1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10)
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4)
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.39.cfs8
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: don't use GFP_* in generic Lustre code.
18 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
22 Description: obdclass is limited by single OBD_ALLOC(idarray)
23 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
27 Description: warnings with build patchless client with vanila 2.6.19 and up
28 Details : change old ctl_table style and replace ctl_table/ctl_table_header
29 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
33 Frequency : Only for SLES
34 Description: sanity_quota fail test_1
35 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
36 file, which makes TSTID[2] inunique.
40 Frequency : Only for SLES9
41 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
42 Details : Update target/ChangeLog/which_patch .
46 Frequency : Only for test_10 in sanity-quota.sh
47 Description: a bug in quota test code
48 Details : checking a test flag in a wrong place
52 Frequency : Only for test_13 in sanity-quota.sh
53 Description: a bug in quota test script
54 Details : shouldn't check overfull allocated quota
58 Description: jbd statistics
59 Details : Port older jbd statistics patch for sles10
61 Severity : enhancement
63 Description: Add printing inode info into message about error in writepage.
68 Description: reply_lock_interpret crash due to race with it and lock cancel.
69 Details : Do not replay locks that are being cancelled. Do not reference
70 locks by their address during replay, just by their handle.
74 Frequency : When flocks are used.
75 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
76 Details : Flock locks might destroy just granted lock if it could be merged
77 with another existing flock, this is done in completion handler,
78 so teach ldlm_cli_enquque_fini that this is a valid case for
84 Description: Oops in read and write path when failing to allocate lock.
85 Details : Check if lock allocation failed and return error back.
87 --------------------------------------------------------------------------------
89 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
91 * Support for kernels:
94 2.6.16.46-0.14 (SLES 10)
95 2.6.18.8 vanilla (kernel.org)
96 * Client support for unpatched kernels:
97 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
98 2.6.16 - 2.6.22 vanilla (kernel.org)
99 * Due to recently discovered recovery problems, we do not recommend
100 using patchless RHEL 4 clients with this or any earlier release.
101 * Recommended e2fsprogs version: 1.39.cfs8
102 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
103 * Starting with this release, the ldiskfs backing filesystem required
104 by Lustre is now in its own package, lustre-ldiskfs. This package
105 should be installed. It is versioned separately from Lustre and
106 may be released separately in future.
110 Description: block reactivating mgc import until all deactivates complete
111 Details : Fix race when failing back MDT/MGS to itself (testing)
113 Severity : enhancement
115 Description: add optional extra BUILD_VERSION info
116 Details : add a new environment variable (namely LUSTRE_VERS) which allows
117 to override the lustre version.
120 Frequency : 2.6.18 servers only
122 Description: ll_kern_mount() doesn't release the module reference
123 Details : The ldiskfs module reference count never drops down to 0
124 because ll_kern_mount() doesn't release the module reference.
129 Description: server LBUG when using old ost_num_threads parameter
130 Details : Accept the old ost_num_threads parameter but warn that it
131 is deprecated, and fix an off-by-one error that caused an LBUG.
136 Description: Transient SCSI error results in persistent IO issue
137 Details : iobuf->dr_error is not reinitialized to 0 between two
141 Frequency : sometimes when underlying device returns I/O errors
143 Description: OSTs not going read-only during write failures
144 Details : OSTs are not remounted read-only when the journal commit threads
145 get I/O errors because fsfilt_ext3 calls journal_start/stop()
146 instead of the ext3 wrappers.
150 Description: poor connect scaling with increasing client count
151 Details : Don't run filter_grant_sanity_check for more than 100 exports
152 to improve scaling for large numbers of clients.
155 Frequency : SLES10 only
157 Description: sanity-quota.sh quotacheck failed: rc = -22
158 Details : Quotas cannot be enabled on SLES10.
161 Frequency : liblustre clients only
163 Description: getdirentries does not give error when run on compute nodes
164 Details : getdirentries does not fail when the size specified as an argument
165 is too small to contain at least one entry
167 Severity : enhancement
169 Description: Add LNET router traceability for debug purposes
170 Details : If a checksum failure occurs with a router as part of the
171 IO path, the NID of the last router that forwarded the bulk data
172 is printed so it can be identified.
177 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
178 Details : Due to a race condition, liblustre clients were occasionally
181 Severity : enhancement
183 Description: lfs setstripe use optional parameters instead of postional
186 Severity : enhancement
188 Description: Nanosecond timestamp support for ldiskfs
189 Details : The on-disk ldiskfs filesystem has added support for nanosecond
190 resolution timestamps. There is not yet support for this at
191 the Lustre filesystem level.
194 Frequency : during server recovery
196 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
197 Details : request with rq_no_resend flag not awake l_wait_event if they get a
201 Frequency : nfs export on patchless client
203 Description: connectathon hang when test nfs export over patchless client
204 Details : Disconnected dentry cannot be found with lookup, so we do not need
205 to unhash it or make it invalid
208 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
209 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
210 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
215 Description: open req refcounting wrong on reconnect
216 Details : If reconnect happened between getting open reply from server and
217 call to mdc_set_replay_data in ll_file_open, we will schedule
218 replay for unreferenced request that we are about to free.
219 Subsequent close will crash in variety of ways.
220 Check that request is still eligible for replay in
221 mdc_set_replay_data().
226 Description: disable writes to filesystem when reading health_check file
227 Details : the default for reading the health_check proc file has changed
228 to NOT do a journal transaction and write to disk, because this
229 can cause reads of the /proc file to hang and block HA state
230 checking on a healthy but otherwise heavily loaded system. It
231 is possible to return to the previous behaviour during configure
232 with --enable-health-write.
234 Severity : enhancement
236 Description: 64-bit inode version
237 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
238 to the inode. This will be required for version-based recovery.
243 Description: MDS fails to start if a duplicate client export is detected
244 Details : in some rare cases it was possible for a client to connect to
245 an MDS multiple times. Upon recovery the MDS would detect this
246 and fail during startup. Handle this more gracefully.
248 Severity : enhancement
250 Description: Add -o localflock option to simulate old noflock
252 Details : This will achieve local-only flock/fcntl locks
258 Description: log_commit_thread vs filter_destroy race leads to crash
259 Details : Take import reference before releasing llog record semaphore
264 Description: Wrong request locking in request set processing
265 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
266 imp_delayed_list, in this place should be used imp_lock.
269 Frequency : when reconnection
271 Description: Grant Leak when osc reconnect to OST
272 Details : When osc reconnect ost, OST(filter) should check whether it
273 should grant more space to client by comparing fed_grant and
274 cl_avail_grant, and return the granted space to client instead
275 of "new granted" space, because client will call osc_init_grant
276 to update the client grant space info.
279 Frequency : when client reconnect to OST
281 Description: Grant Leak when osc do resend and replay bulk write
282 Details : When osc reconnect to OST, OST(filter)should clear grant info of
283 bulk write request, because the grant info while be sync between
284 OSC and OST when reconnect, and we should ignore the grant info
285 these of resend/replay write req.
290 Description: Grant space more than avaiable left space sometimes.
291 Details : When then OST is about to be full, if two bulk writing from
292 different clients came to OST. Accord the avaliable space of the
293 OST, the first req should be permitted, and the second one
294 should be denied by ENOSPC. But if the seconde arrived before
295 the first one is commited. The OST might wrongly permit second
296 writing, which will cause grant space > avaiable space.
299 Frequency : when client is evicted
301 Description: Grant might be wrongly erased when osc is evicted by OST
302 Details : when the import is evicted by server, it will fork another
303 thread ptlrpc_invalidate_import_thread to invalidate the
304 import, where the grant will be set to 0. While the original
305 thread will update the grant it got when connecting. So if
306 the former happened latter, the grant will be wrongly errased
307 because of this race.
312 Description: Checking Stale with correct fid
313 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
314 to check whether it is stale, because sometimes, we want the
315 enqueue happened anyway, and op_data.fid2 will not be initialized.
317 Severity : enhancement
319 Description: update patchless client
320 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
323 Frequency : only with 2.4 kernel
325 Description: random memory corruption
326 Details : size of struct ll_inode_info is to big for union inode.u and this
327 can be cause of random memory corruption.
332 Description: Memory leak in recovery
333 Details : Lov_mds_md was not free in an error handler in mds_create_object.
334 It should also check obd_fail before fsfilt_start, otherwise if
335 fsfilt_start return -EROFS,(failover mds during mds recovery).
336 then the req will return with repmsg->transno = 0 and rc = EROFS.
337 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
338 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
339 be freed no matter whether fsfilt_commit success or not.
342 Frequency : only with huge count clients
344 Description: Prevents from taking the superblock lock in llap_from_page for
346 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
347 for llap_from_page call in ll_removepage prevents from taking the
348 superblock lock for a soon died page.
353 Description: Not check open intent error before release open handle
354 Details : in some rare cases, the open intent error is not checked before
355 release open handle, which may cause
356 ASSERTION(open_req->rq_transno != 0), because it tries to release
357 the failed open handle.
362 Description: Set cat log bitmap only after create log success.
363 Details : in some rare cases, the cat log bitmap is set too early. and it
364 should be set only after create log success.
368 Description: Accessing a block bevice can re-enable I/O when Lustre is
369 tearing down a device.
370 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
375 Description: service threads may hog cpus when there are a lot of requests
377 Details : Insert cond_resched to give other threads a chance to use some of
383 Description: the cat log was not initialized in recovery
384 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
385 unlink log on mds will not be initialized until mds post
386 recovery. And also in mds post recovery, the unlink log will
387 initialization will be done asynchronausly, so there will be race
388 between add unlink log and unlink log initialization.
392 Description: brw_stats were being printed incorrectly
393 Details : brw_stats were being printed as log2 but all of them were not
394 recorded as log2. Also remove some code duplication arising from
395 filter_tally_{read,write}.
399 Frequency : rare, only in recovery.
400 Description: ASSERTION(req->rq_type != LI_POISON) failed
401 Details : imp_lock should be held while iterating over imp_sending_list for
402 prevent destroy request after get timeout in ptlrpc_queue_wait.
406 Description: replay-single.sh test 52 fails
407 Details : A lock's skiplist need to be cleanup when it being unlinked
408 from its resource list.
412 Description: Short directio read returns full requested size rather than
414 Details : Direct I/O operations should return actual amount of bytes
415 transferred rather than requested size.
417 Severity : enhancement
419 Description: metadata RPC reduction (e.g. for rm performance)
420 Details : decrease the amount of synchronous RPC between clients and servers
421 by canceling conflicing lock before the operation on the client side
422 and packing thier handles into the main operation RPC to server.
424 --------------------------------------------------------------------------------
426 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
431 Frequency : on some architectures
433 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
434 Details : Uninitialized flags sometimes cause configuration commands to
438 Frequency : patchless clients only
440 Description: missing __iget() symbol export
441 Details : The __iget() symbol export is missing. To avoid the need for
442 this on patchless clients the deathrow inode reaper is turned
443 off, and we depend on the VM to clean up old inodes. This
444 dependency was during via the fix for bug 12181.
446 --------------------------------------------------------------------------------
448 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
450 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
451 INTEROPERATE with older versions automatically. In many cases a
452 special upgrade step is needed. Please read the
453 user documentation before upgrading any part of a live system.
454 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
455 version will not interoperate with 1.6 betas before beta5 (1.5.95).
456 * WARNING: Lustre configuration and startup changes are required with
457 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
462 Severity : enhancement
464 Description: MountConf
465 Details : Lustre configuration is now managed via mkfs and mount
466 commands instead of lmc and lconf. New obd types (MGS, MGC)
467 are added for dynamic configuration management. See
468 https://mail.clusterfs.com/wikis/lustre/MountConf for
471 Severity : enhancement
473 Description: dynamic OST addition
474 Details : OSTs can now be added to a live filesystem
476 Severity : enhancement
478 Description: startup order invariance
479 Details : MDTs and OSTs can be started in any order. Clients only
480 require the MDT to complete startup.
482 Severity : enhancement
484 Description: parallel, asynchronous orphan cleanup
485 Details : orphan cleanup is now performed in separate threads for each
486 OST, allowing parallel non-blocking operation.
488 Severity : enhancement
490 Description: optimized stripe assignment
491 Details : stripe assignments are now made based on ost space available,
492 ost previous usage, and OSS previous usage, in order to try
493 to optimize storage space and networking resources.
495 Severity : enhancement
497 Description: Permanently set tunables
498 Details : All writable /proc/fs/lustre tunables can now be permanently
499 set on a per-server basis, at mkfs time or on a live
502 Severity : enhancement
504 Description: Lustre message v2
505 Details : Add lustre message format v2.
507 Severity : enhancement
509 Description: client OST exclusion list
510 Details : Clients can be started with a list of OSTs that should be
511 declared "inactive" for known non-responsive OSTs.
515 Description: SPEC SFS validation failure on NFS v2 over lustre.
516 Details : Changes the blocksize for regular files to be 2x RPC size,
517 and not depend on stripe size.
519 Severity : enhancement
521 Description: Multiple MD RPCs in flight.
522 Details : Further unserialise some read-only MDS RPCs - learn about intents.
523 To avoid overly-overloading MDS, introduce a limit on number of
524 MDS RPCs in flight for a single client and add /proc controls
525 to adjust this limit.
527 Severity : enhancement
529 Description: client read/write statistics
530 Details : Add client read/write call usage stats for performance
531 analysis of user processes.
532 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
533 file access. extents_stats shows chunk size distribution.
534 extents_stats_per_process show chunk size distribution per
537 Severity : enhancement
539 Description: mds statistics
540 Details : Add detailed mds operations statistics in
541 /proc/fs/lustre/mds/*/stats.
545 Description: Failure of copying files with lustre special EAs.
546 Details : Client side always return success for setxattr call for lustre
547 special xattr (currently only "trusted.lov").
552 Description: Refcount LNET uuids
553 Details : The global LNET uuid list grew linearly with every startup;
554 refcount repeated list entries instead of always adding to
557 Severity : enhancement
559 Description: Dynamic service threads
560 Details : Within a small range, start extra service threads
561 automatically when the request queue builds up.
564 Frequency : mixed-endian client/server environments
566 Description: mixed-endian crashes
567 Details : The new msg_v2 system had some failures in mixed-endian
570 Severity : enhancement
572 Description: Easy OST removal
573 Details : OSTs can be permanently deactivated with e.g. 'lctl
574 conf_param lustre-OST0001.osc.active=0'
576 Severity : enhancement
578 Description: MGS proc entries
579 Details : Added basic proc entries for the MGS showing what filesystems
582 Severity : enhancement
584 Description: provide MGS failover
585 Details : Added config lock reacquisition after MGS server failover.
587 Severity : enhancement
589 Description: add Linux 2.4 support
590 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
594 Description: a large application tries to do I/O to the same resource and dies
596 Details : Check the req->rq_arrival time after the call to
597 ost_brw_lock_get(), but before we do anything about
598 processing it & sending the BULK transfer request. This
599 should help move old stale pending locks off the queue as
600 quickly as obd_timeout.
603 Frequency : when an incorrect nid is specified during startup
605 Description: ptlrpc connect to non-existant node causes kernel crash
606 Details : LNET can't be re-entered from an event callback, which
607 happened when we expire a message after the export has been
608 cleaned up. Instead, hand the zombie cleanup off to another
611 Severity : enhancement
613 Description: plain/inodebits lock performance improvement
614 Details : Grouping plain/inodebits in granted list by their request modes
615 and bits policy, thus improving the performance of search through
619 Frequency : only if OST filesystem is corrupted
621 Description: client incorrectly hits assertion in ptlrpc_replay_req()
622 Details : for a short time RPCs with bulk IO are in the replay list,
623 but replay of bulk IOs is unimplemented. If the OST filesystem
624 is corrupted due to disk cache incoherency and then replay is
625 started it is possible to trip an assertion. Avoid putting
626 committed RPCs into the replay list at all to avoid this issue.
629 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
631 Description: System hang on startup
632 Details : This bug allowed the liblustre (e.g. catamount) client to
633 return to the app before handling all startup RPCs. This
634 could leave the node unresponsive to lustre network traffic
635 and manifested as a server ptllnd timeout.
637 Severity : enhancement
639 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
640 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
641 boolean to control whether to print peer debug info when a
642 client's RPC times out.
645 Frequency : only for kernels with patches from Lustre below 1.4.3
647 Description: Remove old rdonly API
648 Details : Remove old rdonly API which unused from at least lustre 1.4.3
651 Frequency : only for devices with external journals
653 Description: Set external device read-only also
654 Details : During a commanded failover stop, we set the disk device
655 read-only while the server shuts down. We now also set any
656 external journal device read-only at the same time.
659 Frequency : when upgrading from 1.4 while trying to change parameters
661 Description: The wrong (new) MDC name was used when setting parameters for
662 upgraded MDT's. Also allows changing of OSC (and MDC)
663 parameters if --writeconf is specified at tunefs upgrade time.
666 Frequency : when setting specific ost indicies
668 Description: QOS code breaks on skipped indicies
669 Details : Add checks for missing OST indicies in the QOS code, so OSTs
670 created with --index need not be sequential.
672 Severity : enhancement
674 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
675 Details : The uninit_groups feature works in conjunction with the kernel
676 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
677 up the pass1 processing of e2fsck. This is a read-only feature
678 in ldiskfs2 only, so older kernels and current ldiskfs cannot
679 mount filesystems that have had this feature enabled.
681 Severity : enhancement
683 Description: Improve multi-block allocation algorithm to avoid fragmentation
684 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
685 allocation locality and avoid filesystem fragmentation.
687 ------------------------------------------------------------------------------
689 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
691 * Support for kernels:
692 2.4.21-47.0.1.EL (RHEL 3)
694 2.6.9-42.0.10.EL (RHEL 4)
695 2.6.12.6 vanilla (kernel.org)
696 2.6.16.27-0.9 (SLES 10)
697 * Recommended e2fsprogs version: 1.39.cfs5
699 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
703 Frequency : occasional, depends on client load and configuration
704 Bugzilla : 12181, 12203
705 Description: data loss for recently-modified files
707 Details : In some cases it is possible that recently written or created
708 files may not be written to disk in a timely manner (this should
709 normally be within 30s unless client IO load is very high).
710 The problem appears as zero-length files or files that are a
711 multiple of 1MB in size after a client crash or client eviction
712 that are missing data at the end of the file.
714 This problem is more likely to be hit on clients where files are
715 repeatedly created and unlinked in the same directory, clients
716 have a large amount of RAM, have many CPUs, the filesystem has
717 many OSTs, the clients are rebooted frequently, and/or the files
718 are not accessed by other nodes after being written.
720 The presence of the problem can be detected by looking at
721 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
722 smaller than the second (nr_unused) then dirty files will not
723 be flushed automatically to disk. "sync; sleep 10" should be
724 run several times on the node before unmounting it to update
725 Lustre (this is also safe to run on nodes without this problem).
727 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
728 that can cause this same problem, so customers using that kernel
729 also need to update the kernel in addition to Lustre. In order
730 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
732 It is normal that files written just before a client crash (less
733 than 30s) may not yet have been flushed to disk, even for local
737 Frequency : frequent on thin XT3 nodes
739 Description: UUID collision on thin XT3 Linux nodes
740 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
741 randomly, since we relied on an insufficiently-seeded PRNG.
746 Description: OSS hangs after "All ost request buffers busy"
747 Details : A deadlock between quota and journal operations caused OSS
748 hangs after printing "All ost request buffers busy."
751 Frequency : always on liblustre builds
753 Description: Cleanup compiler warnings on liblustre
756 Frequency : always on liblustre builds on XT3
758 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
762 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
763 > 32000 subdirectories
764 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
765 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
766 subdirectory count crosses 32000. This will aid e2fsck to
767 correctly handle more than 32000 subdirectories.
770 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
772 Description: System hang on startup
773 Details : This bug allowed the liblustre (e.g. catamount) client to
774 return to the app before handling all startup RPCs. This
775 could leave the node unresponsive to lustre network traffic
776 and manifested as a server ptllnd timeout.
778 Severity : enhancement
780 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
781 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
782 boolean to control whether to print peer debug info when a
783 client's RPC times out.
788 Description: make O_SYNC working on 2.6 kernels
789 Details : 2.6 kernels use different method for mark pages for write,
790 so need add a code to lustre for O_SYNC work.
795 Description: Failure to close file and release space on NFS
796 Details : Put inode details into lock acquired in ll_intent_file_open.
797 Use mdc_intent_lock in ll_intent_open to properly
798 detect all kind of errors unhandled by mdc_enqueue.
803 Description: proc file read during shutdown sometimes raced obd removal,
805 Details : Add lock to prevent obd access after proc file removal.
808 Frequency : Only for files larger than 4GB on 32-bit clients.
810 Description: improperly doing page alignment of locks
811 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
812 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
817 Description: RPCs being resent when they shouldn't be
818 Details : Some RPCs that should not be resent are being resent. This
819 can cause inconsistencies in the RPC state machine. Do not
820 resend such requests.
823 Frequency : rare, only with NFS export
825 Description: Crash on NFS re-export node
826 Details : under very unusual load conditions an assertion is hit in
827 ll_intent_file_open()
830 Frequency : only if OST filesystem is corrupted
832 Description: client incorrectly hits assertion in ptlrpc_replay_req()
833 Details : for a short time RPCs with bulk IO are in the replay list,
834 but replay of bulk IOs is unimplemented. If the OST filesystem
835 is corrupted due to disk cache incoherency and then replay is
836 started it is possible to trip an assertion. Avoid putting
837 committed RPCs into the replay list at all to avoid this issue.
842 Description: large O_DIRECT requests fail under memory pressure/fragmentation
843 Details : Large single O_DIRECT read and write calls can fail to allocate
844 a sufficiently large buffer to process the request. In case of
845 allocation failure the allocation is retried with a smaller
846 buffer and broken into smaller requests.
848 Severity : enhancement
850 Description: Add -o localflock option to simulate old noflock behaviour.
851 Details : This will achieve local-only flock/fcntl locks coherentness.
856 Description: versioning check is incomplete
857 Details : Checking the version difference of client vs. server, report
858 error if the gap is too big.
863 Description: add support PG_writeback bit
864 Details : add support for PG_writeback bit for Lustre, for more carefull
865 work with page cache in 2.6 kernel. This also fix some deadlocks
866 and remove hack for work O_SYNC with 2.6 kernel.
868 Severity : enhancement
870 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
871 Details : The uninit_groups feature works in conjunction with the kernel
872 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
873 up the pass1 processing of e2fsck. This is a read-only feature
874 in ldiskfs2 only, so older kernels and current ldiskfs cannot
875 mount filesystems that have had this feature enabled.
877 Severity : enhancement
879 Description: Improve multi-block allocation algorithm to avoid fragmentation
880 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
881 allocation locality and avoid filesystem fragmentation.
883 ------------------------------------------------------------------------------
885 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
887 * Support for kernels:
888 2.6.9-42.0.3.EL (RHEL 4)
890 2.4.21-47.0.1.EL (RHEL 3)
891 2.6.12.6 vanilla (kernel.org)
892 2.6.16.21-0.8 (SLES10)
893 * Recommended e2fsprogs version: 1.39.cfs2-0
895 * The backwards-compatible /proc/sys/portals symlink has been removed
896 in this release. Before upgrading, please ensure that you change
897 any configuration scripts or /etc/sysctl.conf files that access
898 /proc/sys/portals/* or sysctl portals.* to use the corresponding
899 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
900 in advance of the upgrade on any system running Lustre 1.4.6 or
901 newer, since /proc/sys/lnet was added in that version.
902 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
906 Frequency : only when quota is used
908 Description: avoid scanning export list for quota master
909 Details : Change the algorithms to avoid scanning export list in order
910 to improve the efficiency.
913 Frequency : MDS failover only, very rarely
915 Description: "went back in time" messages on mds failover
916 Details : The greatest transno may be lost when the current operation
917 finishes with an error (transno==0) and the client's last_rcvd
918 record is over-written. Save the greatest transno in the
919 mds_last_transno for this case.
922 Frequency : always for specific kernels and striping counts
924 Description: client may get "Matching packet too big" without ACL support
925 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
926 when trying to access files in certain configurations. The
927 clients should in fact be denied when mounting because they do
931 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
933 Description: many clients connecting with IO in progress causes connect timeouts
934 Details : Avoid synchronous journal commits to avoid delays caused by many
935 clients connecting/disconnecting when bulk IO is in progress.
936 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
937 OST_IO_PORTAL to avoid delays behind potentially many pending
941 Frequency : occasionally with multiple writers to a single file
943 Description: shared writes to file may result in wrong size reported by stat()
944 Details : Allow growing of kms when extent lock is cancelled
947 Frequency : always with random mmap IO to multi-striped file
949 Description: mmap write might be lost if we are writing to a 'hole' in stripe
950 Details : Only if the hole is at the end of OST object so that kms is too
951 small. Fix is to increase kms accordingly in ll_nopage.
954 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
956 Description: writes to a missing object would leak memory on the OST
957 Details : If there is an inconsistency between the MDS and OST filesystems,
958 such that the MDS references an object that doesn't exist, writes
959 to that object will leak memory due to incorrect cleanup in the
960 error handling path, eventually running out of memory on the OST.
965 Description: Creating too long symlink causes lustre errors
966 Details : Check symlink and name lengths before sending requests to MDS.
969 Frequency : only if flock is enabled (not on by default)
971 Description: posix locks not released on fd closure on 2.6.9+
972 Details : We failed to add posix locks to list of inode locks on 2.6.9+
973 kernels, this caused such locks not to be released on fd close and
974 then assertions on fs unmount about still used locks.
977 Frequency : MDS failover only, very rarely
979 Description: clients may get ASSERTION(granted_lock != NULL)
980 Details : When request was taking a long time, and a client was resending
981 a getattr by name lock request. The were multiple lock requests
982 with the same client lock handle and
983 mds_getattr_name->fixup_handle_for_resent_request found one of the
984 lock handles but later failed with ASSERTION(granted_lock != NULL).
989 Description: handle->h_buffer_credits > 0, assertion failure
990 Details : h_buffer_credits is zero after truncate, causing assertion
991 failure. This patch extends the transaction or creates a new
995 Frequency : NFS re-export or patchless client
996 Bugzilla : 11179, 10796
997 Description: Crash on NFS re-export node (__d_move)
998 Details : We do not want to hash the dentry if we don't have a lock.
999 But if this dentry is later used in d_move, we'd hit uninitialised
1000 list head d_hash, so we just do this to init d_hash field but
1001 leave dentry unhashed.
1004 Frequency : NFS re-export or patchless client
1006 Description: NFS exports has problem with symbolic link
1007 Details : lustre client didn't properly install dentry when re-exported
1008 to NFS or running patchless client.
1011 Frequency : NFS re-export or patchless client
1013 Description: Various nfs/patchless fixes.
1014 Details : fixes reuse disconected alias for lookup process - this fixes
1015 warning "find_exported_dentry: npd != pd",
1016 fix permission error with open files at nfs.
1017 fix apply umask when do revalidate.
1020 Frequency : occasional
1022 Description: Crash on NFS re-export node
1023 Details : calling clear_page() on the wrong pointer triggered oops in
1024 generic_mapping_read().
1027 Frequency : rarely, using O_DIRECT IO
1029 Description: unaligned directio crashes client with LASSERT
1030 Details : check for unaligned buffers before trying any requests.
1033 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1035 Description: stale data returned from RAID cache
1036 Details : If only a small amount of IO is done to the RAID device before
1037 reading it again it is possible to get stale data from the RAID
1038 cache instead of reading it from disk.
1041 Frequency : always for sles10 kernel
1043 Description: sles10 support
1044 Details : ll_follow_link: compile fixes and using of nd_set_link
1045 under newer kernels.
1048 Frequency : depends on arch, kernel and compiler version, always on sles10
1051 Description: recursive or deep enough symlinks cause stack overflow
1052 Details : getting rid of large stack-allocated variable in
1056 Frequency : depends on hardware
1058 Description: lustre write performance loss in the SLES10 kernel
1059 Details : the performance loss is caused by using of write barriers in the
1060 ext3 code. The SLES10 kernel turns barrier support on by
1061 default. The fix is to undo that change for ldiskfs.
1063 ------------------------------------------------------------------------------
1065 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1067 * Support for kernels:
1068 2.6.9-42.0.3EL (RHEL 4)
1069 2.6.5-7.276 (SLES 9)
1070 2.4.21-47.0.1.EL (RHEL 3)
1071 2.6.12.6 vanilla (kernel.org)
1075 Frequency : quota enabled and large files being deleted
1077 Description: releasing more than 4GB of quota at once hangs OST
1078 Details : If a user deletes more than 4GB of files on a single OST it
1079 will cause the OST to spin in an infinite loop. Release
1080 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1085 Description: statfs data retrieved from /proc may be stale or zero
1086 Details : When reading per-device statfs data from /proc, in the
1087 {kbytes,files}_{total,free,avail} files, it may appear
1088 as zero or be out of date.
1091 Frequency : systems with MD RAID1 external journal devices
1093 Description: lconf's call to blkid is confused by RAID1 journal devices
1094 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1095 of returning all block devices that match the journal UUID.
1098 Frequency : always, for aggregate stripe size over 4GB
1100 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1101 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1102 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1105 Frequency : always if "lfs find" used on a local file/directory
1107 Description: "lfs find" segfaults if used on a local file/directory
1108 Details : The case where a directory component was not specified wasn't
1109 handled correctly. Handle this properly.
1112 Frequency : always on ppc64
1114 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1115 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1116 perform endianness conversion.
1119 Frequency : rarely (truncate to non-zero file size after write under load)
1120 Bugzilla : 10730, 10687
1121 Description: Files padded with zeros to next 4K multiple
1122 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1123 it is possible that files that are truncated to a non-zero size
1124 immediately after being written are filled with zero bytes beyond
1125 the truncated size. No file data is lost.
1127 Severity : enhancement
1129 Description: Allow recovery/failover for liblustre clients.
1130 Details : liblustre clients were unaware of failover configurations until
1133 Severity : enhancement
1135 Description: user file locks should fail when not mounting with flock option
1136 Details : Set up an error-returning stub in ll_file_operations.lock field
1137 to prevent incorrect behaviour when client is mounted without
1138 flock option. Also, set up properly f_op->flock field for
1142 Frequency : always on ia64
1144 Description: "lfs df" loops on printing out MDS statfs information
1145 Details : The obd_ioctl_data was not initialized and in some systems
1146 this caused a failure during the ioctl that did not return
1147 an error. Initialize the struct and return an error on failure.
1150 Frequency : SLES 9 only
1152 Description: Error of copying files with lustre special EAs as root
1153 Details : Client side always return success for setxattr call for lustre
1154 special xattr (currently only "trusted.lov").
1157 Frequency : rarely on clusters with both ia64+i386 clients
1159 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1160 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1161 clients are doing concurrent writes to the same file it is
1162 possible that the ia64 clients may LASSERT because the OST
1163 extent locks are not PAGE_SIZE aligned. Ensure that grown
1164 locks are always aligned on the request boundary.
1167 Frequency : specific use, occasional
1169 Description: Overwriting in use executable truncates on-disk binary image
1170 Details : If one node attempts to overwrite an executable in use by
1171 another node, we now correctly return ETXTBSY instead of
1172 truncating the file.
1174 Severity : enhancement
1176 Description: Async OSC create to avoid the blocking unnecessarily.
1177 Details : If a OST has no remain object, system will block on the creating
1178 when need to create a new object on this OST. Now, ways use
1179 pre-created objects when available, instead of blocking on an
1180 empty osc while others are not empty. If we must block, we block
1181 for the shortest possible period of time.
1186 Description: chmod on Lustre root is propagated to other clients
1187 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1188 invalid by the follow_mount time.
1191 Frequency : liblustre clients only
1193 Description: Race in 'instant cancel' lock handling could lead to such locks
1194 never to be granted in case of SMP MDS
1195 Details : Do not destroy not yet granted but cbpending locks in
1199 Frequency : replay/resend of open
1201 Description: non null lock assetion failure in mds_intent_policy
1202 Details : Trying to replay/resend lockless open requests resulted in
1203 mds_open() returning 0 with no lock. Now it sets a flag if
1204 a lock is going to be returned.
1206 Severity : enhancement
1208 Description: Checksum enhancements
1209 Details : New checksum enhancements allow for resending RPCs that failed
1212 Severity : enhancement
1214 Description: Tunables on number of dirty pages in cacche
1215 Details : Allow to set limit on number of dirty pages cached.
1220 Description: client crash on unmount - lock still has references
1221 Details : In some error handling cases it was possible to leak a lock
1222 reference on a client while accessing a file. This was not
1223 harmful to the client during operation, but would cause the
1224 client to crash when the filesystem is unmounted.
1227 Frequency : specific case, rare
1229 Description: ETXTBSY on mds though file not in use by client
1230 Details : ETXTBSY is no longer incorrectly returned when attempting to
1231 chmod or chown a directory that the user previously tried to
1232 execute or a currently-executing binary.
1235 Frequency : extremely rare except on liblustre-based clients
1237 Description: Lustre space not freed when files are deleted
1238 Details : Clean up open-unlinked files after client eviction. Previously
1239 the unlink was skipped and the files remained as orphans.
1244 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1245 Details : In some cases it was possible to send a blocking callback to
1246 a client doing a glimpse, even though that client didn't get
1247 a lock granted. When the glimpse lock is cancelled on the OST
1248 the freed lock is left on the waiting list and corrupted the list.
1251 Frequency : all core dumps
1253 Description: Broke core dumps to lustre
1254 Details : Negative dentry may be unhashed if parent does not have UPDATE
1255 lock, but some callers, e.g. do_coredump, expect dentry to be
1256 hashed after successful create, hash it in ll_create_it.
1258 ------------------------------------------------------------------------------
1260 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1262 * Support for kernels:
1263 2.6.9-42.0.2.EL (RHEL 4)
1264 2.6.5-7.276 (SLES 9)
1265 2.4.21-40.EL (RHEL 3)
1266 2.6.12.6 vanilla (kernel.org)
1270 Frequency : always on RHEL 3
1272 Description: Number of open files grows over time
1273 Details : The number of open files grows over time, whether or not
1274 Lustre is started. This was due to a filp leak introduced
1275 by one of our kernel patches.
1277 ------------------------------------------------------------------------------
1279 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1281 * Support for kernels:
1282 2.6.9-42.EL (RHEL 4)
1283 2.6.5-7.276 (SLES 9)
1284 2.4.21-40.EL (RHEL 3)
1285 2.6.12.6 vanilla (kernel.org)
1290 Bugzilla : 5719, 9635, 9792, 9684
1291 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1292 Details : If a server is under heavy load and cannot reply to new
1293 connection requests before the client resends the (re)connect,
1294 the connection handling code can behave badly if two service
1295 threads are concurrently handing separate (re)connections from
1296 the same client. Add better locking to the connection handling
1297 code, and ensure that only a single connection will be processed
1298 for a given client UUID, even if the lock is dropped.
1300 Severity : enhancement
1302 Description: add TCP zero-copy support to kernel
1303 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1304 sends if the hardware supports scatter-gather and checksumming.
1305 This allows socklnd to do client-write and server-read more
1306 efficiently and reduce CPU utilization from skbuf copying.
1309 Frequency : only if NFS exporting from client
1311 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1312 Details : If setting attributes on a file created under NFS that had
1313 never been opened it would be possible to oops the client
1314 if the file had no objects.
1317 Frequency : always for liblustre
1319 Description: liblustre client does MDS+OSTs setattr RPC for each write
1320 Details : When doing a write from a liblustre client, the client
1321 incorrectly issued an RPC to the MDS and each OST the file was
1322 striped over in order to update the timestamps. When writing
1323 with small chunks and many clients this could overwhelm the MDS
1324 with RPCs. In all cases it would slow down the write because
1325 these RPCs are unnecessary.
1327 Severity : enhancement
1329 Description: allow number of MDS service threads to be changed at module load
1330 Details : It is now possible to change the number of MDS service threads
1331 running. Adding "options mds mds_num_threads={N}" to the MDS's
1332 /etc/modprobe.conf will set the number of threads for the next
1333 time Lustre is restarted (assuming the "mds" module is also
1334 reloaded at that time). The default number of threads will
1335 stay the same, 32 for most systems.
1340 Description: OST crash if filesystem is unformatted or corrupt
1341 Details : If an OST is started on a device that has never been formatted
1342 or if the filesystem is corrupt and cannot even mount then the
1343 error handling cleanup routines would dereference a NULL pointer.
1348 Description: NULL pointer deref in llap_from_page.
1349 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1350 incorrect) mapping if the page was truncated/reclaimed while it was
1351 searched for. Check for this condition and skip such pages when
1352 doing readahead. Introduce extra check to llap_from_page() to
1353 verify page->mapping->host is non-NULL (so page is not anonymous).
1356 Frequency : Sometimes when using sys_sendfile
1358 Description: "page not covered by a lock" warnings from ll_readpage
1359 Details : sendfile called ll_readpage without right page locks present.
1360 Now we introduced ll_file_sendfile that does necessary locking
1361 around call to generic_file_sendfile() much like we do in
1365 Frequency : with certain MDS communication failures at client mount time
1367 Description: NULL pointer deref after failed client mount
1368 Details : a client connection request may delayed by the network layer
1369 and not be sent until after the PTLRPC layer has timed out the
1370 request. If the client fails the mount immediately it will try
1371 to clean up before the network times out the request. Add a
1372 reference from the request import to the obd device and delay
1373 the cleanup until the network drops the request.
1376 Frequency : occasionally during client (re)connect
1378 Description: assertion failure during client (re)connect
1379 Details : processing a client connection request may be delayed by the
1380 client or server longer than the client connect timeout. This
1381 causes the client to resend the connection request. If the
1382 original connection request is replied in this interval, the
1383 client may trip an assertion failure in ptlrpc_connect_interpret()
1384 which thought it would be the only running connect process.
1387 Frequency : only with obd_echo servers and clients that are rebooted
1389 Description: kernel BUG accessing uninitialized data structure
1390 Details : When running an obd_echo server it did not start the ping_evictor
1391 thread, and when a client was evicted an uninitialized data
1392 structure was accessed. Start the ping_evictor in the RPC
1393 service startup instead of the OBD startup.
1395 Severity : enhancement
1396 Bugzilla : 10193 (patchless)
1397 Description: Remove dependency on various unexported kernel interfaces.
1398 Details : No longer need reparent_to_init, exit_mm, exit_files,
1399 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1402 Frequency : rare (only users of deprecated and unsupported LDAP config)
1404 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1405 Details : LDAP apparently contains 'lustreName' attributes instead of
1406 'name'. A simple remapping of the name is sufficient.
1409 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1411 Description: waiting_locks_callback trips kernel BUG if client is evicted
1412 Details : Running with the dump_on_timeout debug flag turned on makes
1413 it possible that the waiting_locks_callback() can try to dump
1414 the Lustre kernel debug logs from an interrupt handler. Defer
1415 this log dumping to the expired_lock_main() thread.
1417 Severity : enhancement
1419 Description: Support NFS exporting on 2.6 kernels.
1420 Details : Implement non-rawops metadata methods for NFS server to use without
1421 changing NFS server code.
1424 Frequency : very rare (synthetic metadata workload only)
1426 Description: two racing renames might cause an MDS thread to deadlock
1427 Details : Running the "racer" program may cause one MDS thread to rename
1428 a file from being the source of a rename to being the target of
1429 a rename at exactly the same time that another thread is doing
1430 so, and the second thread has already enqueued these locks after
1431 doing a lookup of the target and is trying to relock them in
1432 order. Ensure that we don't try to re-lock the same resource.
1435 Frequency : only very large systems with liblustre clients
1437 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1438 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1439 all OSTs in parallel. This allows the eviction of stale liblustre
1440 clients to proceed much faster than if they were done in series,
1441 and also offers similar improvements for other set_info RPCs.
1446 Description: excessive CPU usage during initial read phase on client
1447 Details : During the initial read phase on a client, it would agressively
1448 retry readahead on the file, consuming too much CPU and impacting
1449 performance (since 1.4.5.8). Improve the readahead algorithm
1450 to avoid this, and also improve some other common cases (read
1451 of small files in particular, where "small" is files smaller than
1452 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1457 Description: MDS crash when receiving packet with unknown intent.
1458 Details : Do not LBUG in unknown intent case, just return -EFAULT
1460 Severity : enhancement
1461 Bugzilla : 9293, 9385
1462 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1463 Details : Do not serialize getattr (non-intent version) and statfs.
1466 Frequency : occasional, when OST network is overloaded/intermittent
1468 Description: client evicted by OST after bulk IO timeout
1469 Details : If a client sends a bulk IO request (read or write) the OST
1470 may evict the client if it is unresposive to its data GET/PUT
1471 request. This is incorrect if the network is overloaded (takes
1472 too long to transfer the RPC data) or dropped the OST GET/PUT
1473 request. There is no need to evict the client at all, since
1474 the pinger and/or lock callbacks will handle this, and the
1475 client can restart the bulk request.
1478 Frequency : Always when mmapping file with no objects
1480 Description: client crashes when mmapping file with no objects
1481 Details : Check that we actually have objects in a file before doing any
1482 operations on objects in ll_vm_open, ll_vm_close and
1488 Description: Request leak when working with deleted CWD
1489 Details : Introduce advanced request refcount tracking for requests
1490 referenced from lustre intent.
1492 Severity : Enhancement
1494 Description: Cache open file handles on client.
1495 Details : MDS now will return special lock along with openhandle, if
1496 requested and client is allowed to hold openhandle, even if unused,
1497 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1498 closing files for every read/write openration.
1500 Severity : Enhancement
1502 Description: Cache open negative dentries on client when possible.
1503 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1504 negative dentries on lock revocation.
1509 Description: Remounting a client read-only wasn't possible with a zconf mount
1510 Details : It wasn't possible to remount a client read-only with llmount.
1512 Severity : enhancement
1513 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1514 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1515 MPICH is included to add Lustre-specific ADIO interfaces.
1516 This is based closely on the UFS ADIO layer and only differs
1517 in file creation, in order to allow the OST striping to be set.
1518 This is user-contributed code and not supported by CFS.
1523 Description: extended inode attributes (immutable, append-only) work improperly
1524 when 2.4 and 2.6 kernels are used on client/server or vice versa
1525 Details : Introduce kernel-independent values for these flags.
1527 Severity : enhancement
1530 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1531 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1532 granularity. Now, Fractional MB granularity is be supported,
1533 this is very useful for low memory system.
1535 Severity : enhancement
1537 Description: Getattr by fid
1538 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1539 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1540 have locks and avoids getattr rpc requests for it on every stat.
1543 Frequency : Always, for filesystems larger than 2TB
1545 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1546 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1547 filesystems larger than 2TB to be mounted with mballoc enabled.
1550 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1552 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1553 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1554 filesystems larger than 2TB is mounted then the kernel will
1555 silently corrupt the start of the filesystem. CONFIG_LBD is
1556 enabled for all CFS-supported kernels, but the possibility of
1557 this happening with a modified kernel config exists.
1559 Severity : enhancement
1561 Description: add client O_DIRECT support for 2.6 kernels
1562 Details : It is now possible to do O_DIRECT reads and writes to files
1563 in the Lustre client mountpoint on 2.6 kernel clients.
1565 Severity : enhancement
1567 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1568 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1569 parallel, not waiting for response from every OST before sending
1570 a rpc to the next OST.
1575 Description: setattr vs write race when updating file timestamps
1576 Details : Client processes that update a file timestamp into the past
1577 right after writing to the file (e.g. tar) it is possible that
1578 the updated file modification time can be reset to the current
1579 time due to a race between processing the setattr and write RPC.
1581 Severity : enhancement
1583 Description: Bring 'lfs find' closer in line with regular Linux find.
1584 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1585 -print0 options and obtains all the needed info through the lustre
1588 Severity : enhancement
1590 Description: support up to 1024 configured devices on one node
1591 Details : change obd_dev array from statically allocated to dynamically
1592 allocated structs as they are first used to reduce memory usage
1597 Description: Flush dirty partially truncated pages during truncate
1598 Details : Immediatelly flush partially truncated pages in filter_setattr,
1599 this way we completely avoid having any pages in page cache on OST
1600 and can retire ugly workarounds during writes to flush such pages.
1605 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1606 Details : protect inode from truncation within vfs_unlink() context
1607 just take a reference before calling vfs_unlink() and release it
1608 when parent's i_sem is free.
1613 Description: last_id value checked outside lock on OST caused LASSERT failure
1614 Details : If there were multiple MDS->OST object precreate requests in
1615 flight, it was possible that the OST's last object id was checked
1616 outside a lock and incorrectly tripped an assertion. Move checks
1617 inside locks, and discard old precreate requests.
1620 Frequency : always, if extents are used on OSTs
1622 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1623 Details : OSTs using the extents format would not zero the high 16 bits of
1624 the index physical block number. This is not a problem for any
1625 OST filesystems smaller than 16TB, and no kernels support ext3
1626 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1627 new/modified files) and can be fixed for existing filesystems
1628 with e2fsprogs-1.39-cfs1.
1633 Description: import connection selection may be incorrect if timer wraps
1634 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1635 import connections to be ignored if the 32-bit jiffies counter
1636 wraps. Use a 64-bit jiffies counter.
1639 Frequency : very large clusters immediately after boot
1641 Description: LNET request buffers exhausted under heavy short-term load
1642 Details : If a large number of client requests are generated on a service
1643 that has previously never seen so many requests it is possible
1644 that the request buffer growth cannot keep up with the spike in
1645 demand. Instead of dropping incoming requests, they are held in
1646 the LND until the RPC service can accept more requests.
1649 Frequency : Sometimes during replay
1651 Description: Assertion failure in ll_local_open after replay.
1652 Details : If replay happened on an open request reply before we were able
1653 to set replay handler, reply will become not swabbed tripping the
1654 assertion in ll_local_open. Now we set the handler right after
1655 recognising of open request
1658 Frequency : very rare
1660 Description: kernel reports "badness in vsnprintf"
1661 Details : Reading from the "recovery_status" /proc file in small chunks
1662 may cause a negative length in lprocfs_obd_rd_recovery_status()
1663 call to vsnprintf() (which is otherwise harmless). Exit early
1664 if there is no more space in the output buffer.
1666 Severity : enhancement
1668 Description: clear OBD RPC statistics by writing to them
1669 Details : It is now possible to clear the OBD RPC statistics by writing
1670 to the "stats" file.
1675 Description: Client mtime is not the same on different clients after utimes
1676 Details : In some cases, the client was using the utimes() syscall on
1677 a file cached on another node. The clients now validate the
1678 ctime from the MDS + OSTs to determine which one is right.
1683 Description: Inability to activate failout mode
1684 Details : lconf script incorrectly assumed that in pythong string's numeric
1685 value is used in comparisons.
1688 Frequency : always with multiple stripes per file
1690 Description: Inefficient object allocation for mutli-stripe files
1691 Details : When selecting which OSTs to stripe files over, for files with
1692 a stripe count that divides evenly into the number of OSTs,
1693 the MDS is always picking the same starting OST for each file.
1694 Return the OST selection heuristic to the original design.
1699 Description: mount failures may take full timeout to return an error
1700 Details : Under some heavy load conditions it is possible that a
1701 failed mount can wait for the full obd_timeout interval,
1702 possibly several minutes, before reporting an error.
1703 Instead return an error as soon as the status is known.
1705 Frequency : quota enabled and large files being deleted
1707 Description: releasing more than 4GB of quota at once hangs OST
1708 Details : If a user deletes more than 4GB of files on a single OST it
1709 will cause the OST to spin in an infinite loop. Release
1710 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1715 Description: statfs data retrieved from /proc may be stale or zero
1716 Details : When reading per-device statfs data from /proc, in the
1717 {kbytes,files}_{total,free,avail} files, it may appear
1718 as zero or be out of date.
1721 Frequency : systems with MD RAID1 external journal devices
1723 Description: lconf's call to blkid is confused by RAID1 journal devices
1724 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1725 of returning all block devices that match the journal UUID.
1728 Frequency : always, for aggregate stripe size over 4GB
1730 Description: assertion fails when trying to use 4GB stripe size
1731 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1732 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1735 Frequency : always on ppc64
1737 Description: the first write on an ext3 filesystem with mballoc got stuck
1738 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1739 perform endianness conversion.
1741 ------------------------------------------------------------------------------
1743 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1745 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1746 INTEROPERATE with older versions automatically. Please read the
1747 user documentation before upgrading any part of a live system.
1748 * WARNING: Lustre networking configuration changes are required with
1749 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1752 * Support for kernels:
1753 2.6.9-22.0.2.EL (RHEL 4)
1754 2.6.5-7.244 (SLES 9)
1755 2.6.12.6 vanilla (kernel.org)
1758 Severity : enhancement
1759 Bugzilla : 7981/8208
1760 Description: Introduced Lustre Networking (LNET)
1761 Details : LNET is new networking infrastructure for Lustre, it includes
1762 a reorganized network configuration mode (see the user
1763 documentation for full details) as well as support for routing
1764 between different network fabrics. Lustre Networking Devices
1765 (LNDs) for the supported network fabrics have also been
1766 created for this new infrastructure.
1768 Severity : enhancement
1769 Description: Introduced Access control lists
1770 Details : clients can set ACLs on files and directories in order to have
1771 more fine-grained permissions than the standard Unix UGO+RWX.
1772 The MDS must be started with the "-o acl" mount option.
1774 Severity : enhancement
1775 Description: Introduced filesystem quotas
1776 Details : Administrators may now establish per-user quotas on the
1779 Severity : enhancement
1781 Description: Configuration change for the XT3
1782 The PTLLND is now used to run Lustre over Portals on the XT3
1783 The configure option(s) --with-cray-portals are no longer used.
1784 Rather --with-portals=<path-to-portals-includes> is used to
1785 enable building on the XT3. In addition to enable XT3 specific
1786 features the option --enable-cray-xt3 must be used.
1791 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1792 Details : A race between allocating a new llcd and re-getting the llcd_lock
1793 allowed another thread to grab newly-allocated llcd.
1795 Severity : enhancement
1797 Description: 2.6 OST async journal commit and locking fix to improve performance
1798 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1799 2.6 kernels are now async as they already were in 2.4 kernels so
1800 that they can commit concurrently with the network bulk transfer.
1801 For block-allocated files the filter allocation semaphore is held
1802 to avoid filesystem fragmentation during allocation. BKL lock
1803 removed for 2.6 xattr operations where it is no longer needed.
1808 Description: lconf incorrectly determined whether two IP networks could talk
1809 Details : In some more complicated routing and multiple-network
1810 configurations, lconf will avoid trying to make a network
1811 connection to a disjoint part of the IP space. It was doing the
1812 math incorrectly for one set of cases.
1817 Description: Fix for potential infinite loop processing records in an llog.
1818 Details : If an llog record is corrupted/zeroed, it is possible to loop
1819 forever in llog_process(). Validate the llog record length
1820 and skip the remainder of the block on error.
1823 Frequency : occasional (liblustre only)
1825 Description: liblustre could not open files whose last component is a symlink
1826 Details : sysio_path_walk() would incorrectly pass the open intent to
1827 intermediate path components.
1830 Frequency : rare (liblustre only with non-standard tuning)
1831 Bugzilla : 7201 (7350)
1832 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1833 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1834 after it has been released, only do this for referenced locks
1836 Severity : enhancement
1838 Description: specifying an (invalid) directory default stripe_size of -1
1839 would reset the directory default striping
1840 Details : stripe_size -1 was used internally to signal directory stripe
1841 removal, now use "all default" to signal dir stripe removal
1842 as a directory striping of "all default" is not useful
1845 Frequency : common for large clusters running liblustre clients
1847 Description: doing an ls when liblustre clients are running is slow
1848 Details : sending a glimpse AST to a liblustre client waits for every AST
1849 to time out, as liblustre clients will not respond. Since they
1850 cannot cache data we refresh the OST lock LVB from disk instead.
1852 Severity : enhancement
1854 Description: doing an ls at the same time as file IO can be slow
1855 Details : enqueue and other "small" requests can be blocked behind many
1856 large IO requests. Create a new OST IO portal for non-IO
1857 requests so they can be processed faster.
1860 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1862 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1863 created file with mode 000
1864 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1865 allow an ftruncate() as a non-root user to a file with mode 000.
1866 Lustre now respects this flag to disable mode checks when
1867 truncating a file owned by the user
1870 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1873 Description: Revoking locks from clients that went dead or catatonic might take
1875 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1876 cancellation of such locks instant on servers without waiting for
1877 any reply from clients. Clients drops these locks when cancel
1878 notification from server is received without replying.
1881 Frequency : liblustre-only, when liblustre client dies or becomes busy
1883 Description: Doing ls on Linux clients can take a long time with active
1885 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1886 granting such locks to it in the first place if possible. Locks
1887 are taken by proxy on the OST during the read or write and
1888 dropped immediately afterward. Add connect flags handling, do
1889 not grant locks to liblustre clients for glimpse ASTs.
1891 Severity : enhancement
1893 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1894 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1895 sequential reads and behaves badly with applications doing random
1896 reads. Improve it by reading ahead at least read region, and
1897 avoiding excessive large RPC for small reads.
1899 Severity : enhancement
1901 Description: Creating more than 1000 files for a single job may cause a load
1902 imbalance on the OSTs if there are also a large number of OSTs.
1903 Details : qos_prep_create() uses an OST index reseed value that is an
1904 even multiple of the number of available OSTs so that if the
1905 reseed happens in the middle of the object allocation it will
1906 still utilize the OSTs as uniformly as possible.
1911 Description: OST or MDS may oops in ping_evictor_main()
1912 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1913 but doesn't restart at beginning of obd_exports_timed list
1916 Severity : enhancement
1918 Description: improve by-nid export eviction on the MDS and OST
1919 Details : allow multiple exports with the same NID to be evicted at one
1920 time without re-searching the exports list.
1923 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1925 Description: MDS may oops in groups_free()
1926 Details : in rare race conditions a newly allocated group_info struct is
1927 freed again, and this can be NULL. The 2.4 compatibility code
1928 for groups_free() checked for a NULL pointer, but 2.6 did not.
1931 Frequency : common for liblustre clients doing little filesystem IO
1932 Bugzilla : 9352, 7313
1933 Description: server may evict liblustre clients accessing contended locks
1934 Details : if a client is granted a lock or receives a completion AST
1935 with a blocking AST already set it would not reply to the AST
1936 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1939 Frequency : lfs setstripe, only systems with more than 160 OSTs
1941 Description: unable to set striping with a starting offset beyond OST 160
1942 Details : llapi_create_file() incorrectly limited the starting stripe
1943 index to the maximum single-file stripe count.
1946 Frequency : LDAP users only
1948 Description: lconf did not handle in-kernel recovery with LDAP properly
1949 Details : lconf/LustreDB get_refs() is searching the wrong namespace
1951 Severity : enhancement
1953 Description: bind OST threads to NUMA nodes to improve performance
1954 Details : all OST threads are uniformly bound to CPUs on a single NUMA
1955 node and do their allocations there to localize memory access
1957 Severity : enhancement
1959 Description: llmount can determine client NID directly from Myrinet (GM)
1960 Details : the client NID code from gmnalnid was moved directly into
1961 llmount, removing the need to use this or specifying the
1962 client NID explicitly when mounting GM clients with zeroconf
1965 Frequency : if client is started with down MDS
1967 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
1968 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
1969 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
1970 ptlrpc_import_delayed_req() because we didn't check if the
1971 request was interrupted, and we also didn't break out of the
1972 event loop if there was no timeout
1977 Description: data loss during non-page-aligned writes to a single file from
1978 both multiple nodes and multiple threads on one node at same time
1979 Details : updates to KMS and lsm weren't protected by common lock. Resulting
1980 inconsistency led to false short-reads, that were cached and later
1981 used by ->prepare_write() to fill in partially written page,
1982 leading to data loss.
1985 Frequency : always, if lconf --abort_recovery used
1987 Description: lconf --abort_recovery fails with 'Operation not supported'
1988 Details : lconf was attempting to abort recovery on the MDT device and not
1991 Severity : enhancement
1993 Description: remove cleanup logs
1994 Details : replace lconf-generated cleanup logs with lustre internal
1995 cleanup routines. Eliminates the need for client-cleanup and
1998 Severity : enhancement
2000 Description: add support for EAs (user and system) on lustre filesystems
2001 Details : it is now possible to store extended attributes in the Lustre
2002 client filesystem, and with the user_xattr mount option it
2003 is possible to allow users to store EAs on their files also
2005 Severity : enhancement
2007 Description: Add possibility (config option) to show minimal available OST free
2009 Details : When compiled with --enable-mindf configure option, statfs(2)
2010 (and so, df) will return least minimal free space available from
2011 all OSTs as amount of free space on FS, instead of summary of
2012 free spaces of all OSTs.
2014 Severity : enhancement
2016 Description: do not expand extent locks acquired on OST-side
2017 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2018 by server: they are not cached anyway.
2021 Frequency : when mmap is used/binaries executed from Lustre
2023 Description: Unmmap pages before throwing them away from read cache.
2024 Details : llap_shrink cache now attempts to unmap pages before discarding
2025 them (if unmapping failed - do not discard). SLES9 kernel has
2026 extra checks that trigger if this unmapping is not done first.
2031 Description: lconf didn't resolve symlinks before checking to see whether a
2032 given mountpoint was already in use
2035 Frequency : when migrating failover services
2036 Bugzilla : 6395, 9514
2037 Description: When migrating a subset of services from a node (e.g. failback
2038 from a failover service node) the remaining services would
2039 time out and evict clients.
2040 Details : lconf --force (implied by --failover) sets the global obd_timeout
2041 to 5 seconds in order to quickly disconnect, but this caused
2042 other RPCs to time out too quickly. Do not change the global
2043 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2045 Severity : enhancement
2046 Frequency : if MDS is started with down OST
2047 Bugzilla : 9439,5706
2048 Description: Allow startup/shutdown of an MDS without depending on the
2049 availability of the OSTs.
2050 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2051 Add appropriate locking and lov-osc refcounts for safe
2052 cleaning. Add osc abort_inflight calls in case the
2053 synchronize never started.
2056 Frequency : occasional (Cray XT3 only)
2058 Description: root not authorized to access files in CRAY_PORTALS environment
2059 Details : The client process capabilities were not honoured on the MDS in
2060 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2061 been accessed by an authorized user then root was able to access
2062 the file on the local client also. The root user capabilities
2063 are now allowed on the MDS, as this environment has secure UID.
2066 Frequency : occasional
2068 Description: ldiskfs "too long searching" message happens too often
2069 Details : A debugging message (otherwise harmless) prints too often on
2070 the OST console. This has been reduced to only happen when
2071 there are fragmentation problems on the filesystem.
2076 Description: Division by zero in statfs when all OSCs are inactive
2077 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2078 lov_statfs divides by value returned by lov_get_stripecnt().
2082 Bugzilla : 9489, 3273
2083 Description: First write from each client to each OST was only 4kB in size,
2084 to initialize client writeback cache, which caused sub-optimal
2085 RPCs and poor layout on disk for the first writen file.
2086 Details : Clients now request an initial cache grant at (re)connect time
2087 and so that they can start streaming writes to the cache right
2088 away and always do full-sized RPCs if there is enough data.
2089 If the OST is rebooted the client also re-establishes its grant
2090 so that client cached writes will be honoured under the grant.
2095 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2096 Details : Now I/O RPCs go to different portal number and (presumably) fast
2097 lock requests (and glimses) and other RPCs get their own service
2098 threads pool that should be able to service those RPCs
2101 Severity : enhancement
2103 Description: Ability to exchange lustre version between client and servers and
2104 issue warnings at client side if client is too old. Also for
2105 liblustre clients there is ability to refuse connection of too old
2107 Details : New 'version' field is added to connect data structure that is
2108 filled with version info. That info is later checked by server and
2112 Frequency : rare, liblustre only.
2113 Bugzilla : 9296, 9581
2114 Description: Two simultaneous writes from liblustre at offset within same page
2115 might proceed at the same time overwriting eachother with stale
2117 Details : I/O lock withing llu_file_prwv was released too early, before data
2118 actually was hitting the wire. Extended lock-holding time until
2119 server acknowledges receiving data.
2122 Frequency : extremely rare. Never observed in practice.
2124 Description: avoid generating lustre_handle cookie of 0.
2125 Details : class_handle_hash() generates handle cookies by incrementing
2126 global counter, and can hit 0 occasionaly (this is unlikely, but
2127 not impossible, because initial value of cookie counter is
2128 selected randonly). Value of 0 is used as a sentinel meaning
2129 "unassigned handle" --- avoid it. Also coalesce two critical
2130 sections in this function into one.
2132 Severity : enhancement
2134 Description: allow liblustre clients to delegate truncate locking to OST
2135 Details : To avoid overhead of locking, liblustre client instructs OST to
2136 take extent lock in ost_punch() on client's behalf. New connection
2137 flag is added to handle backward compatibility.
2139 Severity : enhancement
2140 Bugzilla : 4928, 7341, 9758
2141 Description: allow number of OST service threads to be specified
2142 Details : a module parameter allows the number of OST service threads
2143 to be specified via "options ost ost_num_threads={N}" in the
2144 OSS's /etc/modules.conf or /etc/modprobe.conf.
2148 Bugzilla : 6146, 9635, 9895
2149 Description: servers crash with bad pointer in target_handle_connect()
2150 Details : In rare cases when a client is reconnecting it was possible that
2151 the connection request was the last reference for that export.
2152 We would temporarily drop the export reference and get a new
2153 one, but this may have been the last reference and the export
2154 was just destroyed. Get new reference before dropping old one.
2156 Severity : enhancement
2157 Frequency : if client is started with failover MDS
2159 Description: Allow multiple MDS hostnames in the mount command
2160 Details : Try to read the configuration from all specified MDS
2161 hostnames during a client mount in case the "primary"
2164 Severity : enhancement
2166 Description: Stop sending data to evicted clients as soon as possible.
2167 Details : Check if the client we are about to send or are sending data to
2168 was evicted already. (Check is done every second of waiting,
2169 for which l_wait_event interface was extended to allow checking
2170 of exit condition at specified intervals).
2173 Frequency : rare, normally only when NFS exporting is done from client
2175 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2177 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2180 Frequency : always, if srand() is called before liblustre initialization
2182 Description: Liblustre uses system PRNG disturbing its usage by user application
2183 Details : Introduce internal to lustre fast and high-quality PRNG for
2184 lustre usage and make liblustre and some other places in generic
2185 lustre code to use it.
2187 Severity : enhancement
2188 Bugzilla : 9477, 9557, 9870
2189 Description: Verify that the MDS configuration logs are updated when xml is
2190 Details : Check if the .xml configuration logs are newer than the config
2191 logs stored on the MDS and report an error if this is the case.
2192 Request --write-conf, or allow starting with --old_conf.
2194 Severity : enhancement
2196 Description: Handle symlinks in the path when checking if Lustre is mounted.
2197 Details : Resolve intermediate symlinks when checking if a client has
2198 mounted a filesystem to avoid duplicate client mounts.
2203 Description: lconf can hit an error exception but still return success.
2204 Details : The lconf command catches the Command error exception at the top
2205 level script context and will exit with the associated exit
2206 status, but doesn't ensure that this exit status is non-zero.
2211 Description: failure of ptlrpc thread startup can cause oops
2212 Details : Starting a ptlrpc service thread can fail if there are a large
2213 number of threads or the server memory is very fragmented.
2214 Handle this without oopsing.
2217 Frequency : always, only if liblustre and non-default acceptor port was used
2219 Description: liblustre cannot connect to servers with non-default acceptor port
2220 Details : tcpnal_set_default_params() was not called and was therefore
2221 ignoring the environment varaible TCPNAL_PORT, as well as other
2222 TCPNAL_ environment variables
2227 Description: two objects could be created on the same OST for a single file
2228 Details : If an OST is down, in some cases it was possible to create two
2229 objects on a single OST for a single file. No problems other
2230 than potential performance impact and spurious error messages.
2234 Bugzilla : 5681, 9562
2235 Description: Client may oops in ll_unhash_aliases
2236 Details : Client dcache may become inconsistent in race condition.
2237 In some cases "getcwd" can fail if the current directory is
2243 Description: Inode refcounting problems in NFS export code
2244 Details : link_raw functions used to call d_instantiate without obtaining
2245 extra inode reference first.
2249 Bugzilla : 9942, 9903
2250 Description: Referencing freed requests leading to crash, memleaks with NFS.
2251 Details : We used to require that call to ll_revalidate_it was always
2252 followed by ll_lookup_it. Also with revalidate_special() it is
2253 possible to call ll_revalidate_it() twice for the same dentry
2254 even if first occurence returned success. This fix changes semantic
2255 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2256 reference on a request referred from the intent.
2257 ll_intent_release() then releases such a request.
2260 Frequency : rare, normally benchmark loads only
2262 Description: unlinked inodes were kept in memory on the client
2263 Details : If a client is repeatedly creating and unlinking files it
2264 can accumulate a lot of stale inodes in the inode slab cache.
2265 If there is no other client load running this can cause the
2266 client node to run out of memory. Instead flush old inodes
2267 from client cache that have the same inode number as a new inode.
2270 Frequency : SLES9 2.6.5 kernel and long filenames only
2271 Bugzilla : 9969, 10379
2272 Description: utime reports stale NFS file handle
2273 Details : SLES9 uses out-of-dentry names in some cases, which confused
2274 the lustre dentry revalidation. Change it to always use the
2278 Frequency : rare, unless heavy write-truncate concurrency is continuous
2279 Bugzilla : 4180, 6984, 7171, 9963, 9331
2280 Description: OST becomes very slow and/or deadlocked during object unlink
2281 Details : filter_destroy() was holding onto the parent directory lock
2282 while truncating+unlinking objects. For very large objects this
2283 may block other threads for a long time and slow overall OST
2284 responsiveness. It may also be possible to get a lock ordering
2285 deadlock in this case, or run out of journal credits because of
2286 the combined truncate+unlink. Solution is to do object truncate
2287 first in one transaction without parent lock, and then do the
2288 final unlink in a new transaction with the parent lock. This
2289 reduces the lock hold time dramatically.
2292 Frequency : rare, 2.4 kernels only
2294 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2295 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2296 before unmounting their filesystem in order to avoid deadlock.
2297 The kernel_locked() function in 2.4 kernels only checks whether
2298 the kernel lock is held, not whether it is this process that is
2299 holding it as 2.6 kernels do.
2304 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2305 Details : The client ptlrpc code may be trying to reconnect to a down
2306 server before a previous connection attempt has timed out.
2307 Increase the reconnect interval to be longer than the connection
2308 timeout interval to avoid sending duplicate connections to
2312 Frequency : echo_client brw_test command
2314 Description: fix echo_client to work with OST preallocated code
2315 Details : OST preallocation code (5137) didn't take echo_client IO path
2316 into account: echo_client calls filter methods outside of any
2317 OST thread and, hence, there is no per-thread preallocated
2318 pages and buffers to use. Solution: hijack pga pages for IO. As
2319 a byproduct, this avoids unnecessary data copying.
2323 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2324 Description: Client can oops in mdc_commit_close() after open replay
2325 Details : It was possible for the MDS to return an open request with no
2326 transaction number in mds_finish_transno() if the client was
2327 evicted, but without actually returning an error. Clients
2328 would later try to replay that open and may trip an assertion
2329 Simplify the client close codepath, and always return an error
2330 from the MDS in case the open is not successful.
2333 Frequency : rare, 2.6 OSTs only
2335 Description: OST may deadlock under high load on fragmented files
2336 Details : If there was a heavy load and highly-fragmented OST filesystems
2337 it was possible to have all the OST threads deadlock waiting on
2338 allocation of biovecs, because the biovecs were not released
2339 until the entire RPC IO was completed. Instead, release biovecs
2340 as soon as they are complete to ensure forward IO progress.
2342 Severity : enhancement
2344 Description: Support for specifying external journal device at mount
2345 Details : If an OST or MDS device is formatted with an external journal
2346 device, this device major/minor is stored in the ext3 superblock
2347 and may not be valid for failover. Allow detecting and
2348 specifying the external journal at mount time.
2353 Description: Mounting an MDS with pending unlinked files may cause oops
2354 Details : target_finish_recovery() calls mds_postrecov() which returned
2355 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2356 considers this an error and immediately begins cleaning up the
2357 lov, just after starting the mds_lov process
2359 Severity : enhancement
2361 Description: Implement 'lfs df' to report actual free space on per-OST basis
2362 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2363 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2364 usage of objects; '-h' to report in human readable format.
2366 ------------------------------------------------------------------------------
2368 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2375 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2376 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2377 at startup during orphan unlinking. Instead, skip buddy update
2378 before setup, as it will be regenerated after recovery is complete.
2383 Description: If an OST is inactive, its locks might reference stale inodes.
2384 Details : lov_change_cbdata() must iterate over all namespaces, even if
2385 they are inactive to clear inode references from the lock.
2387 Severity : enhancement
2388 Frequency : occasional, if non-standard max_dirty_mb used
2390 Description: Client will block write RPCs if not enough grant
2391 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2392 then the client will block writes while waiting for another RPC
2393 to complete instead of consuming its dirty limit. With change
2394 we get improved performance when max_dirty_mb is small.
2396 Severity : enhancement
2397 Bugzilla : 3389, 6253
2398 Description: Add support for supplementary groups on the MDS.
2399 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2400 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2401 lookups for user supplementary groups into a cache.
2405 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2406 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2407 for liblustre O_EXCL is enforced by MDS.
2411 Description: readdir never returns NULL in liblustre.
2412 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2413 dentry in struct dirent.
2417 Frequency : liblustre only, depends on application IO pattern
2418 Description: liblustre clients evicted if not contacting servers
2419 Details : Don't put liblustre clients into the ping_evictor list, so
2420 they will not be evicted by the pinger ever.
2422 Severity : enhancement
2424 Description: Add ability to evict clients by NID from MDS.
2425 Details : By echoing "nid:$NID" string into
2426 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2427 $NID would be instantly evicted from this MDS and from all active
2428 OSTs connected to it.
2432 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2433 Details : lookup_it_finish() used to query file size from OSTs that was not
2438 Description: service threads change working directory to that of init
2439 Details : Starting lustre service threads may pin the working directory
2440 of the parent thread, making that filesystem busy. Threads
2441 now change to the working directory of init to avoid this.
2445 Frequency : during shutdown only
2446 Description: shutdown with a failed MDS or OST can cause unmount to hang
2447 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2452 Frequency : 2.6 only
2453 Description: chmod/chown may include an extra supplementary group
2454 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2455 supplementary group and if none is specified this is used.
2458 Bugzilla : 5479 (6816)
2460 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2461 Details : If lookup is in progress on a file that is unlinked we might try
2462 to revalidate the inode and fail in revalidate after lookup is
2463 complete and ll_file_open() enqueues the open again but
2464 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2467 Frequency : always, if lconf --abort_recovery used
2469 Description: lconf --abort_recovery fails with 'Operation not supported'
2470 Details : lconf was attempting to abort recovery on the MDT device and not
2473 ------------------------------------------------------------------------------
2475 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2480 Frequency : rare (only unsupported configurations with a node running as an
2482 Bugzilla : 6514, 5137
2483 Description: Mounting a Lustre file system on a node running as an OST could
2485 Details : OSTs now preallocates memory needed to write out data at
2486 startup, instead of when needed, to avoid having to
2487 allocate memory in possibly low memory situations.
2488 Specifically, if the file system is mounted on on OST,
2489 memory pressure could force it to try to write out data,
2490 which it needed to allocate memory to do. Due to the low
2491 memory, it would be unable to do so and the node would
2492 become unresponsive.
2494 Severity : enhancement
2496 Description: Addition of lconf --service command line option
2497 Details : lconf now accepts a '--service <arg>' option, which is
2498 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2500 Severity : enhancement
2502 Description: Failover mode is now the default for OSTs.
2503 Details : By default, OSTs will now run in failover mode. To return to
2504 the old behaviour, add '--failout' to the lmc line for OSTs.
2506 Severity : enhancement
2508 Description: Health checks are now provided for MDS and OSTs
2509 Details : Additional detailed health check information on MSD and OSTs
2510 is now provided through the procfs health_check value.
2513 Frequency : occasional, depends on IO load
2515 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2516 after numerous create/delete cycles
2517 Details : The ext3 inode allocation policy would not allocate new inodes
2518 very well on the OSTs because there are no new directories
2519 being created. Instead we look for groups with free space if
2520 the parent directories are nearly full.
2525 Description: Network or server problems during mount may cause partially
2526 mounted clients instead of returning an error.
2527 Details : The config llog parsing code may overwrite the error return
2528 code during mount error handling, returning success instead
2534 Description: MDS can fail to allocate large reply buffers
2535 Details : After long uptimes the MDS can fail to allocate large reply
2536 buffers (e.g. zconf client mount config records) due to memory
2537 fragmentation or consumption by the buffer cache. Preallocate
2538 some large reply buffers so that these replies can be sent even
2539 under memory pressure.
2543 Frequency : rare (liblustre)
2544 Description: fsx running with liblustre complained that using truncate() to
2545 extend the file doesn't work. This patch corrects that issue.
2546 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2547 ATTR_SIZE and lsm use in llu_setattr_raw.
2551 Frequency : rare, only 2.6 kernels
2552 Description: Unusual file access patterns on the MDS may result in inode
2553 data being lost in very rare circumstances.
2554 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2555 code in the 2.6 kernel caused the inode and/or EA data not to
2556 be read from disk, causing single-file corruption.
2560 Frequency : rare, only 2.6 filesystems using extents
2561 Description: Heavy concurrent write and delete load may cause data corruption.
2562 Details : It was possible under high-load situations to have an extent
2563 metadata block in the block device cache from a just-unlinked
2564 file overwrite a newly-allocated data block. We now unmap any
2565 metadata buffers that alias just-allocated data blocks.
2569 Frequency : filesystems with default stripe_count larger than 77
2570 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2571 Details : lconf specifies an inode size of 4096 bytes when the default
2572 stripe_count is larger than 77. This conflicts with the default
2573 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2574 this case to avoid pinning too much memory for large EAs.
2576 ------------------------------------------------------------------------------
2578 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2583 Frequency : rare (extremely heavy IO load with hundreds of clients)
2585 Description: Client is evicted, gets IO error writing to file
2586 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2587 caused clients to be evicted for AST timeouts. The fixes in
2588 bug 5192 mean we no longer need to have such short AST timeouts
2589 so ldlm_timeout has been increased.
2592 Frequency : occasional during --force or --failover shutdown under load
2593 Bugzilla : 5949, 4834
2594 Description: Server oops/LBUG if stopped with --force or --failover under load
2595 Details : a collection of import/export refcount and cleanup ordering
2596 issues fixed for safer force cleanup
2599 Frequency : only filesystems larger than 120 OSTs
2600 Bugzilla : 5990, 6223
2601 Description: lfs getstripe would oops on a very large filesystem
2602 Details : lov_getconfig used kfree on vmalloc'd memory
2605 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2606 Bugzilla : 6242, 6243
2607 Description: reading from files that had been truncated to a non-zero size
2608 but never opened returned no data
2609 Details : ll_file_read() reads zeros from no-object files to EOF
2614 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2615 Details : There's little harm in aborting MDS/OSS recovery and letting it
2616 try again, so I removed the LASSERT and return an error instead.
2618 Severity : enhancement
2620 Description: New debugging infrastructure for tracking down data corruption
2621 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2622 (b) cover more of the client-side code path, and (c) try to narrow
2623 down where problems occurred
2627 Bugzilla : 3819, 4364, 4397, 6313
2628 Description: Racing close and eviction MDS could cause assertion in mds_close
2629 Details : It was possible to get multiple mfd references during close and
2630 client eviction, leading to one thread referencing a freed mfd.
2632 Severity: : enhancement
2633 Bugzilla : 3262, 6359
2634 Description: Attempts to reconnect to servers are now more aggressive.
2635 Details : This builds on the enhanced upcall-less recovery that was added
2636 in 1.4.2. When trying to reconnect to servers, clients will
2637 now try each server in the failover group every 10 seconds. By
2638 default, clients would previously try one server every 25 seconds.
2643 Description: After recovery, certain operations trigger a failed
2644 assertion on a client.
2645 Details : Failing over an mds, using lconf -d --failover, while a
2646 client was doing a readdir() call would cause the client to
2647 LBUG after recovery completed and the readdir() was resent.
2649 Severity : enhancement
2651 Description: Default groups are now added by lconf
2652 Details : You can now run lconf --group <servicename> without having to
2653 manually add groups with lmc.
2656 Frequency : occasional
2658 Description: Nodes with an elan id of 0 trigger a failed assertion
2661 Frequency : always when accessing e.g. tty/console device nodes
2663 Description: tty and some other devices nodes cannot be used on lustre
2664 Details : file's private_data field is used by device data and lustre
2665 values in there got lost. New field was added to struct file to
2666 store fs-specific private data.
2669 Frequency : when exporting Lustre via NFS
2671 Description: NFSD failed occasionally when looking up a path component
2672 Details : NFSD is looking up ".." which was broken in ext3 directories
2673 that had grown large enough to become hashed.
2676 Frequency : Clusters with multiple interfaces not on the same subnet
2678 Description: Nodes will repeatedly try to reconnect to an interface which it
2679 cannot reach and report an error to the log.
2680 Details : Extra peer list entries will be created by lconf with some peers
2681 unreachable. lconf now validates the peer before adding it.
2684 Frequency : Only if a default stripe is set on the filesystem root.
2686 Description: Setting a default stripe on the filesystem root prevented the
2687 filesystem from being remounted.
2688 Details : The client was sending extra request flags in the root getattr
2689 request and did not allocate a reply buffer for the dir EA.
2692 Frequency : occasional, higher if lots of files are accessed by one client
2693 Bugzilla : 6159, 6097
2694 Description: Client trips assertion regarding lsm mismatch/magic
2695 Details : While revalidating inodes the VFS looks up inodes with ifind()
2696 and in rare cases can find an inode that is being freed.
2697 The ll_test_inode() code will free the lsm during ifind()
2698 when it finds an existing inode and then the VFS later attaches
2699 this free lsm to a new inode.
2703 Bugzilla : 6422, 7030
2704 Description: MDS deadlock between mkdir and client eviction
2705 Details : Creating a new file via mkdir or mknod (starting a transaction
2706 and getting the ns lock) can deadlock with client eviction
2707 (gets ns lock and trying to finish a synchronous transaction).
2710 Frequency : occasional
2711 Description: While starting a server, the fsfilt_ext3 module could not be
2713 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2714 kernels. Previously, lconf would still use the ext3 name
2715 when trying to load modules. Now, it will correctly use
2716 ext3 on 2.4 and ldiskfs on 2.6.
2718 Severity : enhancement
2719 Description: The default stripe count has been changed to 1
2720 Details : The interpretation of the default stripe count (0, to lfs
2721 or lmc) has been changed to mean striping across a single
2722 OST, rather than all available. For general usage we have
2723 found a stripe count of 1 or 2 works best.
2725 Severity : enhancement
2726 Description: Add support for compiling against Cray portals.
2727 Details : Conditional compiling for some areas that are different
2731 Frequency : occasional
2732 Bugzilla : 6409, 6834
2733 Description: Creating files with an explicit stripe count may lead to
2734 a failed assertion on the MDS
2735 Details : If some OSTs are full or unavailable, creating files may
2736 trigger a failed assertion on the MDS. Now, Lustre will
2737 try to use other servers or return an error to the
2741 Frequency : occasional
2743 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2744 caused lock timeout during readahead (since 1.4.2).
2745 Details : Processes doing readahead might match a lock that hasn't been
2746 granted yet if there are overlapping and conflicting lock
2747 requests. The readahead process waits on ungranted lock
2748 (original lock is CBPENDING), while OST waits for that process
2749 to cancel CBPENDING read lock and eventually evicts client.
2751 Severity : enhancement
2753 Description: Initial enabling of flock support for clients
2754 Details : Implements fcntl advisory locking and file status functions.
2755 This feature is provided as an optional mount flag (default
2756 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2757 locking are implemented yet, and those that are are not guaranteed
2758 to be completely correct in production environments.
2759 mount -t lustre -o [flock|noflock] ...
2762 Frequency : occasional
2764 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2765 assertion in the ext3 JBD (journaling) layer.
2766 Details : The b_committed_data struct is protected by the big kernel lock
2767 in 2.4 kernels, serializing journal_commit_transaction() and
2768 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2769 access to this struct. In 2.6 kernels there is finer grained
2770 locking to improve SMP performance of the JBD layer.
2774 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2776 -----------------------------------------------------------------------------
2778 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2780 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2781 versions of Lustre. Please update all servers and clients to
2782 version 1.4.2 or later at the same time. You must also run
2783 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2784 to update the configuration logs.
2786 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2787 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2788 - swab llog records of type '0' so we get proper header size/idx (5861)
2789 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2790 - fix rename of one directory over another leaking an inode (5953)
2791 - avoid SetPageDirty on 2.6 (5981)
2792 - don't re-add just-being-destroyed locks to the waiting list (5653)
2793 - when creating new directories, inherit the parent's custom
2794 striping settings if present parent (3048)
2795 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2796 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2797 - don't hold client locks on temporary worklist from l_lru (5666)
2798 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2799 - automatically evict dead clients (5921)
2800 - Update file size properly in create+truncate+fstat case (6196)
2801 - Do not unhash mountpoint dentries, do not allow removal of
2803 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2804 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2805 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2806 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2807 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2808 - Correctly handle reads of files with no objects (6243)
2809 - lctl recover will also mark a device active if deactivate used (5933)
2811 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2812 - allow --write-conf on an MDS with different nettype than client (5619)
2813 - don't write config llogs to MDS for mounts not from that MDS (5617)
2814 - lconf should create multiple TCP connections from a client (5201)
2815 - init scripts are now turned off by default; run chkconfig --on
2816 lustre and chkconfig --on lustrefs to use them
2817 - upcalls are no longer needed for clients to recover to failover
2819 - add --abort-recovery option to lconf to abort recovery on device
2821 - add support for an arbitrary number of OSTs (3026)
2822 - Quota support protocol changes.
2823 - forward compatibility changes to wire structs (6007)
2824 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2825 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2826 - improved "lustre status" script
2827 - initialize blocksize for non-regular files (6062)
2828 - added --disable-server and --disable-client configure options (5782)
2829 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2830 - Vanilla 2.4.29 support
2831 - increase maximum number of obd devices to 520 (6242)
2832 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2833 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2834 SLES 9 2.6.5 kernels
2835 - limit stripes per file to 160 (the maximum EA size) (6093)
2837 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2840 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2841 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2842 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2843 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2844 - small ext3 extents cleanups and fixes (5733)
2845 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2846 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2847 - avoid needless client->OST connect, fix handle mismatch (5317)
2848 - fix DLM error path that led to out-of-sync client, long delays (5779)
2849 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2850 - fix several locking issues related to i_size (5492,5624,5654,5672)
2851 - don't move pending lock onto export if it is already evicted (5683)
2852 - fix kernel oops when creating .foo in unlinked directory (5548)
2853 - fix deadlock in obdfilter statistics vs. object create (5811)
2854 - use time_{before,after} to avoid timer jiffies wrap (5882)
2855 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2856 - Do not leak request if server was not able to process it (5154)
2857 - If mds_open unable to find parent dir, make that negative lookup(5154)
2858 - don't create new directories with extent-mapping (5909, 5936)
2860 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2861 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2862 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2863 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2864 - support for RHEL4 kernel on i686 (5773)
2865 - provide error messages when incompatible logs are encountered (5898)
2867 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2868 * version 1.4.0.10 (1.4.1 release candidate 1)
2870 - don't keep a lock reference when lock is not granted (4238)
2871 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2872 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2873 - drop import inflight refcount on signal_completed_replay error (5255)
2874 - unlock page after async write error during send (3677)
2875 - handle missing objects in filter_preprw_read properly (5265)
2876 - no transno return for symlink open, don't save no-trasno open (3440)
2877 - don't try to complete elan receive that already failed (4012)
2878 - free RPC server reply state on error (5406)
2879 - clean up thread from ptlrpc_start_thread() on error (5160)
2880 - readahead could read extra page into cache that wasn't ejected (5388)
2881 - prevent races in class_attach/setup/cleanup/detach (5260)
2882 - don't dereference de->d_inode after l_dput of de (5458)
2883 - use "int" for stripe value returned from lock_to_stripe (5544)
2884 - mballoc allocation and error-checking fixes in 2.6 (5504)
2885 - block device patches to fix I/O request sizes in 2.6 (5482)
2886 - look up hostnames for IB nals (5602)
2887 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2888 - don't start multiple acceptors for the same port (5277)
2889 - fix incorrect LASSERT in mds_getattr_name (5635)
2890 - export a proc file for general "ping" checking (5628)
2891 - fix "lfs check" to not block when the MDS is down (5628)
2893 - service request history (4965)
2894 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2895 - create an "evict_client" /proc entry on OSTs, like the MDS has
2896 - fix mount usage message, return errors per mount(8) (5168)
2897 - change grep [] to grep "[]" in tests so they work in more UMLs
2898 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2899 - remove ancient LOV_MAGIC_V0 EA support (5047)
2900 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2901 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2902 - allow readahead allocations to fail when low on memory (5383)
2903 - mmap locking landed again, after considerable improvement (2828)
2904 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2906 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2909 - send OST transaction number in read/write reply to free req (4966)
2910 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2911 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2912 - the watchdog thread now runs as interruptible (5246)
2913 - flock/lockf fixes (but it's still disabled, pending 5135)
2914 - don't use EXT3 constants in llite code (5094)
2915 - memory shortage at startup could cause assertion (5176)
2917 - reorganization of lov code
2918 - single portals codebase
2920 - add extents/mballoc support (5025)
2921 - direct I/O reads in the obdfilter (4048)
2922 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2924 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2927 - send OST transaction number in read/write reply to free req (4966)
2928 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2929 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2930 - the watchdog thread now runs as interruptible (5246)
2931 - handle missing objects in filter_preprw_read properly (5265)
2932 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2933 - drop import inflight refcount on signal_completed_replay error (5255)
2934 - unlock page after async write error during send (3677)
2935 - return original error code on reconstructed replies (3761)
2936 - no transno return for symlink open, don't save no-trasno open (3440)
2938 - add pid to ldlm debugging output (4922)
2939 - bump the watchdog timeouts -- we can't handle 30sec yet
2940 - extra debugging for orphan dentry/inode bug (5259)
2942 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
2945 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
2946 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
2947 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
2948 - fix NULL dereference in /proc/sys/portals/routes (4827)
2949 - allow failed mdc_close() operations to be interrupted (4561)
2950 - stop precreate on OST before MDS would time out on it (4778)
2951 - don't send partial-page writes before EOF from client (4410)
2952 - discard client grant for sub-page writes on large-page clients (4520)
2953 - don't free dentries not owned by NFS code, check generation (4806)
2954 - fix lsm leak if mds_create_objects() fails (4801)
2955 - limit debug_daemon file size, always print CERROR messages (4789)
2956 - use transno after validating reply (3892)
2957 - process timed out requests if import state changes (3754)
2958 - update mtime on OST during writes, return in glimpse (4829)
2959 - add mkfsoptions to LDAP (4679)
2960 - use ->max_readahead method instead of zapping global ra (5039)
2961 - don't interrupt __l_wait_event() during strace
2963 - add software watchdogs to catch hung threads quickly (4941)
2964 - make lustrefs init script start after nfs is mounted
2965 - fix CWARN/ERROR duplication (4930)
2966 - return async write errors to application if possible (2248)
2967 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
2968 - print NAL number in %x format (4645)
2969 - update barely-supported suse-2.4.21-171 series (4842)
2970 - support for sles 9 %post scripts
2971 - support for building 2.6 kernel-source packages
2972 - support for sles km_* packages
2974 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
2977 - ignore -ENOENT errors in osc_destroy (3639)
2978 - notify osc create thread that OSC is being cleaned up (4600)
2979 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
2980 - reconstruct ost_handle() like mds_handle() (4657)
2981 - create a new thread to do import eviction to avoid deadlock (3969)
2982 - let lconf resolve symlinked-to devices (4629)
2983 - don't unlink "objects" from directory with default EA (4554)
2984 - hold socknal file ref over connect in case target is down (4394)
2985 - allow more than 32000 subdirectories in a single directory (3244)
2986 - fix blocks count for O_DIRECT writes (3751)
2987 - OST returns ENOSPC from object create when no space left (4539)
2988 - don't send truncate RPC if file size isn't changing (4410)
2989 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
2990 - bind to privileged port in socknal and tcpnal (3689)
2992 - rate limit CERROR/CWARN console message to avoid overload (4519)
2993 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
2994 - basic mmap support (3918)
2995 - kernel patch series update from b1_4 (4711)
2997 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3000 - avoid crash during MDS cleanup with OST shut down (2775)
3001 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3002 - don't use bad inodes on the MDS (3744)
3003 - dynamic object preallocation to improve recovery speed (4236)
3004 - don't hold spinlock over lock dumping or change debug flags (4401)
3005 - don't zero obd_dev when it is force cleaned (3651)
3006 - print grants to console if they go negative (4431)
3007 - "lctl deactivate" will stop automatic recovery attempts (3406)
3008 - look for existing locks in ldlm_handle_enqueue() (3764)
3009 - don't resolve lock handle twice in recovery avoiding race (4401)
3010 - revalidate should check working dir is a directory (4134)
3012 - don't always mark "slow" obdfilter messages as errors (4418)
3014 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3017 - don't close LustreDB during write_conf until it is done (3860)
3018 - fix typo in lconf for_each_profile (3821)
3019 - allow dumping logs from multiple threads at one time (3820)
3020 - don't allow multiple threads in OSC recovery (3812)
3021 - fix debug_size parameters (3864)
3022 - fix mds_postrecov to initialize import for llog ctxt (3121)
3023 - replace config semaphore with spinlock (3306)
3024 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3025 - don't allow enqueue to complete on a destroyed export (3822)
3026 - down write_lock before checking llog header bitmap (3825)
3027 - recover from lock replay timeout (3764)
3028 - up llog sem before sending rpc (3652)
3029 - reduce ns lock hold times when setting kms (3267)
3030 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3031 - fix NULL deref and obd_dev leak on setup error (3312)
3032 - replace some LBUG about llog ops with error handling (3841)
3033 - don't match INVALID dentries from d_lookup and spin (3784)
3034 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3035 - fix invalid assertion in ptlrpc_set_wait (3880)
3037 - add libwrap support for the TCP acceptor (3996)
3038 - add /proc/sys/portals/routes for non-root route listing (3994)
3039 - allow setting MDS UUID in .xml (2580)
3040 - print the stack of a process that LBUGs (4228)
3042 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3045 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3046 - make sure to unset replay flag from failed open requests (3440)
3047 - if default stripe count is 0, use OST count for inode size (3636)
3048 - update parent mtime/ctime on client for create/unlink (2611)
3049 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3050 - free recovery state on server during a forced cleanup (3571)
3051 - unregister_reply for resent reqs (3063)
3052 - loop back devices mounting and status check on 2.6 (3563)
3053 - fix resource-creation race that can provoke i_size == 0 (3513)
3054 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3055 - more debugging for page-accounting assertion (3746)
3056 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3057 - avoid deadlock after precreation failure (3758)
3058 - fix race and lock order deadlock in orphan handling (3450, 3750)
3059 - add validity checks when grabbing inodes from l_ast_data (3599)
3061 - add /proc/.../recovery_status to obdfilter (3428)
3062 - lightweight CDEBUG infrastructure, debug daemon (3668)
3063 - change default OSC RPC parameters to be better on small clusters
3064 - turn off OST read cache for files smaller than 32MB
3065 - install man pages and include them in rpms (3100)
3066 - add new init script for (un)mounting lustre filesystems (2593)
3067 - run chkconfig in %post for init scripts (3701)
3068 - drop scimac NAL (unmaintained)
3070 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3073 - clean kiobufs before and after use (3485)
3074 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3075 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3076 - change init script to fail more clearly if not run as root (1528)
3077 - allow clients to reconnect during replay (1742)
3078 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3079 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3080 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3081 - fix cond_resched (3554)
3082 - only evict unfinished clients after recovery (3515)
3083 - allow bulk resend, prevent data loss (3570)
3084 - dynamic ptlrpc request buffer allocation (2102)
3085 - don't allow unlinking open directory if it isn't empty (2904)
3086 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3087 - remove extraneous obd dereference causing LASSERT failure (3334)
3088 - don't use get_cycles() when creating temp. files on the mds (3156)
3089 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3090 - handle EEXIST for set-stripe, set proper directory name (3336)
3092 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3093 - fix ksocknal_fmb_callback() error messages (2918)
3095 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3098 - don't copy lvb into (possibly NULL) reply on error (2983)
3099 - don't deref dentry after dput, don't free lvb on error (2922)
3100 - use the kms to determine writeback rpc length (2947)
3101 - increment oti_logcookies when osc is inactive (2948)
3102 - update client's i_blocks count via lvb messages (2543)
3103 - handle intent open/close of special files properly (1557)
3104 - mount MDS with errors=remount-ro, like obdfilter (2009)
3105 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3106 - don't use cancelling-locks' kms values (2947)
3107 - use highest lock extent for kms, not last one (2925)
3108 - don't dereference ERR_PTR() dentry in error handling path (3107)
3109 - fix thread race in portals_debug_dumplog() (3122)
3110 - create lprocfs device entries at setup instead of at attach (1519)
3111 - common AST error handler, don't evict client on completion race (3145)
3112 - zero nameidata in detach_mnt in 2.6 (3118)
3113 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3114 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3115 - initialize RPC timeout timer earlier for 2.6 (3219)
3116 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3117 - print nal/nid for unknown nid (3258)
3118 - additional checks for oscc recovery before doing precreate (3284)
3119 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3120 - don't crash in mdc_close for bad permissions on open (3285)
3121 - zero i_rdev for non-device files (3147)
3122 - clear page->private before handing to FS, better assertion (3119)
3123 - tune the read pipeline (3236)
3124 - fix incorrect decref of invalidated dentry (2350)
3125 - provide read-ahead stats and refine rpc in flight stats (3328)
3126 - don't hold journal transaction open across create RPC (3313)
3127 - update atime on MDS at close time (3265)
3128 - close LDAP connection when recovering to avoid server load (3315)
3129 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3130 - don't leak open file on MDS after open resend (3325)
3131 - serialize filter_precreate and filter_destroy_precreated (3329)
3132 - loop device shouldn't call sync_dev() for nul device (3092)
3133 - clear page cache after eviction (2766)
3134 - resynchronize MDS->OST in background (2824)
3135 - refuse to mount the same filesystem twice on same mountpoint (3394)
3136 - allow llmount to create routes for mounting behind routers (3320)
3137 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3138 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3139 - fix rare problem with rename on htree directories (3417)
3141 - allow default OST striping configuration per directory (1414)
3142 - fix compilation for qswnal for 2.6 kernels (3125)
3143 - increase maximum number of MDS request buffers for large systems
3144 - change liblustreapi to be useful for external progs like lfsck (3098)
3145 - increase local configuration timeout for slow disks (3353)
3146 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3148 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3151 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3152 - don't overwrite extent policy data in reply if lock was blocked (2901)
3153 - drop filter export grants atomically with removal from device (2663)
3154 - del obd_self_export from work_list in class_disconnect_exports (2908)
3155 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3156 - swab reply message in mdc_close, other PPC fixes (2464)
3157 - fix destroying of named logs (2325)
3158 - overwrite old logs when running lconf --write_conf (2264)
3159 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3160 - fix race in target_handle_connect (2898)
3161 - mds_reint_create() should take same inode create lock (2926)
3162 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3163 - don't close files for self_export to avoid uninitialized obd (2936)
3164 - allow MDS with the same name as client node (2939)
3165 - hold dentry reference for closed log files for unlink (2325)
3166 - reserve space for all logs during transactions (2059)
3167 - don't evict page beyond end of stripe extent (2925)
3168 - don't oops on a deleted current working directory (2399)
3169 - handle hard links to targets without a parent properly (2517)
3170 - don't dereference NULL lock when racing during eviction (2867)
3171 - don't grow lock extents when lots of conflicting locks (2919)
3173 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3176 - account for cache space usage on clients to avoid data loss (974)
3177 - lfsck support in lustre kernel code (2349)
3178 - reduce journal credits needed for BRW writes (2370)
3179 - orphan handling to avoid losing space on client/server crashes
3180 - ptlrpcd can be blocked, stopping ALL progress (2477)
3181 - use lock value blocks to assist in proper KMS, faster stat (1021)
3182 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3183 - recovery for initial connections (2355)
3184 - fixes for mds_cleanup_orphans (1934)
3185 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3186 - block all file creations until orphan recovery completes (1901)
3187 - client remove rq_connection from request struct (2423)
3188 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3189 - recovery timer race (2670)
3190 - mdc_close recovey bug (2532)
3191 - ptlrpc cleanup bug (2710)
3192 - mds timeout on local locks (2588)
3193 - namespace lock held during RPCs (2431)
3194 - handle interrupted sync write properly (2503)
3195 - don't try to handle a message that hasn't been replied to (2699)
3196 - client assert failure during cleanup after abort recovery (2701)
3197 - leak mdc device after failed mount (2712)
3198 - ptlrpc_check_set allows timedout requests to complete (2714)
3199 - wait for inflight reqs when ptlrpcd finishes (2710)
3200 - make sure unregistered services are removed from the srv_list
3201 - reset bulk XID's when resending them (caught by 1138 test)
3202 - unregister_bulk after timeout
3203 - fix lconf error (2694)
3204 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3205 - readahead locks pages, leaves pending causing memory pressure (2673)
3206 - increase OST request buffers to 4096 on large machines (2729)
3207 - fix up permission of existing directories in simple_mkdir (2661)
3208 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3209 - don't assign transno to errored transactions (2742)
3210 - don't delete objects on OST if given a bogus objid from MDS (2751)
3211 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3212 - if rq_no_resend, then timeout request after recovery (2432)
3213 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3214 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3215 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3216 - MDS recovery shouldn't delete objects that it has given out (2730)
3217 - if enqueue arrives after completion, don't clobber LVB (2819)
3218 - don't unlock pages twice when trigger_group_io returns error (2814)
3219 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3220 - don't write pages to disk if there was an error (1450)
3221 - don't ping imports that have recovery disabled (2676)
3222 - take buffered bytes into account when balancing socknal conn (2817)
3223 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3224 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3225 - remove little-endian swabbing of llog records (1987)
3226 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3227 - retry reposting request buffers if they fail (1191)
3228 - grow extent at grant time to avoid granting a revoked lock (2809)
3229 - lock revoke doesn't evict page if covered by a second lock (2765)
3230 - disable VM readahead to avoid reading outside lock extents (2805)
3232 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3233 - updated kernel patches for hp-2.4.20 kernel (2681)
3235 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3238 - fix truncated write corruption (2366)
3239 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3241 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3242 - ensure len(uuid) < 37 in lmc (1171)
3243 - fix ia64 OOPS in llog_test (2255)
3244 - zero end of page at obdfilter for partial page writes (2648)
3245 - don't leave stale dentries around after renames (bug 2428)
3246 - fix timeouts when evicting a client with a single lock held (2642)
3247 - set deadline for the initial HELLO message to drain (2634)
3248 - print out dotted-quad IP addresses in the socknal (2302)
3250 - additional debugging for MDS client eviction problem (2443)
3251 - fix mkfsoptions support for osts (2603, 2604)
3253 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3256 - add series for the vanilla 2.6.0 kernel
3257 - add series for the vanilla 2.4.24 kernel
3258 - add series for a cray x86/64 UL kernel drop
3259 - fix xattr patches for the vanilla 2.4.19 series
3261 - generate true UUIDs in lmc (1171)
3262 - have portals stack dumping break in UML (2466)
3263 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3264 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3265 - wake callback waiting threads on client eviction (2460)
3266 - Add --ptldebug and --subsystem to lmc (1719)
3267 - update assertion to allow safe interrupt allocation
3268 - set rq_no_resend for cancel requests (2432)
3269 - recalculate ptlrpcd timeout after resend (2494)
3270 - call vfs_rmdir when removing pending directories (2368)
3271 - fix renaming a file to itself (2429)
3272 - lmc creates a default one-stripe lov (2454)
3273 - expand procfs space to handle large clusters (2326)
3274 - increase UML stack to avoid overflow
3275 - update lconf's list of debug and subsystem masks
3276 - fix lfs find --obd (2510)
3277 - /proc tunable for disabling filter read caching (2591)
3278 - stop rpm packages from altering slapd.conf (2301)
3279 - disable nagle in the socknal under 0conf (2578)
3280 - choose mds inode size based on stripe count (2572)
3281 - fix kernel-source rpm problems (2516)
3283 - add --disable-doc to avoid pdf generation (2421)
3284 - update documentation, tests, type-os, comments
3285 - avoid format warnings on ia64
3286 - remove the TOE NAL
3287 - tiny code cleanups by removing unused fields
3289 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3292 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3293 - remove the most blatant lies from BUILDING file (2371)
3294 - change default debug level to reasonable production setting
3295 - reduce client side cache size to reduce cache flush time
3296 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3297 - make TCP zerocopy and pinger support enabled by default (2476)
3298 - sync writes completed after process exits caused crashes (2319)
3299 - maintain correct mount count on the MDS (2356)
3300 - backout 1557, because 2316 wasn't really fixed
3301 - better file I/O statistics gathering in /proc
3302 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3303 - another kernel patch to fix zero-copy TCP function export
3304 - don't take duplicate lock when processing re-sent getattr (2420)
3305 - lctl uses obd_self_export instead of creating new conn (2353)
3306 - MDS/OST recovery case which requires object creation asserted (2425)
3307 - move lfs from /usr/sbin to /usr/bin in packages
3308 - fix race between mds_client_add and mds_client_free (2417)
3309 - use kmalloc instead of slabs in portals (2430)
3310 - don't create duplicate records when a failover MDS is present (2442)
3311 - remove unnecessary mount age check (2332)
3312 - don't remove directory inodes from locks prematurely (2451)
3313 - don't break if MDS service name is the same as hostname (2103)
3314 - fix races in client write RPC generation when cache full (2482)
3316 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3319 - remove now-unused request->rq_obd (278)
3320 - if an allocation fails, print out how much memory we've used (1933)
3321 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3322 - add the "configurable stack size" patch to most series files (1256)
3323 - ability to write large log records, for 100+ OST configs (2306)
3324 - fix NULL deref when filter_prep fails (2314)
3325 - fix operator precedence error in filter_sync
3326 - dynamic allocation of socknal TX descriptors (2315)
3327 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3328 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3329 - ensure that CWARN messages in Portals always get to the syslog
3330 - __init/__exit are not for prototype decls (ldlm_init/exit)
3331 - x86-64 compile warning fixes
3332 - fix gateway LMC keyword conflict (2318)
3333 - fix MDS lock inversions in getattr/reint paths (1844)
3334 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3335 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3336 - DLM race condition prevented some lock evictions (2328)
3337 - ENOMEM detection and retry on socknal sends (2230)
3338 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3339 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3340 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3341 - better error messages when a client is rejected during recovery (1505)
3342 - avoid cancelling locks which were never granted, after failure (2330)
3343 - fix i_sem/journal inversion in mds_client_add (2333)
3344 - fix truncate/getattr lock cycle deadlock (2334)
3345 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3346 - fix two rare exit paths which could leak an l_lock() ref (2321)
3347 - fencepost error in MDS/OST orphan recovery (2226)
3348 - make log record alignment 8 bytes (1988)
3349 - lstripe now fails when requested offset > ost_count (2237)
3350 - ensure that all kernel series have a complete list.h (1607)
3351 - fix crashes in special-file operations (2316)
3352 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3354 - allow configurable automake binary, for testing new versions
3355 - small update to the lfs documentation
3357 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3359 * fix negative export reference count in fsfilt_sync (2312)
3361 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3362 * release candidate 0.9.1
3364 - orphans are moved into the PENDING directory for possible recovery
3365 - replayed opens now open by fid for orphan/rename safety (1042)
3366 - last close of an orphan inode generates a transno (683)
3367 - chdir() and mount() now pin the directory entry (1020)
3368 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3369 - discard very old requests without processing them (1502)
3370 - remove some common, well-understood CERRORs (1505)
3371 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3372 - clear "grant" flags in OST replies until OST grant code lands (1644)
3373 - fix read performance by not clobbering i_blksize on client (1598)
3374 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3375 - make LRU size a /proc tunable, clears locks when reduced (707)
3376 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3377 - server locks take references on exports now (1558)
3378 - build fixes for 2.4.20-rh trees (1663)
3379 - return an error from lov_create if all OSCs are inactive (1751)
3380 - fix import levels when a reconnect happens without a timeout (1597)
3381 - exit early from mds_open if we get a lookup error (1749)
3382 - partial page read at EOF wouldn't wait for disk before sending (1642)
3383 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3384 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3385 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3386 - write/truncate lock inversion (1639)
3387 - Don't auto-load obdclass, portals modules during cleanup (1495)
3388 - fix timestamps from jumping to "now" (1763)
3389 - extra journal assertions (1648)
3390 - add an extra multiunlink test (1771)
3391 - fix read_record/write_record API (1776)
3392 - fix leak of offset_extent, possible incorrect i_size later (1772)
3393 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3394 - Debugging for the kqswnal_get_idle_tx problems (1820)
3395 - Allow recovery to be attempted multiple times (1536)
3396 - Write out MDS last_rcvd file after it is first created (1600)
3397 - Fix tx_descriptor leak in failed transmit situations (1827)
3398 - ext3 journaling fixes for assertion failure after IO error (1871)
3399 - class_export_put() on freed export after completion AST error (1896)
3400 - Fix revalidate looping in VFS (1322)
3401 - Don't access a freed export during MDS_REINT timeout (1521)
3402 - Add open-unlink recovery support on the MDS (1673,1764)
3403 - Return an error if no MDS data was read from last_rcvd (1946)
3404 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3405 - Don't setup a disk device that doesn't match exported UUID (317)
3406 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3407 - avoid committing NULL handle in force close
3408 - local.sh is now a one-stripe LOV configuration
3409 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3410 - fix ext3 htree duplicate directory entry corruption (1516)
3411 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3412 - update to new LOV EA format (2097)
3413 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3415 2003-06-15 Phil Schwan <phil@clusterfs.com>
3418 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3419 - per-import/export recovery handling (958, 931, 959)
3420 - multiple last-rcvd slots, for serving multiple FSes (949)
3421 - connections are again shared between multiple imp/exports (963, 964)
3422 - "umount -f" would hang if any requests needed to be sent (393, 978)
3423 - avoid pinning large req buffer by copying for queued messages (989)
3424 - add "uuid" to "lctl device" command to help upcalls (991)
3425 - "open" RPCs with transnos would confuse recovery counters (1037)
3426 - do proper endian conversion of all wire messages (288, 340, 891)
3427 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3428 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3429 - fix ptlrpc_connection leak in target_handle_connect (1174)
3430 - fix import refcounting bug in OST and MDS cleanup (1134)
3431 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3432 - fix very unlikely obd_types race condition (501)
3433 - remove osc_open hack for echo_client (1187)
3434 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3435 - a failure in read_inode2 leads to deadlock (1139)
3436 - cancel ack-locks as soon as transaction is committed (1072)
3437 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3438 - make sure to commitrw after any preprw to avoid deadlock (1162)
3439 - failing to execute a file in a lustre FS would lock inode (1203)
3440 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3441 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3442 - obd_uuid2tgt would walk off the end of the list (1255)
3443 - on IA64 the osc would give portals incorrect bulk size (1258)
3444 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3445 - fix lock inversion caused by new llite matching code (1282)
3446 - limit the number of dirty pages on a client to 10MB (1286)
3447 - timed out locks were not being corrected cancelled (1289)
3448 - fix O_DIRECT above 4GB on IA-32 (1292)
3449 * major user-visible changes
3450 - fail out/fail over policy now controlled by the upcall (993)
3452 - add OBD_PING to check server availability and failure (954)
3453 - lustre messages are now sent in sending host order (288, 340, 891)
3454 - add eadatalen to MDS getattr reply (340)
3455 - OST read replies may contain second buffer, with per-page status (593)
3457 2003-03-11 Phil Schwan <phil@clusterfs.com>
3460 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3461 - failed open()s could cause deadlock; fixed (867, 869)
3462 - stop cancelling OST locks when files are closed (481)
3463 - overlapping XID spaces caused network corruption (851, 853)
3464 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3465 - setattr_raw added, to do single-RPC, server-side setattrs
3466 - lmc/lconf syntax change for OST UUIDs
3467 - fix crashy race condition between ptlrpc_free_req and osc_close
3468 - don't use request in mdc_enqueue if we hit a timeout (889)
3469 - don't set the inode i_size for regular files from the MDS (896)
3470 - handle out of order completion AST (842)
3471 - don't LBUG if a lock request times out after receiving AST (913)
3472 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3473 - if a bad lock AST arrives, send an error instead of dropping entirely
3474 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3475 - fix leak in bulk IO when only partially completed (899, 900, 926)
3476 - fix O_DIRECT for ia64 (55)
3477 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3478 - C-z after timeout could hang a process forever; fixed (977)
3480 - client-side I/O cache (678, 924, 929, 941, 970)
3482 - READPAGE and SETATTRs which don't take server-side locks get
3485 2003-02-11 Phil Schwan <phil@clusterfs.com>
3488 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3489 - Fix fsfilt_extN_readpage() to read a full page of directory
3490 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3491 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3492 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3493 - Most intent operations were removed, in favour of a new RPC mode
3494 that does a single RPC to the server and bypasses most of the VFS
3495 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3496 - Aggressively cancel local locks on DLM servers
3497 - mds_reint_unlink sends EA to the client if it's the last nlink.
3498 client uses that EA to unlink OST objects.
3499 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3500 - recursive symlinks were fixed (439)
3501 - fixed NULL deref in DEBUG_REQ
3502 - filter_update_lastobjid no longer calls sync, which annoyed extN
3503 - fixed multi-client small-writes to a single file problem (445)
3504 - fixed mtime updates during file writes (607)
3505 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3506 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3507 - fixed Linux OST size reporting problem (444, 656)
3508 - OST now updates object mtime with writes or setattr (607, 619)
3509 - client verifies file size before zeroing page past EOF (445)
3510 - OST now writes last allocated objid to disk with allocation (108)
3511 - LOV on echo now works (409)
3513 - mds_reint_unlink sends a new buffer, with the EA included. this
3514 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3515 set if a regular file was being unlinked, and it was the last link
3516 - use PtlGet from the target for bulk writes (315)
3517 - OST now updates object mtime with writes or setattr (607, 619)
3518 - LDLM now has a grant-time callback to revalidate locked items, if
3520 - Many MDS operations were reorganized to combat race conditions
3522 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3523 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3525 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3528 - Fully reactivate OST imports after reconnection (512, others)
3529 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3530 - More graceful error handling for truncating on dead OST (515)
3531 - Don't error out unless we're actually accessing dead stripes (474)
3532 - Fix garbage sizes when stripes are missing (410)
3533 - LRU counters were broken, causing constant lock purge (433, 432)
3534 - garbage on read from stripes with failed OSTs (441)
3535 - mark OSCs as active before reconnecting during recovery (438)
3536 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3537 - lfind did not preserve OST order in output (443)
3538 - symlinks cause hung clients, incorrect data (439)
3539 - stop dereferencing request after dropping refcount (457)
3540 - don't LASSERT(spin_is_locked) on non-SMP (455)
3541 - fixes for many rename() bugs
3542 - fstat didn't correctly synchronize attributes (399)
3543 - server must handle lock cancellation during blocking AST prep (487)
3544 - bulk descriptors were free()d too soon (511)
3545 - fix paths in lconf, which would load incorrect modules (451, 507)
3546 - fix confusing lconf 'host not found' error message (386)
3547 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3548 - fix race condition in mdc_blocking_ast() for inode access (526)
3549 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3550 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3551 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3552 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3553 - fix lstripe in several ways (kernel side) (527)
3554 - fix request leak in ldlm_cli_enqueue (262)
3555 - incorrect OSC was marked inactive after OST failure
3556 - call mds_fs_cleanup before unmounting filesystem (524)
3557 - fix races between taking ns_lock and ldlm_lock_change_resource
3558 - fix races updating LOV export open file list
3559 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3560 - fix recovery NULL deref in ldlm_cli_cancel_unused
3561 - fix some DLM races by using new hash table for lock handles (419)
3562 - permit the client to specify desired inodes, at replay
3563 - duplicate requests when we queue them for replay reintegration
3564 - fix last_rcvd offset calculation
3565 - sync after each recovered transaction, so we always make progress
3566 - never, not always, ERESTART requests without transnos
3567 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3569 - skip replay if the MDS says that the client is already connected
3570 - don't check for a recovery-enabled export to match lctl's UUID
3571 - don't INC_USE_COUNT for phantom exports
3572 - don't crash when cleaning up phantom exports (567)
3573 - don't double-finish or set replay data for errored mdc_open requests
3574 - abort requests when they time out, so we don't get old replies
3575 - send/receive replies for AST messages again
3576 - if the client says that it doesn't have the lock, cancel it on the
3578 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3579 mark it as destroyed, it will all work out when decref is called
3580 - fix module use counts (22, 581)
3582 - ASTs now expect a reply (server cancels lock on error reply)
3584 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3587 - fix many simultaneous client startup (392)
3588 - fix dentry->d_it clobbering
3589 - credentials weren't being shipped for readdir/getattr operations
3590 - remove invalid assertions triggered during some concurrent MD
3592 - proper Lustre versions added (336, 389)
3593 - fix memory leak for create error case (398)
3594 - fix LOV locking bug that would get cli/srv out of sync
3595 - fix echo client over LOV (409)
3596 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3597 - fix double-O_EXCL intent crash (424)
3598 - avoid sending multiple lock CANCELs (352)
3600 - MDS can do multi-client recovery (modulo bugs in new code)
3602 - many updates, edits, cleanups
3604 2002-11-18 Phil Schwan <phil@clusterfs.com>
3607 - fix null d_it dereference (346)
3608 - fix full OST/dbench hang (333)
3609 - fix permission problem with file removal (286)
3610 - fix removal of OSCs from LOV when they fail
3611 - fix NULL deref during bulk timeout (214)
3612 - fix problems related to multiple filesystems on one MDS (241)
3613 - fixed serious subtle metadata locking bugs
3614 - free locks on clients when inodes are removed due to memory
3616 - fix inode pointer in lock data (285)
3617 - partial support for multiple MDS on a single host (241)
3618 - data locks weren't cancelled at clear_inode time (290, 311)
3619 - intent locks could lead to unbounded lock growth (205)
3620 - added a maximum lock count, an LRU list, and a flusher
3621 - fix multiple rename (365)
3622 - properly abstracted the echo client
3623 - OSC locked 1 byte too many; fixed
3624 - rewrote brw callback code:
3625 - fixed recovery bugs related to LOVs (306)
3626 - fixed too-many-pages-in-one-write crash (191)
3627 - fixed (again) crash in sync_io_timeout (214)
3628 - probably fixed callback-related race (385)
3630 - Add capability to MDS protocol
3631 - LDLM cancellations and callbacks on different portals
3633 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3636 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3637 - defer open object destruction to close time (601981, 138)
3638 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3639 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3640 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3641 - don't call obd_setattr in ll_file_release for destroyed objects
3642 * protocol change to lustre_msg: move |version| and add |flags|
3643 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3644 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3645 * added replay of create, unlink, link and rename operations during
3646 MDS failover; recovery should be much more robust now
3647 * remove failed OSCs from LOVs (only lov_create uses this so far)
3648 * the lustre-HOWTO was brought (more) up to date (582544)
3650 2002-10-23 Phil Schwan <phil@clusterfs.com>
3653 - in-use dentries weren't being reused properly (617851)
3654 - prevent multiple LDLM setup (599178)
3655 - fix LOV size calculations for truncate (617853)
3656 - fix client handling of MDS intent errors (POSIX)
3657 - fix permission bug in lovstripe.c test (624321)
3658 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3659 - truncate past end of file could corrupt data
3660 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3661 - a race in recovery could return ETIMEDOUT to apps (623947)
3662 - building outside the source directory was fixed
3663 * the lustre-HOWTO was brought (more) up to date (582544)
3664 * major progress was made on recovery functionality
3666 2002-10-10 Phil Schwan <phil@clusterfs.com>
3669 - recovery deadlock fix
3670 - rm -rf causes LBUG fix (617817)
3671 - file open by multiple tasks fix (618962)
3672 - directory permissions bugs (602707 and 620007)
3673 - journal_stop fixed with locking (611313)
3674 - O_APPEND failures resolved (618273, perhaps 614459)
3675 - lconf PATH fix (619770)
3676 - IA64 build fix (621450)
3677 - RPC buffer sizes scale with amount of memory
3679 2002-10-01 Phil Schwan <phil@clusterfs.com>
3682 - locks would be cancelled without throwing away data pages,
3683 resulting in inconsistent data (605627)
3684 - inode attributes were not always being refreshed (605627, 612449)
3685 - lconf now continues to cleanup after lctl reports an error
3686 - MDS now enforces user permissions (602707)
3687 - lprocfs cleanup fixed, but not yet enabled (614157)
3688 - fixed infinite server hang, should a client not respond to an AST
3689 - avoid going into recovery if user calls readlink() with a buffer
3690 that's too small (613941)
3691 - AST RPCs no longer require replies (614867) -- this may be changed
3692 - don't crash server if client sends an IOV that's too big (611336)
3693 - fixed lock conversion deadlock (611892)
3694 - fixed the following of symlinks (614622)
3695 * recovery: the server can remove locks from a client that dies, other
3696 clients can make progress
3697 * more extN patch fixes
3698 * compile-time configurable ptlrpc buffer allocations
3700 - collaborative read cache document
3701 - Lustre Lite Performance CDR document-in-progress
3703 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3706 - fix typo in patch-2.4.18
3708 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3711 - clear ptlrpc request each time in handle_incoming_request()
3712 - unlink of files now destroys the object on the OST
3714 2002-09-19 Peter Braam <braam@clusterfs.com>
3716 * add hard link support
3717 * change obdfile creation method
3718 * kernel patch changed
3720 2002-09-19 Peter Braam <braam@clusterfs.com>
3723 - stack overflow bug in extN fixed
3725 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3727 * documentation updates
3728 - add man pages for config tools
3729 - update tests/README to describe testing with new config tools
3730 - finish metadata API descriptions
3731 * bug fixes and cleanups
3732 - statfs workaround for 16TB limit
3733 - LOV stripe allocation improved, can stripe on subset of OSTs
3734 - LOV file size/IO offset was wrong for files > 4GB in size
3735 - object EA data was being dropped, caused files to be unreadable
3736 - memory overflow with non-LOV OST caused memory corruption
3737 - fixed regression tests to work with new config tools, obdfilter
3738 - fixed bug when directory size became larger than 1 block
3739 - fixed bug (for single client case) when PWD was deleted
3740 - invalidate local directory pages when doing intent-based ops
3741 - avoid LDLM oops when lock callback contained bad data
3743 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3745 * documentation updates
3746 * bug fixes and cleanups
3747 - configuration tools
3750 - 64-bit compile warnings
3751 - 64-bit internal statfs data
3753 * test_brw on persistent OST devices
3755 * lprocfs (disabled)
3757 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3759 * documentation updates
3760 * bug fixes and cleanups
3761 * configuration tools
3763 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3766 * many small fixes to 0_5_4
3767 * io/network handling
3768 * thinkos in MDS operations
3770 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3773 * crucial basic fixes to 0.5.3
3774 * IOR, Iozone work over Elan
3777 2002-08-07 Phil Schwan <phil@clusterfs.com>
3778 * version 0_5_3, our first alpha
3779 * we use the new Portals iovs
3780 * documentation updates
3781 * bug fixes and cleanups
3782 * small changes in the DLM wire protocol
3784 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3785 * version 0_5_1 with some initial stability,
3786 * locking on MD and file I/O.
3787 * documentation updates
3788 * several bug fixes since 0.5.0
3789 * small changes in wire protocol
3791 2002-07-18 Phil Schwan <phil@clusterfs.com>
3793 * delivered as Lustre Light Alpha
3794 * fixed a crash after handling invalid MDS requests
3795 * fixed directory pages for architectures with non-4k pages sizes
3797 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3798 * release version v0_4_4
3799 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3801 * Fixes a number of interruption problems with OST operations.
3802 * Update documentation for portals header changes
3803 * Move all wire protocol structs/defines to lustre_idl.h
3804 * Fixes symlink length bug.
3805 * Add tcpdump to repository.
3807 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3808 * release version v0_4_3
3809 * Fixes statfs for inodes on extN.
3810 * Fixes bug in runtests which would delete /etc/hosts.
3811 * Use 64-bit object IDs wherever possible (not into VFS though)
3812 Remove ost_get_info, which is unused by lustre, and out of date.
3814 2002-07-03 Peter Braam <braam@clusterfs.com>
3815 * release version v0_4_2 Fixes a lookup error (type not passed)
3816 * move forward to head of Portals
3817 * move forward to latest Lustre kernel
3819 2002-06-25 Peter Braam <braam@clusterfs.com>
3820 * release version v0_4_1. Hopefully stable on single node use.