
1636 Description: add check for >8TB ldiskfs filesystems
1637 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1638 Don't allow >8TB ldiskfs filesystems to be mounted without
1639 force_over_8tb mount option
1643 Description: Client locked up when running multiple instances of an app. on
1644 multiple mount points
1645 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1646 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1650 Description: Cannot acces an NFS-mounted Lustre filesystem
1651 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1652 from a Lustre-client exporting the Lustre filesystem via NFS.
1656 Description: panic in ll_statahead_thread
1657 Details : grab dentry reference in parent process.
1660 -------------------------------------------------------------------------------
1662 tbd Sun Microsystems, Inc.
1664 * Support for kernels:
1665 2.6.16.60-0.31 (SLES 10),
1666 2.6.18-128.1.6.el5 (RHEL 5),
1667 2.6.22.14 vanilla (kernel.org)
1668 * Client support for unpatched kernels:
1669 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1670 2.6.16 - 2.6.22 vanilla (kernel.org)
1671 * Recommended e2fsprogs version: 1.40.11-sun1
1672 * File join has been disabled, refer to Bugzilla 16929.
1673 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1674 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1675 Lustre filesystem with 4K stack may cause a stack overflow. For more
1676 information, please refer to bugzilla 17630.
1678 Severity : enhancement
1680 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1682 Severity : enhancement
1684 Description: Update OFED release to 1.4.1 RC4
1686 Severity : major, only with big OST
1688 Description: Very poor metadata performance on Infiniband lustre configuration
1689 Details : OST object precreation becomes very slow on big OSTs. This is due
1690 to the ialloc patch spending too much time scanning groups.
1693 Frequency : during recovery
1695 Description: don't mix llog inodes with normal.
1696 Details : allocate inodes for log in last inode group
1701 Description: fix lqs' reference which won't be put in some situations
1702 Details : This patch fixes:
1703 1. In quota_check_common(), this function will check quota
1704 for user and group, but only send one return via "pending".
1705 In most cases, the pendings should be same. But that is not
1707 2. If quotaoff runs between lquota_chkquota() and
1708 lquota_pending_commit(), the same thing will happen too.
1709 That is why it comes:
1710 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1713 -------------------------------------------------------------------------------
1715 2008-12-31 Sun Microsystems, Inc.
1717 * Support for kernels:
1718 2.6.5-7.314 (SLES 9),
1719 2.6.9-67.0.22.EL (RHEL 4),
1720 2.6.16.60-0.31 (SLES 10),
1721 2.6.18-92.1.17.el5 (RHEL 5),
1722 2.6.22.14 vanilla (kernel.org)
1723 * Client support for unpatched kernels:
1724 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1725 2.6.16 - 2.6.22 vanilla (kernel.org)
1726 * Client support for unpatched kernels:
1727 we do not recommend using patchless RHEL4 clients with kernels
1728 prior to 2.6.9-55EL (RHEL4U5).
1729 * Recommended e2fsprogs version: 1.40.11-sun1
1730 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1731 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1732 removed cwd "./" (refer to Bugzilla 14399).
1733 * A new quota file format has been introduced in 1.6.5.
1734 The format conversion from prior releases is handled transparently,
1735 but releases older than 1.4.12/1.6.5 don't understand this new
1736 format. The automatic format conversion can be avoided by running
1737 the following command on the MDS:
1738 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1739 For more information, please refer to bugzilla 13904.
1740 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1741 The format conversion from prior releases is handled transparently,
1742 but releases older than 1.6.6/1.8.0 don't understand this new
1743 format. The automatic format conversion can be avoided by running
1744 the following commands on the MDS and OSS servers (for
1745 pre 1.4.12-1.6.5 quota files):
1746 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1747 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1748 or (for 1.4.12/1.6.5 quota files)
1749 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1750 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1751 For more information, please refer to bugzilla 13904.
1752 * Output of lfs quota has been made less detailed by default,
1753 old (verbose) output can be obtained by using -v option.
1754 * File join has been disabled in this release, refer to Bugzilla 16929.
1755 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1756 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1757 Lustre filesystem with 4K stack may cause a stack overflow. For more
1758 information, please refer to bugzilla 17630.
1760 Severity : enhancement
1762 Description: Caching OSS
1763 Details : introduce data caching on the OSS. The OSS now relies on the linux
1764 kernel page cache to keep recently accessed data in memory.
1765 It is worth noting that all write requests are still flushed
1766 synchronously as in lustre 1.6.
1768 Severity : enhancement
1770 Description: Version based recovery
1771 Details : introduce finer grained recovery able to detect transaction
1772 dependencies and can deal with transaction gaps caused by clients
1773 failing at the same time as the server.
1775 Severity : enhancement
1777 Description: Enable adaptive timeouts by default
1778 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1779 managed dynamically based on server load and should not need
1780 to be tuned manually based on cluster size. This allows Lustre
1781 to work under a wider variety of system sizes and loads, without
1782 unnecessarily causing lengthy recovery times.
1784 Severity : enhancement
1786 Description: Add OST Pools support
1787 Details : File striping can now be set to use an arbitrary pool of OSTs
1789 Severity : enhancement
1791 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1792 Details : allow skip disconnected ost for send statfs request and hide error
1796 Frequency : rare, on llog test 6
1798 Descriptoin: don't allow connect to already connected import
1799 Details : allowing connect to already connected import is hide connecting problem.
1802 Frequency : rare, on failed llog setup
1804 Descriptoin: don't leak obd reference on failed llog setup
1805 Details : for failed llog setup - mgc forget call class_destroy_import for
1806 client import, move destroy import to more generic place.
1811 Descriptoin: allow kill process which wait statahead result
1812 Details : for some reasons 'ls' can stick in waiting result from statahead,
1813 in this case need way for kill this process.
1816 Frequency : rare, at shutdown
1818 Descriptoin: panic at umount
1819 Details : llap_shrinker can be raced with killing super block from list and
1820 this produce panic with access to already freeded pointer
1825 Descriptoin: don't lose wakeup for imp_recovery_waitq
1826 Details : recover_import_no_retry or invalidate_import and import_close can
1827 both sleep on imp_recovery_waitq, but we was send only one wakeup
1833 Descriptoin: panic in mds_open
1834 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1839 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1841 Details : release lock refernce only after add page to pages list.
1844 Frequency : always with long access acl
1846 Descriptoin: mds can't pack reply with long acl.
1847 Details : mds don't control size of acl but they limited by reint/getattr
1852 Frequency : on remount
1854 Description: external journal device not working after the remount
1855 Details : clear dev_rdonly flag for external journal devices in
1861 Description: shutdown vs evict race
1862 Details : client_disconnect_export vs connect request race.
1863 if client will evicted at this time - we start invalidate
1864 thread without referece to import and import can be freed
1870 Description: shrink LOV EAs before replying
1871 Details : correctly adjust LOV EA buffer for reply.
1876 Description: don't skip ost target if they assigned to file
1877 Details : Drop slow OSCs if we can, but not for requested start idx.
1878 This means "if OSC is slow and it is not the requested
1879 start OST, then it can be skipped, otherwise skip it only
1880 if it is inactive/recovering/out-of-space.
1882 Severity : enhancement
1884 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1887 Frequency : rare, need acl's on inode.
1889 Description: client can't handle ost additional correctly
1890 Details : if ost was added after client connected to mds client can have
1891 hit lnet_try_match_md ... to big messages to wide striped files.
1892 in this case need teach client to handle config events about add
1893 lov target and update client max ea size at that event.
1895 Severity : enhancement
1897 Description: Update to sles9 kernel-2.6.5-7.314.
1899 Severity : enhancement
1901 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1904 Frequency : Create a symlink file with a very long name
1906 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1907 Details : If there is no extra space in the request for early cancels,
1908 ldlm_req_handles_avail() returns 0 instead of a negative value.
1913 Description: mds is deadlocked
1914 Details : in rare cases, inode in catalog can have i_no less than have parent
1915 i_no, this produce wrong order for locking during open, and
1916 parallel unlink can be lock open. this need teach mds_open to grab
1917 locks in resource id order, not at parent -> child order.
1919 Severity : enhancement
1921 Description: Add /proc entry for import status
1922 Details : The mdc, osc, and mgc import directories now have
1923 an import directory that contains useful import data for debugging
1924 connection problems.
1926 Severity : enhancement
1928 Description: Re-disable certain /proc logging
1929 Details : Enable and disable client's offset_stats, extents_stats and
1930 extents_stats_per_process stats logging on the fly.
1933 Frequency : Only on FC kernels 2.6.22+
1935 Description: oops in statahead
1936 Details : Do not drop reference count for the dentry from VFS when lookup,
1937 VFS will do that by itself.
1939 Severity : enhancement
1941 Description: Generic /proc file permissions
1942 Details : Set /Proc file permissions in a more generic way to enable non-
1943 root users operate on some /proc files.
1947 Description: Hitting mdc_commit_close() ASSERTION
1948 Details : Properly handle request reference release in
1949 ll_release_openhandle().
1953 Frequency : only patchless client
1954 Description: add workaround for race between add/remove dentry from hash
1956 Severity : enhancement
1958 Description: Allow OST glimpses to return PW locks
1962 Description: LBUG when llog conf file is full
1963 Details : When llog bitmap is full, ENOSPC should be returned for plain
1968 Description: Prevent import from entering FULL state when server in recovery
1972 Description: service mount cannot take device name with ":"
1973 Details : Only when device name contains ":/" will mount treat it as
1979 Description: replace ptlrpcd with the statahead thread to interpret the async
1980 statahead RPC callback
1984 Frequency : on recovery
1985 Description: I/O failures after umount during fail back
1986 Details : if client reconnected to restarted server we need join to recovery
1987 instead of find server handler is changed and process self eviction
1988 with cancel all locks.
1990 Severity : enhancement
1992 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1996 Description: Kernel BUG tries to release flock
1997 Details : Lustre does not destroy flock lock before last reference goes
1998 away. So always drop flock locks when client is evicted and
1999 perform unlock regardless of successfulness of speaking to MDS.
2001 Severity : enhancement
2003 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
2005 Severity : enhancement
2007 Description: Upcall on Lustre log has been dumped
2008 Details : Allow for a user mode script to be called once a Lustre log has
2009 been dumped. It passes the filename of the dumped log to the
2010 script, the location of the script can be specified via
2011 /proc/sys/lnet/debug_log_upcall.
2016 Description: avoid messages about idr_remove called for id that is not allocated
2017 Details : Move assigment s_dev for clustered nfs to end of initialization,
2018 for avoid problem with error handling.
2023 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
2024 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
2025 CONN_USED_HASH and this prodice warning when put connection again
2031 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
2032 Details : release reference to stats when client disconnected, not
2033 when export destroyed for avoid races when client destroyed
2034 after main ost export.
2038 Description: more cleanup in mds_lov
2039 Details : add workaround for get valid ost count for avoid warnings about
2040 drop too big messages, not init llog cat under semphore which
2041 can be blocked on reconnect and break normal replay, fix access
2044 Severity : enhancement
2046 Description: Export bytes_read/bytes_write count on OSC/OST.
2050 Description: Early reply size mismatch, MGC loses connection
2051 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
2052 the connect flags are properly negotiated.
2056 Description: Properly propagate oinfo flags from lov to osc for statfs
2057 Details : restore missing copy oi_flags to lov requests.
2061 Description: exports in /proc are broken
2062 Details : recreate /proc entries for clients when they reconnect.
2064 Severity : enhancement
2066 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
2067 : l_getgroups(8), lst(8), routerstat(8)
2068 Details : included man pages for llobdstat(8), llstat(8),
2069 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
2071 Severity : enhancement
2073 Description: Implement lustre ll_show_options method.
2075 Severity : enhancement
2077 Description: Update to SLES9 kernel-2.6.5-7.312.
2079 Severity : enhancement
2081 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
2085 Description: exports in /proc are broken
2086 Details : recreate /proc entries for clients when they reconnect.
2090 Description: don't fail open with -ERANGE
2091 Details : if client connected until mds will be know about real ost count
2092 get LOV EA can be fail because mds not allocate enougth buffer
2097 Description: Resolve device initialization race
2098 Details : Prevent proc handler from accessing devices added to the
2099 obd_devs array but yet be intialized.
2101 Severity : enhancement
2103 Description: configure's --enable-quota should check the
2104 : kernel .config for CONFIG_QUOTA
2105 Details : configure is terminated if --enable-quota is passed but
2106 : no quota support is in kernel
2108 Severity : enhancement
2110 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
2112 Severity : enhancement
2114 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
2118 Frequency : rare, on PPC clients
2119 Description: don't swab ost objects in response about directory, because
2121 Details : bug similar bug 14856, but in different function.
2123 Severity : enhancement
2125 Description: lfs quota tool enhancement
2126 Details : added units specifiers support for setquota, default to
2127 current uid/gid for quota report, short quota stats by
2128 default, nonpositional parameters for setquota, added
2129 llapi_quotactl manual page.
2131 Severity : enhancement
2133 Description: *optional* service tags registration
2134 Details : if the "service tags" package is installed on a Lustre node
2135 When the filesystem is mounted, a local-node service tag will
2136 be created. See http://inventory.sun.com/ for more information
2137 about the Service Tags asset management system.
2141 Description: Client runs out of low memory
2142 Details : Consider only lowmem when counting initial number of llap pages
2145 Frequency : occasional
2147 Description: add refcount for osc callbacks, so avoid panic on shutdown
2149 Severity : enhancement
2151 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
2154 Frequency : testing only
2156 Description: sanity test 65a fails if stripecount of -1 is set
2157 Details : handle -1 striping on filesystem in ll_dirstripe_verify
2160 Frequency : only in unusual configurations
2162 Description: Kernel panic with find ost index.
2163 Details : lov_obd have panic if some OST's have sparse indexes.
2165 Severity : enhancement
2167 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
2170 Frequency : rarely, if filesystem is mounted with -o flock
2172 Description: do not process already freed flock
2173 Details : flock can possibly be freed by another thread before it reaches
2174 to ldlm_flock_completion_ast.
2177 Frequency : rarely, if filesystem is mounted with -o flock
2179 Description: LBUG during stress test
2180 Details : Need properly lock accesses the flock deadlock detection list.
2183 Frequency : rarely, if binaries are being run from Lustre
2185 Description: oops in page fault handler
2186 Details : kernel page fault handler can return two special 'pages' in
2187 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
2190 Frequency : rarely, during shutdown
2192 Description: timeout with invalidate import.
2193 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
2194 should be handled by ptlrpcd. This produce long age waiting and
2195 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
2201 Description: ASSERTION(CheckWriteback(page,cmd)) failed
2202 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
2206 Frequency : only with broken builds/installations
2208 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
2209 Details : just return an error to a user, put a console error message
2211 Severity : enhancement
2213 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
2215 Severity : enhancement
2217 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
2219 Severity : enhancement
2221 Description: enable MGS and MDT services start separately
2222 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
2223 a co-located MGS without starting the MGS, which is a complement
2224 to 'nosvc' mount option.
2227 Frequency : always, on big-endian systems
2229 Description: cleanup in ptlrpc code, related to PPC platform
2230 Details : store magic in native order avoid panic's in recovery on PPC
2231 node and forbid from this error in future. Also fix posibily
2232 of twice swab data. Fix get lov striping to userpace.
2235 Frequency : rarely, if replay get lost on server
2237 Description: server incorrectly drop resent replays lead to recovery failure.
2238 Details : do not drop replay according to msg flags, instead we check the
2239 per-export recovery request queue for duplication of transno.
2242 Frequency : after recovery
2244 Description: precreate to many object's after del orphan.
2245 Details : del orphan st in oscc last_id == next_id and this triger growing
2246 count of precreated objects. Set flag LOW to skip increase count
2247 of precreated objects.
2250 Frequency : rare, on clear nid stats
2252 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
2253 Details : when clean nid stats sometimes try destroy live entry,
2254 and this produce panic in free.
2257 Frequency : occasionally since 1.6.4
2259 Description: Stack overflow during MDS log replay
2260 Details : ease stack pressure by using a thread dealing llog_process.
2263 Frequency : very rare
2265 Description: MDT cannot be unmounted, reporting "Mount still busy"
2266 Details : Mountpoint references were being leaked during open reply
2267 reconstruction after an MDS restart. Drop mountpoint reference
2268 in reconstruct_open() and free dentry reference also.
2273 Description: wait until IO finished before start new when do lock cancel.
2274 Details : VM protocol want old IO finished before start new, in this case
2275 need wait until PG_writeback is cleared until check dirty flag
2276 and call writepages in lock cancel callback.
2281 Description: mds_mfd_close() ASSERTION(rc == 0)
2282 Details : In mds_mfd_close(), we need protect inode's writecount change
2283 within its orphan write semaphore to prevent possible races.
2286 Frequency : rare, on shutdown ost
2288 Description: don't hit live lock with umount ost.
2289 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2290 use shrink_dcache_sb instead.
2293 Frequency : only when echo_client is used
2295 Description: don't panic with use echo_client
2296 Details : echo client pass NULL as client nid pointer and this produce NULL
2297 pointer dereference.
2300 Frequency : Always on 32-bit PowerPC systems
2302 Description: fix build on PPC32
2303 Details : compile code with -m64 flag produce wrong object file for PPC32.
2308 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2309 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2310 by a client, and the data on disk are dangerous and incorrect.
2311 Change them with client eviction.
2313 Severity : enhancement
2315 Description: skiplist implementation simplification
2316 Details : skiplists are used to group compatible locks on granted list
2317 that was implemented as tracking first and last lock of each lock
2318 group the patch changes that to using doubly linked lists
2322 Description: delete compatibility for 32bit qdata
2323 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2324 qunit. That means servers of b1_4 and servers of b1_8 can't be
2325 used together if users want to use quota.
2328 Frequency : only with administrator action
2330 Description: mount failure if config log has invalid conf_param setting
2331 Details : If administrator specified an incorrect configuration parameter
2332 with "lctl conf_param" this would cause an error during future
2333 client mounts. Instead, ignore the bad configuration parameter.
2336 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2338 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2339 Details : If blocks per group is less than blocksize*8, set rest of the
2343 Frequency : Application do stride read on lustre
2345 Description: The read performance will drop a lot if the application does
2347 Details : Because the stride_start_offset are missing in stride read-ahead,
2348 it will cause clients read a lot of unused pages in read-ahead,
2349 then the read-performance drops.
2353 Description: more ldlm soft lockups
2354 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2355 starve other threads from the resource lock for a long time in
2356 case of long waiting queue, so change the debug level from
2357 D_OTHER to the less frequently used D_INFO.
2359 Severity : enhancement
2361 Description: add -gid, -group, -uid, -user options to lfs find
2363 Severity : enhancement
2365 Description: ll_recover_lost_found_objs - recover objects in lost+found
2366 Details : OST corruption and subsequent e2fsck can leave objects in the
2367 lost+found directory. Using the "ll_recover_lost_found_objs"
2368 tool, these objects can be retrieved and data can be salvaged
2369 by using the object ID saved in the fid EA on each object.
2374 Description: this bug _only_ happens when inode quota limitation is very low
2375 (less than 12), so that inode quota unit is 1 at initialization.
2376 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2377 is effective now. So least quota qunit should be 2.
2381 Description: Hung threads in invalidate_inode_pages2_range
2382 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2383 one is completed. As a result, some RPCs are stuck in the queue
2388 Description: Procfs and llog threads access destoryed import sometimes.
2389 Details : Sync the import destoryed process with procfs and llog threads by
2390 the import refcount and semaphore.
2394 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2395 Details : Sort source/child resource pair after updating child resource.
2400 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2401 Details : If insertion of an extent fails, then discard the inode
2402 preallocation and free data blocks else it can lead to duplicate
2407 Description: don't always update ctime in ext3_xattr_set_handle()
2408 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2409 In some cases the ctime should not be updated, for example for
2410 2.0->1.8 compatibility it is necessary to delete an xattr and it
2411 should not update the ctime.
2415 Description: add quota statistics
2416 Details : 1. sort out quota proc entries and proc code.
2417 2. add quota statistics
2422 Description: quotas are not honored with O_DIRECT
2423 Details : all writes with the flag O_DIRECT will use grants which leads to
2424 this problem. Now using OBD_BRW_SYNC to guard this.
2428 Bugzilla : 15713/16362
2429 Description: Assertion in iopen_connect_dentry in 1.6.3
2430 Details : looking up an inode via iopen with the wrong generation number can
2431 populate the dcache with a disconneced dentry while the inode
2432 number is in the process of being reallocated. This causes an
2433 assertion failure in iopen since the inode's dentry list contains
2434 both a connected and disconnected dentry.
2438 Description: assertion failure in ldlm_handle2lock()
2439 Details : fix a race between class_handle_unhash() and class_handle2object()
2440 introduced in lustre 1.6.5 by bug 13622.
2442 Severity : enhancement
2444 Description: superblock lock contention with many SMP cores on one client
2445 Details : several client filesystem locks were highly contended on SMP
2446 NUMA systems with 8 or more cores. Per-CPU datastructures
2447 and more efficient locking implemented to reduce contention.
2452 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2453 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2458 Description: oops in sd_iostats_seq_show()
2459 Details : unloading/reloading the scsi low level driver triggers a kernel
2460 bug when trying to access the sd iostat file.
2465 Description: Kernel panics during QLogic driver reload
2466 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2467 patch, causing memory corruption.
2472 Description: journal_dev option does not work in b1_6
2473 Details : pass mount option during pre-mount.
2475 Severity : enhancement
2477 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2478 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2479 extent information of a file. It can be used to map logical blocks
2480 in a file to physical blocks in the block device.
2483 Frequency : only with adaptive timeout enabled
2485 Description: DEBUG_REQ() bad paging request
2486 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2487 because it can be accessed by reply_in_callback() without the
2491 Frequency : only on Cray X2
2493 Description: X2 build failures
2494 Details : fix build failures on Cray X2.
2498 Description: xid & resent requests
2499 Details : Initialize RPC XID from clock at startup (randomly if clock is
2504 Description: quota recovery deadlock during mds failover
2505 Details : This patch includes att18982, att18236, att18237 in bz14840.
2507 1. fix osts hang when mds does failover with quotaon
2508 2. prevent watchdog storm when osts threads wait for the
2513 Description: kernel panic on racer
2514 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2516 Severity : enhancement
2518 Description: Add lustre_start utility to start or stop multiple Lustre servers
2523 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2524 Details : In case of memory pressure, list_del() can be called twice on
2525 req->rq_history_list, causing a kernel oops.
2529 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2530 Details : fix stack overflow in the distributed lock manager by defering
2531 export eviction after a failed ast to the elt thread instead of
2532 handling it in the dlm interpret routine.
2534 Severity : enhancement
2536 Description: More exported tunables for mballoc
2537 Details : Add support for tunable preallocation window and new tunables for
2538 large/small requests
2542 Description: Detect corruption of block bitmap and checking for preallocations
2543 Details : Checks validity of on-disk block bitmap. Also it does better
2544 checking of number of applied preallocations. When corruption is
2545 found, it turns filesystem readonly to prevent further corruptions.
2549 Frequency : only for big-endian servers
2550 Description: Check for big-endian system while mounting fs with extents feature
2551 Details : Mounting a filesystem with extents feature will fail on big-endian
2552 systems since ext3-based ldiskfs is not supported on big-endian
2553 systems. Can be over-ridden with "bigendian_extents" mount option.
2557 Description: Excessive recovery window
2558 Details : With AT enabled, the recovery window can be excessively long (6000+
2559 seconds). To address this problem, we no longer use
2560 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2561 timeout no longer depends on the service time, it is set to
2562 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2563 time via pb_service_time.
2566 Descriptoin: Don't sync journal after every i/o
2567 Details : Implement write RPC replay to allow server replies for write RPCs
2568 before data is on disk. However, this feature is disabled by
2569 default since some issues leading to data corruptions have been
2570 found during recovery (e.g. bug 19128). This feature can be enabled
2571 by running the following command on the OSSs:
2572 lctl set_param obdfilter.*.sync_journal=0
2576 Description: Watchdog triggered on MDS failover
2577 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2578 the OSTs know that the MDS "UUID" can be reused for the same export
2579 from a different NID, so we do not need to wait for the export to
2584 Descriptoin: Lustre detected file system corruption with inode out of bounds
2585 Details : don't update i_size on MDS_CLOSE for directories. This causes
2586 directory corruptions on the MDT.
2588 -------------------------------------------------------------------------------
2590 2008-05-26 Sun Microsystems, Inc.
2592 * Support for kernels:
2593 2.6.5-7.311 (SLES 9),
2594 2.6.9-67.0.7.EL (RHEL 4),
2595 2.6.16.54-0.2.5 (SLES 10),
2596 2.6.18-53.1.14.el5 (RHEL 5),
2597 2.6.22.14 vanilla (kernel.org)
2598 * Client support for unpatched kernels:
2599 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2600 2.6.16 - 2.6.22 vanilla (kernel.org)
2601 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2602 we do not recommend using patchless RHEL4 clients with kernels
2603 prior to 2.6.9-55EL (RHEL4U5).
2604 * Recommended e2fsprogs version: 1.40.7-sun1
2605 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2606 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2607 removed cwd "./" (refer to Bugzilla 14399).
2608 * A new quota file format has been introduced in 1.6.5.
2609 The format conversion from prior releases is handled transparently,
2610 but releases older than 1.4.12/1.6.5 will not understand this new
2611 format. The automatic format conversion can be avoided by running
2612 the following command on the MDS before upgrading:
2613 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2614 For more information, please refer to bugzilla 13904.
2618 Description: quota performance fix
2619 Details : quota data is written in journalled mode instead of ordered to
2620 increase performance
2624 Description: lfs support for human-readable quota grace time strings
2625 Details : lfs setquota -t and lfs quota -t represent quota grace times
2626 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2629 Frequency : always with o2ib 1.3 and sles10
2631 Description: fix build with SLES10 and o2ib v3.
2632 Details : sles10 uses diffrent name for Module.symver file but configure
2633 assume this file has same name on RHEL/SLES/vanila kernels.
2636 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2638 Description: MDS may lose file striping (and hence file data) in some cases
2639 Details : If there are additional extended attributes stored on the MDS,
2640 in particular ACLs, SELinux, or user attributes (if user_xattr
2641 is specified for the client mount options) then there is a risk
2642 of attribute loss. Additionally, the Lustre file striping
2643 needs to be larger than default (e.g. striped over all OSTs),
2644 and an additional attribute must be stored initially in the
2645 inode and then increase in size enough to be moved to the
2646 external attribute block (e.g. ACL growing in size) for file
2649 Severity : enhancement
2651 Description: add message levels for liblustreapi
2656 Description: MDT cannot be unmounted, reporting "Mount still busy"
2657 Details : Mountpoint references were being leaked during open reply
2658 reconstruction after an MDS restart. Drop mountpoint reference
2659 in reconstruct_open() and free dentry reference also.
2664 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2665 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2666 will result in an -EAGAIN and in the exit call path we call
2667 alloc_rr() which will with increasing aggressiveness attempt to
2668 aquire precreated objects on the minimum number of required OSCs.
2672 Description: Use old size assignment to avoid deadlock
2673 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2674 the scheduling while holding a spinlock. We do not need locking
2675 for size in ll_update_inode() because size is only updated from
2676 the MDS for directories or files without objects, so there is no
2677 other place to do the update, and concurrent access to such inodes
2678 are protected by the inode lock.
2682 Description: Use __u64 instead of int for valid bits
2686 Description: resolve "_IOWR redefined" build error on SLES10
2690 Description: dump the memory debugging after all modules are unloaded to
2691 suppress false negative in conf_sanity test 39
2695 Description: the recovery timer never expires
2696 Details : for new client connect request, the recovery timer should not be
2697 reset, otherwise recovery timer will never expired, if the old
2698 client never come. Only old client connect and first connection
2699 req should trigger recovery timer reset.
2703 Description: the min numbers of lproc stats are wrong
2704 Details : adding a new constant LC_MIN_INIT and use it for initialization
2708 Frequency : always with interactive lfs
2710 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2713 Frequency : with multiple concurrent readdir processes in same directory
2714 Bugzilla : 15406, 15169, 15175
2715 Description: misc fixes for directory readahead.
2716 Details : prevent previous statahead async RPC callback from processing the
2717 current "statahead_info", race condition between async RPC callback
2718 add dentry into dentry hash table and "ls" thread revalidate such
2719 dentry, statahead his/miss control for hidden items, and so on.
2721 Severity : enhancement
2723 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2727 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2728 Details : several u32 variables declarations are replaced with u64 declarations
2733 Description: fix SLES kernel versioning
2734 Details : the kernel version for our SLES 10 kernel did not include a "-"
2735 before the "smp" at the end. while this was not a problem in
2736 general, it did mean that software trying to use the kernel
2737 version to try to detect a vendor specific kernel would fail.
2738 this was most evident by the OFED build scripts.
2743 Description: Don't update lov_desc members until making sure they are valid
2744 Details : When updating lov_desc members via proc fs, need fix their
2745 validities before doing the real update.
2748 Frequency : very rare
2750 Description: don't put request into delay list while invalidate in flight.
2751 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2752 import in flight. this produce timeout for invalidate and sometimes
2753 can cause stale data.
2755 Severity : enhancement
2757 Description: Update kernel to SLES9 2.6.5-7.311.
2759 Severity : enhancement
2761 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2766 Frequency : on PPC only
2767 Description: not convert ost objects for directory because it's not exist.
2768 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2770 Severity : enhancement
2772 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2773 __u64 as usingied long long type.
2776 Frequency : rare, on shutdown
2778 Description: race process ast vs remove callback
2779 Details : removing callback before disconnect import open race with
2780 processing callback.
2782 Severity : enhancement
2784 Description: Update kernel to SLES9 2.6.5-7.311.
2786 Severity : enhancement
2788 Description: Files open for execute are not marked busy on SLES10
2789 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2791 Severity : enhancement
2793 Description: Add server support for vanilla-2.6.22.14.
2796 Frequency : occasional
2798 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2799 Details : If an OST is down the MDS will hang indefinitely in
2800 obd_statfs_rqset() waiting for the statfs data. While for
2801 MDS QOS usage of statfs, it should not stuck in waiting.
2803 Severity : enhancement
2805 Description: Disable adaptive timeouts by default
2808 Frequency : on network error
2810 Description: panic with double free request if network error
2811 Details : mdc_finish_enqueue is finish request if any network error occuring,
2812 but it's true only for synchronus enqueue, for async enqueue
2813 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2817 Frequency : rare, on recovery
2819 Description: read procfs can produce deadlock in some situation
2820 Details : Holding lprocfs lock with send rpc can produce block for destroy
2821 obd objects and this also block reconnect with -EALREADY.
2822 This isn't fix all lprocfs bugs - but make it rare.
2824 Severity : enhancement
2826 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2829 Frequency : frequent on X2 node
2831 Description: mdc_set_open_replay_data LBUG
2832 Details : Set replay data for requests that are eligible for replay.
2837 Description: lustre_mgs: operation 101 on unconnected MGS
2838 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2839 MGC, and late on MGC cannot successfully connect to MGS and a lot
2840 of the error messages complaining that MGS is not connected.
2843 Frequency : rare, depends on device drivers and load
2845 Description: MDS or OSS nodes crash due to stack overflow
2846 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2847 In some cases, in conjunction with device drivers that use a lot
2848 of stack, the MDS (or possibly OSS) service threads could overflow
2849 the stack. One change which was identified to consume additional
2850 stack has been reworked to avoid the extra stack usage.
2852 Severity : enhancement
2854 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2856 Severity : enhancement
2858 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2860 Severity : enhancement
2862 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2864 Serverity : enhancement
2866 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2869 Frequency : rare on shutdown OST
2871 Description: Don't allow skipping OSTs if index has been specified.
2872 Details : Don't allow skipping OSTs if index has been specified, make locking
2873 in internal create lots better.
2878 Description: ASSERTION(!PageDirty(page)) failed
2879 Details : Wrong check could lead to an assertion failure under specific
2885 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2886 Details : access to bitfield in structure is always rounded to long
2887 and this produce problem with not atomic change any bit.
2892 Description: Lustre make rpms failed.
2893 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2894 builds Lustre rpms from tarball.
2896 Severity : enhancement
2898 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2901 Frequency : rare on shutdown OST
2903 Description: If llog cancel was not send before clean_exports phase, this can
2904 produce deadlock in llog code.
2905 Details : If llog thread has last reference to obd and call class_import_put
2906 this produce deadlock because llog_cleanup_commit_master wait when
2907 last llog_commit_thread exited, but this never success because was
2908 called from llog_commit_thread.
2911 Frequency : only if OST index is skipped
2913 Description: NULL lov_tgts causing MDS oops
2914 Details : more safe checks for NULL lov_tgts for avoid oops.
2916 Severity : enhancement
2918 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2920 Severity : enhancement
2922 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2927 Description: make mgs_setparam() handle fsname containing dash
2928 Details : fsname containing a dash does not work with lctl conf_param
2930 Severity : enhancement
2932 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2935 Frequency : rare, in recovery and (or) destroy lovobjid file.
2937 Description: rewrite lov objid code.
2938 Details : Cleanup for lov objid code, remove scability problems and wrong
2939 locking. Fix sending last_id into ost.
2941 Severity : enhancement
2943 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2945 Severity : enhancement
2947 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2948 Details : Use d_move_locked instead of __d_move.
2951 Frequency : rare, at shutdown
2953 Description: access already free / zero obd_namespace.
2954 Details : if client_disconnect_export was called without force flag set,
2955 and exist connect request in flight, this can produce access to
2956 NULL pointer (or already free pointer) when connect_interpret
2957 store ocd flags in obd_namespace.
2960 Frequency : only at startup
2962 Description: not alloc memory with spinlock held.
2963 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2964 if any spinlock held.
2969 Description: lfs find does not continue on file error
2970 Details : Continue other files processing when a file/dir is absent.
2974 Description: Inconsistent usage of lustre_pack_reply()
2975 Details : Standardize the usage of lustre_pack_reply() such that it
2976 always generate a CERROR on failure.
2979 Frequency : very rare
2981 Description: Fix replay if there is an un-replied request and open
2982 Details : In some cases, older replay request will revert the
2983 mcd->mcd_last_xid on MDS which is used to record the client's
2984 latest sent request.
2986 Severity : enhancement
2988 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2990 Severity : enhancement
2992 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2994 Severity : enhancement
2996 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2998 Severity : enhancement
3000 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
3002 Severity : enhancement
3004 Description: Build SLES10 patchless client fails
3005 Details : The configure was broken by run ./configure with
3006 --with-linux-obj=.... argument for patchless client. When the
3007 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
3008 can't search header adequately. Use absolute path such as
3009 -I($LINUX)/include instead.
3011 Severity : enhancement
3013 Description: Lustre Page Accounting
3014 Details : New macros for page alloc and free which enable accounting
3015 of page allocation of Lustre. Use percpu counters to store memory
3016 and page statistics.
3019 Frequency : only if debugging is disabled
3021 Description: LASSERT_{REQ,REP}SWAB macros are buggy
3022 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
3023 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
3024 these macros and replace them with their definitions instead.
3029 Description: interrupt oig_wait produce painc on resend.
3030 Details : brw_redo_request can be used for resend requests from ptlrpcd and
3031 private set, and this produce situation when rq_ptlrpcd_data not
3032 copyed to new allocated request and triggered LBUG on assert
3033 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
3034 ptlrpcd set if request is changed and can be safety changed to use
3037 Severity : enhancement
3039 Description: organize the server-side client stats on per-nid basis
3040 Details : Change the structure of stats under obdfilter and mds to
3048 The "uuid"s file would list the uuids of _active_ exports.
3049 And the clear entry is to clear all stats and stale nids.
3054 Description: Processes looping in ll_readdir() on Lustre clients finally causing
3055 a full node pseudo-hang
3056 Details : Concurrent access to the same directory from multiple clients with
3057 intensive file creation/removal can cause a client node to spin in
3058 ll_readdir(). i_version must be increased every time the lock is
3059 cancelled to ensure a revalidate is done.
3064 Description: touch file failed when fs is not full
3065 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
3066 otherwise we can get ENOSP while fs is not full.
3069 Frequency : only for Cray XT3
3070 Bugzilla : 12829/13455
3071 Description: Changing primary group doesn't change the group lustre assigns to
3073 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
3074 overridden with the primary group provided by the group upcall,
3075 whereas the supplied fsgid can be trusted if it is in the list of
3076 supplementary groups returned by the group upcall.
3078 Severity : enhancement
3080 Description: Root Squash Functionality
3081 Details : Implementation of NFS-like root squash capability. Specifically,
3082 don't allow someone with root access on a client node to be able
3083 to manipulate files owned by root on a server node.
3085 Severity : enhancement
3087 Description: Slow trucate/writes to huge files at high offsets.
3088 Details : Directly associate cached pages to lock that protect those pages,
3089 this allows us to quickly find what pages to write and remove
3090 once lock callback is received.
3095 Description: Too many locks accumulating on client during NFS usage
3096 Details : mds_open improperly used accmode to find out access mode to a
3097 file. Also mdc_intent_lock logic to find out if we already have
3098 lock similar to just received was flawed since introduction of
3099 skiplists - locks are now added to the front of the granted
3104 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
3106 Details : In obd setup failure handler, obd_stopping will not necessarily to
3107 be 1, and obd_set_up should also be checked to make sure whether
3108 obd is completely setup.
3110 Severity : enhancement
3112 Description: Allow masking D_WARNING, D_ERROR messages from console
3113 Details : Console messages can now be disabled via lnet.printk.
3118 Description: User code with malformed file open parameter crashes client node
3119 Details : Before packing join_file req, all the related reference should be
3120 checked carefully in case some malformed flags cause fake join_file
3126 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
3127 grain of quota doing harm to quota's accuracy
3128 Details : qunit size will be changed when quota limitation is too low/high;
3129 record the pending quota write in order to get more accureate
3130 quota; delete the patch for bug12588, which is unnecessary when
3131 this patch is landed. This bug also contains fixes for bug 14526,
3132 14299, 14601 and 13794.
3136 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
3137 Details : ldlm_completion_ast() assumes that a lock is granted when the req
3138 mode is equal to the granted mode. However, it should also check
3139 that LDLM_FL_CP_REQD is not set.
3143 Description: Heavy nfs access might result in deadlocks
3144 Details : After ELC code landed, it is now improper to enqueue any mds
3145 locks under och_sem, because enqueue might want to decide to
3146 cancel open locks for same inode we are holding och_sem for.
3150 Description: 35% write performance drop with ldiskfs2 when quotas are on
3151 Details : Enable ext3 journalled quota by default to improve performance
3152 when quotas are turned on.
3156 Description: Client eviction while running blogbench
3157 Details : A lot of unlink operations with concurrent I/O can lead to a
3158 deadlock causing evictions. To address the problem, the number of
3159 oustanding OST_DESTROY requests is now throttled to
3160 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
3161 callbacks are processed in priority.
3164 Frequency : RHEL4 only
3166 Description: mkfs is very slow on IA64/RHEL4
3167 Details : A performance regression has been discovered in the MPT Fusion
3168 driver between versions 3.02.73rh and 3.02.99.00rh. As a
3169 consequence, we have downgraded the MPT Fusion driver in the RHEL4
3170 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
3173 Frequency : PPC/PPC64 only
3175 Description: conflicts between asm-ppc64/types.h and lustre_types.h
3176 Details : fix duplicated definitions between asm-ppc64/types.h and
3177 lustre_types.h on PPC.
3180 Frequency : PPC/PPC64 only
3182 Description: asm-ppc/segment.h does not exist
3183 Details : fix compile issue on PPC.
3187 Description: data checksumming impacts single node performance
3188 Details : add support for several checksum algorithms. Currently, CRC32 and
3189 Adler-32 are supported. The checksum type can be changed on the fly
3190 through /proc/fs/lustre/osc/*/checksum_type.
3194 Description: use adler32 for page checksums
3195 Details : when available, use the Adler-32 algorithm instead of CRC32 for
3200 Description: better handle error messages in extents code
3202 Severity : enhancement
3204 Description: SNMP support enhancement
3205 Details : Adding total number of sampled request for an MDS node in snmp
3208 Severity : enhancement
3210 Description: Optimize ldlm waiting list processing for PR extent locks
3211 Details : When processing waiting list for read extent lock and meeting read
3212 lock that is same or wider to it that is not contended, skip
3213 processing rest of the list and immediatelly return current
3214 status of conflictness, since we are guaranteed there are no
3215 conflicting locks in the rest of the list.
3219 Description: Time out and refuse to reconnect
3220 Details : When the failover node is the primary node, it is possible
3221 to have two identical connections in imp_conn_list. We must
3222 compare not conn's pointers but NIDs, otherwise we
3223 can defeat connection throttling.
3227 Description: Client not clear own cache if answer to reconnect is lost.
3228 Details : Client gets evicted from server. Now client also thinks it is
3229 disconnected (or gets ENOTCONN on its operation) and decides to
3230 reconnect. Server receives reconnect message, but cannot find
3231 export. New export is created that is fully valid (new cookie!),
3232 but reply is lost and not reported to client. Client reconnects
3233 again and gets back a just-created connection, but it is not new
3234 so client thinks it was not evicted and does not do recovery.
3238 Description: Detect stride IO mode in read-ahead
3239 Details : When a client does stride read, read-ahead should detect that and
3240 read-ahead pages according to the detected stride pattern.
3244 Description: build for x2 fails
3245 Details : fix compile issue on Cray systems.
3247 Severity : enhancement
3249 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
3250 Details : This greatly improves speed of NFS writes on 2.6 kernels.
3253 Frequency : only on PPC/SLES10
3255 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
3256 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
3257 be defined. Add a hack in mkfs_lustre.c to work around this compile
3262 Description: LASSERT on MDS when client holding flock lock dies
3263 Details : ldlm pool logic depends on number of granted locks equal to
3264 number of released locks which is not true for flock locks, so
3265 just exclude such locks from consideration.
3269 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
3270 Details : Use fsfilt_sync() for both the whole filesystem sync and
3271 individual file sync to eliminate dangerous inode locking
3272 with I_LOCK that can lead to a deadlock.
3276 Description: Update an obsolete wirecheck.c generator
3277 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
3281 Description: Client can panic on open sometimes
3282 Details : It is possible that we try to free already freed request in
3283 ll_file_open in some error cases when we send request from
3288 Description: performance in 1.6.3
3289 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3290 sure that 1MB requests are not fragmented by the block layer.
3294 Description: LDLM soft lockups - improvement
3295 Details : It is be possible to send the lock handle along with each read
3296 or write request because the client is already doing a lock match
3297 itself so there isn't any reason the OST should have to re-do that
3303 Description: lfs quota fails with deactivated OSTS
3304 Details : With this patch, three improvements are included:
3305 1. detete the softlimit in mds and osts when use "lfs quota".
3306 2. display the inaccurate data in the output of "lfs quota".
3307 3. try to get quota info when "lfs quota" is executed.
3312 Description: Extent locks not granted with no conflicts sometimes.
3313 Details : When race occurs in glimpse handler and nothing is returned,
3314 we do not reprocess the queue after lock cancel, and that leads
3315 to a stall until next activity on a resource
3318 Frequency : failover with quotaon
3320 Description: during mds failovers with quota on, OSTs got into deadlock state
3321 and causing dumpstack.
3322 Details : for every quota slave, at any time, there is only one quota req
3323 is sent to quota master for every uid/gid. Before that quota req
3324 returns, all the thread relative to the same uid/gid will wait.
3325 So if the quota req is lost because mds failovers or any other
3326 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3327 quota reqs who time out.
3329 Severity : enhancement
3332 Description: when quota slave checks if quota is enough, there is an unnecessary
3334 Details : place this wait on necessary place instead of always waiting.
3336 --------------------------------------------------------------------------------
3338 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3340 * Support for kernels:
3341 2.6.5-7.286 (SLES 9),
3342 2.6.9-55.0.9.EL (RHEL 4),
3343 2.6.16.53-0.8 (SLES 10),
3344 2.6.18-8.1.14.el5 (RHEL 5),
3345 2.6.18.8 vanilla (kernel.org)
3346 * Client support for unpatched kernels:
3347 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3348 2.6.16 - 2.6.22 vanilla (kernel.org)
3349 * Due to recently discovered recovery problems, we do not recommend
3350 using patchless RHEL 4 clients with this or any earlier release.
3351 * Recommended e2fsprogs version: 1.40.2-cfs1
3352 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3355 Frequency : occasional
3357 Description: MDS hang or stay in waiting lock
3358 Details : If client receive lock with CBPENDING flag ldlm need send lock
3359 cancel as separate rpc, to avoid situation when cancel request
3360 can't processed due all i/o threads stay in wait lock.
3363 Frequency : occasional
3365 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3366 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3367 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3368 oscc_has_objects and see OSC has no objests and call
3369 oscc_internal_create to resend create request.
3371 Severity : enhancement
3373 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3374 Details : Update which_patch & target file for SLES10 latest kernel.
3376 Severity : enhancement
3378 Description: add --type and --size parameters to lfs find
3379 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3380 multiple OBDs can now be specified for the --obd option.
3382 Severity : enhancement
3384 Description: eliminate client locks in face of contention
3385 Details : file contention detection and lockless i/o implementation
3386 for contended files.
3388 Severity : enhancement
3390 Description: Remove client patches from SLES 10 kernel.
3391 Details : This causes SLES 10 clients to behave as patchless clients
3392 even on a Lustre-patched (server) kernel.
3394 Severity : enhancement
3396 Description: use i_size_read and i_size_write in 2.6 port
3397 Details : replace inode->i_size access with i_size_read/write()
3400 Frequency : when removing large files
3402 Description: scheduling issue during removal of large Lustre files
3403 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3404 It causes scheduling issues when removing large files (17TB in the
3410 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3411 Details : Impossible (EPROTO is returned) to access a directory that has a
3412 non-default striping and ACLs.
3415 Frequency : only on ppc
3417 Description: /proc/fs/lustre/devices broken on ppc
3418 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3419 We should make sure the type of 'index' is loff_t and then cast
3420 explicitly as needed below. Do not assign an explicitly cast
3424 Frequency : only for rhel5
3426 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3427 Details : Modify the target file & which_kernel.
3430 Frequency : if the uninit_groups feature is enabled on ldiskfs
3432 Description: e2fsck reports "invalid unused inodes count"
3433 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3434 feature and only a single inode is created in a group then the
3435 "bg_unused_inodes" count is incorrectly updated. Creating a
3436 second inode in that group would update it correctly.
3441 Description: buffer overruns could theoretically occur
3442 Details : llapi_semantic_traverse() modifies the "path" argument by
3443 appending values to the end of the origin string, and a buffer
3444 overrun may occur. Adding buffer overrun check in liblustreapi.
3446 Severity : enhancement
3448 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3451 Frequency : only if filesystem is inconsistent
3453 Description: handle "serious error: objid * already exists" more gracefully
3454 Details : If LAST_ID value on disk is smaller than the objects existing in
3455 the O/0/d* directories, it indicates disk corruption and causes an
3456 LBUG(). If the object is 0-length, then we should use the existing
3457 object. This will help to avoid a full fsck in most cases.
3459 Severity : enhancement
3461 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3462 Details : Modify vm-tunables-rhel4.patch.
3464 Severity : enhancement
3466 Description: Kernel config for 2.6.18-vanilla.
3467 Details : Modify targets/2.6-vanilla.target.in.
3468 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3469 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3470 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3471 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3474 Frequency : occasional
3476 Description: improve handling recoverable errors
3477 Details : If request processed with error which can be recoverable on server
3478 request should be resend, otherwise page released from cache and
3484 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3485 Details : Modify target file & which_patch.
3486 A flaw was found in the IA32 system call emulation provided
3487 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3488 value could be stored in the %RAX register, which could trigger an
3489 out-of-bounds system call table access. An untrusted local user
3490 could exploit this flaw to run code in the kernel
3491 (ie a root privilege escalation). (CVE-2007-4573).
3495 Description: change order of libsysio includes
3496 Details : '#include sysio.h' should always come before '#include xtio.h'
3498 Severity : enhancement
3500 Description: adapt the lustre_config script to support the upgrade case
3501 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3502 server targets to 1.6 in parallel.
3507 Description: To avoid grant space > avaible space when the disk is almost
3508 full. Without this patch you might see the error "grant XXXX >
3509 available" or some LBUG about grant, when the disk is almost
3511 Details : In filter_check_grant, for non_grant cache write, we should
3512 check the left space by if (*left > ungranted + bytes), instead
3513 of (*left > ungranted), because only we are sure the left space
3514 is enough for another "bytes", then the ungrant space should be
3515 increase. In client, we should update cl_avail_grant only there is
3516 OBD_MD_FLGRANT in the reply.
3521 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3522 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3523 described in CVE-2007-4573. This problem could allow untrusted
3524 local users to gain root access.
3527 Frequency : when using O_DIRECT and quotas
3529 Description: Incorrect file ownership on O_DIRECT output files
3530 Details : block usage reported by 'lfs quota' does not take into account
3531 files that have been written with O_DIRECT.
3535 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3536 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3537 window to 0. If it does hit cache miss, and read-pattern is still
3538 stride-io mode, does not reset the stride window, but also does not
3539 increase the stride window length in this case.
3541 --------------------------------------------------------------------------------
3543 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3545 * Support for kernels:
3546 2.6.5-7.286 (SLES 9),
3547 2.6.9-55.0.2.EL (RHEL 4),
3548 2.6.16.46-0.14 (SLES 10),
3549 2.6.18-8.1.8.el5 (RHEL 5),
3550 2.6.18.8 vanilla (kernel.org)
3551 * Client support for unpatched kernels:
3552 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3553 2.6.16 - 2.6.21 vanilla (kernel.org)
3554 * Due to recently discovered recovery problems, we do not recommend
3555 using patchless RHEL 4 clients with this or any earlier release.
3556 * Recommended e2fsprogs version: 1.40.2-cfs1
3557 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3561 Description: Fix errors in lfs documentation
3562 Details : Fixes man pages
3564 Severity : enhancement
3566 Description: Adaptive timeouts
3567 Details : RPC timeouts adapt to changing server load and network
3568 conditions to reduce resend attempts and improve recovery time.
3570 Severity : enhancement
3572 Description: llapi_file_create() does not allow some changes
3573 Details : add llapi_file_open() that allows specifying the file creation
3574 mode and open flags, and also returns an open file handle.
3576 Severity : enhancement
3578 Description: df doesn't work properly if diskfs blocksize != 4K
3579 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3582 Frequency : always on directories with default striping set
3584 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3585 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3587 Severity : enhancement
3589 Description: merge and cleanup kernel patches.
3590 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3592 Severity : enhancement
3594 Description: RedHat Update kernel for RHEL5
3595 Details : Kernel config file for RHEL5.
3597 Severity : enhancement
3599 Description: OSS needs mutliple precreate threads
3600 Details : Add ability to start more than one create thread per OSS.
3603 Frequency : only with quota on the root user
3605 Description: mds_obd_create error creating tmp object
3606 Details : When the user sets quota on root, llog will be affected and can't
3607 create files and write files.
3610 Frequency : Always on ia64 patchless client, and possibly others.
3612 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3613 Details : This allows the patchless client to be loaded on architectures
3614 without this export.
3619 Description: disorder of journal start and llog_add cause deadlock.
3620 Details : in llog_origin_connect, journal start should happen before
3621 llog_add keep the same order as other functions to avoid
3624 Severity : enhancement
3626 Description: RedHat Update kernel for RHEL5
3627 Details : Modify the kernel config file more closer RHEL5.
3629 Severity : enhancement
3631 Description: Build failure against Centos5 (RHEL5)
3632 Details : Define PAGE_SIZE when it isn't present.
3635 Frequency : occasionally when using NFS
3637 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3638 Details : with NFS, the anon dentry's parent was set to itself in
3639 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3640 corresponding dentry other than use rec->ur_name.
3642 Severity : enhancement
3644 Description: enable data checksumming by default
3645 Details : enable checksum by default, allow --disable-checksum
3646 configure option and "-o nochecksum" mount option. Checksums
3647 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3650 Frequency : Occasionally with failover
3652 Description: Client eviction due to failover config
3653 Details : after a connection loss, the lustre client should attempt to
3654 reconnect to the last active server first before trying the
3655 other potential connections.
3657 Severity : enhancement
3659 Description: client-side metadata stat-ahead during readdir(directory readahead)
3660 Details : perform client-side metadata stat-ahead when the client detects
3661 readdir and sequential stat of dir entries therein
3666 Description: when mds and osts use different quota unit(32bit and 64bit),
3667 quota will be released repeatly.
3668 Details : void sending multiple quota reqs to mds, which will keep the
3669 status between the reqs.
3672 Frequency : only with liblustre clients on XT3
3674 Description: evictions taking too long
3675 Details : allow llrd to evict clients directly on OSTs
3679 Description: osts not allocated evenly to files
3680 Details : change the condition to increase offset_idx
3682 Severity : enhancement
3684 Description: self-adjustable client's lru lists
3685 Details : use adaptive algorithm for managing client cached locks lru
3686 lists according to current server load, other client's work
3687 pattern, memory activities, etc. Both, server and client
3688 side namespaces provide number of proc tunables for controlling
3693 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3694 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3695 readdir code from ext3) from scratch preserving functionality.
3699 Description: Only those disconnect error should be returned by rq_status.
3700 Details : In open/enqueue processs, Some errors, which will cause client
3701 disconnected, should be returned by rq_status, while other
3702 errors should still be returned by intent, then mdc or llite will
3705 Severity : enhancement
3707 Description: Tune the kernel for good SCSI performance.
3708 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3709 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3713 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3714 Bugzilla : 13547, 13627
3715 Description: Data corruption for OSTs that are formatted larger than 2TB
3717 Details : When generating the bio request for lustre file writes the
3718 sector number would overflow a temporary variable before being
3719 used for the IO. The data reads correctly from Lustre (which
3720 will overflow in a similar manner) but other file data or
3721 filesystem metadata may be corrupted in some cases.
3725 Description: "lfs find -obd UUID" prints directories
3726 Details : "lfs find -obd UUID" will return all directory names instead
3727 of just file names. It is incorrect because the directories
3728 do not reside on the OSTs.
3731 Bugzilla : 13596, 13614
3732 Description: MDS hang after unclean shutdown of lots of clients
3733 Details : Never resend AST requests.
3737 Frequency : Always, for kernels after 2.6.16
3738 Description: Fix warning idr_remove called for id=.. which is not allocated.
3739 Details : Last kernels save old s_dev before kill super and not allow
3740 to restore from callback - restore it before call kill_anon_super.
3744 Description: Fix error on 'ls .' at the top of the Lustre mount.
3745 Details : Don't revalidate dentry if it is a root dentry.
3747 --------------------------------------------------------------------------------
3749 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3751 * Support for kernels:
3752 2.6.5-7.286 (SLES 9),
3753 2.6.9-55.0.2.EL (RHEL 4),
3754 2.6.16.46-0.14 (SLES 10),
3755 2.6.18-8.1.8.el5 (RHEL 5),
3756 2.6.18.8 vanilla (kernel.org)
3757 * Client support for unpatched kernels:
3758 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3759 2.6.16 - 2.6.22 vanilla (kernel.org)
3760 * Due to recently discovered recovery problems, we do not recommend
3761 using patchless RHEL 4 clients with this or any earlier release.
3762 * Recommended e2fsprogs version: 1.39.cfs8
3763 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3768 Description: block reactivating mgc import until all deactivates complete
3769 Details : Fix race when failing back MDT/MGS to itself (testing)
3771 Severity : enhancement
3773 Description: lfs setstripe enhancement
3774 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3777 Frequency : mds/oss recovery
3779 Description: llog ctxt is refrenced after it has been freed.
3780 Details : llog ctxt refcount was added to avoide the race between ctxt free
3781 and llog recovery process. Each llog user must hold ctxt refcount
3782 before it access the llog. And the llog ctxt can only be freed
3783 when its refcount is zero.
3785 Severity : enhancement
3787 Description: randomly memory allocation failure util
3788 Details : Make lustre randomly failed allocating memory for testing purpose.
3791 Frequency : only for SLES10
3793 Description: Update kernel patch for SLES10 SP1
3794 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3796 Severity : enhancement
3798 Description: omit set fsid for export NFS
3799 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3800 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3803 Frequency : after network failures
3805 Description: Add sync option to mount_lustre.c
3806 Details : Client loses data written to lustre after a network interruption.
3808 Severity : enhancement
3810 Description: Error message improvement.
3811 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3813 Severity : enhancement
3815 Description: don't use GFP_* in generic Lustre code.
3816 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3818 Severity : enhancement
3820 Description: obdclass is limited by single OBD_ALLOC(idarray)
3821 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3823 Severity : enhancement
3825 Description: updated patchess for new RHEL4 kernel
3826 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3828 Severity : enhancement
3830 Description: warnings with build patchless client with vanila 2.6.19 and up
3831 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3832 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3834 Severity : enhancement
3836 Description: O_DIRECT bypasses client statistics.
3837 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3838 read_ahead_stats were updated. Copied stats section from
3839 osc_send_oap_rpc() into async_internal().
3842 Frequency : only for Cray XT3
3844 Description: peer credits not enough on many OST per OSS systems.
3845 Details : Use new lnet way to add credits as we need those for pings and ASTs
3848 Frequency : only with liblustre
3850 Description: Liblustre is not releasing flock locks on file close.
3851 Details : Release flock locks on file close.
3854 Frequency : only for RHEL4
3856 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3857 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3858 Update target file and kernel config.
3862 Description: lustre support for RHEL5
3863 Details : Add support for RHEL5.
3868 Description: ASSERTION(export != NULL) failed in target_handle_connect
3869 Details : Assetion hit is result of rare race between disconnect and connet
3870 to same nid. target_handle_connect found old connect cockie and
3871 tried to reconnect, but can't find export for this cockie.
3876 Description: umount blocks forever on error
3877 Details : In result of wrong using obd_no_recov and obd_force flags client
3878 can hand if cancel or some other requests is lost.
3881 Frequency : Only for SLES
3883 Description: sanity_quota fail test_1
3884 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3885 file, which makes TSTID[2] inunique.
3887 Severity : enhancement
3889 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3890 Details : Update target/ChangeLog/which_patch .
3892 Severity : enhancement
3894 Description: jbd statistics
3895 Details : Port older jbd statistics patch for sles10
3900 Description: lvbo_init failed for resource with missing objects.
3901 Details : Fix returning error if we do stat for file with missing/corrupted
3902 objects and i_size set to all sum of size of all avaible objects.
3903 if we to truncate/write to missing object - it is recreated.
3908 Description: Oops in read and write path when failing to allocate lock.
3909 Details : Check if lock allocation failed and return error back.
3912 Frequency : When flocks are used.
3914 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3915 Details : Flock locks might destroy just granted lock if it could be merged
3916 with another existing flock, this is done in completion handler,
3917 so teach ldlm_cli_enquque_fini that this is a valid case for
3923 Description: reply_lock_interpret crash due to race with it and lock cancel.
3924 Details : Do not replay locks that are being cancelled. Do not reference
3925 locks by their address during replay, just by their handle.
3928 Frequency : only with deactivated OSTs
3930 Description: lstripe command fails for valid OST index
3931 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3932 instead of lov->desc.ld_active_tgt_count.
3934 Severity : enhancement
3936 Description: Build failure against Centos5 (RHEL5)
3937 Details : Use getpagesize() instead of PAGE_SIZE.
3939 --------------------------------------------------------------------------------
3941 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3943 * Support for kernels:
3944 2.6.5-7.283 (SLES 9),
3945 2.6.9-55.EL (RHEL 4),
3946 2.6.16.46-0.14 (SLES 10),
3947 2.6.18.8 vanilla (kernel.org)
3948 * Client support for unpatched kernels:
3949 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3950 2.6.16 - 2.6.22 vanilla (kernel.org)
3951 * Due to recently discovered recovery problems, we do not recommend
3952 using patchless RHEL 4 clients with this or any earlier release.
3953 * Recommended e2fsprogs version: 1.39.cfs8
3954 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3955 * Starting with this release, the ldiskfs backing filesystem required
3956 by Lustre is now in its own package, lustre-ldiskfs. This package
3957 should be installed. It is versioned separately from Lustre and
3958 may be released separately in future.
3963 Description: server LBUG when shutting down
3964 Details : Block umount forever until the mount refcount is zero rather
3965 than giving up after an arbitrary timeout.
3967 Severity : enhancement
3969 Description: add optional extra BUILD_VERSION info
3970 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3971 to override the lustre version.
3974 Frequency : 2.6.18 servers only
3976 Description: ll_kern_mount() doesn't release the module reference
3977 Details : The ldiskfs module reference count never drops down to 0
3978 because ll_kern_mount() doesn't release the module reference.
3983 Description: server LBUG when using old ost_num_threads parameter
3984 Details : Accept the old ost_num_threads parameter but warn that it
3985 is deprecated, and fix an off-by-one error that caused an LBUG.
3990 Description: Transient SCSI error results in persistent IO issue
3991 Details : iobuf->dr_error is not reinitialized to 0 between two
3995 Frequency : sometimes when underlying device returns I/O errors
3997 Description: OSTs not going read-only during write failures
3998 Details : OSTs are not remounted read-only when the journal commit threads
3999 get I/O errors because fsfilt_ext3 calls journal_start/stop()
4000 instead of the ext3 wrappers.
4003 Frequency : only with 10000 clients or more
4005 Description: poor connect scaling with increasing client count
4006 Details : Don't run filter_grant_sanity_check for more than 100 exports
4007 to improve scaling for large numbers of clients.
4010 Frequency : SLES10 only
4012 Description: sanity-quota.sh quotacheck failed: rc = -22
4013 Details : Quotas cannot be enabled on SLES10.
4016 Frequency : liblustre clients only
4018 Description: getdirentries does not give error when run on compute nodes
4019 Details : getdirentries does not fail when the size specified as an argument
4020 is too small to contain at least one entry
4022 Severity : enhancement
4024 Description: Add LNET router traceability for debug purposes
4025 Details : If a checksum failure occurs with a router as part of the
4026 IO path, the NID of the last router that forwarded the bulk data
4027 is printed so it can be identified.
4032 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
4033 Details : Due to a race condition, liblustre clients were occasionally
4034 evicted incorrectly.
4036 Severity : enhancement
4038 Description: lfs setstripe use optional parameters instead of postional
4041 Severity : enhancement
4043 Description: Nanosecond timestamp support for ldiskfs
4044 Details : The on-disk ldiskfs filesystem has added support for nanosecond
4045 resolution timestamps. There is not yet support for this at
4046 the Lustre filesystem level.
4049 Frequency : during server recovery
4051 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
4052 Details : request with rq_no_resend flag not awake l_wait_event if they get a
4056 Frequency : nfs export on patchless client
4058 Description: connectathon hang when test nfs export over patchless client
4059 Details : Disconnected dentry cannot be found with lookup, so we do not need
4060 to unhash it or make it invalid
4063 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
4064 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
4065 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
4070 Description: open req refcounting wrong on reconnect
4071 Details : If reconnect happened between getting open reply from server and
4072 call to mdc_set_replay_data in ll_file_open, we will schedule
4073 replay for unreferenced request that we are about to free.
4074 Subsequent close will crash in variety of ways.
4075 Check that request is still eligible for replay in
4076 mdc_set_replay_data().
4081 Description: disable writes to filesystem when reading health_check file
4082 Details : the default for reading the health_check proc file has changed
4083 to NOT do a journal transaction and write to disk, because this
4084 can cause reads of the /proc file to hang and block HA state
4085 checking on a healthy but otherwise heavily loaded system. It
4086 is possible to return to the previous behaviour during configure
4087 with --enable-health-write.
4089 Severity : enhancement
4091 Description: 64-bit inode version
4092 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
4093 to the inode. This will be required for version-based recovery.
4098 Description: MDS fails to start if a duplicate client export is detected
4099 Details : in some rare cases it was possible for a client to connect to
4100 an MDS multiple times. Upon recovery the MDS would detect this
4101 and fail during startup. Handle this more gracefully.
4103 Severity : enhancement
4105 Description: Add -o localflock option to simulate old noflock behaviour.
4106 Details : This will achieve local-only flock/fcntl locks coherentness.
4111 Description: log_commit_thread vs filter_destroy race leads to crash
4112 Details : Take import reference before releasing llog record semaphore
4117 Description: Wrong request locking in request set processing
4118 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
4119 imp_delayed_list, in this place should be used imp_lock.
4122 Frequency : when reconnecting
4124 Description: Grant leak when OSC reconnect to OST
4125 Details : When osc reconnect ost, OST(filter) should check whether it
4126 should grant more space to client by comparing fed_grant and
4127 cl_avail_grant, and return the granted space to client instead
4128 of "new granted" space, because client will call osc_init_grant
4129 to update the client grant space info.
4132 Frequency : when client reconnects to OST
4134 Description: Grant leak when OSC does a resend and replays bulk write
4135 Details : When osc reconnect to OST, OST(filter) should clear grant info of
4136 bulk write request, because the grant info while be sync between
4137 OSC and OST when reconnect, and we should ignore the grant info
4138 these of resend/replay write req.
4143 Description: Grant space more than avaiable space sometimes.
4144 Details : When then OST is about to be full, if two bulk writing from
4145 different clients came to OST. Accord the avaliable space of the
4146 OST, the first req should be permitted, and the second one
4147 should be denied by ENOSPC. But if the seconde arrived before
4148 the first one is commited. The OST might wrongly permit second
4149 writing, which will cause grant space > avaiable space.
4152 Frequency : when client is evicted
4154 Description: Grant might be wrongly erased when osc is evicted by OST
4155 Details : when the import is evicted by server, it will fork another
4156 thread ptlrpc_invalidate_import_thread to invalidate the
4157 import, where the grant will be set to 0. While the original
4158 thread will update the grant it got when connecting. So if
4159 the former happened latter, the grant will be wrongly errased
4160 because of this race.
4165 Description: Checking Stale with correct fid
4166 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
4167 to check whether it is stale, because sometimes, we want the
4168 enqueue happened anyway, and op_data.fid2 will not be initialized.
4170 Severity : enhancement
4172 Description: update patchless client
4173 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
4176 Frequency : only with 2.4 kernel
4178 Description: random memory corruption
4179 Details : size of struct ll_inode_info is to big for union inode.u and this
4180 can be cause of random memory corruption.
4185 Description: Memory leak in recovery
4186 Details : Lov_mds_md was not free in an error handler in mds_create_object.
4187 It should also check obd_fail before fsfilt_start, otherwise if
4188 fsfilt_start return -EROFS,(failover mds during mds recovery).
4189 then the req will return with repmsg->transno = 0 and rc = EROFS.
4190 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
4191 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
4192 be freed no matter whether fsfilt_commit success or not.
4195 Frequency : only with large numbers of cores on a single node
4197 Description: Prevents from taking the superblock lock in llap_from_page for
4199 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
4200 for llap_from_page call in ll_removepage() prevents from taking
4201 the superblock lock for a soon died page.
4206 Description: Not check open intent error before release open handle
4207 Details : in some rare cases, the open intent error is not checked before
4208 release open handle, which may cause
4209 ASSERTION(open_req->rq_transno != 0), because it tries to release
4210 the failed open handle.
4215 Description: Set cat log bitmap only after create log success.
4216 Details : in some rare cases, the cat log bitmap is set too early. and it
4217 should be set only after create log success.
4221 Description: Accessing a block bevice can re-enable I/O when Lustre is
4222 tearing down a device.
4223 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
4228 Description: service threads may hog cpus when there are a lot of requests
4229 Details : Insert cond_resched to give other threads a chance to use some CPU
4234 Description: the cat log was not initialized in recovery
4235 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
4236 unlink log on mds will not be initialized until mds post
4237 recovery. And also in mds post recovery, the unlink log will
4238 initialization will be done asynchronausly, so there will be race
4239 between add unlink log and unlink log initialization.
4243 Description: brw_stats were being printed incorrectly
4244 Details : brw_stats were being printed as log2 but all of them were not
4245 recorded as log2. Also remove some code duplication arising from
4246 filter_tally_{read,write}.
4250 Frequency : rare, only in recovery.
4251 Description: ASSERTION(req->rq_type != LI_POISON) failed
4252 Details : imp_lock should be held while iterating over imp_sending_list for
4253 prevent destroy request after get timeout in ptlrpc_queue_wait.
4257 Description: replay-single.sh test 52 fails
4258 Details : A lock's skiplist need to be cleanup when it being unlinked
4259 from its resource list.
4264 Description: Short directio read returns full requested size rather than
4266 Details : Direct I/O operations should return actual amount of bytes
4267 transferred rather than requested size.
4269 Severity : enhancement
4271 Description: metadata RPC reduction (e.g. for rm performance)
4272 Details : decrease the amount of synchronous RPC between clients and servers
4273 by canceling conflicing lock before the operation on the client
4274 and packing thier handles into the main operation RPC to server.
4276 Severity : enhancement
4278 Description: add #ifdef HAVE_KERNEL_CONFIG_H
4279 Details : kernels from 2.6.19 not need include linux/config.h, but add
4280 include linux/autoconf.h in commpiler command line.
4282 Severity : enhancement
4284 Description: patchless client support for 2.6.22 kernel
4285 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
4286 removed. In this case we need drop using os depended interface to
4287 kmem_cache and use cfs_mem_cache API.
4292 Description: fix mal-formatted messages
4293 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4297 Frequency : always in liblustre
4298 Description: wrong IS_ERR implementation in liblustre.h
4299 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4304 Description: Correct condition for output debug message.
4305 Details : inode i_nlink equal zero is not enough for output message about
4306 disk corruption, i_ctime and i_mode should be also checked.
4310 Frequency : always in patchless client
4311 Description: add configure check for truncate_complete_page
4312 Details : improve checks for exported symbols. This allow run check without
4313 sources, but with Module.symvers shipped with kernel distribution.
4314 add check for truncate_complete_page used by patchless client.
4318 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4319 Details : test_77h uses a file which was messed by other test case.
4323 Description: Not Check whether lov_tgts is NULL in some lov functions
4324 Details : Checking whether lov_tgts is NULL in some functions.
4328 Description: replace obdo_alloc() with OBDO_ALLOC macro
4329 Details : nothing special is done in obdo_alloc() function, and for
4330 debugging purpose, it needs to be replaced with macros.
4334 Description: bad return value and errno from fcntl call
4335 Details : In liblustre API, errno should be a negative value if error
4340 Description: ptlrpc_check_set() LBUG
4341 Details : In case of positive reply from server and failed client bulk
4342 callback after bulk transfer shouldn't LBUG, but process this
4343 request as erroneous.
4345 Severity : enhancement
4347 Description: VFS operations stats tool.
4348 Details : Tool which collects stats by tracking value written in pid,
4349 ppid, gid and uses llstat to generate output to plot graph using
4351 Updated lustre/utils/Makefile.am
4352 Added lustre/utils/ltrack_stats.c
4354 Severity : enhancement
4356 Description: 2.6.18 server support (lustre 1.6.1)
4357 Details : Support for 2.6.18 kernels on the server side.
4362 Description: ASSERTION(imp->imp_conn_current) failed
4363 Details : an assertion failure is hit if a client node boots and attempts to
4364 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4367 Frequency : only for i686
4369 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4370 Details : Fixed config variable for build.
4375 Description: Updated patchess for new RHEL4 kernel
4376 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4377 Updated patch jbd-stats-2.6.9.patch
4378 Updated patch qsnet-rhel4-2.6.patch
4379 Updated patch quota-deadlock-on-pagelock-core.patch
4380 Updated patch vfs_intent-2.6-rhel4.patch
4381 Updated patch vfs_races-2.6-rhel4.patch
4382 Updated series file 2.6-rhel4-titech.series
4383 Updated series file 2.6-rhel4.series
4384 Updated kernel config files
4389 Description: lquota slave complains LBUG when reconnecting with mds
4391 Details : quota slave depends qctxt->lqc_import to send its quota request.
4392 This pointer will be invalid if mds did failover or broke its
4393 connect to osts, which leads to LBUG.
4396 Frequency : when qunit size is too small(less than 20M)
4398 Description: write is stopped by improper -EDQUOT
4399 Details : If the master is busy and qunit size is small enough(let's say 1M),
4400 the slave can not get quota from master on time, which will lead
4401 slave to trigger a -EQUOTA to client.
4406 Description: Deadlock during metadata tests
4407 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4408 called with the per-dentry lock held.
4411 Frequency : SLES9 only
4413 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4414 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4416 Severity : enhancement
4418 Description: remove fs_prep_san_write operation and related patches
4419 Details : remove the ext3-san-jdike patches which are no longer useful.
4424 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4425 Details : a disk corruption can cause the mballoc code to assert on a
4426 double free or other extent corruptions. Handle these with
4427 ext3_error() instead of assertions.
4430 Frequency : only with mballoc3 code and deep extent trees
4432 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4433 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4434 can cause the filesystem to be remounted read-only.
4439 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4440 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4441 few external users of this ioctl number is lfsck's call to
4442 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4443 Add the old ioctl number to the handler so both old and new
4448 Description: parallel lock callbacks
4449 Details : Instead of sending blocking and completion callbacks as separated
4450 requests, adding them to a set and sending in parallel.
4454 Description: Disable most debugging by default
4455 Details : To improve performance, disable most logging (for debug purposes)
4456 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4457 default, and HA includes fewer messages.
4461 Frequency : only run on patchless client.
4462 Description: use do_facet on sanity.sh for test handling recoverables errors
4463 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4467 Description: hash tables for lists of nids, connections and uuids
4468 Details : Hash tables noticeably help when a lot of clients connect to a
4469 server, to faster identify duplicate connections or reconnects,
4470 also to faster find export to evict in manual eviction case.
4474 Description: Sometimes, when the server evict a client, and the client will
4475 not be evicted as soon as possible.
4476 Details : In enqueue req, the error was returned by intent, instead of
4477 rq_status which make ptlrpc layer not detect this error, and
4478 does not evict the client. So enqueue error should be returned
4482 Frequency : only at startup
4484 Description: Delay client connections to MDT until fist MDT->OST connect
4485 Details : If a client tried to create a new file before the MDT had
4486 connected to any OSTs, the create would return EIO. Now
4487 the client will simply block until the MDT connects to the
4488 first OST and the create can succeed.
4491 Frequency : only for SLES9
4493 Description: Routinely utilize latest Quadrics drivers in CFS releases
4494 Details : Update patch qsnet-suse-2.6.patch.
4497 Frequency : only for sles10
4499 Description: Update patches for SLES 10 SP1 kernel.
4500 Details : Update the patch vfs_intent-2.6-sles10.patch.
4505 Description: Routinely utilize latest Quadrics drivers in CFS releases
4506 Details : Update patch qsnet-rhel4-2.6.patch.
4509 Frequency : at statup only
4511 Description: mds_lov_synchronize race leads to various problems
4512 Details : simultaneous MDT->OST connections at startup can cause the
4513 sync to abort, leaving the OSC in a bad state.
4518 Description: Using wrong pointer in osc_brw_prep_request
4519 Details : Access to array[-1] can produce panic if kernel compiled with
4520 CONFIG_PAGE_ALLOC enabled
4522 Severity : enhancement
4524 Description: Async OSC create to avoid the blocking unnecessarily.
4525 Details : If a OST has no remain object, system will block on the creating
4526 when need to create a new object on this OST. Now, ways use
4527 pre-created objects when available, instead of blocking on an
4528 empty osc while others are not empty. If we must block, we block
4529 for the shortest possible period of time.
4533 Frequency : only in recovery
4534 Description: Mark OST as early accessible if his start SYNC.
4535 Details : osc_precreate return flag early accessible if oscc marked as
4536 OSCC_FLAG_SYNC_IN_PROGRESS.
4541 Description: Sometimes precreate code can triger create object on wrong ost
4542 Details : Wrong protected or not not restored variables aftre precreate loop
4543 can produce creation object on wrong ost.
4546 Frequency : oss recovery
4548 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4550 Details : llog_commit_thread_count should be synced between llog_commit
4551 start and cleanup, so new llog_commit thread should not be started
4552 when llog_commit threads being stopped to avoid accessing some
4555 Severity : enhancement
4557 Description: Add printing inode info into message about error in writepage.
4559 --------------------------------------------------------------------------------
4561 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4566 Frequency : on some architectures
4568 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4569 Details : Uninitialized flags sometimes cause configuration commands to
4573 Frequency : patchless clients only
4575 Description: missing __iget() symbol export
4576 Details : The __iget() symbol export is missing. To avoid the need for
4577 this on patchless clients the deathrow inode reaper is turned
4578 off, and we depend on the VM to clean up old inodes. This
4579 dependency was during via the fix for bug 12181.
4584 Description: sanity.sh fail: test_52b
4585 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4586 incorrect inode's flags to client.
4588 --------------------------------------------------------------------------------
4590 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4592 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4593 INTEROPERATE with older versions automatically. In many cases a
4594 special upgrade step is needed. Please read the
4595 user documentation before upgrading any part of a 1.4.x system.
4596 * WARNING: Lustre configuration and startup changes are required with
4597 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4599 * Support for kernels:
4600 2.4.21-47.0.1.EL (RHEL 3),
4601 2.6.5-7.283 (SLES 9),
4602 2.6.9-42.0.10.EL (RHEL 4),
4603 2.6.12.6 vanilla (kernel.org),
4604 2.6.16.27-0.9 (SLES10)
4605 * Client support for unpatched kernels:
4606 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4607 2.6.16 - 2.6.19 vanilla (kernel.org),
4608 2.6.9-42.0.8EL (RHEL 4)
4609 * Recommended e2fsprogs version: 1.39.cfs6
4610 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4613 Severity : enhancement
4615 Description: Async OSC create to avoid the blocking unnecessarily.
4616 Details : If a OST has no remain object, system will block on the creating
4617 when need to create a new object on this OST. Now, ways use
4618 pre-created objects when available, instead of blocking on an
4619 empty osc while others are not empty. If we must block, we block
4620 for the shortest possible period of time.
4622 Severity : enhancement
4624 Description: MountConf
4625 Details : Lustre configuration is now managed via mkfs and mount
4626 commands instead of lmc and lconf. New obd types (MGS, MGC)
4627 are added for dynamic configuration management. See
4628 https://mail.clusterfs.com/wikis/lustre/MountConf for
4631 Severity : enhancement
4633 Description: dynamic OST addition
4634 Details : OSTs can now be added to a live filesystem
4636 Severity : enhancement
4638 Description: startup order invariance
4639 Details : MDTs and OSTs can be started in any order. Clients only
4640 require the MDT to complete startup.
4642 Severity : enhancement
4644 Description: parallel, asynchronous orphan cleanup
4645 Details : orphan cleanup is now performed in separate threads for each
4646 OST, allowing parallel non-blocking operation.
4648 Severity : enhancement
4650 Description: optimized stripe assignment
4651 Details : stripe assignments are now made based on ost space available,
4652 ost previous usage, and OSS previous usage, in order to try
4653 to optimize storage space and networking resources.
4655 Severity : enhancement
4657 Description: Permanently set tunables
4658 Details : All writable /proc/fs/lustre tunables can now be permanently
4659 set on a per-server basis, at mkfs time or on a live
4662 Severity : enhancement
4664 Description: Lustre message v2
4665 Details : Add lustre message format v2.
4667 Severity : enhancement
4669 Description: client OST exclusion list
4670 Details : Clients can be started with a list of OSTs that should be
4671 declared "inactive" for known non-responsive OSTs.
4675 Description: ENOENT returned for valid filehandle during dbench.
4676 Details : Check if a directory has children when invalidating dentries
4677 associated with an inode during lock cancellation. This fixes
4678 an incorrect ENOENT sometimes seen for valid filehandles during
4679 testing with dbench.
4682 Frequency : SFS test only (otherwise harmless)
4684 Description: SPEC SFS validation failure on NFS v2 over lustre.
4685 Details : Changes the blocksize for regular files to be 2x RPC size,
4686 and not depend on stripe size.
4688 Severity : enhancement
4690 Description: fine-grained SMP locking inside DLM
4691 Details : Improve DLM performance on SMP systems by removing the single
4692 per-namespace lock and replace it with per-resource locks.
4694 Severity : enhancement
4696 Description: don't hold multiple extent locks at one time
4697 Details : To avoid client eviction during large writes, locks are not
4698 held on multiple stripes at one time or for very large writes.
4699 Otherwise, clients can block waiting for a lock on a failed OST
4700 while holding locks on other OSTs and be evicted.
4702 Severity : enhancement
4704 Description: Multiple MD RPCs in flight.
4705 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4706 To avoid overly-overloading MDT, introduce a limit on number of
4707 MDT RPCs in flight for a single client and add /proc controls
4708 to adjust this limit.
4710 Severity : enhancement
4712 Description: client read/write statistics
4713 Details : Add client read/write call usage stats for performance
4714 analysis of user processes.
4715 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4716 file access. extents_stats shows chunk size distribution.
4717 extents_stats_per_process show chunk size distribution per
4720 Severity : enhancement
4722 Description: per-client statistics on server
4723 Details : Add ldlm and operations statistics for each client in
4724 /proc/fs/lustre/mds|obdfilter/*/exports/
4726 Severity : enhancement
4728 Description: improved MDT statistics
4729 Details : Add detailed MDT operations statistics in
4730 /proc/fs/lustre/mds/*/stats
4732 Severity : enhancement
4734 Description: VFS operations stats
4735 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4736 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4741 Description: Fix client-side osc byte counters
4742 Details : The osc read/write byte counters in
4743 /proc/fs/lustre/osc/*/stats are now working
4746 Frequency : always as root on SLES
4748 Description: Failure of copying files with lustre special EAs.
4749 Details : Client side always return success for setxattr call for lustre
4750 special xattr (currently only "trusted.lov").
4755 Description: Refcount LNET uuids
4756 Details : The global LNET uuid list grew linearly with every startup;
4757 refcount repeated list entries instead of always adding to
4760 Severity : enhancement
4762 Description: Dynamic service threads
4763 Details : Within a small range, start extra service threads
4764 automatically when the request queue builds up.
4767 Frequency : mixed-endian client/server environments
4769 Description: mixed-endian crashes
4770 Details : The new msg_v2 system had some failures in mixed-endian
4773 Severity : enhancement
4775 Description: Easy OST removal
4776 Details : OSTs can be permanently deactivated with e.g. 'lctl
4777 conf_param lustre-OST0001.osc.active=0'
4779 Severity : enhancement
4781 Description: MGS proc entries
4782 Details : Added basic proc entries for the MGS showing what filesystems
4785 Severity : enhancement
4787 Description: provide MGS failover
4788 Details : Added config lock reacquisition after MGS server failover.
4790 Severity : enhancement
4792 Description: add Linux 2.4 support
4793 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4797 Description: a large application tries to do I/O to the same resource and dies
4798 in the middle of it.
4799 Details : Check the req->rq_arrival time after the call to
4800 ost_brw_lock_get(), but before we do anything about
4801 processing it & sending the BULK transfer request. This
4802 should help move old stale pending locks off the queue as
4803 quickly as obd_timeout.
4806 Frequency : when an incorrect nid is specified during startup
4808 Description: ptlrpc connect to non-existant node causes kernel crash
4809 Details : LNET can't be re-entered from an event callback, which
4810 happened when we expire a message after the export has been
4811 cleaned up. Instead, hand the zombie cleanup off to another
4814 Severity : enhancement
4816 Description: plain/inodebits lock performance improvement
4817 Details : Grouping plain/inodebits in granted list by their request modes
4818 and bits policy, thus improving the performance of search through
4822 Frequency : only if OST filesystem is corrupted
4824 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4825 Details : for a short time RPCs with bulk IO are in the replay list,
4826 but replay of bulk IOs is unimplemented. If the OST filesystem
4827 is corrupted due to disk cache incoherency and then replay is
4828 started it is possible to trip an assertion. Avoid putting
4829 committed RPCs into the replay list at all to avoid this issue.
4832 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4834 Description: System hang on startup
4835 Details : This bug allowed the liblustre (e.g. catamount) client to
4836 return to the app before handling all startup RPCs. This
4837 could leave the node unresponsive to lustre network traffic
4838 and manifested as a server ptllnd timeout.
4840 Severity : enhancement
4842 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4843 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4844 boolean to control whether to print peer debug info when a
4845 client's RPC times out.
4848 Frequency : only for kernels with patches from Lustre below 1.4.3
4850 Description: Remove old rdonly API
4851 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4854 Frequency : only for devices with external journals
4856 Description: Set external device read-only also
4857 Details : During a commanded failover stop, we set the disk device
4858 read-only while the server shuts down. We now also set any
4859 external journal device read-only at the same time.
4862 Frequency : when upgrading from 1.4 while trying to change parameters
4864 Description: The wrong (new) MDC name was used when setting parameters for
4865 upgraded MDT's. Also allows changing of OSC (and MDC)
4866 parameters if --writeconf is specified at tunefs upgrade time.
4869 Frequency : when setting specific ost indicies
4871 Description: QOS code breaks on skipped indicies
4872 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4873 created with --index need not be sequential.
4875 Severity : enhancement
4877 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4878 Details : The uninit_groups feature works in conjunction with the kernel
4879 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4880 up the pass1 processing of e2fsck. This is a read-only feature
4881 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4882 mount filesystems that have had this feature enabled.
4884 Severity : enhancement
4886 Description: Improve multi-block allocation algorithm to avoid fragmentation
4887 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4888 allocation locality and avoid filesystem fragmentation.
4890 ------------------------------------------------------------------------------
4892 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4894 * Support for kernels:
4895 2.4.21-47.0.1.EL (RHEL 3)
4896 2.6.5-7.283 (SLES 9)
4897 2.6.9-42.0.10.EL (RHEL 4)
4898 2.6.12.6 vanilla (kernel.org)
4899 2.6.16.27-0.9 (SLES 10)
4900 * Recommended e2fsprogs version: 1.39.cfs5
4902 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4906 Frequency : occasional, depends on client load and configuration
4907 Bugzilla : 12181, 12203
4908 Description: data loss for recently-modified files
4910 Details : In some cases it is possible that recently written or created
4911 files may not be written to disk in a timely manner (this should
4912 normally be within 30s unless client IO load is very high).
4913 The problem appears as zero-length files or files that are a
4914 multiple of 1MB in size after a client crash or client eviction
4915 that are missing data at the end of the file.
4917 This problem is more likely to be hit on clients where files are
4918 repeatedly created and unlinked in the same directory, clients
4919 have a large amount of RAM, have many CPUs, the filesystem has
4920 many OSTs, the clients are rebooted frequently, and/or the files
4921 are not accessed by other nodes after being written.
4923 The presence of the problem can be detected by looking at
4924 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4925 smaller than the second (nr_unused) then dirty files will not
4926 be flushed automatically to disk. "sync; sleep 10" should be
4927 run several times on the node before unmounting it to update
4928 Lustre (this is also safe to run on nodes without this problem).
4930 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4931 that can cause this same problem, so customers using that kernel
4932 also need to update the kernel in addition to Lustre. In order
4933 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4935 It is normal that files written just before a client crash (less
4936 than 30s) may not yet have been flushed to disk, even for local
4940 Frequency : frequent on thin XT3 nodes
4942 Description: UUID collision on thin XT3 Linux nodes
4943 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4944 randomly, since we relied on an insufficiently-seeded PRNG.
4949 Description: OSS hangs after "All ost request buffers busy"
4950 Details : A deadlock between quota and journal operations caused OSS
4951 hangs after printing "All ost request buffers busy."
4954 Frequency : always on liblustre builds
4956 Description: Cleanup compiler warnings on liblustre
4959 Frequency : always on liblustre builds on XT3
4961 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4965 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4966 > 32000 subdirectories
4967 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4968 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4969 subdirectory count crosses 32000. This will aid e2fsck to
4970 correctly handle more than 32000 subdirectories.
4973 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4975 Description: System hang on startup
4976 Details : This bug allowed the liblustre (e.g. catamount) client to
4977 return to the app before handling all startup RPCs. This
4978 could leave the node unresponsive to lustre network traffic
4979 and manifested as a server ptllnd timeout.
4981 Severity : enhancement
4983 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4984 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4985 boolean to control whether to print peer debug info when a
4986 client's RPC times out.
4991 Description: make O_SYNC working on 2.6 kernels
4992 Details : 2.6 kernels use different method for mark pages for write,
4993 so need add a code to lustre for O_SYNC work.
4998 Description: Failure to close file and release space on NFS
4999 Details : Put inode details into lock acquired in ll_intent_file_open.
5000 Use mdc_intent_lock in ll_intent_open to properly
5001 detect all kind of errors unhandled by mdc_enqueue.
5006 Description: proc file read during shutdown sometimes raced obd removal,
5008 Details : Add lock to prevent obd access after proc file removal.
5011 Frequency : Only for files larger than 4GB on 32-bit clients.
5013 Description: improperly doing page alignment of locks
5014 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
5015 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
5020 Description: RPCs being resent when they shouldn't be
5021 Details : Some RPCs that should not be resent are being resent. This
5022 can cause inconsistencies in the RPC state machine. Do not
5023 resend such requests.
5026 Frequency : rare, only with NFS export
5028 Description: Crash on NFS re-export node
5029 Details : under very unusual load conditions an assertion is hit in
5030 ll_intent_file_open()
5033 Frequency : only if OST filesystem is corrupted
5035 Description: client incorrectly hits assertion in ptlrpc_replay_req()
5036 Details : for a short time RPCs with bulk IO are in the replay list,
5037 but replay of bulk IOs is unimplemented. If the OST filesystem
5038 is corrupted due to disk cache incoherency and then replay is
5039 started it is possible to trip an assertion. Avoid putting
5040 committed RPCs into the replay list at all to avoid this issue.
5045 Description: large O_DIRECT requests fail under memory pressure/fragmentation
5046 Details : Large single O_DIRECT read and write calls can fail to allocate
5047 a sufficiently large buffer to process the request. In case of
5048 allocation failure the allocation is retried with a smaller
5049 buffer and broken into smaller requests.
5051 Severity : enhancement
5053 Description: Add -o localflock option to simulate old noflock behaviour.
5054 Details : This will achieve local-only flock/fcntl locks coherentness.
5059 Description: versioning check is incomplete
5060 Details : Checking the version difference of client vs. server, report
5061 error if the gap is too big.
5066 Description: add support PG_writeback bit
5067 Details : add support for PG_writeback bit for Lustre, for more carefull
5068 work with page cache in 2.6 kernel. This also fix some deadlocks
5069 and remove hack for work O_SYNC with 2.6 kernel.
5071 Severity : enhancement
5073 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
5074 Details : The uninit_groups feature works in conjunction with the kernel
5075 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
5076 up the pass1 processing of e2fsck. This is a read-only feature
5077 in ldiskfs2 only, so older kernels and current ldiskfs cannot
5078 mount filesystems that have had this feature enabled.
5080 Severity : enhancement
5082 Description: Improve multi-block allocation algorithm to avoid fragmentation
5083 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
5084 allocation locality and avoid filesystem fragmentation.
5086 ------------------------------------------------------------------------------
5088 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
5090 * Support for kernels:
5091 2.6.9-42.0.3.EL (RHEL 4)
5092 2.6.5-7.276 (SLES 9)
5093 2.4.21-47.0.1.EL (RHEL 3)
5094 2.6.12.6 vanilla (kernel.org)
5095 2.6.16.21-0.8 (SLES10)
5096 * Recommended e2fsprogs version: 1.39.cfs2-0
5098 * The backwards-compatible /proc/sys/portals symlink has been removed
5099 in this release. Before upgrading, please ensure that you change
5100 any configuration scripts or /etc/sysctl.conf files that access
5101 /proc/sys/portals/* or sysctl portals.* to use the corresponding
5102 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
5103 in advance of the upgrade on any system running Lustre 1.4.6 or
5104 newer, since /proc/sys/lnet was added in that version.
5105 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
5109 Frequency : only when quota is used
5111 Description: avoid scanning export list for quota master
5112 Details : Change the algorithms to avoid scanning export list in order
5113 to improve the efficiency.
5116 Frequency : MDS failover only, very rarely
5118 Description: "went back in time" messages on mds failover
5119 Details : The greatest transno may be lost when the current operation
5120 finishes with an error (transno==0) and the client's last_rcvd
5121 record is over-written. Save the greatest transno in the
5122 mds_last_transno for this case.
5125 Frequency : always for specific kernels and striping counts
5127 Description: client may get "Matching packet too big" without ACL support
5128 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
5129 when trying to access files in certain configurations. The
5130 clients should in fact be denied when mounting because they do
5131 not understand ACLs.
5134 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
5136 Description: many clients connecting with IO in progress causes connect timeouts
5137 Details : Avoid synchronous journal commits to avoid delays caused by many
5138 clients connecting/disconnecting when bulk IO is in progress.
5139 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
5140 OST_IO_PORTAL to avoid delays behind potentially many pending
5144 Frequency : occasionally with multiple writers to a single file
5146 Description: shared writes to file may result in wrong size reported by stat()
5147 Details : Allow growing of kms when extent lock is cancelled
5150 Frequency : always with random mmap IO to multi-striped file
5152 Description: mmap write might be lost if we are writing to a 'hole' in stripe
5153 Details : Only if the hole is at the end of OST object so that kms is too
5154 small. Fix is to increase kms accordingly in ll_nopage.
5157 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
5159 Description: writes to a missing object would leak memory on the OST
5160 Details : If there is an inconsistency between the MDS and OST filesystems,
5161 such that the MDS references an object that doesn't exist, writes
5162 to that object will leak memory due to incorrect cleanup in the
5163 error handling path, eventually running out of memory on the OST.
5168 Description: Creating too long symlink causes lustre errors
5169 Details : Check symlink and name lengths before sending requests to MDS.
5172 Frequency : only if flock is enabled (not on by default)
5174 Description: posix locks not released on fd closure on 2.6.9+
5175 Details : We failed to add posix locks to list of inode locks on 2.6.9+
5176 kernels, this caused such locks not to be released on fd close and
5177 then assertions on fs unmount about still used locks.
5180 Frequency : MDS failover only, very rarely
5182 Description: clients may get ASSERTION(granted_lock != NULL)
5183 Details : When request was taking a long time, and a client was resending
5184 a getattr by name lock request. The were multiple lock requests
5185 with the same client lock handle and
5186 mds_getattr_name->fixup_handle_for_resent_request found one of the
5187 lock handles but later failed with ASSERTION(granted_lock != NULL).
5192 Description: handle->h_buffer_credits > 0, assertion failure
5193 Details : h_buffer_credits is zero after truncate, causing assertion
5194 failure. This patch extends the transaction or creates a new
5198 Frequency : NFS re-export or patchless client
5199 Bugzilla : 11179, 10796
5200 Description: Crash on NFS re-export node (__d_move)
5201 Details : We do not want to hash the dentry if we don't have a lock.
5202 But if this dentry is later used in d_move, we'd hit uninitialised
5203 list head d_hash, so we just do this to init d_hash field but
5204 leave dentry unhashed.
5207 Frequency : NFS re-export or patchless client
5209 Description: NFS exports has problem with symbolic link
5210 Details : lustre client didn't properly install dentry when re-exported
5211 to NFS or running patchless client.
5214 Frequency : NFS re-export or patchless client
5216 Description: Various nfs/patchless fixes.
5217 Details : fixes reuse disconected alias for lookup process - this fixes
5218 warning "find_exported_dentry: npd != pd",
5219 fix permission error with open files at nfs.
5220 fix apply umask when do revalidate.
5223 Frequency : occasional
5225 Description: Crash on NFS re-export node
5226 Details : calling clear_page() on the wrong pointer triggered oops in
5227 generic_mapping_read().
5230 Frequency : rarely, using O_DIRECT IO
5232 Description: unaligned directio crashes client with LASSERT
5233 Details : check for unaligned buffers before trying any requests.
5236 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
5238 Description: stale data returned from RAID cache
5239 Details : If only a small amount of IO is done to the RAID device before
5240 reading it again it is possible to get stale data from the RAID
5241 cache instead of reading it from disk.
5244 Frequency : always for sles10 kernel
5246 Description: sles10 support
5247 Details : ll_follow_link: compile fixes and using of nd_set_link
5248 under newer kernels.
5251 Frequency : depends on arch, kernel and compiler version, always on sles10
5254 Description: recursive or deep enough symlinks cause stack overflow
5255 Details : getting rid of large stack-allocated variable in
5259 Frequency : depends on hardware
5261 Description: lustre write performance loss in the SLES10 kernel
5262 Details : the performance loss is caused by using of write barriers in the
5263 ext3 code. The SLES10 kernel turns barrier support on by
5264 default. The fix is to undo that change for ldiskfs.
5266 ------------------------------------------------------------------------------
5268 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
5270 * Support for kernels:
5271 2.6.9-42.0.3EL (RHEL 4)
5272 2.6.5-7.276 (SLES 9)
5273 2.4.21-47.0.1.EL (RHEL 3)
5274 2.6.12.6 vanilla (kernel.org)
5278 Frequency : quota enabled and large files being deleted
5280 Description: releasing more than 4GB of quota at once hangs OST
5281 Details : If a user deletes more than 4GB of files on a single OST it
5282 will cause the OST to spin in an infinite loop. Release
5283 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5288 Description: statfs data retrieved from /proc may be stale or zero
5289 Details : When reading per-device statfs data from /proc, in the
5290 {kbytes,files}_{total,free,avail} files, it may appear
5291 as zero or be out of date.
5294 Frequency : systems with MD RAID1 external journal devices
5296 Description: lconf's call to blkid is confused by RAID1 journal devices
5297 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5298 of returning all block devices that match the journal UUID.
5301 Frequency : always, for aggregate stripe size over 4GB
5303 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5304 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5305 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5308 Frequency : always if "lfs find" used on a local file/directory
5310 Description: "lfs find" segfaults if used on a local file/directory
5311 Details : The case where a directory component was not specified wasn't
5312 handled correctly. Handle this properly.
5315 Frequency : always on ppc64
5317 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5318 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5319 perform endianness conversion.
5322 Frequency : rarely (truncate to non-zero file size after write under load)
5323 Bugzilla : 10730, 10687
5324 Description: Files padded with zeros to next 4K multiple
5325 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5326 it is possible that files that are truncated to a non-zero size
5327 immediately after being written are filled with zero bytes beyond
5328 the truncated size. No file data is lost.
5330 Severity : enhancement
5332 Description: Allow recovery/failover for liblustre clients.
5333 Details : liblustre clients were unaware of failover configurations until
5336 Severity : enhancement
5338 Description: user file locks should fail when not mounting with flock option
5339 Details : Set up an error-returning stub in ll_file_operations.lock field
5340 to prevent incorrect behaviour when client is mounted without
5341 flock option. Also, set up properly f_op->flock field for
5345 Frequency : always on ia64
5347 Description: "lfs df" loops on printing out MDS statfs information
5348 Details : The obd_ioctl_data was not initialized and in some systems
5349 this caused a failure during the ioctl that did not return
5350 an error. Initialize the struct and return an error on failure.
5353 Frequency : SLES 9 only
5355 Description: Error of copying files with lustre special EAs as root
5356 Details : Client side always return success for setxattr call for lustre
5357 special xattr (currently only "trusted.lov").
5360 Frequency : rarely on clusters with both ia64+i386 clients
5362 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5363 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5364 clients are doing concurrent writes to the same file it is
5365 possible that the ia64 clients may LASSERT because the OST
5366 extent locks are not PAGE_SIZE aligned. Ensure that grown
5367 locks are always aligned on the request boundary.
5370 Frequency : specific use, occasional
5372 Description: Overwriting in use executable truncates on-disk binary image
5373 Details : If one node attempts to overwrite an executable in use by
5374 another node, we now correctly return ETXTBSY instead of
5375 truncating the file.
5377 Severity : enhancement
5379 Description: Async OSC create to avoid the blocking unnecessarily.
5380 Details : If a OST has no remain object, system will block on the creating
5381 when need to create a new object on this OST. Now, ways use
5382 pre-created objects when available, instead of blocking on an
5383 empty osc while others are not empty. If we must block, we block
5384 for the shortest possible period of time.
5389 Description: chmod on Lustre root is propagated to other clients
5390 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5391 invalid by the follow_mount time.
5394 Frequency : liblustre clients only
5396 Description: Race in 'instant cancel' lock handling could lead to such locks
5397 never to be granted in case of SMP MDS
5398 Details : Do not destroy not yet granted but cbpending locks in
5402 Frequency : replay/resend of open
5404 Description: non null lock assetion failure in mds_intent_policy
5405 Details : Trying to replay/resend lockless open requests resulted in
5406 mds_open() returning 0 with no lock. Now it sets a flag if
5407 a lock is going to be returned.
5409 Severity : enhancement
5411 Description: Checksum enhancements
5412 Details : New checksum enhancements allow for resending RPCs that failed
5415 Severity : enhancement
5417 Description: Tunables on number of dirty pages in cacche
5418 Details : Allow to set limit on number of dirty pages cached.
5423 Description: client crash on unmount - lock still has references
5424 Details : In some error handling cases it was possible to leak a lock
5425 reference on a client while accessing a file. This was not
5426 harmful to the client during operation, but would cause the
5427 client to crash when the filesystem is unmounted.
5430 Frequency : specific case, rare
5432 Description: ETXTBSY on mds though file not in use by client
5433 Details : ETXTBSY is no longer incorrectly returned when attempting to
5434 chmod or chown a directory that the user previously tried to
5435 execute or a currently-executing binary.
5438 Frequency : extremely rare except on liblustre-based clients
5440 Description: Lustre space not freed when files are deleted
5441 Details : Clean up open-unlinked files after client eviction. Previously
5442 the unlink was skipped and the files remained as orphans.
5447 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5448 Details : In some cases it was possible to send a blocking callback to
5449 a client doing a glimpse, even though that client didn't get
5450 a lock granted. When the glimpse lock is cancelled on the OST
5451 the freed lock is left on the waiting list and corrupted the list.
5454 Frequency : all core dumps
5456 Description: Broke core dumps to lustre
5457 Details : Negative dentry may be unhashed if parent does not have UPDATE
5458 lock, but some callers, e.g. do_coredump, expect dentry to be
5459 hashed after successful create, hash it in ll_create_it.
5461 ------------------------------------------------------------------------------
5463 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5465 * Support for kernels:
5466 2.6.9-42.0.2.EL (RHEL 4)
5467 2.6.5-7.276 (SLES 9)
5468 2.4.21-40.EL (RHEL 3)
5469 2.6.12.6 vanilla (kernel.org)
5473 Frequency : always on RHEL 3
5475 Description: Number of open files grows over time
5476 Details : The number of open files grows over time, whether or not
5477 Lustre is started. This was due to a filp leak introduced
5478 by one of our kernel patches.
5480 ------------------------------------------------------------------------------
5482 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5484 * Support for kernels:
5485 2.6.9-42.EL (RHEL 4)
5486 2.6.5-7.267 (SLES 9)
5487 2.4.21-40.EL (RHEL 3)
5488 2.6.12.6 vanilla (kernel.org)
5493 Bugzilla : 5719, 9635, 9792, 9684
5494 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5495 Details : If a server is under heavy load and cannot reply to new
5496 connection requests before the client resends the (re)connect,
5497 the connection handling code can behave badly if two service
5498 threads are concurrently handing separate (re)connections from
5499 the same client. Add better locking to the connection handling
5500 code, and ensure that only a single connection will be processed
5501 for a given client UUID, even if the lock is dropped.
5503 Severity : enhancement
5505 Description: add TCP zero-copy support to kernel
5506 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5507 sends if the hardware supports scatter-gather and checksumming.
5508 This allows socklnd to do client-write and server-read more
5509 efficiently and reduce CPU utilization from skbuf copying.
5512 Frequency : only if NFS exporting from client
5514 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5515 Details : If setting attributes on a file created under NFS that had
5516 never been opened it would be possible to oops the client
5517 if the file had no objects.
5521 Bugzilla : 9326, 10402, 10897
5522 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5523 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5524 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5525 before ptlrpcd_wake() was called and the request was freed.
5528 Frequency : always for liblustre
5530 Description: liblustre client does MDS+OSTs setattr RPC for each write
5531 Details : When doing a write from a liblustre client, the client
5532 incorrectly issued an RPC to the MDS and each OST the file was
5533 striped over in order to update the timestamps. When writing
5534 with small chunks and many clients this could overwhelm the MDS
5535 with RPCs. In all cases it would slow down the write because
5536 these RPCs are unnecessary.
5538 Severity : enhancement
5540 Description: allow number of MDS service threads to be changed at module load
5541 Details : It is now possible to change the number of MDS service threads
5542 running. Adding "options mds mds_num_threads={N}" to the MDS's
5543 /etc/modprobe.conf will set the number of threads for the next
5544 time Lustre is restarted (assuming the "mds" module is also
5545 reloaded at that time). The default number of threads will
5546 stay the same, 32 for most systems.
5551 Description: OST crash if filesystem is unformatted or corrupt
5552 Details : If an OST is started on a device that has never been formatted
5553 or if the filesystem is corrupt and cannot even mount then the
5554 error handling cleanup routines would dereference a NULL pointer.
5559 Description: NULL pointer deref in llap_from_page.
5560 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5561 incorrect) mapping if the page was truncated/reclaimed while it was
5562 searched for. Check for this condition and skip such pages when
5563 doing readahead. Introduce extra check to llap_from_page() to
5564 verify page->mapping->host is non-NULL (so page is not anonymous).
5567 Frequency : Sometimes when using sys_sendfile
5569 Description: "page not covered by a lock" warnings from ll_readpage
5570 Details : sendfile called ll_readpage without right page locks present.
5571 Now we introduced ll_file_sendfile that does necessary locking
5572 around call to generic_file_sendfile() much like we do in
5576 Frequency : with certain MDS communication failures at client mount time
5578 Description: NULL pointer deref after failed client mount
5579 Details : a client connection request may delayed by the network layer
5580 and not be sent until after the PTLRPC layer has timed out the
5581 request. If the client fails the mount immediately it will try
5582 to clean up before the network times out the request. Add a
5583 reference from the request import to the obd device and delay
5584 the cleanup until the network drops the request.
5587 Frequency : occasionally during client (re)connect
5589 Description: assertion failure during client (re)connect
5590 Details : processing a client connection request may be delayed by the
5591 client or server longer than the client connect timeout. This
5592 causes the client to resend the connection request. If the
5593 original connection request is replied in this interval, the
5594 client may trip an assertion failure in ptlrpc_connect_interpret()
5595 which thought it would be the only running connect process.
5598 Frequency : only with obd_echo servers and clients that are rebooted
5600 Description: kernel BUG accessing uninitialized data structure
5601 Details : When running an obd_echo server it did not start the ping_evictor
5602 thread, and when a client was evicted an uninitialized data
5603 structure was accessed. Start the ping_evictor in the RPC
5604 service startup instead of the OBD startup.
5606 Severity : enhancement
5607 Bugzilla : 10193 (patchless)
5608 Description: Remove dependency on various unexported kernel interfaces.
5609 Details : No longer need reparent_to_init, exit_mm, exit_files,
5610 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5613 Frequency : rare (only users of deprecated and unsupported LDAP config)
5615 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5616 Details : LDAP apparently contains 'lustreName' attributes instead of
5617 'name'. A simple remapping of the name is sufficient.
5620 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5622 Description: waiting_locks_callback trips kernel BUG if client is evicted
5623 Details : Running with the dump_on_timeout debug flag turned on makes
5624 it possible that the waiting_locks_callback() can try to dump
5625 the Lustre kernel debug logs from an interrupt handler. Defer
5626 this log dumping to the expired_lock_main() thread.
5628 Severity : enhancement
5630 Description: Support NFS exporting on 2.6 kernels.
5631 Details : Implement non-rawops metadata methods for NFS server to use without
5632 changing NFS server code.
5635 Frequency : very rare (synthetic metadata workload only)
5637 Description: two racing renames might cause an MDS thread to deadlock
5638 Details : Running the "racer" program may cause one MDS thread to rename
5639 a file from being the source of a rename to being the target of
5640 a rename at exactly the same time that another thread is doing
5641 so, and the second thread has already enqueued these locks after
5642 doing a lookup of the target and is trying to relock them in
5643 order. Ensure that we don't try to re-lock the same resource.
5646 Frequency : only very large systems with liblustre clients
5648 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5649 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5650 all OSTs in parallel. This allows the eviction of stale liblustre
5651 clients to proceed much faster than if they were done in series,
5652 and also offers similar improvements for other set_info RPCs.
5657 Description: excessive CPU usage during initial read phase on client
5658 Details : During the initial read phase on a client, it would agressively
5659 retry readahead on the file, consuming too much CPU and impacting
5660 performance (since 1.4.5.8). Improve the readahead algorithm
5661 to avoid this, and also improve some other common cases (read
5662 of small files in particular, where "small" is files smaller than
5663 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5668 Description: MDS crash when receiving packet with unknown intent.
5669 Details : Do not LBUG in unknown intent case, just return -EFAULT
5671 Severity : enhancement
5672 Bugzilla : 9293, 9385
5673 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5674 Details : Do not serialize getattr (non-intent version) and statfs.
5677 Frequency : occasional, when OST network is overloaded/intermittent
5679 Description: client evicted by OST after bulk IO timeout
5680 Details : If a client sends a bulk IO request (read or write) the OST
5681 may evict the client if it is unresposive to its data GET/PUT
5682 request. This is incorrect if the network is overloaded (takes
5683 too long to transfer the RPC data) or dropped the OST GET/PUT
5684 request. There is no need to evict the client at all, since
5685 the pinger and/or lock callbacks will handle this, and the
5686 client can restart the bulk request.
5689 Frequency : Always when mmapping file with no objects
5691 Description: client crashes when mmapping file with no objects
5692 Details : Check that we actually have objects in a file before doing any
5693 operations on objects in ll_vm_open, ll_vm_close and
5699 Description: Request leak when working with deleted CWD
5700 Details : Introduce advanced request refcount tracking for requests
5701 referenced from lustre intent.
5703 Severity : Enhancement
5705 Description: Cache open file handles on client.
5706 Details : MDS now will return special lock along with openhandle, if
5707 requested and client is allowed to hold openhandle, even if unused,
5708 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5709 closing files for every read/write openration.
5711 Severity : Enhancement
5713 Description: Cache open negative dentries on client when possible.
5714 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5715 negative dentries on lock revocation.
5720 Description: Remounting a client read-only wasn't possible with a zconf mount
5721 Details : It wasn't possible to remount a client read-only with llmount.
5723 Severity : enhancement
5724 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5725 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5726 MPICH is included to add Lustre-specific ADIO interfaces.
5727 This is based closely on the UFS ADIO layer and only differs
5728 in file creation, in order to allow the OST striping to be set.
5729 This is user-contributed code and not supported by CFS.
5734 Description: extended inode attributes (immutable, append-only) work improperly
5735 when 2.4 and 2.6 kernels are used on client/server or vice versa
5736 Details : Introduce kernel-independent values for these flags.
5738 Severity : enhancement
5741 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5742 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5743 granularity. Now, Fractional MB granularity is be supported,
5744 this is very useful for low memory system.
5746 Severity : enhancement
5748 Description: Getattr by fid
5749 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5750 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5751 have locks and avoids getattr rpc requests for it on every stat.
5754 Frequency : Always, for filesystems larger than 2TB
5756 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5757 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5758 filesystems larger than 2TB to be mounted with mballoc enabled.
5761 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5763 Description: filesystem corruption for non-standard kernels and very large OSTs
5764 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5765 filesystems larger than 2TB is mounted then the kernel will
5766 silently corrupt the start of the filesystem. CONFIG_LBD is
5767 enabled for all CFS-supported kernels, but the possibility of
5768 this happening with a modified kernel config exists.
5770 Severity : enhancement
5772 Description: add client O_DIRECT support for 2.6 kernels
5773 Details : It is now possible to do O_DIRECT reads and writes to files
5774 in the Lustre client mountpoint on 2.6 kernel clients.
5776 Severity : enhancement
5778 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5779 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5780 parallel, not waiting for response from every OST before sending
5781 a rpc to the next OST.
5786 Description: setattr vs write race when updating file timestamps
5787 Details : Client processes that update a file timestamp into the past
5788 right after writing to the file (e.g. tar) it is possible that
5789 the updated file modification time can be reset to the current
5790 time due to a race between processing the setattr and write RPC.
5792 Severity : enhancement
5794 Description: Bring 'lfs find' closer in line with regular Linux find.
5795 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5796 -print0 options and obtains all the needed info through the lustre
5799 Severity : enhancement
5801 Description: support up to 1024 configured devices on one node
5802 Details : change obd_dev array from statically allocated to dynamically
5803 allocated structs as they are first used to reduce memory usage
5808 Description: Flush dirty partially truncated pages during truncate
5809 Details : Immediatelly flush partially truncated pages in filter_setattr,
5810 this way we completely avoid having any pages in page cache on OST
5811 and can retire ugly workarounds during writes to flush such pages.
5816 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5817 Details : protect inode from truncation within vfs_unlink() context
5818 just take a reference before calling vfs_unlink() and release it
5819 when parent's i_sem is free.
5822 Frequency : always, if extents are used on OSTs
5824 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5825 Details : OSTs using the extents format would not zero the high 16 bits of
5826 the index physical block number. This is not a problem for any
5827 OST filesystems smaller than 16TB, and no kernels support ext3
5828 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5829 new/modified files) and can be fixed for existing filesystems
5830 with e2fsprogs-1.39-cfs1.
5835 Description: import connection selection may be incorrect if timer wraps
5836 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5837 import connections to be ignored if the 32-bit jiffies counter
5838 wraps. Use a 64-bit jiffies counter.
5841 Frequency : during server recovery
5843 Description: crash after server is denying duplicate export
5844 Details : If clients are resending connect requests to the server, the
5845 server refuses to allow a client to connect multiple times.
5846 Fixed a bug in the handling of this case.
5849 Frequency : very large clusters immediately after boot
5851 Description: LNET request buffers exhausted under heavy short-term load
5852 Details : If a large number of client requests are generated on a service
5853 that has previously never seen so many requests it is possible
5854 that the request buffer growth cannot keep up with the spike in
5855 demand. Instead of dropping incoming requests, they are held in
5856 the LND until the RPC service can accept more requests.
5859 Frequency : Sometimes during replay
5861 Description: Assertion failure in ll_local_open after replay.
5862 Details : If replay happened on an open request reply before we were able
5863 to set replay handler, reply will become not swabbed tripping the
5864 assertion in ll_local_open. Now we set the handler right after
5865 recognising of open request
5868 Frequency : very rare
5870 Description: kernel reports "badness in vsnprintf"
5871 Details : Reading from the "recovery_status" /proc file in small chunks
5872 may cause a negative length in lprocfs_obd_rd_recovery_status()
5873 call to vsnprintf() (which is otherwise harmless). Exit early
5874 if there is no more space in the output buffer.
5876 Severity : enhancement
5878 Description: clear OBD RPC statistics by writing to them
5879 Details : It is now possible to clear the OBD RPC statistics by writing
5880 to the "stats" file.
5885 Description: Client mtime is not the same on different clients after utimes
5886 Details : In some cases, the client was using the utimes() syscall on
5887 a file cached on another node. The clients now validate the
5888 ctime from the MDS + OSTs to determine which one is right.
5893 Description: Inability to activate failout mode
5894 Details : lconf script incorrectly assumed that in python string's numeric
5895 value is used in comparisons.
5898 Frequency : always with multiple stripes per file
5900 Description: Inefficient object allocation for mutli-stripe files
5901 Details : When selecting which OSTs to stripe files over, for files with
5902 a stripe count that divides evenly into the number of OSTs,
5903 the MDS is always picking the same starting OST for each file.
5904 Return the OST selection heuristic to the original design.
5909 Description: mount failures may take full timeout to return an error
5910 Details : Under some heavy load conditions it is possible that a
5911 failed mount can wait for the full obd_timeout interval,
5912 possibly several minutes, before reporting an error.
5913 Instead return an error as soon as the status is known.
5915 ------------------------------------------------------------------------------
5917 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5919 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5920 INTEROPERATE with older versions automatically. Please read the
5921 user documentation before upgrading any part of a live system.
5922 * WARNING: Lustre networking configuration changes are required with
5923 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5926 * Support for kernels:
5927 2.6.9-22.0.2.EL (RHEL 4)
5928 2.6.5-7.244 (SLES 9)
5929 2.6.12.6 vanilla (kernel.org)
5932 Severity : enhancement
5933 Bugzilla : 7981/8208
5934 Description: Introduced Lustre Networking (LNET)
5935 Details : LNET is new networking infrastructure for Lustre, it includes
5936 a reorganized network configuration mode (see the user
5937 documentation for full details) as well as support for routing
5938 between different network fabrics. Lustre Networking Devices
5939 (LNDs) for the supported network fabrics have also been
5940 created for this new infrastructure.
5942 Severity : enhancement
5943 Description: Introduced Access control lists
5944 Details : clients can set ACLs on files and directories in order to have
5945 more fine-grained permissions than the standard Unix UGO+RWX.
5946 The MDS must be started with the "-o acl" mount option.
5948 Severity : enhancement
5949 Description: Introduced filesystem quotas
5950 Details : Administrators may now establish per-user quotas on the
5953 Severity : enhancement
5955 Description: Configuration change for the XT3
5956 The PTLLND is now used to run Lustre over Portals on the XT3
5957 The configure option(s) --with-cray-portals are no longer used.
5958 Rather --with-portals=<path-to-portals-includes> is used to
5959 enable building on the XT3. In addition to enable XT3 specific
5960 features the option --enable-cray-xt3 must be used.
5965 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5966 Details : A race between allocating a new llcd and re-getting the llcd_lock
5967 allowed another thread to grab newly-allocated llcd.
5969 Severity : enhancement
5971 Description: 2.6 OST async journal commit and locking fix to improve performance
5972 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5973 2.6 kernels are now async as they already were in 2.4 kernels so
5974 that they can commit concurrently with the network bulk transfer.
5975 For block-allocated files the filter allocation semaphore is held
5976 to avoid filesystem fragmentation during allocation. BKL lock
5977 removed for 2.6 xattr operations where it is no longer needed.
5982 Description: lconf incorrectly determined whether two IP networks could talk
5983 Details : In some more complicated routing and multiple-network
5984 configurations, lconf will avoid trying to make a network
5985 connection to a disjoint part of the IP space. It was doing the
5986 math incorrectly for one set of cases.
5991 Description: Fix for potential infinite loop processing records in an llog.
5992 Details : If an llog record is corrupted/zeroed, it is possible to loop
5993 forever in llog_process(). Validate the llog record length
5994 and skip the remainder of the block on error.
5997 Frequency : occasional (liblustre only)
5999 Description: liblustre could not open files whose last component is a symlink
6000 Details : sysio_path_walk() would incorrectly pass the open intent to
6001 intermediate path components.
6004 Frequency : rare (liblustre only with non-standard tuning)
6005 Bugzilla : 7201 (7350)
6006 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
6007 Details : llu_lookup_finish_locks() tries to set lock data on a lock
6008 after it has been released, only do this for referenced locks
6010 Severity : enhancement
6012 Description: specifying an (invalid) directory default stripe_size of -1
6013 would reset the directory default striping
6014 Details : stripe_size -1 was used internally to signal directory stripe
6015 removal, now use "all default" to signal dir stripe removal
6016 as a directory striping of "all default" is not useful
6019 Frequency : common for large clusters running liblustre clients
6021 Description: doing an ls when liblustre clients are running is slow
6022 Details : sending a glimpse AST to a liblustre client waits for every AST
6023 to time out, as liblustre clients will not respond. Since they
6024 cannot cache data we refresh the OST lock LVB from disk instead.
6026 Severity : enhancement
6028 Description: doing an ls at the same time as file IO can be slow
6029 Details : enqueue and other "small" requests can be blocked behind many
6030 large IO requests. Create a new OST IO portal for non-IO
6031 requests so they can be processed faster.
6034 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
6036 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
6037 created file with mode 000
6038 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
6039 allow an ftruncate() as a non-root user to a file with mode 000.
6040 Lustre now respects this flag to disable mode checks when
6041 truncating a file owned by the user
6044 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
6047 Description: Revoking locks from clients that went dead or catatonic might take
6049 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
6050 cancellation of such locks instant on servers without waiting for
6051 any reply from clients. Clients drops these locks when cancel
6052 notification from server is received without replying.
6055 Frequency : liblustre-only, when liblustre client dies or becomes busy
6057 Description: Doing ls on Linux clients can take a long time with active
6059 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
6060 granting such locks to it in the first place if possible. Locks
6061 are taken by proxy on the OST during the read or write and
6062 dropped immediately afterward. Add connect flags handling, do
6063 not grant locks to liblustre clients for glimpse ASTs.
6065 Severity : enhancement
6067 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
6068 Details : Existing read-ahead algorithm is tuned for the case of streamlined
6069 sequential reads and behaves badly with applications doing random
6070 reads. Improve it by reading ahead at least read region, and
6071 avoiding excessive large RPC for small reads.
6073 Severity : enhancement
6075 Description: Creating more than 1000 files for a single job may cause a load
6076 imbalance on the OSTs if there are also a large number of OSTs.
6077 Details : qos_prep_create() uses an OST index reseed value that is an
6078 even multiple of the number of available OSTs so that if the
6079 reseed happens in the middle of the object allocation it will
6080 still utilize the OSTs as uniformly as possible.
6085 Description: OST or MDS may oops in ping_evictor_main()
6086 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
6087 but doesn't restart at beginning of obd_exports_timed list
6090 Severity : enhancement
6092 Description: improve by-nid export eviction on the MDS and OST
6093 Details : allow multiple exports with the same NID to be evicted at one
6094 time without re-searching the exports list.
6097 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
6099 Description: MDS may oops in groups_free()
6100 Details : in rare race conditions a newly allocated group_info struct is
6101 freed again, and this can be NULL. The 2.4 compatibility code
6102 for groups_free() checked for a NULL pointer, but 2.6 did not.
6105 Frequency : common for liblustre clients doing little filesystem IO
6106 Bugzilla : 9352, 7313
6107 Description: server may evict liblustre clients accessing contended locks
6108 Details : if a client is granted a lock or receives a completion AST
6109 with a blocking AST already set it would not reply to the AST
6110 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
6113 Frequency : lfs setstripe, only systems with more than 160 OSTs
6115 Description: unable to set striping with a starting offset beyond OST 160
6116 Details : llapi_create_file() incorrectly limited the starting stripe
6117 index to the maximum single-file stripe count.
6120 Frequency : LDAP users only
6122 Description: lconf did not handle in-kernel recovery with LDAP properly
6123 Details : lconf/LustreDB get_refs() is searching the wrong namespace
6125 Severity : enhancement
6127 Description: bind OST threads to NUMA nodes to improve performance
6128 Details : all OST threads are uniformly bound to CPUs on a single NUMA
6129 node and do their allocations there to localize memory access
6131 Severity : enhancement
6133 Description: llmount can determine client NID directly from Myrinet (GM)
6134 Details : the client NID code from gmnalnid was moved directly into
6135 llmount, removing the need to use this or specifying the
6136 client NID explicitly when mounting GM clients with zeroconf
6139 Frequency : if client is started with down MDS
6141 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
6142 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
6143 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
6144 ptlrpc_import_delayed_req() because we didn't check if the
6145 request was interrupted, and we also didn't break out of the
6146 event loop if there was no timeout
6151 Description: data loss during non-page-aligned writes to a single file from
6152 both multiple nodes and multiple threads on one node at same time
6153 Details : updates to KMS and lsm weren't protected by common lock. Resulting
6154 inconsistency led to false short-reads, that were cached and later
6155 used by ->prepare_write() to fill in partially written page,
6156 leading to data loss.
6159 Frequency : always, if lconf --abort_recovery used
6161 Description: lconf --abort_recovery fails with 'Operation not supported'
6162 Details : lconf was attempting to abort recovery on the MDT device and not
6165 Severity : enhancement
6167 Description: remove cleanup logs
6168 Details : replace lconf-generated cleanup logs with lustre internal
6169 cleanup routines. Eliminates the need for client-cleanup and
6172 Severity : enhancement
6174 Description: add support for EAs (user and system) on lustre filesystems
6175 Details : it is now possible to store extended attributes in the Lustre
6176 client filesystem, and with the user_xattr mount option it
6177 is possible to allow users to store EAs on their files also
6179 Severity : enhancement
6181 Description: Add possibility (config option) to show minimal available OST free
6183 Details : When compiled with --enable-mindf configure option, statfs(2)
6184 (and so, df) will return least minimal free space available from
6185 all OSTs as amount of free space on FS, instead of summary of
6186 free spaces of all OSTs.
6188 Severity : enhancement
6190 Description: do not expand extent locks acquired on OST-side
6191 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
6192 by server: they are not cached anyway.
6195 Frequency : when mmap is used/binaries executed from Lustre
6197 Description: Unmmap pages before throwing them away from read cache.
6198 Details : llap_shrink cache now attempts to unmap pages before discarding
6199 them (if unmapping failed - do not discard). SLES9 kernel has
6200 extra checks that trigger if this unmapping is not done first.
6205 Description: lconf didn't resolve symlinks before checking to see whether a
6206 given mountpoint was already in use
6209 Frequency : when migrating failover services
6210 Bugzilla : 6395, 9514
6211 Description: When migrating a subset of services from a node (e.g. failback
6212 from a failover service node) the remaining services would
6213 time out and evict clients.
6214 Details : lconf --force (implied by --failover) sets the global obd_timeout
6215 to 5 seconds in order to quickly disconnect, but this caused
6216 other RPCs to time out too quickly. Do not change the global
6217 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
6219 Severity : enhancement
6220 Frequency : if MDS is started with down OST
6221 Bugzilla : 9439,5706
6222 Description: Allow startup/shutdown of an MDS without depending on the
6223 availability of the OSTs.
6224 Details : Asynchronously call mds_lov_synchronize during MDS startup.
6225 Add appropriate locking and lov-osc refcounts for safe
6226 cleaning. Add osc abort_inflight calls in case the
6227 synchronize never started.
6230 Frequency : occasional (Cray XT3 only)
6232 Description: root not authorized to access files in CRAY_PORTALS environment
6233 Details : The client process capabilities were not honoured on the MDS in
6234 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
6235 been accessed by an authorized user then root was able to access
6236 the file on the local client also. The root user capabilities
6237 are now allowed on the MDS, as this environment has secure UID.
6240 Frequency : occasional
6242 Description: ldiskfs "too long searching" message happens too often
6243 Details : A debugging message (otherwise harmless) prints too often on
6244 the OST console. This has been reduced to only happen when
6245 there are fragmentation problems on the filesystem.
6250 Description: Division by zero in statfs when all OSCs are inactive
6251 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
6252 lov_statfs divides by value returned by lov_get_stripecnt().
6256 Bugzilla : 9489, 3273
6257 Description: First write from each client to each OST was only 4kB in size,
6258 to initialize client writeback cache, which caused sub-optimal
6259 RPCs and poor layout on disk for the first writen file.
6260 Details : Clients now request an initial cache grant at (re)connect time
6261 and so that they can start streaming writes to the cache right
6262 away and always do full-sized RPCs if there is enough data.
6263 If the OST is rebooted the client also re-establishes its grant
6264 so that client cached writes will be honoured under the grant.
6269 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
6270 Details : Now I/O RPCs go to different portal number and (presumably) fast
6271 lock requests (and glimses) and other RPCs get their own service
6272 threads pool that should be able to service those RPCs
6275 Severity : enhancement
6277 Description: Ability to exchange lustre version between client and servers and
6278 issue warnings at client side if client is too old. Also for
6279 liblustre clients there is ability to refuse connection of too old
6281 Details : New 'version' field is added to connect data structure that is
6282 filled with version info. That info is later checked by server and
6286 Frequency : rare, liblustre only.
6287 Bugzilla : 9296, 9581
6288 Description: Two simultaneous writes from liblustre at offset within same page
6289 might proceed at the same time overwriting eachother with stale
6291 Details : I/O lock withing llu_file_prwv was released too early, before data
6292 actually was hitting the wire. Extended lock-holding time until
6293 server acknowledges receiving data.
6296 Frequency : extremely rare. Never observed in practice.
6298 Description: avoid generating lustre_handle cookie of 0.
6299 Details : class_handle_hash() generates handle cookies by incrementing
6300 global counter, and can hit 0 occasionaly (this is unlikely, but
6301 not impossible, because initial value of cookie counter is
6302 selected randonly). Value of 0 is used as a sentinel meaning
6303 "unassigned handle" --- avoid it. Also coalesce two critical
6304 sections in this function into one.
6306 Severity : enhancement
6308 Description: allow liblustre clients to delegate truncate locking to OST
6309 Details : To avoid overhead of locking, liblustre client instructs OST to
6310 take extent lock in ost_punch() on client's behalf. New connection
6311 flag is added to handle backward compatibility.
6313 Severity : enhancement
6314 Bugzilla : 4928, 7341, 9758
6315 Description: allow number of OST service threads to be specified
6316 Details : a module parameter allows the number of OST service threads
6317 to be specified via "options ost ost_num_threads={N}" in the
6318 OSS's /etc/modules.conf or /etc/modprobe.conf.
6322 Bugzilla : 6146, 9635, 9895
6323 Description: servers crash with bad pointer in target_handle_connect()
6324 Details : In rare cases when a client is reconnecting it was possible that
6325 the connection request was the last reference for that export.
6326 We would temporarily drop the export reference and get a new
6327 one, but this may have been the last reference and the export
6328 was just destroyed. Get new reference before dropping old one.
6330 Severity : enhancement
6331 Frequency : if client is started with failover MDS
6333 Description: Allow multiple MDS hostnames in the mount command
6334 Details : Try to read the configuration from all specified MDS
6335 hostnames during a client mount in case the "primary"
6338 Severity : enhancement
6340 Description: Stop sending data to evicted clients as soon as possible.
6341 Details : Check if the client we are about to send or are sending data to
6342 was evicted already. (Check is done every second of waiting,
6343 for which l_wait_event interface was extended to allow checking
6344 of exit condition at specified intervals).
6347 Frequency : rare, normally only when NFS exporting is done from client
6349 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6351 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6354 Frequency : always, if srand() is called before liblustre initialization
6356 Description: Liblustre uses system PRNG disturbing its usage by user application
6357 Details : Introduce internal to lustre fast and high-quality PRNG for
6358 lustre usage and make liblustre and some other places in generic
6359 lustre code to use it.
6361 Severity : enhancement
6362 Bugzilla : 9477, 9557, 9870
6363 Description: Verify that the MDS configuration logs are updated when xml is
6364 Details : Check if the .xml configuration logs are newer than the config
6365 logs stored on the MDS and report an error if this is the case.
6366 Request --write-conf, or allow starting with --old_conf.
6368 Severity : enhancement
6370 Description: Handle symlinks in the path when checking if Lustre is mounted.
6371 Details : Resolve intermediate symlinks when checking if a client has
6372 mounted a filesystem to avoid duplicate client mounts.
6377 Description: lconf can hit an error exception but still return success.
6378 Details : The lconf command catches the Command error exception at the top
6379 level script context and will exit with the associated exit
6380 status, but doesn't ensure that this exit status is non-zero.
6385 Description: failure of ptlrpc thread startup can cause oops
6386 Details : Starting a ptlrpc service thread can fail if there are a large
6387 number of threads or the server memory is very fragmented.
6388 Handle this without oopsing.
6391 Frequency : always, only if liblustre and non-default acceptor port was used
6393 Description: liblustre cannot connect to servers with non-default acceptor port
6394 Details : tcpnal_set_default_params() was not called and was therefore
6395 ignoring the environment varaible TCPNAL_PORT, as well as other
6396 TCPNAL_ environment variables
6401 Description: two objects could be created on the same OST for a single file
6402 Details : If an OST is down, in some cases it was possible to create two
6403 objects on a single OST for a single file. No problems other
6404 than potential performance impact and spurious error messages.
6408 Bugzilla : 5681, 9562
6409 Description: Client may oops in ll_unhash_aliases
6410 Details : Client dcache may become inconsistent in race condition.
6411 In some cases "getcwd" can fail if the current directory is
6417 Description: Inode refcounting problems in NFS export code
6418 Details : link_raw functions used to call d_instantiate without obtaining
6419 extra inode reference first.
6423 Bugzilla : 9942, 9903
6424 Description: Referencing freed requests leading to crash, memleaks with NFS.
6425 Details : We used to require that call to ll_revalidate_it was always
6426 followed by ll_lookup_it. Also with revalidate_special() it is
6427 possible to call ll_revalidate_it() twice for the same dentry
6428 even if first occurence returned success. This fix changes semantic
6429 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6430 reference on a request referred from the intent.
6431 ll_intent_release() then releases such a request.
6434 Frequency : rare, normally benchmark loads only
6436 Description: unlinked inodes were kept in memory on the client
6437 Details : If a client is repeatedly creating and unlinking files it
6438 can accumulate a lot of stale inodes in the inode slab cache.
6439 If there is no other client load running this can cause the
6440 client node to run out of memory. Instead flush old inodes
6441 from client cache that have the same inode number as a new inode.
6444 Frequency : SLES9 2.6.5 kernel and long filenames only
6445 Bugzilla : 9969, 10379
6446 Description: utime reports stale NFS file handle
6447 Details : SLES9 uses out-of-dentry names in some cases, which confused
6448 the lustre dentry revalidation. Change it to always use the
6452 Frequency : rare, unless heavy write-truncate concurrency is continuous
6453 Bugzilla : 4180, 6984, 7171, 9963, 9331
6454 Description: OST becomes very slow and/or deadlocked during object unlink
6455 Details : filter_destroy() was holding onto the parent directory lock
6456 while truncating+unlinking objects. For very large objects this
6457 may block other threads for a long time and slow overall OST
6458 responsiveness. It may also be possible to get a lock ordering
6459 deadlock in this case, or run out of journal credits because of
6460 the combined truncate+unlink. Solution is to do object truncate
6461 first in one transaction without parent lock, and then do the
6462 final unlink in a new transaction with the parent lock. This
6463 reduces the lock hold time dramatically.
6466 Frequency : rare, 2.4 kernels only
6468 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6469 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6470 before unmounting their filesystem in order to avoid deadlock.
6471 The kernel_locked() function in 2.4 kernels only checks whether
6472 the kernel lock is held, not whether it is this process that is
6473 holding it as 2.6 kernels do.
6478 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6479 Details : The client ptlrpc code may be trying to reconnect to a down
6480 server before a previous connection attempt has timed out.
6481 Increase the reconnect interval to be longer than the connection
6482 timeout interval to avoid sending duplicate connections to
6486 Frequency : echo_client brw_test command
6488 Description: fix echo_client to work with OST preallocated code
6489 Details : OST preallocation code (5137) didn't take echo_client IO path
6490 into account: echo_client calls filter methods outside of any
6491 OST thread and, hence, there is no per-thread preallocated
6492 pages and buffers to use. Solution: hijack pga pages for IO. As
6493 a byproduct, this avoids unnecessary data copying.
6497 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6498 Description: Client can oops in mdc_commit_close() after open replay
6499 Details : It was possible for the MDS to return an open request with no
6500 transaction number in mds_finish_transno() if the client was
6501 evicted, but without actually returning an error. Clients
6502 would later try to replay that open and may trip an assertion
6503 Simplify the client close codepath, and always return an error
6504 from the MDS in case the open is not successful.
6507 Frequency : rare, 2.6 OSTs only
6509 Description: OST may deadlock under high load on fragmented files
6510 Details : If there was a heavy load and highly-fragmented OST filesystems
6511 it was possible to have all the OST threads deadlock waiting on
6512 allocation of biovecs, because the biovecs were not released
6513 until the entire RPC IO was completed. Instead, release biovecs
6514 as soon as they are complete to ensure forward IO progress.
6516 Severity : enhancement
6518 Description: Support for specifying external journal device at mount
6519 Details : If an OST or MDS device is formatted with an external journal
6520 device, this device major/minor is stored in the ext3 superblock
6521 and may not be valid for failover. Allow detecting and
6522 specifying the external journal at mount time.
6527 Description: Mounting an MDS with pending unlinked files may cause oops
6528 Details : target_finish_recovery() calls mds_postrecov() which returned
6529 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6530 considers this an error and immediately begins cleaning up the
6531 lov, just after starting the mds_lov process
6533 Severity : enhancement
6535 Description: Implement 'lfs df' to report actual free space on per-OST basis
6536 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6537 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6538 usage of objects; '-h' to report in human readable format.
6540 ------------------------------------------------------------------------------
6542 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6549 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6550 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6551 at startup during orphan unlinking. Instead, skip buddy update
6552 before setup, as it will be regenerated after recovery is complete.
6557 Description: If an OST is inactive, its locks might reference stale inodes.
6558 Details : lov_change_cbdata() must iterate over all namespaces, even if
6559 they are inactive to clear inode references from the lock.
6561 Severity : enhancement
6562 Frequency : occasional, if non-standard max_dirty_mb used
6564 Description: Client will block write RPCs if not enough grant
6565 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6566 then the client will block writes while waiting for another RPC
6567 to complete instead of consuming its dirty limit. With change
6568 we get improved performance when max_dirty_mb is small.
6570 Severity : enhancement
6571 Bugzilla : 3389, 6253
6572 Description: Add support for supplementary groups on the MDS.
6573 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6574 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6575 lookups for user supplementary groups into a cache.
6579 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6580 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6581 for liblustre O_EXCL is enforced by MDS.
6585 Description: readdir never returns NULL in liblustre.
6586 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6587 dentry in struct dirent.
6591 Frequency : liblustre only, depends on application IO pattern
6592 Description: liblustre clients evicted if not contacting servers
6593 Details : Don't put liblustre clients into the ping_evictor list, so
6594 they will not be evicted by the pinger ever.
6596 Severity : enhancement
6598 Description: Add ability to evict clients by NID from MDS.
6599 Details : By echoing "nid:$NID" string into
6600 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6601 $NID would be instantly evicted from this MDS and from all active
6602 OSTs connected to it.
6606 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6607 Details : lookup_it_finish() used to query file size from OSTs that was not
6612 Description: service threads change working directory to that of init
6613 Details : Starting lustre service threads may pin the working directory
6614 of the parent thread, making that filesystem busy. Threads
6615 now change to the working directory of init to avoid this.
6619 Frequency : during shutdown only
6620 Description: shutdown with a failed MDS or OST can cause unmount to hang
6621 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6626 Frequency : 2.6 only
6627 Description: chmod/chown may include an extra supplementary group
6628 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6629 supplementary group and if none is specified this is used.
6632 Bugzilla : 5479 (6816)
6634 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6635 Details : If lookup is in progress on a file that is unlinked we might try
6636 to revalidate the inode and fail in revalidate after lookup is
6637 complete and ll_file_open() enqueues the open again but
6638 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6641 Frequency : always, if lconf --abort_recovery used
6643 Description: lconf --abort_recovery fails with 'Operation not supported'
6644 Details : lconf was attempting to abort recovery on the MDT device and not
6647 ------------------------------------------------------------------------------
6649 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6654 Frequency : rare (only unsupported configurations with a node running as an
6656 Bugzilla : 6514, 5137
6657 Description: Mounting a Lustre file system on a node running as an OST could
6659 Details : OSTs now preallocates memory needed to write out data at
6660 startup, instead of when needed, to avoid having to
6661 allocate memory in possibly low memory situations.
6662 Specifically, if the file system is mounted on on OST,
6663 memory pressure could force it to try to write out data,
6664 which it needed to allocate memory to do. Due to the low
6665 memory, it would be unable to do so and the node would
6666 become unresponsive.
6668 Severity : enhancement
6670 Description: Addition of lconf --service command line option
6671 Details : lconf now accepts a '--service <arg>' option, which is
6672 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6674 Severity : enhancement
6676 Description: Failover mode is now the default for OSTs.
6677 Details : By default, OSTs will now run in failover mode. To return to
6678 the old behaviour, add '--failout' to the lmc line for OSTs.
6680 Severity : enhancement
6682 Description: Health checks are now provided for MDS and OSTs
6683 Details : Additional detailed health check information on MSD and OSTs
6684 is now provided through the procfs health_check value.
6687 Frequency : occasional, depends on IO load
6689 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6690 after numerous create/delete cycles
6691 Details : The ext3 inode allocation policy would not allocate new inodes
6692 very well on the OSTs because there are no new directories
6693 being created. Instead we look for groups with free space if
6694 the parent directories are nearly full.
6699 Description: Network or server problems during mount may cause partially
6700 mounted clients instead of returning an error.
6701 Details : The config llog parsing code may overwrite the error return
6702 code during mount error handling, returning success instead
6708 Description: MDS can fail to allocate large reply buffers
6709 Details : After long uptimes the MDS can fail to allocate large reply
6710 buffers (e.g. zconf client mount config records) due to memory
6711 fragmentation or consumption by the buffer cache. Preallocate
6712 some large reply buffers so that these replies can be sent even
6713 under memory pressure.
6717 Frequency : rare (liblustre)
6718 Description: fsx running with liblustre complained that using truncate() to
6719 extend the file doesn't work. This patch corrects that issue.
6720 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6721 ATTR_SIZE and lsm use in llu_setattr_raw.
6725 Frequency : rare, only 2.6 kernels
6726 Description: Unusual file access patterns on the MDS may result in inode
6727 data being lost in very rare circumstances.
6728 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6729 code in the 2.6 kernel caused the inode and/or EA data not to
6730 be read from disk, causing single-file corruption.
6734 Frequency : rare, only 2.6 filesystems using extents
6735 Description: Heavy concurrent write and delete load may cause data corruption.
6736 Details : It was possible under high-load situations to have an extent
6737 metadata block in the block device cache from a just-unlinked
6738 file overwrite a newly-allocated data block. We now unmap any
6739 metadata buffers that alias just-allocated data blocks.
6743 Frequency : filesystems with default stripe_count larger than 77
6744 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6745 Details : lconf specifies an inode size of 4096 bytes when the default
6746 stripe_count is larger than 77. This conflicts with the default
6747 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6748 this case to avoid pinning too much memory for large EAs.
6750 ------------------------------------------------------------------------------
6752 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6757 Frequency : rare (extremely heavy IO load with hundreds of clients)
6759 Description: Client is evicted, gets IO error writing to file
6760 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6761 caused clients to be evicted for AST timeouts. The fixes in
6762 bug 5192 mean we no longer need to have such short AST timeouts
6763 so ldlm_timeout has been increased.
6766 Frequency : occasional during --force or --failover shutdown under load
6767 Bugzilla : 5949, 4834
6768 Description: Server oops/LBUG if stopped with --force or --failover under load
6769 Details : a collection of import/export refcount and cleanup ordering
6770 issues fixed for safer force cleanup
6773 Frequency : only filesystems larger than 120 OSTs
6774 Bugzilla : 5990, 6223
6775 Description: lfs getstripe would oops on a very large filesystem
6776 Details : lov_getconfig used kfree on vmalloc'd memory
6779 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6780 Bugzilla : 6242, 6243
6781 Description: reading from files that had been truncated to a non-zero size
6782 but never opened returned no data
6783 Details : ll_file_read() reads zeros from no-object files to EOF
6788 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6789 Details : There's little harm in aborting MDS/OSS recovery and letting it
6790 try again, so I removed the LASSERT and return an error instead.
6792 Severity : enhancement
6794 Description: New debugging infrastructure for tracking down data corruption
6795 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6796 (b) cover more of the client-side code path, and (c) try to narrow
6797 down where problems occurred
6801 Bugzilla : 3819, 4364, 4397, 6313
6802 Description: Racing close and eviction MDS could cause assertion in mds_close
6803 Details : It was possible to get multiple mfd references during close and
6804 client eviction, leading to one thread referencing a freed mfd.
6806 Severity: : enhancement
6807 Bugzilla : 3262, 6359
6808 Description: Attempts to reconnect to servers are now more aggressive.
6809 Details : This builds on the enhanced upcall-less recovery that was added
6810 in 1.4.2. When trying to reconnect to servers, clients will
6811 now try each server in the failover group every 10 seconds. By
6812 default, clients would previously try one server every 25 seconds.
6817 Description: After recovery, certain operations trigger a failed
6818 assertion on a client.
6819 Details : Failing over an mds, using lconf -d --failover, while a
6820 client was doing a readdir() call would cause the client to
6821 LBUG after recovery completed and the readdir() was resent.
6823 Severity : enhancement
6825 Description: Default groups are now added by lconf
6826 Details : You can now run lconf --group <servicename> without having to
6827 manually add groups with lmc.
6830 Frequency : occasional
6832 Description: Nodes with an elan id of 0 trigger a failed assertion
6835 Frequency : always when accessing e.g. tty/console device nodes
6837 Description: tty and some other devices nodes cannot be used on lustre
6838 Details : file's private_data field is used by device data and lustre
6839 values in there got lost. New field was added to struct file to
6840 store fs-specific private data.
6843 Frequency : when exporting Lustre via NFS
6845 Description: NFSD failed occasionally when looking up a path component
6846 Details : NFSD is looking up ".." which was broken in ext3 directories
6847 that had grown large enough to become hashed.
6850 Frequency : Clusters with multiple interfaces not on the same subnet
6852 Description: Nodes will repeatedly try to reconnect to an interface which it
6853 cannot reach and report an error to the log.
6854 Details : Extra peer list entries will be created by lconf with some peers
6855 unreachable. lconf now validates the peer before adding it.
6858 Frequency : Only if a default stripe is set on the filesystem root.
6860 Description: Setting a default stripe on the filesystem root prevented the
6861 filesystem from being remounted.
6862 Details : The client was sending extra request flags in the root getattr
6863 request and did not allocate a reply buffer for the dir EA.
6866 Frequency : occasional, higher if lots of files are accessed by one client
6867 Bugzilla : 6159, 6097
6868 Description: Client trips assertion regarding lsm mismatch/magic
6869 Details : While revalidating inodes the VFS looks up inodes with ifind()
6870 and in rare cases can find an inode that is being freed.
6871 The ll_test_inode() code will free the lsm during ifind()
6872 when it finds an existing inode and then the VFS later attaches
6873 this free lsm to a new inode.
6877 Bugzilla : 6422, 7030
6878 Description: MDS deadlock between mkdir and client eviction
6879 Details : Creating a new file via mkdir or mknod (starting a transaction
6880 and getting the ns lock) can deadlock with client eviction
6881 (gets ns lock and trying to finish a synchronous transaction).
6884 Frequency : occasional
6885 Description: While starting a server, the fsfilt_ext3 module could not be
6887 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6888 kernels. Previously, lconf would still use the ext3 name
6889 when trying to load modules. Now, it will correctly use
6890 ext3 on 2.4 and ldiskfs on 2.6.
6892 Severity : enhancement
6893 Description: The default stripe count has been changed to 1
6894 Details : The interpretation of the default stripe count (0, to lfs
6895 or lmc) has been changed to mean striping across a single
6896 OST, rather than all available. For general usage we have
6897 found a stripe count of 1 or 2 works best.
6899 Severity : enhancement
6900 Description: Add support for compiling against Cray portals.
6901 Details : Conditional compiling for some areas that are different
6905 Frequency : occasional
6906 Bugzilla : 6409, 6834
6907 Description: Creating files with an explicit stripe count may lead to
6908 a failed assertion on the MDS
6909 Details : If some OSTs are full or unavailable, creating files may
6910 trigger a failed assertion on the MDS. Now, Lustre will
6911 try to use other servers or return an error to the
6915 Frequency : occasional
6917 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6918 caused lock timeout during readahead (since 1.4.2).
6919 Details : Processes doing readahead might match a lock that hasn't been
6920 granted yet if there are overlapping and conflicting lock
6921 requests. The readahead process waits on ungranted lock
6922 (original lock is CBPENDING), while OST waits for that process
6923 to cancel CBPENDING read lock and eventually evicts client.
6925 Severity : enhancement
6927 Description: Initial enabling of flock support for clients
6928 Details : Implements fcntl advisory locking and file status functions.
6929 This feature is provided as an optional mount flag (default
6930 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6931 locking are implemented yet, and those that are are not guaranteed
6932 to be completely correct in production environments.
6933 mount -t lustre -o [flock|noflock] ...
6936 Frequency : occasional
6938 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6939 assertion in the ext3 JBD (journaling) layer.
6940 Details : The b_committed_data struct is protected by the big kernel lock
6941 in 2.4 kernels, serializing journal_commit_transaction() and
6942 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6943 access to this struct. In 2.6 kernels there is finer grained
6944 locking to improve SMP performance of the JBD layer.
6948 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6950 -----------------------------------------------------------------------------
6952 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6954 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6955 versions of Lustre. Please update all servers and clients to
6956 version 1.4.2 or later at the same time. You must also run
6957 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6958 to update the configuration logs.
6960 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6961 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6962 - swab llog records of type '0' so we get proper header size/idx (5861)
6963 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6964 - fix rename of one directory over another leaking an inode (5953)
6965 - avoid SetPageDirty on 2.6 (5981)
6966 - don't re-add just-being-destroyed locks to the waiting list (5653)
6967 - when creating new directories, inherit the parent's custom
6968 striping settings if present parent (3048)
6969 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6970 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6971 - don't hold client locks on temporary worklist from l_lru (5666)
6972 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6973 - automatically evict dead clients (5921)
6974 - Update file size properly in create+truncate+fstat case (6196)
6975 - Do not unhash mountpoint dentries, do not allow removal of
6977 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6978 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6979 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6980 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6981 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6982 - Correctly handle reads of files with no objects (6243)
6983 - lctl recover will also mark a device active if deactivate used (5933)
6985 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6986 - allow --write-conf on an MDS with different nettype than client (5619)
6987 - don't write config llogs to MDS for mounts not from that MDS (5617)
6988 - lconf should create multiple TCP connections from a client (5201)
6989 - init scripts are now turned off by default; run chkconfig --on
6990 lustre and chkconfig --on lustrefs to use them
6991 - upcalls are no longer needed for clients to recover to failover
6993 - add --abort-recovery option to lconf to abort recovery on device
6995 - add support for an arbitrary number of OSTs (3026)
6996 - Quota support protocol changes.
6997 - forward compatibility changes to wire structs (6007)
6998 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6999 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
7000 - improved "lustre status" script
7001 - initialize blocksize for non-regular files (6062)
7002 - added --disable-server and --disable-client configure options (5782)
7003 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
7004 - Vanilla 2.4.29 support
7005 - increase maximum number of obd devices to 520 (6242)
7006 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
7007 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
7008 SLES 9 2.6.5 kernels
7009 - limit stripes per file to 160 (the maximum EA size) (6093)
7011 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7014 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
7015 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
7016 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
7017 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
7018 - small ext3 extents cleanups and fixes (5733)
7019 - improved mballoc code, several small races and bugs fixed (5733, 5638)
7020 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
7021 - avoid needless client->OST connect, fix handle mismatch (5317)
7022 - fix DLM error path that led to out-of-sync client, long delays (5779)
7023 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
7024 - fix several locking issues related to i_size (5492,5624,5654,5672)
7025 - don't move pending lock onto export if it is already evicted (5683)
7026 - fix kernel oops when creating .foo in unlinked directory (5548)
7027 - fix deadlock in obdfilter statistics vs. object create (5811)
7028 - use time_{before,after} to avoid timer jiffies wrap (5882)
7029 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
7030 - Do not leak request if server was not able to process it (5154)
7031 - If mds_open unable to find parent dir, make that negative lookup(5154)
7032 - don't create new directories with extent-mapping (5909, 5936)
7034 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
7035 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
7036 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
7037 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
7038 - support for RHEL4 kernel on i686 (5773)
7039 - provide error messages when incompatible logs are encountered (5898)
7041 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
7042 * version 1.4.0.10 (1.4.1 release candidate 1)
7044 - don't keep a lock reference when lock is not granted (4238)
7045 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7046 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
7047 - drop import inflight refcount on signal_completed_replay error (5255)
7048 - unlock page after async write error during send (3677)
7049 - handle missing objects in filter_preprw_read properly (5265)
7050 - no transno return for symlink open, don't save no-trasno open (3440)
7051 - don't try to complete elan receive that already failed (4012)
7052 - free RPC server reply state on error (5406)
7053 - clean up thread from ptlrpc_start_thread() on error (5160)
7054 - readahead could read extra page into cache that wasn't ejected (5388)
7055 - prevent races in class_attach/setup/cleanup/detach (5260)
7056 - don't dereference de->d_inode after l_dput of de (5458)
7057 - use "int" for stripe value returned from lock_to_stripe (5544)
7058 - mballoc allocation and error-checking fixes in 2.6 (5504)
7059 - block device patches to fix I/O request sizes in 2.6 (5482)
7060 - look up hostnames for IB nals (5602)
7061 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
7062 - don't start multiple acceptors for the same port (5277)
7063 - fix incorrect LASSERT in mds_getattr_name (5635)
7064 - export a proc file for general "ping" checking (5628)
7065 - fix "lfs check" to not block when the MDS is down (5628)
7067 - service request history (4965)
7068 - put {ll,lov,osc}_async_page structs in a single slab (4699)
7069 - create an "evict_client" /proc entry on OSTs, like the MDS has
7070 - fix mount usage message, return errors per mount(8) (5168)
7071 - change grep [] to grep "[]" in tests so they work in more UMLs
7072 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
7073 - remove ancient LOV_MAGIC_V0 EA support (5047)
7074 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
7075 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
7076 - allow readahead allocations to fail when low on memory (5383)
7077 - mmap locking landed again, after considerable improvement (2828)
7078 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
7080 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
7083 - send OST transaction number in read/write reply to free req (4966)
7084 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7085 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7086 - the watchdog thread now runs as interruptible (5246)
7087 - flock/lockf fixes (but it's still disabled, pending 5135)
7088 - don't use EXT3 constants in llite code (5094)
7089 - memory shortage at startup could cause assertion (5176)
7091 - reorganization of lov code
7092 - single portals codebase
7094 - add extents/mballoc support (5025)
7095 - direct I/O reads in the obdfilter (4048)
7096 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
7098 tbd Cluster File Systems, Inc. <info@clusterfs.com>
7101 - send OST transaction number in read/write reply to free req (4966)
7102 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
7103 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
7104 - the watchdog thread now runs as interruptible (5246)
7105 - handle missing objects in filter_preprw_read properly (5265)
7106 - unsafe list practices (rarely) led to infinite eviction loop (4908)
7107 - drop import inflight refcount on signal_completed_replay error (5255)
7108 - unlock page after async write error during send (3677)
7109 - return original error code on reconstructed replies (3761)
7110 - no transno return for symlink open, don't save no-trasno open (3440)
7112 - add pid to ldlm debugging output (4922)
7113 - bump the watchdog timeouts -- we can't handle 30sec yet
7114 - extra debugging for orphan dentry/inode bug (5259)
7116 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
7119 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
7120 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
7121 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
7122 - fix NULL dereference in /proc/sys/portals/routes (4827)
7123 - allow failed mdc_close() operations to be interrupted (4561)
7124 - stop precreate on OST before MDS would time out on it (4778)
7125 - don't send partial-page writes before EOF from client (4410)
7126 - discard client grant for sub-page writes on large-page clients (4520)
7127 - don't free dentries not owned by NFS code, check generation (4806)
7128 - fix lsm leak if mds_create_objects() fails (4801)
7129 - limit debug_daemon file size, always print CERROR messages (4789)
7130 - use transno after validating reply (3892)
7131 - process timed out requests if import state changes (3754)
7132 - update mtime on OST during writes, return in glimpse (4829)
7133 - add mkfsoptions to LDAP (4679)
7134 - use ->max_readahead method instead of zapping global ra (5039)
7135 - don't interrupt __l_wait_event() during strace
7137 - add software watchdogs to catch hung threads quickly (4941)
7138 - make lustrefs init script start after nfs is mounted
7139 - fix CWARN/ERROR duplication (4930)
7140 - return async write errors to application if possible (2248)
7141 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
7142 - print NAL number in %x format (4645)
7143 - update barely-supported suse-2.4.21-171 series (4842)
7144 - support for sles 9 %post scripts
7145 - support for building 2.6 kernel-source packages
7146 - support for sles km_* packages
7148 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
7151 - ignore -ENOENT errors in osc_destroy (3639)
7152 - notify osc create thread that OSC is being cleaned up (4600)
7153 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
7154 - reconstruct ost_handle() like mds_handle() (4657)
7155 - create a new thread to do import eviction to avoid deadlock (3969)
7156 - let lconf resolve symlinked-to devices (4629)
7157 - don't unlink "objects" from directory with default EA (4554)
7158 - hold socknal file ref over connect in case target is down (4394)
7159 - allow more than 32000 subdirectories in a single directory (3244)
7160 - fix blocks count for O_DIRECT writes (3751)
7161 - OST returns ENOSPC from object create when no space left (4539)
7162 - don't send truncate RPC if file size isn't changing (4410)
7163 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
7164 - bind to privileged port in socknal and tcpnal (3689)
7166 - rate limit CERROR/CWARN console message to avoid overload (4519)
7167 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
7168 - basic mmap support (3918)
7169 - kernel patch series update from b1_4 (4711)
7171 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
7174 - avoid crash during MDS cleanup with OST shut down (2775)
7175 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
7176 - don't use bad inodes on the MDS (3744)
7177 - dynamic object preallocation to improve recovery speed (4236)
7178 - don't hold spinlock over lock dumping or change debug flags (4401)
7179 - don't zero obd_dev when it is force cleaned (3651)
7180 - print grants to console if they go negative (4431)
7181 - "lctl deactivate" will stop automatic recovery attempts (3406)
7182 - look for existing locks in ldlm_handle_enqueue() (3764)
7183 - don't resolve lock handle twice in recovery avoiding race (4401)
7184 - revalidate should check working dir is a directory (4134)
7186 - don't always mark "slow" obdfilter messages as errors (4418)
7188 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
7191 - don't close LustreDB during write_conf until it is done (3860)
7192 - fix typo in lconf for_each_profile (3821)
7193 - allow dumping logs from multiple threads at one time (3820)
7194 - don't allow multiple threads in OSC recovery (3812)
7195 - fix debug_size parameters (3864)
7196 - fix mds_postrecov to initialize import for llog ctxt (3121)
7197 - replace config semaphore with spinlock (3306)
7198 - be sure to send a reply for a CANCEL rpc with bad export (3863)
7199 - don't allow enqueue to complete on a destroyed export (3822)
7200 - down write_lock before checking llog header bitmap (3825)
7201 - recover from lock replay timeout (3764)
7202 - up llog sem before sending rpc (3652)
7203 - reduce ns lock hold times when setting kms (3267)
7204 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
7205 - fix NULL deref and obd_dev leak on setup error (3312)
7206 - replace some LBUG about llog ops with error handling (3841)
7207 - don't match INVALID dentries from d_lookup and spin (3784)
7208 - hold dcache_lock while marking dentries INVALID and hashing (4255)
7209 - fix invalid assertion in ptlrpc_set_wait (3880)
7211 - add libwrap support for the TCP acceptor (3996)
7212 - add /proc/sys/portals/routes for non-root route listing (3994)
7213 - allow setting MDS UUID in .xml (2580)
7214 - print the stack of a process that LBUGs (4228)
7216 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
7219 - don't cleanup request in ll_file_open() on failed MDS open (3430)
7220 - make sure to unset replay flag from failed open requests (3440)
7221 - if default stripe count is 0, use OST count for inode size (3636)
7222 - update parent mtime/ctime on client for create/unlink (2611)
7223 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
7224 - free recovery state on server during a forced cleanup (3571)
7225 - unregister_reply for resent reqs (3063)
7226 - loop back devices mounting and status check on 2.6 (3563)
7227 - fix resource-creation race that can provoke i_size == 0 (3513)
7228 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
7229 - more debugging for page-accounting assertion (3746)
7230 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
7231 - avoid deadlock after precreation failure (3758)
7232 - fix race and lock order deadlock in orphan handling (3450, 3750)
7233 - add validity checks when grabbing inodes from l_ast_data (3599)
7235 - add /proc/.../recovery_status to obdfilter (3428)
7236 - lightweight CDEBUG infrastructure, debug daemon (3668)
7237 - change default OSC RPC parameters to be better on small clusters
7238 - turn off OST read cache for files smaller than 32MB
7239 - install man pages and include them in rpms (3100)
7240 - add new init script for (un)mounting lustre filesystems (2593)
7241 - run chkconfig in %post for init scripts (3701)
7242 - drop scimac NAL (unmaintained)
7244 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
7247 - clean kiobufs before and after use (3485)
7248 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
7249 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7250 - change init script to fail more clearly if not run as root (1528)
7251 - allow clients to reconnect during replay (1742)
7252 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
7253 - don't do DNS lookups on NIDs too small for IP addresses (3442)
7254 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
7255 - fix cond_resched (3554)
7256 - only evict unfinished clients after recovery (3515)
7257 - allow bulk resend, prevent data loss (3570)
7258 - dynamic ptlrpc request buffer allocation (2102)
7259 - don't allow unlinking open directory if it isn't empty (2904)
7260 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
7261 - remove extraneous obd dereference causing LASSERT failure (3334)
7262 - don't use get_cycles() when creating temp. files on the mds (3156)
7263 - hold i_sem when setting i_size in ll_extent_lock() (3564)
7264 - handle EEXIST for set-stripe, set proper directory name (3336)
7266 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
7267 - fix ksocknal_fmb_callback() error messages (2918)
7269 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
7272 - don't copy lvb into (possibly NULL) reply on error (2983)
7273 - don't deref dentry after dput, don't free lvb on error (2922)
7274 - use the kms to determine writeback rpc length (2947)
7275 - increment oti_logcookies when osc is inactive (2948)
7276 - update client's i_blocks count via lvb messages (2543)
7277 - handle intent open/close of special files properly (1557)
7278 - mount MDS with errors=remount-ro, like obdfilter (2009)
7279 - initialize lock handle to avoid ASSERT on error cleanup (3057)
7280 - don't use cancelling-locks' kms values (2947)
7281 - use highest lock extent for kms, not last one (2925)
7282 - don't dereference ERR_PTR() dentry in error handling path (3107)
7283 - fix thread race in portals_debug_dumplog() (3122)
7284 - create lprocfs device entries at setup instead of at attach (1519)
7285 - common AST error handler, don't evict client on completion race (3145)
7286 - zero nameidata in detach_mnt in 2.6 (3118)
7287 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7288 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7289 - initialize RPC timeout timer earlier for 2.6 (3219)
7290 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7291 - print nal/nid for unknown nid (3258)
7292 - additional checks for oscc recovery before doing precreate (3284)
7293 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7294 - don't crash in mdc_close for bad permissions on open (3285)
7295 - zero i_rdev for non-device files (3147)
7296 - clear page->private before handing to FS, better assertion (3119)
7297 - tune the read pipeline (3236)
7298 - fix incorrect decref of invalidated dentry (2350)
7299 - provide read-ahead stats and refine rpc in flight stats (3328)
7300 - don't hold journal transaction open across create RPC (3313)
7301 - update atime on MDS at close time (3265)
7302 - close LDAP connection when recovering to avoid server load (3315)
7303 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7304 - don't leak open file on MDS after open resend (3325)
7305 - serialize filter_precreate and filter_destroy_precreated (3329)
7306 - loop device shouldn't call sync_dev() for nul device (3092)
7307 - clear page cache after eviction (2766)
7308 - resynchronize MDS->OST in background (2824)
7309 - refuse to mount the same filesystem twice on same mountpoint (3394)
7310 - allow llmount to create routes for mounting behind routers (3320)
7311 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7312 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7313 - fix rare problem with rename on htree directories (3417)
7315 - allow default OST striping configuration per directory (1414)
7316 - fix compilation for qswnal for 2.6 kernels (3125)
7317 - increase maximum number of MDS request buffers for large systems
7318 - change liblustreapi to be useful for external progs like lfsck (3098)
7319 - increase local configuration timeout for slow disks (3353)
7320 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7322 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7325 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7326 - don't overwrite extent policy data in reply if lock was blocked (2901)
7327 - drop filter export grants atomically with removal from device (2663)
7328 - del obd_self_export from work_list in class_disconnect_exports (2908)
7329 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7330 - swab reply message in mdc_close, other PPC fixes (2464)
7331 - fix destroying of named logs (2325)
7332 - overwrite old logs when running lconf --write_conf (2264)
7333 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7334 - fix race in target_handle_connect (2898)
7335 - mds_reint_create() should take same inode create lock (2926)
7336 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7337 - don't close files for self_export to avoid uninitialized obd (2936)
7338 - allow MDS with the same name as client node (2939)
7339 - hold dentry reference for closed log files for unlink (2325)
7340 - reserve space for all logs during transactions (2059)
7341 - don't evict page beyond end of stripe extent (2925)
7342 - don't oops on a deleted current working directory (2399)
7343 - handle hard links to targets without a parent properly (2517)
7344 - don't dereference NULL lock when racing during eviction (2867)
7345 - don't grow lock extents when lots of conflicting locks (2919)
7347 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7350 - account for cache space usage on clients to avoid data loss (974)
7351 - lfsck support in lustre kernel code (2349)
7352 - reduce journal credits needed for BRW writes (2370)
7353 - orphan handling to avoid losing space on client/server crashes
7354 - ptlrpcd can be blocked, stopping ALL progress (2477)
7355 - use lock value blocks to assist in proper KMS, faster stat (1021)
7356 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7357 - recovery for initial connections (2355)
7358 - fixes for mds_cleanup_orphans (1934)
7359 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7360 - block all file creations until orphan recovery completes (1901)
7361 - client remove rq_connection from request struct (2423)
7362 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7363 - recovery timer race (2670)
7364 - mdc_close recovey bug (2532)
7365 - ptlrpc cleanup bug (2710)
7366 - mds timeout on local locks (2588)
7367 - namespace lock held during RPCs (2431)
7368 - handle interrupted sync write properly (2503)
7369 - don't try to handle a message that hasn't been replied to (2699)
7370 - client assert failure during cleanup after abort recovery (2701)
7371 - leak mdc device after failed mount (2712)
7372 - ptlrpc_check_set allows timedout requests to complete (2714)
7373 - wait for inflight reqs when ptlrpcd finishes (2710)
7374 - make sure unregistered services are removed from the srv_list
7375 - reset bulk XID's when resending them (caught by 1138 test)
7376 - unregister_bulk after timeout
7377 - fix lconf error (2694)
7378 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7379 - readahead locks pages, leaves pending causing memory pressure (2673)
7380 - increase OST request buffers to 4096 on large machines (2729)
7381 - fix up permission of existing directories in simple_mkdir (2661)
7382 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7383 - don't assign transno to errored transactions (2742)
7384 - don't delete objects on OST if given a bogus objid from MDS (2751)
7385 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7386 - if rq_no_resend, then timeout request after recovery (2432)
7387 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7388 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7389 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7390 - MDS recovery shouldn't delete objects that it has given out (2730)
7391 - if enqueue arrives after completion, don't clobber LVB (2819)
7392 - don't unlock pages twice when trigger_group_io returns error (2814)
7393 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7394 - don't write pages to disk if there was an error (1450)
7395 - don't ping imports that have recovery disabled (2676)
7396 - take buffered bytes into account when balancing socknal conn (2817)
7397 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7398 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7399 - remove little-endian swabbing of llog records (1987)
7400 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7401 - retry reposting request buffers if they fail (1191)
7402 - grow extent at grant time to avoid granting a revoked lock (2809)
7403 - lock revoke doesn't evict page if covered by a second lock (2765)
7404 - disable VM readahead to avoid reading outside lock extents (2805)
7406 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7407 - updated kernel patches for hp-2.4.20 kernel (2681)
7409 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7412 - fix truncated write corruption (2366)
7413 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7415 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7416 - ensure len(uuid) < 37 in lmc (1171)
7417 - fix ia64 OOPS in llog_test (2255)
7418 - zero end of page at obdfilter for partial page writes (2648)
7419 - don't leave stale dentries around after renames (bug 2428)
7420 - fix timeouts when evicting a client with a single lock held (2642)
7421 - set deadline for the initial HELLO message to drain (2634)
7422 - print out dotted-quad IP addresses in the socknal (2302)
7424 - additional debugging for MDS client eviction problem (2443)
7425 - fix mkfsoptions support for osts (2603, 2604)
7427 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7430 - add series for the vanilla 2.6.0 kernel
7431 - add series for the vanilla 2.4.24 kernel
7432 - add series for a cray x86/64 UL kernel drop
7433 - fix xattr patches for the vanilla 2.4.19 series
7435 - generate true UUIDs in lmc (1171)
7436 - have portals stack dumping break in UML (2466)
7437 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7438 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7439 - wake callback waiting threads on client eviction (2460)
7440 - Add --ptldebug and --subsystem to lmc (1719)
7441 - update assertion to allow safe interrupt allocation
7442 - set rq_no_resend for cancel requests (2432)
7443 - recalculate ptlrpcd timeout after resend (2494)
7444 - call vfs_rmdir when removing pending directories (2368)
7445 - fix renaming a file to itself (2429)
7446 - lmc creates a default one-stripe lov (2454)
7447 - expand procfs space to handle large clusters (2326)
7448 - increase UML stack to avoid overflow
7449 - update lconf's list of debug and subsystem masks
7450 - fix lfs find --obd (2510)
7451 - /proc tunable for disabling filter read caching (2591)
7452 - stop rpm packages from altering slapd.conf (2301)
7453 - disable nagle in the socknal under 0conf (2578)
7454 - choose mds inode size based on stripe count (2572)
7455 - fix kernel-source rpm problems (2516)
7457 - add --disable-doc to avoid pdf generation (2421)
7458 - update documentation, tests, type-os, comments
7459 - avoid format warnings on ia64
7460 - remove the TOE NAL
7461 - tiny code cleanups by removing unused fields
7463 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7466 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7467 - remove the most blatant lies from BUILDING file (2371)
7468 - change default debug level to reasonable production setting
7469 - reduce client side cache size to reduce cache flush time
7470 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7471 - make TCP zerocopy and pinger support enabled by default (2476)
7472 - sync writes completed after process exits caused crashes (2319)
7473 - maintain correct mount count on the MDS (2356)
7474 - backout 1557, because 2316 wasn't really fixed
7475 - better file I/O statistics gathering in /proc
7476 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7477 - another kernel patch to fix zero-copy TCP function export
7478 - don't take duplicate lock when processing re-sent getattr (2420)
7479 - lctl uses obd_self_export instead of creating new conn (2353)
7480 - MDS/OST recovery case which requires object creation asserted (2425)
7481 - move lfs from /usr/sbin to /usr/bin in packages
7482 - fix race between mds_client_add and mds_client_free (2417)
7483 - use kmalloc instead of slabs in portals (2430)
7484 - don't create duplicate records when a failover MDS is present (2442)
7485 - remove unnecessary mount age check (2332)
7486 - don't remove directory inodes from locks prematurely (2451)
7487 - don't break if MDS service name is the same as hostname (2103)
7488 - fix races in client write RPC generation when cache full (2482)
7490 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7493 - remove now-unused request->rq_obd (278)
7494 - if an allocation fails, print out how much memory we've used (1933)
7495 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7496 - add the "configurable stack size" patch to most series files (1256)
7497 - ability to write large log records, for 100+ OST configs (2306)
7498 - fix NULL deref when filter_prep fails (2314)
7499 - fix operator precedence error in filter_sync
7500 - dynamic allocation of socknal TX descriptors (2315)
7501 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7502 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7503 - ensure that CWARN messages in Portals always get to the syslog
7504 - __init/__exit are not for prototype decls (ldlm_init/exit)
7505 - x86-64 compile warning fixes
7506 - fix gateway LMC keyword conflict (2318)
7507 - fix MDS lock inversions in getattr/reint paths (1844)
7508 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7509 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7510 - DLM race condition prevented some lock evictions (2328)
7511 - ENOMEM detection and retry on socknal sends (2230)
7512 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7513 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7514 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7515 - better error messages when a client is rejected during recovery (1505)
7516 - avoid cancelling locks which were never granted, after failure (2330)
7517 - fix i_sem/journal inversion in mds_client_add (2333)
7518 - fix truncate/getattr lock cycle deadlock (2334)
7519 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7520 - fix two rare exit paths which could leak an l_lock() ref (2321)
7521 - fencepost error in MDS/OST orphan recovery (2226)
7522 - make log record alignment 8 bytes (1988)
7523 - lstripe now fails when requested offset > ost_count (2237)
7524 - ensure that all kernel series have a complete list.h (1607)
7525 - fix crashes in special-file operations (2316)
7526 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7528 - allow configurable automake binary, for testing new versions
7529 - small update to the lfs documentation
7531 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7533 * fix negative export reference count in fsfilt_sync (2312)
7535 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7536 * release candidate 0.9.1
7538 - orphans are moved into the PENDING directory for possible recovery
7539 - replayed opens now open by fid for orphan/rename safety (1042)
7540 - last close of an orphan inode generates a transno (683)
7541 - chdir() and mount() now pin the directory entry (1020)
7542 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7543 - discard very old requests without processing them (1502)
7544 - remove some common, well-understood CERRORs (1505)
7545 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7546 - clear "grant" flags in OST replies until OST grant code lands (1644)
7547 - fix read performance by not clobbering i_blksize on client (1598)
7548 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7549 - make LRU size a /proc tunable, clears locks when reduced (707)
7550 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7551 - server locks take references on exports now (1558)
7552 - build fixes for 2.4.20-rh trees (1663)
7553 - return an error from lov_create if all OSCs are inactive (1751)
7554 - fix import levels when a reconnect happens without a timeout (1597)
7555 - exit early from mds_open if we get a lookup error (1749)
7556 - partial page read at EOF wouldn't wait for disk before sending (1642)
7557 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7558 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7559 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7560 - write/truncate lock inversion (1639)
7561 - Don't auto-load obdclass, portals modules during cleanup (1495)
7562 - fix timestamps from jumping to "now" (1763)
7563 - extra journal assertions (1648)
7564 - add an extra multiunlink test (1771)
7565 - fix read_record/write_record API (1776)
7566 - fix leak of offset_extent, possible incorrect i_size later (1772)
7567 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7568 - Debugging for the kqswnal_get_idle_tx problems (1820)
7569 - Allow recovery to be attempted multiple times (1536)
7570 - Write out MDS last_rcvd file after it is first created (1600)
7571 - Fix tx_descriptor leak in failed transmit situations (1827)
7572 - ext3 journaling fixes for assertion failure after IO error (1871)
7573 - class_export_put() on freed export after completion AST error (1896)
7574 - Fix revalidate looping in VFS (1322)
7575 - Don't access a freed export during MDS_REINT timeout (1521)
7576 - Add open-unlink recovery support on the MDS (1673,1764)
7577 - Return an error if no MDS data was read from last_rcvd (1946)
7578 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7579 - Don't setup a disk device that doesn't match exported UUID (317)
7580 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7581 - avoid committing NULL handle in force close
7582 - local.sh is now a one-stripe LOV configuration
7583 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7584 - fix ext3 htree duplicate directory entry corruption (1516)
7585 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7586 - update to new LOV EA format (2097)
7587 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7589 2003-06-15 Phil Schwan <phil@clusterfs.com>
7592 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7593 - per-import/export recovery handling (958, 931, 959)
7594 - multiple last-rcvd slots, for serving multiple FSes (949)
7595 - connections are again shared between multiple imp/exports (963, 964)
7596 - "umount -f" would hang if any requests needed to be sent (393, 978)
7597 - avoid pinning large req buffer by copying for queued messages (989)
7598 - add "uuid" to "lctl device" command to help upcalls (991)
7599 - "open" RPCs with transnos would confuse recovery counters (1037)
7600 - do proper endian conversion of all wire messages (288, 340, 891)
7601 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7602 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7603 - fix ptlrpc_connection leak in target_handle_connect (1174)
7604 - fix import refcounting bug in OST and MDS cleanup (1134)
7605 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7606 - fix very unlikely obd_types race condition (501)
7607 - remove osc_open hack for echo_client (1187)
7608 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7609 - a failure in read_inode2 leads to deadlock (1139)
7610 - cancel ack-locks as soon as transaction is committed (1072)
7611 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7612 - make sure to commitrw after any preprw to avoid deadlock (1162)
7613 - failing to execute a file in a lustre FS would lock inode (1203)
7614 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7615 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7616 - obd_uuid2tgt would walk off the end of the list (1255)
7617 - on IA64 the osc would give portals incorrect bulk size (1258)
7618 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7619 - fix lock inversion caused by new llite matching code (1282)
7620 - limit the number of dirty pages on a client to 10MB (1286)
7621 - timed out locks were not being corrected cancelled (1289)
7622 - fix O_DIRECT above 4GB on IA-32 (1292)
7623 * major user-visible changes
7624 - fail out/fail over policy now controlled by the upcall (993)
7626 - add OBD_PING to check server availability and failure (954)
7627 - lustre messages are now sent in sending host order (288, 340, 891)
7628 - add eadatalen to MDS getattr reply (340)
7629 - OST read replies may contain second buffer, with per-page status (593)
7631 2003-03-11 Phil Schwan <phil@clusterfs.com>
7634 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7635 - failed open()s could cause deadlock; fixed (867, 869)
7636 - stop cancelling OST locks when files are closed (481)
7637 - overlapping XID spaces caused network corruption (851, 853)
7638 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7639 - setattr_raw added, to do single-RPC, server-side setattrs
7640 - lmc/lconf syntax change for OST UUIDs
7641 - fix crashy race condition between ptlrpc_free_req and osc_close
7642 - don't use request in mdc_enqueue if we hit a timeout (889)
7643 - don't set the inode i_size for regular files from the MDS (896)
7644 - handle out of order completion AST (842)
7645 - don't LBUG if a lock request times out after receiving AST (913)
7646 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7647 - if a bad lock AST arrives, send an error instead of dropping entirely
7648 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7649 - fix leak in bulk IO when only partially completed (899, 900, 926)
7650 - fix O_DIRECT for ia64 (55)
7651 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7652 - C-z after timeout could hang a process forever; fixed (977)
7654 - client-side I/O cache (678, 924, 929, 941, 970)
7656 - READPAGE and SETATTRs which don't take server-side locks get
7659 2003-02-11 Phil Schwan <phil@clusterfs.com>
7662 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7663 - Fix fsfilt_extN_readpage() to read a full page of directory
7664 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7665 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7666 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7667 - Most intent operations were removed, in favour of a new RPC mode
7668 that does a single RPC to the server and bypasses most of the VFS
7669 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7670 - Aggressively cancel local locks on DLM servers
7671 - mds_reint_unlink sends EA to the client if it's the last nlink.
7672 client uses that EA to unlink OST objects.
7673 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7674 - recursive symlinks were fixed (439)
7675 - fixed NULL deref in DEBUG_REQ
7676 - filter_update_lastobjid no longer calls sync, which annoyed extN
7677 - fixed multi-client small-writes to a single file problem (445)
7678 - fixed mtime updates during file writes (607)
7679 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7680 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7681 - fixed Linux OST size reporting problem (444, 656)
7682 - OST now updates object mtime with writes or setattr (607, 619)
7683 - client verifies file size before zeroing page past EOF (445)
7684 - OST now writes last allocated objid to disk with allocation (108)
7685 - LOV on echo now works (409)
7687 - mds_reint_unlink sends a new buffer, with the EA included. this
7688 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7689 set if a regular file was being unlinked, and it was the last link
7690 - use PtlGet from the target for bulk writes (315)
7691 - OST now updates object mtime with writes or setattr (607, 619)
7692 - LDLM now has a grant-time callback to revalidate locked items, if
7694 - Many MDS operations were reorganized to combat race conditions
7696 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7697 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7699 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7702 - Fully reactivate OST imports after reconnection (512, others)
7703 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7704 - More graceful error handling for truncating on dead OST (515)
7705 - Don't error out unless we're actually accessing dead stripes (474)
7706 - Fix garbage sizes when stripes are missing (410)
7707 - LRU counters were broken, causing constant lock purge (433, 432)
7708 - garbage on read from stripes with failed OSTs (441)
7709 - mark OSCs as active before reconnecting during recovery (438)
7710 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7711 - lfind did not preserve OST order in output (443)
7712 - symlinks cause hung clients, incorrect data (439)
7713 - stop dereferencing request after dropping refcount (457)
7714 - don't LASSERT(spin_is_locked) on non-SMP (455)
7715 - fixes for many rename() bugs
7716 - fstat didn't correctly synchronize attributes (399)
7717 - server must handle lock cancellation during blocking AST prep (487)
7718 - bulk descriptors were free()d too soon (511)
7719 - fix paths in lconf, which would load incorrect modules (451, 507)
7720 - fix confusing lconf 'host not found' error message (386)
7721 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7722 - fix race condition in mdc_blocking_ast() for inode access (526)
7723 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7724 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7725 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7726 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7727 - fix lstripe in several ways (kernel side) (527)
7728 - fix request leak in ldlm_cli_enqueue (262)
7729 - incorrect OSC was marked inactive after OST failure
7730 - call mds_fs_cleanup before unmounting filesystem (524)
7731 - fix races between taking ns_lock and ldlm_lock_change_resource
7732 - fix races updating LOV export open file list
7733 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7734 - fix recovery NULL deref in ldlm_cli_cancel_unused
7735 - fix some DLM races by using new hash table for lock handles (419)
7736 - permit the client to specify desired inodes, at replay
7737 - duplicate requests when we queue them for replay reintegration
7738 - fix last_rcvd offset calculation
7739 - sync after each recovered transaction, so we always make progress
7740 - never, not always, ERESTART requests without transnos
7741 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7743 - skip replay if the MDS says that the client is already connected
7744 - don't check for a recovery-enabled export to match lctl's UUID
7745 - don't INC_USE_COUNT for phantom exports
7746 - don't crash when cleaning up phantom exports (567)
7747 - don't double-finish or set replay data for errored mdc_open requests
7748 - abort requests when they time out, so we don't get old replies
7749 - send/receive replies for AST messages again
7750 - if the client says that it doesn't have the lock, cancel it on the
7752 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7753 mark it as destroyed, it will all work out when decref is called
7754 - fix module use counts (22, 581)
7756 - ASTs now expect a reply (server cancels lock on error reply)
7758 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7761 - fix many simultaneous client startup (392)
7762 - fix dentry->d_it clobbering
7763 - credentials weren't being shipped for readdir/getattr operations
7764 - remove invalid assertions triggered during some concurrent MD
7766 - proper Lustre versions added (336, 389)
7767 - fix memory leak for create error case (398)
7768 - fix LOV locking bug that would get cli/srv out of sync
7769 - fix echo client over LOV (409)
7770 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7771 - fix double-O_EXCL intent crash (424)
7772 - avoid sending multiple lock CANCELs (352)
7774 - MDS can do multi-client recovery (modulo bugs in new code)
7776 - many updates, edits, cleanups
7778 2002-11-18 Phil Schwan <phil@clusterfs.com>
7781 - fix null d_it dereference (346)
7782 - fix full OST/dbench hang (333)
7783 - fix permission problem with file removal (286)
7784 - fix removal of OSCs from LOV when they fail
7785 - fix NULL deref during bulk timeout (214)
7786 - fix problems related to multiple filesystems on one MDS (241)
7787 - fixed serious subtle metadata locking bugs
7788 - free locks on clients when inodes are removed due to memory
7790 - fix inode pointer in lock data (285)
7791 - partial support for multiple MDS on a single host (241)
7792 - data locks weren't cancelled at clear_inode time (290, 311)
7793 - intent locks could lead to unbounded lock growth (205)
7794 - added a maximum lock count, an LRU list, and a flusher
7795 - fix multiple rename (365)
7796 - properly abstracted the echo client
7797 - OSC locked 1 byte too many; fixed
7798 - rewrote brw callback code:
7799 - fixed recovery bugs related to LOVs (306)
7800 - fixed too-many-pages-in-one-write crash (191)
7801 - fixed (again) crash in sync_io_timeout (214)
7802 - probably fixed callback-related race (385)
7804 - Add capability to MDS protocol
7805 - LDLM cancellations and callbacks on different portals
7807 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7810 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7811 - defer open object destruction to close time (601981, 138)
7812 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7813 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7814 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7815 - don't call obd_setattr in ll_file_release for destroyed objects
7816 * protocol change to lustre_msg: move |version| and add |flags|
7817 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7818 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7819 * added replay of create, unlink, link and rename operations during
7820 MDS failover; recovery should be much more robust now
7821 * remove failed OSCs from LOVs (only lov_create uses this so far)
7822 * the lustre-HOWTO was brought (more) up to date (582544)
7824 2002-10-23 Phil Schwan <phil@clusterfs.com>
7827 - in-use dentries weren't being reused properly (617851)
7828 - prevent multiple LDLM setup (599178)
7829 - fix LOV size calculations for truncate (617853)
7830 - fix client handling of MDS intent errors (POSIX)
7831 - fix permission bug in lovstripe.c test (624321)
7832 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7833 - truncate past end of file could corrupt data
7834 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7835 - a race in recovery could return ETIMEDOUT to apps (623947)
7836 - building outside the source directory was fixed
7837 * the lustre-HOWTO was brought (more) up to date (582544)
7838 * major progress was made on recovery functionality
7840 2002-10-10 Phil Schwan <phil@clusterfs.com>
7843 - recovery deadlock fix
7844 - rm -rf causes LBUG fix (617817)
7845 - file open by multiple tasks fix (618962)
7846 - directory permissions bugs (602707 and 620007)
7847 - journal_stop fixed with locking (611313)
7848 - O_APPEND failures resolved (618273, perhaps 614459)
7849 - lconf PATH fix (619770)
7850 - IA64 build fix (621450)
7851 - RPC buffer sizes scale with amount of memory
7853 2002-10-01 Phil Schwan <phil@clusterfs.com>
7856 - locks would be cancelled without throwing away data pages,
7857 resulting in inconsistent data (605627)
7858 - inode attributes were not always being refreshed (605627, 612449)
7859 - lconf now continues to cleanup after lctl reports an error
7860 - MDS now enforces user permissions (602707)
7861 - lprocfs cleanup fixed, but not yet enabled (614157)
7862 - fixed infinite server hang, should a client not respond to an AST
7863 - avoid going into recovery if user calls readlink() with a buffer
7864 that's too small (613941)
7865 - AST RPCs no longer require replies (614867) -- this may be changed
7866 - don't crash server if client sends an IOV that's too big (611336)
7867 - fixed lock conversion deadlock (611892)
7868 - fixed the following of symlinks (614622)
7869 * recovery: the server can remove locks from a client that dies, other
7870 clients can make progress
7871 * more extN patch fixes
7872 * compile-time configurable ptlrpc buffer allocations
7874 - collaborative read cache document
7875 - Lustre Lite Performance CDR document-in-progress
7877 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7880 - fix typo in patch-2.4.18
7882 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7885 - clear ptlrpc request each time in handle_incoming_request()
7886 - unlink of files now destroys the object on the OST
7888 2002-09-19 Peter Braam <braam@clusterfs.com>
7890 * add hard link support
7891 * change obdfile creation method
7892 * kernel patch changed
7894 2002-09-19 Peter Braam <braam@clusterfs.com>
7897 - stack overflow bug in extN fixed
7899 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7901 * documentation updates
7902 - add man pages for config tools
7903 - update tests/README to describe testing with new config tools
7904 - finish metadata API descriptions
7905 * bug fixes and cleanups
7906 - statfs workaround for 16TB limit
7907 - LOV stripe allocation improved, can stripe on subset of OSTs
7908 - LOV file size/IO offset was wrong for files > 4GB in size
7909 - object EA data was being dropped, caused files to be unreadable
7910 - memory overflow with non-LOV OST caused memory corruption
7911 - fixed regression tests to work with new config tools, obdfilter
7912 - fixed bug when directory size became larger than 1 block
7913 - fixed bug (for single client case) when PWD was deleted
7914 - invalidate local directory pages when doing intent-based ops
7915 - avoid LDLM oops when lock callback contained bad data
7917 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7919 * documentation updates
7920 * bug fixes and cleanups
7921 - configuration tools
7924 - 64-bit compile warnings
7925 - 64-bit internal statfs data
7927 * test_brw on persistent OST devices
7929 * lprocfs (disabled)
7931 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7933 * documentation updates
7934 * bug fixes and cleanups
7935 * configuration tools
7937 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7940 * many small fixes to 0_5_4
7941 * io/network handling
7942 * thinkos in MDS operations
7944 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7947 * crucial basic fixes to 0.5.3
7948 * IOR, Iozone work over Elan
7951 2002-08-07 Phil Schwan <phil@clusterfs.com>
7952 * version 0_5_3, our first alpha
7953 * we use the new Portals iovs
7954 * documentation updates
7955 * bug fixes and cleanups
7956 * small changes in the DLM wire protocol
7958 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7959 * version 0_5_1 with some initial stability,
7960 * locking on MD and file I/O.
7961 * documentation updates
7962 * several bug fixes since 0.5.0
7963 * small changes in wire protocol
7965 2002-07-18 Phil Schwan <phil@clusterfs.com>
7967 * delivered as Lustre Light Alpha
7968 * fixed a crash after handling invalid MDS requests
7969 * fixed directory pages for architectures with non-4k pages sizes
7971 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7972 * release version v0_4_4
7973 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7975 * Fixes a number of interruption problems with OST operations.
7976 * Update documentation for portals header changes
7977 * Move all wire protocol structs/defines to lustre_idl.h
7978 * Fixes symlink length bug.
7979 * Add tcpdump to repository.
7981 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7982 * release version v0_4_3
7983 * Fixes statfs for inodes on extN.
7984 * Fixes bug in runtests which would delete /etc/hosts.
7985 * Use 64-bit object IDs wherever possible (not into VFS though)
7986 Remove ost_get_info, which is unused by lustre, and out of date.
7988 2002-07-03 Peter Braam <braam@clusterfs.com>
7989 * release version v0_4_2 Fixes a lookup error (type not passed)
7990 * move forward to head of Portals
7991 * move forward to latest Lustre kernel
7993 2002-06-25 Peter Braam <braam@clusterfs.com>
7994 * release version v0_4_1. Hopefully stable on single node use.