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 --------------------------------------------------------------------------------
63 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
65 * Support for kernels:
68 2.6.16.46-0.14 (SLES 10)
69 2.6.18.8 vanilla (kernel.org)
70 * Client support for unpatched kernels:
71 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
72 2.6.16 - 2.6.22 vanilla (kernel.org)
73 * Due to recently discovered recovery problems, we do not recommend
74 using patchless RHEL 4 clients with this or any earlier release.
75 * Recommended e2fsprogs version: 1.39.cfs8
76 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
77 * Starting with this release, the ldiskfs backing filesystem required
78 by Lustre is now in its own package, lustre-ldiskfs. This package
79 should be installed. It is versioned separately from Lustre and
80 may be released separately in future.
84 Description: block reactivating mgc import until all deactivates complete
85 Details : Fix race when failing back MDT/MGS to itself (testing)
87 Severity : enhancement
89 Description: add optional extra BUILD_VERSION info
90 Details : add a new environment variable (namely LUSTRE_VERS) which allows
91 to override the lustre version.
94 Frequency : 2.6.18 servers only
96 Description: ll_kern_mount() doesn't release the module reference
97 Details : The ldiskfs module reference count never drops down to 0
98 because ll_kern_mount() doesn't release the module reference.
103 Description: server LBUG when using old ost_num_threads parameter
104 Details : Accept the old ost_num_threads parameter but warn that it
105 is deprecated, and fix an off-by-one error that caused an LBUG.
110 Description: Transient SCSI error results in persistent IO issue
111 Details : iobuf->dr_error is not reinitialized to 0 between two
115 Frequency : sometimes when underlying device returns I/O errors
117 Description: OSTs not going read-only during write failures
118 Details : OSTs are not remounted read-only when the journal commit threads
119 get I/O errors because fsfilt_ext3 calls journal_start/stop()
120 instead of the ext3 wrappers.
124 Description: poor connect scaling with increasing client count
125 Details : Don't run filter_grant_sanity_check for more than 100 exports
126 to improve scaling for large numbers of clients.
129 Frequency : SLES10 only
131 Description: sanity-quota.sh quotacheck failed: rc = -22
132 Details : Quotas cannot be enabled on SLES10.
135 Frequency : liblustre clients only
137 Description: getdirentries does not give error when run on compute nodes
138 Details : getdirentries does not fail when the size specified as an argument
139 is too small to contain at least one entry
141 Severity : enhancement
143 Description: Add LNET router traceability for debug purposes
144 Details : If a checksum failure occurs with a router as part of the
145 IO path, the NID of the last router that forwarded the bulk data
146 is printed so it can be identified.
151 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
152 Details : Due to a race condition, liblustre clients were occasionally
155 Severity : enhancement
157 Description: lfs setstripe use optional parameters instead of postional
160 Severity : enhancement
162 Description: Nanosecond timestamp support for ldiskfs
163 Details : The on-disk ldiskfs filesystem has added support for nanosecond
164 resolution timestamps. There is not yet support for this at
165 the Lustre filesystem level.
168 Frequency : during server recovery
170 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
171 Details : request with rq_no_resend flag not awake l_wait_event if they get a
175 Frequency : nfs export on patchless client
177 Description: connectathon hang when test nfs export over patchless client
178 Details : Disconnected dentry cannot be found with lookup, so we do not need
179 to unhash it or make it invalid
182 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
183 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
184 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
189 Description: open req refcounting wrong on reconnect
190 Details : If reconnect happened between getting open reply from server and
191 call to mdc_set_replay_data in ll_file_open, we will schedule
192 replay for unreferenced request that we are about to free.
193 Subsequent close will crash in variety of ways.
194 Check that request is still eligible for replay in
195 mdc_set_replay_data().
200 Description: disable writes to filesystem when reading health_check file
201 Details : the default for reading the health_check proc file has changed
202 to NOT do a journal transaction and write to disk, because this
203 can cause reads of the /proc file to hang and block HA state
204 checking on a healthy but otherwise heavily loaded system. It
205 is possible to return to the previous behaviour during configure
206 with --enable-health-write.
208 Severity : enhancement
210 Description: 64-bit inode version
211 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
212 to the inode. This will be required for version-based recovery.
217 Description: MDS fails to start if a duplicate client export is detected
218 Details : in some rare cases it was possible for a client to connect to
219 an MDS multiple times. Upon recovery the MDS would detect this
220 and fail during startup. Handle this more gracefully.
222 Severity : enhancement
224 Description: Add -o localflock option to simulate old noflock
226 Details : This will achieve local-only flock/fcntl locks
232 Description: log_commit_thread vs filter_destroy race leads to crash
233 Details : Take import reference before releasing llog record semaphore
238 Description: Wrong request locking in request set processing
239 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
240 imp_delayed_list, in this place should be used imp_lock.
243 Frequency : when reconnection
245 Description: Grant Leak when osc reconnect to OST
246 Details : When osc reconnect ost, OST(filter) should check whether it
247 should grant more space to client by comparing fed_grant and
248 cl_avail_grant, and return the granted space to client instead
249 of "new granted" space, because client will call osc_init_grant
250 to update the client grant space info.
253 Frequency : when client reconnect to OST
255 Description: Grant Leak when osc do resend and replay bulk write
256 Details : When osc reconnect to OST, OST(filter)should clear grant info of
257 bulk write request, because the grant info while be sync between
258 OSC and OST when reconnect, and we should ignore the grant info
259 these of resend/replay write req.
264 Description: Grant space more than avaiable left space sometimes.
265 Details : When then OST is about to be full, if two bulk writing from
266 different clients came to OST. Accord the avaliable space of the
267 OST, the first req should be permitted, and the second one
268 should be denied by ENOSPC. But if the seconde arrived before
269 the first one is commited. The OST might wrongly permit second
270 writing, which will cause grant space > avaiable space.
273 Frequency : when client is evicted
275 Description: Grant might be wrongly erased when osc is evicted by OST
276 Details : when the import is evicted by server, it will fork another
277 thread ptlrpc_invalidate_import_thread to invalidate the
278 import, where the grant will be set to 0. While the original
279 thread will update the grant it got when connecting. So if
280 the former happened latter, the grant will be wrongly errased
281 because of this race.
286 Description: Checking Stale with correct fid
287 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
288 to check whether it is stale, because sometimes, we want the
289 enqueue happened anyway, and op_data.fid2 will not be initialized.
291 Severity : enhancement
293 Description: update patchless client
294 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
297 Frequency : only with 2.4 kernel
299 Description: random memory corruption
300 Details : size of struct ll_inode_info is to big for union inode.u and this
301 can be cause of random memory corruption.
306 Description: Memory leak in recovery
307 Details : Lov_mds_md was not free in an error handler in mds_create_object.
308 It should also check obd_fail before fsfilt_start, otherwise if
309 fsfilt_start return -EROFS,(failover mds during mds recovery).
310 then the req will return with repmsg->transno = 0 and rc = EROFS.
311 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
312 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
313 be freed no matter whether fsfilt_commit success or not.
316 Frequency : only with huge count clients
318 Description: Prevents from taking the superblock lock in llap_from_page for
320 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
321 for llap_from_page call in ll_removepage prevents from taking the
322 superblock lock for a soon died page.
327 Description: Not check open intent error before release open handle
328 Details : in some rare cases, the open intent error is not checked before
329 release open handle, which may cause
330 ASSERTION(open_req->rq_transno != 0), because it tries to release
331 the failed open handle.
336 Description: Set cat log bitmap only after create log success.
337 Details : in some rare cases, the cat log bitmap is set too early. and it
338 should be set only after create log success.
342 Description: Accessing a block bevice can re-enable I/O when Lustre is
343 tearing down a device.
344 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
349 Description: service threads may hog cpus when there are a lot of requests
351 Details : Insert cond_resched to give other threads a chance to use some of
357 Description: the cat log was not initialized in recovery
358 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
359 unlink log on mds will not be initialized until mds post
360 recovery. And also in mds post recovery, the unlink log will
361 initialization will be done asynchronausly, so there will be race
362 between add unlink log and unlink log initialization.
366 Description: brw_stats were being printed incorrectly
367 Details : brw_stats were being printed as log2 but all of them were not
368 recorded as log2. Also remove some code duplication arising from
369 filter_tally_{read,write}.
373 Frequency : rare, only in recovery.
374 Description: ASSERTION(req->rq_type != LI_POISON) failed
375 Details : imp_lock should be held while iterating over imp_sending_list for
376 prevent destroy request after get timeout in ptlrpc_queue_wait.
380 Description: replay-single.sh test 52 fails
381 Details : A lock's skiplist need to be cleanup when it being unlinked
382 from its resource list.
386 Description: Short directio read returns full requested size rather than
388 Details : Direct I/O operations should return actual amount of bytes
389 transferred rather than requested size.
391 Severity : enhancement
393 Description: metadata RPC reduction (e.g. for rm performance)
394 Details : decrease the amount of synchronous RPC between clients and servers
395 by canceling conflicing lock before the operation on the client side
396 and packing thier handles into the main operation RPC to server.
398 --------------------------------------------------------------------------------
400 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
405 Frequency : on some architectures
407 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
408 Details : Uninitialized flags sometimes cause configuration commands to
412 Frequency : patchless clients only
414 Description: missing __iget() symbol export
415 Details : The __iget() symbol export is missing. To avoid the need for
416 this on patchless clients the deathrow inode reaper is turned
417 off, and we depend on the VM to clean up old inodes. This
418 dependency was during via the fix for bug 12181.
420 --------------------------------------------------------------------------------
422 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
424 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
425 INTEROPERATE with older versions automatically. In many cases a
426 special upgrade step is needed. Please read the
427 user documentation before upgrading any part of a live system.
428 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
429 version will not interoperate with 1.6 betas before beta5 (1.5.95).
430 * WARNING: Lustre configuration and startup changes are required with
431 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
436 Severity : enhancement
438 Description: MountConf
439 Details : Lustre configuration is now managed via mkfs and mount
440 commands instead of lmc and lconf. New obd types (MGS, MGC)
441 are added for dynamic configuration management. See
442 https://mail.clusterfs.com/wikis/lustre/MountConf for
445 Severity : enhancement
447 Description: dynamic OST addition
448 Details : OSTs can now be added to a live filesystem
450 Severity : enhancement
452 Description: startup order invariance
453 Details : MDTs and OSTs can be started in any order. Clients only
454 require the MDT to complete startup.
456 Severity : enhancement
458 Description: parallel, asynchronous orphan cleanup
459 Details : orphan cleanup is now performed in separate threads for each
460 OST, allowing parallel non-blocking operation.
462 Severity : enhancement
464 Description: optimized stripe assignment
465 Details : stripe assignments are now made based on ost space available,
466 ost previous usage, and OSS previous usage, in order to try
467 to optimize storage space and networking resources.
469 Severity : enhancement
471 Description: Permanently set tunables
472 Details : All writable /proc/fs/lustre tunables can now be permanently
473 set on a per-server basis, at mkfs time or on a live
476 Severity : enhancement
478 Description: Lustre message v2
479 Details : Add lustre message format v2.
481 Severity : enhancement
483 Description: client OST exclusion list
484 Details : Clients can be started with a list of OSTs that should be
485 declared "inactive" for known non-responsive OSTs.
489 Description: SPEC SFS validation failure on NFS v2 over lustre.
490 Details : Changes the blocksize for regular files to be 2x RPC size,
491 and not depend on stripe size.
493 Severity : enhancement
495 Description: Multiple MD RPCs in flight.
496 Details : Further unserialise some read-only MDS RPCs - learn about intents.
497 To avoid overly-overloading MDS, introduce a limit on number of
498 MDS RPCs in flight for a single client and add /proc controls
499 to adjust this limit.
501 Severity : enhancement
503 Description: client read/write statistics
504 Details : Add client read/write call usage stats for performance
505 analysis of user processes.
506 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
507 file access. extents_stats shows chunk size distribution.
508 extents_stats_per_process show chunk size distribution per
511 Severity : enhancement
513 Description: mds statistics
514 Details : Add detailed mds operations statistics in
515 /proc/fs/lustre/mds/*/stats.
519 Description: Failure of copying files with lustre special EAs.
520 Details : Client side always return success for setxattr call for lustre
521 special xattr (currently only "trusted.lov").
526 Description: Refcount LNET uuids
527 Details : The global LNET uuid list grew linearly with every startup;
528 refcount repeated list entries instead of always adding to
531 Severity : enhancement
533 Description: Dynamic service threads
534 Details : Within a small range, start extra service threads
535 automatically when the request queue builds up.
538 Frequency : mixed-endian client/server environments
540 Description: mixed-endian crashes
541 Details : The new msg_v2 system had some failures in mixed-endian
544 Severity : enhancement
546 Description: Easy OST removal
547 Details : OSTs can be permanently deactivated with e.g. 'lctl
548 conf_param lustre-OST0001.osc.active=0'
550 Severity : enhancement
552 Description: MGS proc entries
553 Details : Added basic proc entries for the MGS showing what filesystems
556 Severity : enhancement
558 Description: provide MGS failover
559 Details : Added config lock reacquisition after MGS server failover.
561 Severity : enhancement
563 Description: add Linux 2.4 support
564 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
568 Description: a large application tries to do I/O to the same resource and dies
570 Details : Check the req->rq_arrival time after the call to
571 ost_brw_lock_get(), but before we do anything about
572 processing it & sending the BULK transfer request. This
573 should help move old stale pending locks off the queue as
574 quickly as obd_timeout.
577 Frequency : when an incorrect nid is specified during startup
579 Description: ptlrpc connect to non-existant node causes kernel crash
580 Details : LNET can't be re-entered from an event callback, which
581 happened when we expire a message after the export has been
582 cleaned up. Instead, hand the zombie cleanup off to another
585 Severity : enhancement
587 Description: plain/inodebits lock performance improvement
588 Details : Grouping plain/inodebits in granted list by their request modes
589 and bits policy, thus improving the performance of search through
593 Frequency : only if OST filesystem is corrupted
595 Description: client incorrectly hits assertion in ptlrpc_replay_req()
596 Details : for a short time RPCs with bulk IO are in the replay list,
597 but replay of bulk IOs is unimplemented. If the OST filesystem
598 is corrupted due to disk cache incoherency and then replay is
599 started it is possible to trip an assertion. Avoid putting
600 committed RPCs into the replay list at all to avoid this issue.
603 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
605 Description: System hang on startup
606 Details : This bug allowed the liblustre (e.g. catamount) client to
607 return to the app before handling all startup RPCs. This
608 could leave the node unresponsive to lustre network traffic
609 and manifested as a server ptllnd timeout.
611 Severity : enhancement
613 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
614 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
615 boolean to control whether to print peer debug info when a
616 client's RPC times out.
619 Frequency : only for kernels with patches from Lustre below 1.4.3
621 Description: Remove old rdonly API
622 Details : Remove old rdonly API which unused from at least lustre 1.4.3
625 Frequency : only for devices with external journals
627 Description: Set external device read-only also
628 Details : During a commanded failover stop, we set the disk device
629 read-only while the server shuts down. We now also set any
630 external journal device read-only at the same time.
633 Frequency : when upgrading from 1.4 while trying to change parameters
635 Description: The wrong (new) MDC name was used when setting parameters for
636 upgraded MDT's. Also allows changing of OSC (and MDC)
637 parameters if --writeconf is specified at tunefs upgrade time.
640 Frequency : when setting specific ost indicies
642 Description: QOS code breaks on skipped indicies
643 Details : Add checks for missing OST indicies in the QOS code, so OSTs
644 created with --index need not be sequential.
646 Severity : enhancement
648 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
649 Details : The uninit_groups feature works in conjunction with the kernel
650 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
651 up the pass1 processing of e2fsck. This is a read-only feature
652 in ldiskfs2 only, so older kernels and current ldiskfs cannot
653 mount filesystems that have had this feature enabled.
655 Severity : enhancement
657 Description: Improve multi-block allocation algorithm to avoid fragmentation
658 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
659 allocation locality and avoid filesystem fragmentation.
661 ------------------------------------------------------------------------------
663 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
665 * Support for kernels:
666 2.4.21-47.0.1.EL (RHEL 3)
668 2.6.9-42.0.10.EL (RHEL 4)
669 2.6.12.6 vanilla (kernel.org)
670 2.6.16.27-0.9 (SLES 10)
671 * Recommended e2fsprogs version: 1.39.cfs5
673 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
677 Frequency : occasional, depends on client load and configuration
678 Bugzilla : 12181, 12203
679 Description: data loss for recently-modified files
681 Details : In some cases it is possible that recently written or created
682 files may not be written to disk in a timely manner (this should
683 normally be within 30s unless client IO load is very high).
684 The problem appears as zero-length files or files that are a
685 multiple of 1MB in size after a client crash or client eviction
686 that are missing data at the end of the file.
688 This problem is more likely to be hit on clients where files are
689 repeatedly created and unlinked in the same directory, clients
690 have a large amount of RAM, have many CPUs, the filesystem has
691 many OSTs, the clients are rebooted frequently, and/or the files
692 are not accessed by other nodes after being written.
694 The presence of the problem can be detected by looking at
695 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
696 smaller than the second (nr_unused) then dirty files will not
697 be flushed automatically to disk. "sync; sleep 10" should be
698 run several times on the node before unmounting it to update
699 Lustre (this is also safe to run on nodes without this problem).
701 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
702 that can cause this same problem, so customers using that kernel
703 also need to update the kernel in addition to Lustre. In order
704 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
706 It is normal that files written just before a client crash (less
707 than 30s) may not yet have been flushed to disk, even for local
711 Frequency : frequent on thin XT3 nodes
713 Description: UUID collision on thin XT3 Linux nodes
714 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
715 randomly, since we relied on an insufficiently-seeded PRNG.
720 Description: OSS hangs after "All ost request buffers busy"
721 Details : A deadlock between quota and journal operations caused OSS
722 hangs after printing "All ost request buffers busy."
725 Frequency : always on liblustre builds
727 Description: Cleanup compiler warnings on liblustre
730 Frequency : always on liblustre builds on XT3
732 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
736 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
737 > 32000 subdirectories
738 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
739 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
740 subdirectory count crosses 32000. This will aid e2fsck to
741 correctly handle more than 32000 subdirectories.
744 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
746 Description: System hang on startup
747 Details : This bug allowed the liblustre (e.g. catamount) client to
748 return to the app before handling all startup RPCs. This
749 could leave the node unresponsive to lustre network traffic
750 and manifested as a server ptllnd timeout.
752 Severity : enhancement
754 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
755 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
756 boolean to control whether to print peer debug info when a
757 client's RPC times out.
762 Description: make O_SYNC working on 2.6 kernels
763 Details : 2.6 kernels use different method for mark pages for write,
764 so need add a code to lustre for O_SYNC work.
769 Description: Failure to close file and release space on NFS
770 Details : Put inode details into lock acquired in ll_intent_file_open.
771 Use mdc_intent_lock in ll_intent_open to properly
772 detect all kind of errors unhandled by mdc_enqueue.
777 Description: proc file read during shutdown sometimes raced obd removal,
779 Details : Add lock to prevent obd access after proc file removal.
782 Frequency : Only for files larger than 4GB on 32-bit clients.
784 Description: improperly doing page alignment of locks
785 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
786 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
791 Description: RPCs being resent when they shouldn't be
792 Details : Some RPCs that should not be resent are being resent. This
793 can cause inconsistencies in the RPC state machine. Do not
794 resend such requests.
797 Frequency : rare, only with NFS export
799 Description: Crash on NFS re-export node
800 Details : under very unusual load conditions an assertion is hit in
801 ll_intent_file_open()
804 Frequency : only if OST filesystem is corrupted
806 Description: client incorrectly hits assertion in ptlrpc_replay_req()
807 Details : for a short time RPCs with bulk IO are in the replay list,
808 but replay of bulk IOs is unimplemented. If the OST filesystem
809 is corrupted due to disk cache incoherency and then replay is
810 started it is possible to trip an assertion. Avoid putting
811 committed RPCs into the replay list at all to avoid this issue.
816 Description: large O_DIRECT requests fail under memory pressure/fragmentation
817 Details : Large single O_DIRECT read and write calls can fail to allocate
818 a sufficiently large buffer to process the request. In case of
819 allocation failure the allocation is retried with a smaller
820 buffer and broken into smaller requests.
822 Severity : enhancement
824 Description: Add -o localflock option to simulate old noflock behaviour.
825 Details : This will achieve local-only flock/fcntl locks coherentness.
830 Description: versioning check is incomplete
831 Details : Checking the version difference of client vs. server, report
832 error if the gap is too big.
837 Description: add support PG_writeback bit
838 Details : add support for PG_writeback bit for Lustre, for more carefull
839 work with page cache in 2.6 kernel. This also fix some deadlocks
840 and remove hack for work O_SYNC with 2.6 kernel.
842 Severity : enhancement
844 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
845 Details : The uninit_groups feature works in conjunction with the kernel
846 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
847 up the pass1 processing of e2fsck. This is a read-only feature
848 in ldiskfs2 only, so older kernels and current ldiskfs cannot
849 mount filesystems that have had this feature enabled.
851 Severity : enhancement
853 Description: Improve multi-block allocation algorithm to avoid fragmentation
854 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
855 allocation locality and avoid filesystem fragmentation.
857 ------------------------------------------------------------------------------
859 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
861 * Support for kernels:
862 2.6.9-42.0.3.EL (RHEL 4)
864 2.4.21-47.0.1.EL (RHEL 3)
865 2.6.12.6 vanilla (kernel.org)
866 2.6.16.21-0.8 (SLES10)
867 * Recommended e2fsprogs version: 1.39.cfs2-0
869 * The backwards-compatible /proc/sys/portals symlink has been removed
870 in this release. Before upgrading, please ensure that you change
871 any configuration scripts or /etc/sysctl.conf files that access
872 /proc/sys/portals/* or sysctl portals.* to use the corresponding
873 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
874 in advance of the upgrade on any system running Lustre 1.4.6 or
875 newer, since /proc/sys/lnet was added in that version.
876 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
880 Frequency : only when quota is used
882 Description: avoid scanning export list for quota master
883 Details : Change the algorithms to avoid scanning export list in order
884 to improve the efficiency.
887 Frequency : MDS failover only, very rarely
889 Description: "went back in time" messages on mds failover
890 Details : The greatest transno may be lost when the current operation
891 finishes with an error (transno==0) and the client's last_rcvd
892 record is over-written. Save the greatest transno in the
893 mds_last_transno for this case.
896 Frequency : always for specific kernels and striping counts
898 Description: client may get "Matching packet too big" without ACL support
899 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
900 when trying to access files in certain configurations. The
901 clients should in fact be denied when mounting because they do
905 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
907 Description: many clients connecting with IO in progress causes connect timeouts
908 Details : Avoid synchronous journal commits to avoid delays caused by many
909 clients connecting/disconnecting when bulk IO is in progress.
910 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
911 OST_IO_PORTAL to avoid delays behind potentially many pending
915 Frequency : occasionally with multiple writers to a single file
917 Description: shared writes to file may result in wrong size reported by stat()
918 Details : Allow growing of kms when extent lock is cancelled
921 Frequency : always with random mmap IO to multi-striped file
923 Description: mmap write might be lost if we are writing to a 'hole' in stripe
924 Details : Only if the hole is at the end of OST object so that kms is too
925 small. Fix is to increase kms accordingly in ll_nopage.
928 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
930 Description: writes to a missing object would leak memory on the OST
931 Details : If there is an inconsistency between the MDS and OST filesystems,
932 such that the MDS references an object that doesn't exist, writes
933 to that object will leak memory due to incorrect cleanup in the
934 error handling path, eventually running out of memory on the OST.
939 Description: Creating too long symlink causes lustre errors
940 Details : Check symlink and name lengths before sending requests to MDS.
943 Frequency : only if flock is enabled (not on by default)
945 Description: posix locks not released on fd closure on 2.6.9+
946 Details : We failed to add posix locks to list of inode locks on 2.6.9+
947 kernels, this caused such locks not to be released on fd close and
948 then assertions on fs unmount about still used locks.
951 Frequency : MDS failover only, very rarely
953 Description: clients may get ASSERTION(granted_lock != NULL)
954 Details : When request was taking a long time, and a client was resending
955 a getattr by name lock request. The were multiple lock requests
956 with the same client lock handle and
957 mds_getattr_name->fixup_handle_for_resent_request found one of the
958 lock handles but later failed with ASSERTION(granted_lock != NULL).
963 Description: handle->h_buffer_credits > 0, assertion failure
964 Details : h_buffer_credits is zero after truncate, causing assertion
965 failure. This patch extends the transaction or creates a new
969 Frequency : NFS re-export or patchless client
970 Bugzilla : 11179, 10796
971 Description: Crash on NFS re-export node (__d_move)
972 Details : We do not want to hash the dentry if we don't have a lock.
973 But if this dentry is later used in d_move, we'd hit uninitialised
974 list head d_hash, so we just do this to init d_hash field but
975 leave dentry unhashed.
978 Frequency : NFS re-export or patchless client
980 Description: NFS exports has problem with symbolic link
981 Details : lustre client didn't properly install dentry when re-exported
982 to NFS or running patchless client.
985 Frequency : NFS re-export or patchless client
987 Description: Various nfs/patchless fixes.
988 Details : fixes reuse disconected alias for lookup process - this fixes
989 warning "find_exported_dentry: npd != pd",
990 fix permission error with open files at nfs.
991 fix apply umask when do revalidate.
994 Frequency : occasional
996 Description: Crash on NFS re-export node
997 Details : calling clear_page() on the wrong pointer triggered oops in
998 generic_mapping_read().
1001 Frequency : rarely, using O_DIRECT IO
1003 Description: unaligned directio crashes client with LASSERT
1004 Details : check for unaligned buffers before trying any requests.
1007 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1009 Description: stale data returned from RAID cache
1010 Details : If only a small amount of IO is done to the RAID device before
1011 reading it again it is possible to get stale data from the RAID
1012 cache instead of reading it from disk.
1015 Frequency : always for sles10 kernel
1017 Description: sles10 support
1018 Details : ll_follow_link: compile fixes and using of nd_set_link
1019 under newer kernels.
1022 Frequency : depends on arch, kernel and compiler version, always on sles10
1025 Description: recursive or deep enough symlinks cause stack overflow
1026 Details : getting rid of large stack-allocated variable in
1030 Frequency : depends on hardware
1032 Description: lustre write performance loss in the SLES10 kernel
1033 Details : the performance loss is caused by using of write barriers in the
1034 ext3 code. The SLES10 kernel turns barrier support on by
1035 default. The fix is to undo that change for ldiskfs.
1037 ------------------------------------------------------------------------------
1039 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1041 * Support for kernels:
1042 2.6.9-42.0.3EL (RHEL 4)
1043 2.6.5-7.276 (SLES 9)
1044 2.4.21-47.0.1.EL (RHEL 3)
1045 2.6.12.6 vanilla (kernel.org)
1049 Frequency : quota enabled and large files being deleted
1051 Description: releasing more than 4GB of quota at once hangs OST
1052 Details : If a user deletes more than 4GB of files on a single OST it
1053 will cause the OST to spin in an infinite loop. Release
1054 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1059 Description: statfs data retrieved from /proc may be stale or zero
1060 Details : When reading per-device statfs data from /proc, in the
1061 {kbytes,files}_{total,free,avail} files, it may appear
1062 as zero or be out of date.
1065 Frequency : systems with MD RAID1 external journal devices
1067 Description: lconf's call to blkid is confused by RAID1 journal devices
1068 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1069 of returning all block devices that match the journal UUID.
1072 Frequency : always, for aggregate stripe size over 4GB
1074 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1075 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1076 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1079 Frequency : always if "lfs find" used on a local file/directory
1081 Description: "lfs find" segfaults if used on a local file/directory
1082 Details : The case where a directory component was not specified wasn't
1083 handled correctly. Handle this properly.
1086 Frequency : always on ppc64
1088 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1089 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1090 perform endianness conversion.
1093 Frequency : rarely (truncate to non-zero file size after write under load)
1094 Bugzilla : 10730, 10687
1095 Description: Files padded with zeros to next 4K multiple
1096 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1097 it is possible that files that are truncated to a non-zero size
1098 immediately after being written are filled with zero bytes beyond
1099 the truncated size. No file data is lost.
1101 Severity : enhancement
1103 Description: Allow recovery/failover for liblustre clients.
1104 Details : liblustre clients were unaware of failover configurations until
1107 Severity : enhancement
1109 Description: user file locks should fail when not mounting with flock option
1110 Details : Set up an error-returning stub in ll_file_operations.lock field
1111 to prevent incorrect behaviour when client is mounted without
1112 flock option. Also, set up properly f_op->flock field for
1116 Frequency : always on ia64
1118 Description: "lfs df" loops on printing out MDS statfs information
1119 Details : The obd_ioctl_data was not initialized and in some systems
1120 this caused a failure during the ioctl that did not return
1121 an error. Initialize the struct and return an error on failure.
1124 Frequency : SLES 9 only
1126 Description: Error of copying files with lustre special EAs as root
1127 Details : Client side always return success for setxattr call for lustre
1128 special xattr (currently only "trusted.lov").
1131 Frequency : rarely on clusters with both ia64+i386 clients
1133 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1134 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1135 clients are doing concurrent writes to the same file it is
1136 possible that the ia64 clients may LASSERT because the OST
1137 extent locks are not PAGE_SIZE aligned. Ensure that grown
1138 locks are always aligned on the request boundary.
1141 Frequency : specific use, occasional
1143 Description: Overwriting in use executable truncates on-disk binary image
1144 Details : If one node attempts to overwrite an executable in use by
1145 another node, we now correctly return ETXTBSY instead of
1146 truncating the file.
1148 Severity : enhancement
1150 Description: Async OSC create to avoid the blocking unnecessarily.
1151 Details : If a OST has no remain object, system will block on the creating
1152 when need to create a new object on this OST. Now, ways use
1153 pre-created objects when available, instead of blocking on an
1154 empty osc while others are not empty. If we must block, we block
1155 for the shortest possible period of time.
1160 Description: chmod on Lustre root is propagated to other clients
1161 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1162 invalid by the follow_mount time.
1165 Frequency : liblustre clients only
1167 Description: Race in 'instant cancel' lock handling could lead to such locks
1168 never to be granted in case of SMP MDS
1169 Details : Do not destroy not yet granted but cbpending locks in
1173 Frequency : replay/resend of open
1175 Description: non null lock assetion failure in mds_intent_policy
1176 Details : Trying to replay/resend lockless open requests resulted in
1177 mds_open() returning 0 with no lock. Now it sets a flag if
1178 a lock is going to be returned.
1180 Severity : enhancement
1182 Description: Checksum enhancements
1183 Details : New checksum enhancements allow for resending RPCs that failed
1186 Severity : enhancement
1188 Description: Tunables on number of dirty pages in cacche
1189 Details : Allow to set limit on number of dirty pages cached.
1194 Description: client crash on unmount - lock still has references
1195 Details : In some error handling cases it was possible to leak a lock
1196 reference on a client while accessing a file. This was not
1197 harmful to the client during operation, but would cause the
1198 client to crash when the filesystem is unmounted.
1201 Frequency : specific case, rare
1203 Description: ETXTBSY on mds though file not in use by client
1204 Details : ETXTBSY is no longer incorrectly returned when attempting to
1205 chmod or chown a directory that the user previously tried to
1206 execute or a currently-executing binary.
1209 Frequency : extremely rare except on liblustre-based clients
1211 Description: Lustre space not freed when files are deleted
1212 Details : Clean up open-unlinked files after client eviction. Previously
1213 the unlink was skipped and the files remained as orphans.
1218 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1219 Details : In some cases it was possible to send a blocking callback to
1220 a client doing a glimpse, even though that client didn't get
1221 a lock granted. When the glimpse lock is cancelled on the OST
1222 the freed lock is left on the waiting list and corrupted the list.
1225 Frequency : all core dumps
1227 Description: Broke core dumps to lustre
1228 Details : Negative dentry may be unhashed if parent does not have UPDATE
1229 lock, but some callers, e.g. do_coredump, expect dentry to be
1230 hashed after successful create, hash it in ll_create_it.
1232 ------------------------------------------------------------------------------
1234 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1236 * Support for kernels:
1237 2.6.9-42.0.2.EL (RHEL 4)
1238 2.6.5-7.276 (SLES 9)
1239 2.4.21-40.EL (RHEL 3)
1240 2.6.12.6 vanilla (kernel.org)
1244 Frequency : always on RHEL 3
1246 Description: Number of open files grows over time
1247 Details : The number of open files grows over time, whether or not
1248 Lustre is started. This was due to a filp leak introduced
1249 by one of our kernel patches.
1251 ------------------------------------------------------------------------------
1253 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1255 * Support for kernels:
1256 2.6.9-42.EL (RHEL 4)
1257 2.6.5-7.276 (SLES 9)
1258 2.4.21-40.EL (RHEL 3)
1259 2.6.12.6 vanilla (kernel.org)
1264 Bugzilla : 5719, 9635, 9792, 9684
1265 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1266 Details : If a server is under heavy load and cannot reply to new
1267 connection requests before the client resends the (re)connect,
1268 the connection handling code can behave badly if two service
1269 threads are concurrently handing separate (re)connections from
1270 the same client. Add better locking to the connection handling
1271 code, and ensure that only a single connection will be processed
1272 for a given client UUID, even if the lock is dropped.
1274 Severity : enhancement
1276 Description: add TCP zero-copy support to kernel
1277 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1278 sends if the hardware supports scatter-gather and checksumming.
1279 This allows socklnd to do client-write and server-read more
1280 efficiently and reduce CPU utilization from skbuf copying.
1283 Frequency : only if NFS exporting from client
1285 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1286 Details : If setting attributes on a file created under NFS that had
1287 never been opened it would be possible to oops the client
1288 if the file had no objects.
1291 Frequency : always for liblustre
1293 Description: liblustre client does MDS+OSTs setattr RPC for each write
1294 Details : When doing a write from a liblustre client, the client
1295 incorrectly issued an RPC to the MDS and each OST the file was
1296 striped over in order to update the timestamps. When writing
1297 with small chunks and many clients this could overwhelm the MDS
1298 with RPCs. In all cases it would slow down the write because
1299 these RPCs are unnecessary.
1301 Severity : enhancement
1303 Description: allow number of MDS service threads to be changed at module load
1304 Details : It is now possible to change the number of MDS service threads
1305 running. Adding "options mds mds_num_threads={N}" to the MDS's
1306 /etc/modprobe.conf will set the number of threads for the next
1307 time Lustre is restarted (assuming the "mds" module is also
1308 reloaded at that time). The default number of threads will
1309 stay the same, 32 for most systems.
1314 Description: OST crash if filesystem is unformatted or corrupt
1315 Details : If an OST is started on a device that has never been formatted
1316 or if the filesystem is corrupt and cannot even mount then the
1317 error handling cleanup routines would dereference a NULL pointer.
1322 Description: NULL pointer deref in llap_from_page.
1323 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1324 incorrect) mapping if the page was truncated/reclaimed while it was
1325 searched for. Check for this condition and skip such pages when
1326 doing readahead. Introduce extra check to llap_from_page() to
1327 verify page->mapping->host is non-NULL (so page is not anonymous).
1330 Frequency : Sometimes when using sys_sendfile
1332 Description: "page not covered by a lock" warnings from ll_readpage
1333 Details : sendfile called ll_readpage without right page locks present.
1334 Now we introduced ll_file_sendfile that does necessary locking
1335 around call to generic_file_sendfile() much like we do in
1339 Frequency : with certain MDS communication failures at client mount time
1341 Description: NULL pointer deref after failed client mount
1342 Details : a client connection request may delayed by the network layer
1343 and not be sent until after the PTLRPC layer has timed out the
1344 request. If the client fails the mount immediately it will try
1345 to clean up before the network times out the request. Add a
1346 reference from the request import to the obd device and delay
1347 the cleanup until the network drops the request.
1350 Frequency : occasionally during client (re)connect
1352 Description: assertion failure during client (re)connect
1353 Details : processing a client connection request may be delayed by the
1354 client or server longer than the client connect timeout. This
1355 causes the client to resend the connection request. If the
1356 original connection request is replied in this interval, the
1357 client may trip an assertion failure in ptlrpc_connect_interpret()
1358 which thought it would be the only running connect process.
1361 Frequency : only with obd_echo servers and clients that are rebooted
1363 Description: kernel BUG accessing uninitialized data structure
1364 Details : When running an obd_echo server it did not start the ping_evictor
1365 thread, and when a client was evicted an uninitialized data
1366 structure was accessed. Start the ping_evictor in the RPC
1367 service startup instead of the OBD startup.
1369 Severity : enhancement
1370 Bugzilla : 10193 (patchless)
1371 Description: Remove dependency on various unexported kernel interfaces.
1372 Details : No longer need reparent_to_init, exit_mm, exit_files,
1373 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1376 Frequency : rare (only users of deprecated and unsupported LDAP config)
1378 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1379 Details : LDAP apparently contains 'lustreName' attributes instead of
1380 'name'. A simple remapping of the name is sufficient.
1383 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1385 Description: waiting_locks_callback trips kernel BUG if client is evicted
1386 Details : Running with the dump_on_timeout debug flag turned on makes
1387 it possible that the waiting_locks_callback() can try to dump
1388 the Lustre kernel debug logs from an interrupt handler. Defer
1389 this log dumping to the expired_lock_main() thread.
1391 Severity : enhancement
1393 Description: Support NFS exporting on 2.6 kernels.
1394 Details : Implement non-rawops metadata methods for NFS server to use without
1395 changing NFS server code.
1398 Frequency : very rare (synthetic metadata workload only)
1400 Description: two racing renames might cause an MDS thread to deadlock
1401 Details : Running the "racer" program may cause one MDS thread to rename
1402 a file from being the source of a rename to being the target of
1403 a rename at exactly the same time that another thread is doing
1404 so, and the second thread has already enqueued these locks after
1405 doing a lookup of the target and is trying to relock them in
1406 order. Ensure that we don't try to re-lock the same resource.
1409 Frequency : only very large systems with liblustre clients
1411 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1412 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1413 all OSTs in parallel. This allows the eviction of stale liblustre
1414 clients to proceed much faster than if they were done in series,
1415 and also offers similar improvements for other set_info RPCs.
1420 Description: excessive CPU usage during initial read phase on client
1421 Details : During the initial read phase on a client, it would agressively
1422 retry readahead on the file, consuming too much CPU and impacting
1423 performance (since 1.4.5.8). Improve the readahead algorithm
1424 to avoid this, and also improve some other common cases (read
1425 of small files in particular, where "small" is files smaller than
1426 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1431 Description: MDS crash when receiving packet with unknown intent.
1432 Details : Do not LBUG in unknown intent case, just return -EFAULT
1434 Severity : enhancement
1435 Bugzilla : 9293, 9385
1436 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1437 Details : Do not serialize getattr (non-intent version) and statfs.
1440 Frequency : occasional, when OST network is overloaded/intermittent
1442 Description: client evicted by OST after bulk IO timeout
1443 Details : If a client sends a bulk IO request (read or write) the OST
1444 may evict the client if it is unresposive to its data GET/PUT
1445 request. This is incorrect if the network is overloaded (takes
1446 too long to transfer the RPC data) or dropped the OST GET/PUT
1447 request. There is no need to evict the client at all, since
1448 the pinger and/or lock callbacks will handle this, and the
1449 client can restart the bulk request.
1452 Frequency : Always when mmapping file with no objects
1454 Description: client crashes when mmapping file with no objects
1455 Details : Check that we actually have objects in a file before doing any
1456 operations on objects in ll_vm_open, ll_vm_close and
1462 Description: Request leak when working with deleted CWD
1463 Details : Introduce advanced request refcount tracking for requests
1464 referenced from lustre intent.
1466 Severity : Enhancement
1468 Description: Cache open file handles on client.
1469 Details : MDS now will return special lock along with openhandle, if
1470 requested and client is allowed to hold openhandle, even if unused,
1471 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1472 closing files for every read/write openration.
1474 Severity : Enhancement
1476 Description: Cache open negative dentries on client when possible.
1477 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1478 negative dentries on lock revocation.
1483 Description: Remounting a client read-only wasn't possible with a zconf mount
1484 Details : It wasn't possible to remount a client read-only with llmount.
1486 Severity : enhancement
1487 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1488 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1489 MPICH is included to add Lustre-specific ADIO interfaces.
1490 This is based closely on the UFS ADIO layer and only differs
1491 in file creation, in order to allow the OST striping to be set.
1492 This is user-contributed code and not supported by CFS.
1497 Description: extended inode attributes (immutable, append-only) work improperly
1498 when 2.4 and 2.6 kernels are used on client/server or vice versa
1499 Details : Introduce kernel-independent values for these flags.
1501 Severity : enhancement
1504 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1505 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1506 granularity. Now, Fractional MB granularity is be supported,
1507 this is very useful for low memory system.
1509 Severity : enhancement
1511 Description: Getattr by fid
1512 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1513 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1514 have locks and avoids getattr rpc requests for it on every stat.
1517 Frequency : Always, for filesystems larger than 2TB
1519 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1520 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1521 filesystems larger than 2TB to be mounted with mballoc enabled.
1524 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1526 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1527 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1528 filesystems larger than 2TB is mounted then the kernel will
1529 silently corrupt the start of the filesystem. CONFIG_LBD is
1530 enabled for all CFS-supported kernels, but the possibility of
1531 this happening with a modified kernel config exists.
1533 Severity : enhancement
1535 Description: add client O_DIRECT support for 2.6 kernels
1536 Details : It is now possible to do O_DIRECT reads and writes to files
1537 in the Lustre client mountpoint on 2.6 kernel clients.
1539 Severity : enhancement
1541 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1542 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1543 parallel, not waiting for response from every OST before sending
1544 a rpc to the next OST.
1549 Description: setattr vs write race when updating file timestamps
1550 Details : Client processes that update a file timestamp into the past
1551 right after writing to the file (e.g. tar) it is possible that
1552 the updated file modification time can be reset to the current
1553 time due to a race between processing the setattr and write RPC.
1555 Severity : enhancement
1557 Description: Bring 'lfs find' closer in line with regular Linux find.
1558 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1559 -print0 options and obtains all the needed info through the lustre
1562 Severity : enhancement
1564 Description: support up to 1024 configured devices on one node
1565 Details : change obd_dev array from statically allocated to dynamically
1566 allocated structs as they are first used to reduce memory usage
1571 Description: Flush dirty partially truncated pages during truncate
1572 Details : Immediatelly flush partially truncated pages in filter_setattr,
1573 this way we completely avoid having any pages in page cache on OST
1574 and can retire ugly workarounds during writes to flush such pages.
1579 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1580 Details : protect inode from truncation within vfs_unlink() context
1581 just take a reference before calling vfs_unlink() and release it
1582 when parent's i_sem is free.
1587 Description: last_id value checked outside lock on OST caused LASSERT failure
1588 Details : If there were multiple MDS->OST object precreate requests in
1589 flight, it was possible that the OST's last object id was checked
1590 outside a lock and incorrectly tripped an assertion. Move checks
1591 inside locks, and discard old precreate requests.
1594 Frequency : always, if extents are used on OSTs
1596 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1597 Details : OSTs using the extents format would not zero the high 16 bits of
1598 the index physical block number. This is not a problem for any
1599 OST filesystems smaller than 16TB, and no kernels support ext3
1600 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1601 new/modified files) and can be fixed for existing filesystems
1602 with e2fsprogs-1.39-cfs1.
1607 Description: import connection selection may be incorrect if timer wraps
1608 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1609 import connections to be ignored if the 32-bit jiffies counter
1610 wraps. Use a 64-bit jiffies counter.
1613 Frequency : very large clusters immediately after boot
1615 Description: LNET request buffers exhausted under heavy short-term load
1616 Details : If a large number of client requests are generated on a service
1617 that has previously never seen so many requests it is possible
1618 that the request buffer growth cannot keep up with the spike in
1619 demand. Instead of dropping incoming requests, they are held in
1620 the LND until the RPC service can accept more requests.
1623 Frequency : Sometimes during replay
1625 Description: Assertion failure in ll_local_open after replay.
1626 Details : If replay happened on an open request reply before we were able
1627 to set replay handler, reply will become not swabbed tripping the
1628 assertion in ll_local_open. Now we set the handler right after
1629 recognising of open request
1632 Frequency : very rare
1634 Description: kernel reports "badness in vsnprintf"
1635 Details : Reading from the "recovery_status" /proc file in small chunks
1636 may cause a negative length in lprocfs_obd_rd_recovery_status()
1637 call to vsnprintf() (which is otherwise harmless). Exit early
1638 if there is no more space in the output buffer.
1640 Severity : enhancement
1642 Description: clear OBD RPC statistics by writing to them
1643 Details : It is now possible to clear the OBD RPC statistics by writing
1644 to the "stats" file.
1649 Description: Client mtime is not the same on different clients after utimes
1650 Details : In some cases, the client was using the utimes() syscall on
1651 a file cached on another node. The clients now validate the
1652 ctime from the MDS + OSTs to determine which one is right.
1657 Description: Inability to activate failout mode
1658 Details : lconf script incorrectly assumed that in pythong string's numeric
1659 value is used in comparisons.
1662 Frequency : always with multiple stripes per file
1664 Description: Inefficient object allocation for mutli-stripe files
1665 Details : When selecting which OSTs to stripe files over, for files with
1666 a stripe count that divides evenly into the number of OSTs,
1667 the MDS is always picking the same starting OST for each file.
1668 Return the OST selection heuristic to the original design.
1673 Description: mount failures may take full timeout to return an error
1674 Details : Under some heavy load conditions it is possible that a
1675 failed mount can wait for the full obd_timeout interval,
1676 possibly several minutes, before reporting an error.
1677 Instead return an error as soon as the status is known.
1679 Frequency : quota enabled and large files being deleted
1681 Description: releasing more than 4GB of quota at once hangs OST
1682 Details : If a user deletes more than 4GB of files on a single OST it
1683 will cause the OST to spin in an infinite loop. Release
1684 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1689 Description: statfs data retrieved from /proc may be stale or zero
1690 Details : When reading per-device statfs data from /proc, in the
1691 {kbytes,files}_{total,free,avail} files, it may appear
1692 as zero or be out of date.
1695 Frequency : systems with MD RAID1 external journal devices
1697 Description: lconf's call to blkid is confused by RAID1 journal devices
1698 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1699 of returning all block devices that match the journal UUID.
1702 Frequency : always, for aggregate stripe size over 4GB
1704 Description: assertion fails when trying to use 4GB stripe size
1705 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1706 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1709 Frequency : always on ppc64
1711 Description: the first write on an ext3 filesystem with mballoc got stuck
1712 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1713 perform endianness conversion.
1715 ------------------------------------------------------------------------------
1717 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1719 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1720 INTEROPERATE with older versions automatically. Please read the
1721 user documentation before upgrading any part of a live system.
1722 * WARNING: Lustre networking configuration changes are required with
1723 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1726 * Support for kernels:
1727 2.6.9-22.0.2.EL (RHEL 4)
1728 2.6.5-7.244 (SLES 9)
1729 2.6.12.6 vanilla (kernel.org)
1732 Severity : enhancement
1733 Bugzilla : 7981/8208
1734 Description: Introduced Lustre Networking (LNET)
1735 Details : LNET is new networking infrastructure for Lustre, it includes
1736 a reorganized network configuration mode (see the user
1737 documentation for full details) as well as support for routing
1738 between different network fabrics. Lustre Networking Devices
1739 (LNDs) for the supported network fabrics have also been
1740 created for this new infrastructure.
1742 Severity : enhancement
1743 Description: Introduced Access control lists
1744 Details : clients can set ACLs on files and directories in order to have
1745 more fine-grained permissions than the standard Unix UGO+RWX.
1746 The MDS must be started with the "-o acl" mount option.
1748 Severity : enhancement
1749 Description: Introduced filesystem quotas
1750 Details : Administrators may now establish per-user quotas on the
1753 Severity : enhancement
1755 Description: Configuration change for the XT3
1756 The PTLLND is now used to run Lustre over Portals on the XT3
1757 The configure option(s) --with-cray-portals are no longer used.
1758 Rather --with-portals=<path-to-portals-includes> is used to
1759 enable building on the XT3. In addition to enable XT3 specific
1760 features the option --enable-cray-xt3 must be used.
1765 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1766 Details : A race between allocating a new llcd and re-getting the llcd_lock
1767 allowed another thread to grab newly-allocated llcd.
1769 Severity : enhancement
1771 Description: 2.6 OST async journal commit and locking fix to improve performance
1772 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1773 2.6 kernels are now async as they already were in 2.4 kernels so
1774 that they can commit concurrently with the network bulk transfer.
1775 For block-allocated files the filter allocation semaphore is held
1776 to avoid filesystem fragmentation during allocation. BKL lock
1777 removed for 2.6 xattr operations where it is no longer needed.
1782 Description: lconf incorrectly determined whether two IP networks could talk
1783 Details : In some more complicated routing and multiple-network
1784 configurations, lconf will avoid trying to make a network
1785 connection to a disjoint part of the IP space. It was doing the
1786 math incorrectly for one set of cases.
1791 Description: Fix for potential infinite loop processing records in an llog.
1792 Details : If an llog record is corrupted/zeroed, it is possible to loop
1793 forever in llog_process(). Validate the llog record length
1794 and skip the remainder of the block on error.
1797 Frequency : occasional (liblustre only)
1799 Description: liblustre could not open files whose last component is a symlink
1800 Details : sysio_path_walk() would incorrectly pass the open intent to
1801 intermediate path components.
1804 Frequency : rare (liblustre only with non-standard tuning)
1805 Bugzilla : 7201 (7350)
1806 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1807 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1808 after it has been released, only do this for referenced locks
1810 Severity : enhancement
1812 Description: specifying an (invalid) directory default stripe_size of -1
1813 would reset the directory default striping
1814 Details : stripe_size -1 was used internally to signal directory stripe
1815 removal, now use "all default" to signal dir stripe removal
1816 as a directory striping of "all default" is not useful
1819 Frequency : common for large clusters running liblustre clients
1821 Description: doing an ls when liblustre clients are running is slow
1822 Details : sending a glimpse AST to a liblustre client waits for every AST
1823 to time out, as liblustre clients will not respond. Since they
1824 cannot cache data we refresh the OST lock LVB from disk instead.
1826 Severity : enhancement
1828 Description: doing an ls at the same time as file IO can be slow
1829 Details : enqueue and other "small" requests can be blocked behind many
1830 large IO requests. Create a new OST IO portal for non-IO
1831 requests so they can be processed faster.
1834 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1836 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1837 created file with mode 000
1838 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1839 allow an ftruncate() as a non-root user to a file with mode 000.
1840 Lustre now respects this flag to disable mode checks when
1841 truncating a file owned by the user
1844 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1847 Description: Revoking locks from clients that went dead or catatonic might take
1849 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1850 cancellation of such locks instant on servers without waiting for
1851 any reply from clients. Clients drops these locks when cancel
1852 notification from server is received without replying.
1855 Frequency : liblustre-only, when liblustre client dies or becomes busy
1857 Description: Doing ls on Linux clients can take a long time with active
1859 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1860 granting such locks to it in the first place if possible. Locks
1861 are taken by proxy on the OST during the read or write and
1862 dropped immediately afterward. Add connect flags handling, do
1863 not grant locks to liblustre clients for glimpse ASTs.
1865 Severity : enhancement
1867 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1868 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1869 sequential reads and behaves badly with applications doing random
1870 reads. Improve it by reading ahead at least read region, and
1871 avoiding excessive large RPC for small reads.
1873 Severity : enhancement
1875 Description: Creating more than 1000 files for a single job may cause a load
1876 imbalance on the OSTs if there are also a large number of OSTs.
1877 Details : qos_prep_create() uses an OST index reseed value that is an
1878 even multiple of the number of available OSTs so that if the
1879 reseed happens in the middle of the object allocation it will
1880 still utilize the OSTs as uniformly as possible.
1885 Description: OST or MDS may oops in ping_evictor_main()
1886 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1887 but doesn't restart at beginning of obd_exports_timed list
1890 Severity : enhancement
1892 Description: improve by-nid export eviction on the MDS and OST
1893 Details : allow multiple exports with the same NID to be evicted at one
1894 time without re-searching the exports list.
1897 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1899 Description: MDS may oops in groups_free()
1900 Details : in rare race conditions a newly allocated group_info struct is
1901 freed again, and this can be NULL. The 2.4 compatibility code
1902 for groups_free() checked for a NULL pointer, but 2.6 did not.
1905 Frequency : common for liblustre clients doing little filesystem IO
1906 Bugzilla : 9352, 7313
1907 Description: server may evict liblustre clients accessing contended locks
1908 Details : if a client is granted a lock or receives a completion AST
1909 with a blocking AST already set it would not reply to the AST
1910 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1913 Frequency : lfs setstripe, only systems with more than 160 OSTs
1915 Description: unable to set striping with a starting offset beyond OST 160
1916 Details : llapi_create_file() incorrectly limited the starting stripe
1917 index to the maximum single-file stripe count.
1920 Frequency : LDAP users only
1922 Description: lconf did not handle in-kernel recovery with LDAP properly
1923 Details : lconf/LustreDB get_refs() is searching the wrong namespace
1925 Severity : enhancement
1927 Description: bind OST threads to NUMA nodes to improve performance
1928 Details : all OST threads are uniformly bound to CPUs on a single NUMA
1929 node and do their allocations there to localize memory access
1931 Severity : enhancement
1933 Description: llmount can determine client NID directly from Myrinet (GM)
1934 Details : the client NID code from gmnalnid was moved directly into
1935 llmount, removing the need to use this or specifying the
1936 client NID explicitly when mounting GM clients with zeroconf
1939 Frequency : if client is started with down MDS
1941 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
1942 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
1943 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
1944 ptlrpc_import_delayed_req() because we didn't check if the
1945 request was interrupted, and we also didn't break out of the
1946 event loop if there was no timeout
1951 Description: data loss during non-page-aligned writes to a single file from
1952 both multiple nodes and multiple threads on one node at same time
1953 Details : updates to KMS and lsm weren't protected by common lock. Resulting
1954 inconsistency led to false short-reads, that were cached and later
1955 used by ->prepare_write() to fill in partially written page,
1956 leading to data loss.
1959 Frequency : always, if lconf --abort_recovery used
1961 Description: lconf --abort_recovery fails with 'Operation not supported'
1962 Details : lconf was attempting to abort recovery on the MDT device and not
1965 Severity : enhancement
1967 Description: remove cleanup logs
1968 Details : replace lconf-generated cleanup logs with lustre internal
1969 cleanup routines. Eliminates the need for client-cleanup and
1972 Severity : enhancement
1974 Description: add support for EAs (user and system) on lustre filesystems
1975 Details : it is now possible to store extended attributes in the Lustre
1976 client filesystem, and with the user_xattr mount option it
1977 is possible to allow users to store EAs on their files also
1979 Severity : enhancement
1981 Description: Add possibility (config option) to show minimal available OST free
1983 Details : When compiled with --enable-mindf configure option, statfs(2)
1984 (and so, df) will return least minimal free space available from
1985 all OSTs as amount of free space on FS, instead of summary of
1986 free spaces of all OSTs.
1988 Severity : enhancement
1990 Description: do not expand extent locks acquired on OST-side
1991 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
1992 by server: they are not cached anyway.
1995 Frequency : when mmap is used/binaries executed from Lustre
1997 Description: Unmmap pages before throwing them away from read cache.
1998 Details : llap_shrink cache now attempts to unmap pages before discarding
1999 them (if unmapping failed - do not discard). SLES9 kernel has
2000 extra checks that trigger if this unmapping is not done first.
2005 Description: lconf didn't resolve symlinks before checking to see whether a
2006 given mountpoint was already in use
2009 Frequency : when migrating failover services
2010 Bugzilla : 6395, 9514
2011 Description: When migrating a subset of services from a node (e.g. failback
2012 from a failover service node) the remaining services would
2013 time out and evict clients.
2014 Details : lconf --force (implied by --failover) sets the global obd_timeout
2015 to 5 seconds in order to quickly disconnect, but this caused
2016 other RPCs to time out too quickly. Do not change the global
2017 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2019 Severity : enhancement
2020 Frequency : if MDS is started with down OST
2021 Bugzilla : 9439,5706
2022 Description: Allow startup/shutdown of an MDS without depending on the
2023 availability of the OSTs.
2024 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2025 Add appropriate locking and lov-osc refcounts for safe
2026 cleaning. Add osc abort_inflight calls in case the
2027 synchronize never started.
2030 Frequency : occasional (Cray XT3 only)
2032 Description: root not authorized to access files in CRAY_PORTALS environment
2033 Details : The client process capabilities were not honoured on the MDS in
2034 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2035 been accessed by an authorized user then root was able to access
2036 the file on the local client also. The root user capabilities
2037 are now allowed on the MDS, as this environment has secure UID.
2040 Frequency : occasional
2042 Description: ldiskfs "too long searching" message happens too often
2043 Details : A debugging message (otherwise harmless) prints too often on
2044 the OST console. This has been reduced to only happen when
2045 there are fragmentation problems on the filesystem.
2050 Description: Division by zero in statfs when all OSCs are inactive
2051 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2052 lov_statfs divides by value returned by lov_get_stripecnt().
2056 Bugzilla : 9489, 3273
2057 Description: First write from each client to each OST was only 4kB in size,
2058 to initialize client writeback cache, which caused sub-optimal
2059 RPCs and poor layout on disk for the first writen file.
2060 Details : Clients now request an initial cache grant at (re)connect time
2061 and so that they can start streaming writes to the cache right
2062 away and always do full-sized RPCs if there is enough data.
2063 If the OST is rebooted the client also re-establishes its grant
2064 so that client cached writes will be honoured under the grant.
2069 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2070 Details : Now I/O RPCs go to different portal number and (presumably) fast
2071 lock requests (and glimses) and other RPCs get their own service
2072 threads pool that should be able to service those RPCs
2075 Severity : enhancement
2077 Description: Ability to exchange lustre version between client and servers and
2078 issue warnings at client side if client is too old. Also for
2079 liblustre clients there is ability to refuse connection of too old
2081 Details : New 'version' field is added to connect data structure that is
2082 filled with version info. That info is later checked by server and
2086 Frequency : rare, liblustre only.
2087 Bugzilla : 9296, 9581
2088 Description: Two simultaneous writes from liblustre at offset within same page
2089 might proceed at the same time overwriting eachother with stale
2091 Details : I/O lock withing llu_file_prwv was released too early, before data
2092 actually was hitting the wire. Extended lock-holding time until
2093 server acknowledges receiving data.
2096 Frequency : extremely rare. Never observed in practice.
2098 Description: avoid generating lustre_handle cookie of 0.
2099 Details : class_handle_hash() generates handle cookies by incrementing
2100 global counter, and can hit 0 occasionaly (this is unlikely, but
2101 not impossible, because initial value of cookie counter is
2102 selected randonly). Value of 0 is used as a sentinel meaning
2103 "unassigned handle" --- avoid it. Also coalesce two critical
2104 sections in this function into one.
2106 Severity : enhancement
2108 Description: allow liblustre clients to delegate truncate locking to OST
2109 Details : To avoid overhead of locking, liblustre client instructs OST to
2110 take extent lock in ost_punch() on client's behalf. New connection
2111 flag is added to handle backward compatibility.
2113 Severity : enhancement
2114 Bugzilla : 4928, 7341, 9758
2115 Description: allow number of OST service threads to be specified
2116 Details : a module parameter allows the number of OST service threads
2117 to be specified via "options ost ost_num_threads={N}" in the
2118 OSS's /etc/modules.conf or /etc/modprobe.conf.
2122 Bugzilla : 6146, 9635, 9895
2123 Description: servers crash with bad pointer in target_handle_connect()
2124 Details : In rare cases when a client is reconnecting it was possible that
2125 the connection request was the last reference for that export.
2126 We would temporarily drop the export reference and get a new
2127 one, but this may have been the last reference and the export
2128 was just destroyed. Get new reference before dropping old one.
2130 Severity : enhancement
2131 Frequency : if client is started with failover MDS
2133 Description: Allow multiple MDS hostnames in the mount command
2134 Details : Try to read the configuration from all specified MDS
2135 hostnames during a client mount in case the "primary"
2138 Severity : enhancement
2140 Description: Stop sending data to evicted clients as soon as possible.
2141 Details : Check if the client we are about to send or are sending data to
2142 was evicted already. (Check is done every second of waiting,
2143 for which l_wait_event interface was extended to allow checking
2144 of exit condition at specified intervals).
2147 Frequency : rare, normally only when NFS exporting is done from client
2149 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2151 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2154 Frequency : always, if srand() is called before liblustre initialization
2156 Description: Liblustre uses system PRNG disturbing its usage by user application
2157 Details : Introduce internal to lustre fast and high-quality PRNG for
2158 lustre usage and make liblustre and some other places in generic
2159 lustre code to use it.
2161 Severity : enhancement
2162 Bugzilla : 9477, 9557, 9870
2163 Description: Verify that the MDS configuration logs are updated when xml is
2164 Details : Check if the .xml configuration logs are newer than the config
2165 logs stored on the MDS and report an error if this is the case.
2166 Request --write-conf, or allow starting with --old_conf.
2168 Severity : enhancement
2170 Description: Handle symlinks in the path when checking if Lustre is mounted.
2171 Details : Resolve intermediate symlinks when checking if a client has
2172 mounted a filesystem to avoid duplicate client mounts.
2177 Description: lconf can hit an error exception but still return success.
2178 Details : The lconf command catches the Command error exception at the top
2179 level script context and will exit with the associated exit
2180 status, but doesn't ensure that this exit status is non-zero.
2185 Description: failure of ptlrpc thread startup can cause oops
2186 Details : Starting a ptlrpc service thread can fail if there are a large
2187 number of threads or the server memory is very fragmented.
2188 Handle this without oopsing.
2191 Frequency : always, only if liblustre and non-default acceptor port was used
2193 Description: liblustre cannot connect to servers with non-default acceptor port
2194 Details : tcpnal_set_default_params() was not called and was therefore
2195 ignoring the environment varaible TCPNAL_PORT, as well as other
2196 TCPNAL_ environment variables
2201 Description: two objects could be created on the same OST for a single file
2202 Details : If an OST is down, in some cases it was possible to create two
2203 objects on a single OST for a single file. No problems other
2204 than potential performance impact and spurious error messages.
2208 Bugzilla : 5681, 9562
2209 Description: Client may oops in ll_unhash_aliases
2210 Details : Client dcache may become inconsistent in race condition.
2211 In some cases "getcwd" can fail if the current directory is
2217 Description: Inode refcounting problems in NFS export code
2218 Details : link_raw functions used to call d_instantiate without obtaining
2219 extra inode reference first.
2223 Bugzilla : 9942, 9903
2224 Description: Referencing freed requests leading to crash, memleaks with NFS.
2225 Details : We used to require that call to ll_revalidate_it was always
2226 followed by ll_lookup_it. Also with revalidate_special() it is
2227 possible to call ll_revalidate_it() twice for the same dentry
2228 even if first occurence returned success. This fix changes semantic
2229 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2230 reference on a request referred from the intent.
2231 ll_intent_release() then releases such a request.
2234 Frequency : rare, normally benchmark loads only
2236 Description: unlinked inodes were kept in memory on the client
2237 Details : If a client is repeatedly creating and unlinking files it
2238 can accumulate a lot of stale inodes in the inode slab cache.
2239 If there is no other client load running this can cause the
2240 client node to run out of memory. Instead flush old inodes
2241 from client cache that have the same inode number as a new inode.
2244 Frequency : SLES9 2.6.5 kernel and long filenames only
2245 Bugzilla : 9969, 10379
2246 Description: utime reports stale NFS file handle
2247 Details : SLES9 uses out-of-dentry names in some cases, which confused
2248 the lustre dentry revalidation. Change it to always use the
2252 Frequency : rare, unless heavy write-truncate concurrency is continuous
2253 Bugzilla : 4180, 6984, 7171, 9963, 9331
2254 Description: OST becomes very slow and/or deadlocked during object unlink
2255 Details : filter_destroy() was holding onto the parent directory lock
2256 while truncating+unlinking objects. For very large objects this
2257 may block other threads for a long time and slow overall OST
2258 responsiveness. It may also be possible to get a lock ordering
2259 deadlock in this case, or run out of journal credits because of
2260 the combined truncate+unlink. Solution is to do object truncate
2261 first in one transaction without parent lock, and then do the
2262 final unlink in a new transaction with the parent lock. This
2263 reduces the lock hold time dramatically.
2266 Frequency : rare, 2.4 kernels only
2268 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2269 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2270 before unmounting their filesystem in order to avoid deadlock.
2271 The kernel_locked() function in 2.4 kernels only checks whether
2272 the kernel lock is held, not whether it is this process that is
2273 holding it as 2.6 kernels do.
2278 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2279 Details : The client ptlrpc code may be trying to reconnect to a down
2280 server before a previous connection attempt has timed out.
2281 Increase the reconnect interval to be longer than the connection
2282 timeout interval to avoid sending duplicate connections to
2286 Frequency : echo_client brw_test command
2288 Description: fix echo_client to work with OST preallocated code
2289 Details : OST preallocation code (5137) didn't take echo_client IO path
2290 into account: echo_client calls filter methods outside of any
2291 OST thread and, hence, there is no per-thread preallocated
2292 pages and buffers to use. Solution: hijack pga pages for IO. As
2293 a byproduct, this avoids unnecessary data copying.
2297 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2298 Description: Client can oops in mdc_commit_close() after open replay
2299 Details : It was possible for the MDS to return an open request with no
2300 transaction number in mds_finish_transno() if the client was
2301 evicted, but without actually returning an error. Clients
2302 would later try to replay that open and may trip an assertion
2303 Simplify the client close codepath, and always return an error
2304 from the MDS in case the open is not successful.
2307 Frequency : rare, 2.6 OSTs only
2309 Description: OST may deadlock under high load on fragmented files
2310 Details : If there was a heavy load and highly-fragmented OST filesystems
2311 it was possible to have all the OST threads deadlock waiting on
2312 allocation of biovecs, because the biovecs were not released
2313 until the entire RPC IO was completed. Instead, release biovecs
2314 as soon as they are complete to ensure forward IO progress.
2316 Severity : enhancement
2318 Description: Support for specifying external journal device at mount
2319 Details : If an OST or MDS device is formatted with an external journal
2320 device, this device major/minor is stored in the ext3 superblock
2321 and may not be valid for failover. Allow detecting and
2322 specifying the external journal at mount time.
2327 Description: Mounting an MDS with pending unlinked files may cause oops
2328 Details : target_finish_recovery() calls mds_postrecov() which returned
2329 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2330 considers this an error and immediately begins cleaning up the
2331 lov, just after starting the mds_lov process
2333 Severity : enhancement
2335 Description: Implement 'lfs df' to report actual free space on per-OST basis
2336 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2337 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2338 usage of objects; '-h' to report in human readable format.
2340 ------------------------------------------------------------------------------
2342 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2349 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2350 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2351 at startup during orphan unlinking. Instead, skip buddy update
2352 before setup, as it will be regenerated after recovery is complete.
2357 Description: If an OST is inactive, its locks might reference stale inodes.
2358 Details : lov_change_cbdata() must iterate over all namespaces, even if
2359 they are inactive to clear inode references from the lock.
2361 Severity : enhancement
2362 Frequency : occasional, if non-standard max_dirty_mb used
2364 Description: Client will block write RPCs if not enough grant
2365 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2366 then the client will block writes while waiting for another RPC
2367 to complete instead of consuming its dirty limit. With change
2368 we get improved performance when max_dirty_mb is small.
2370 Severity : enhancement
2371 Bugzilla : 3389, 6253
2372 Description: Add support for supplementary groups on the MDS.
2373 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2374 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2375 lookups for user supplementary groups into a cache.
2379 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2380 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2381 for liblustre O_EXCL is enforced by MDS.
2385 Description: readdir never returns NULL in liblustre.
2386 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2387 dentry in struct dirent.
2391 Frequency : liblustre only, depends on application IO pattern
2392 Description: liblustre clients evicted if not contacting servers
2393 Details : Don't put liblustre clients into the ping_evictor list, so
2394 they will not be evicted by the pinger ever.
2396 Severity : enhancement
2398 Description: Add ability to evict clients by NID from MDS.
2399 Details : By echoing "nid:$NID" string into
2400 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2401 $NID would be instantly evicted from this MDS and from all active
2402 OSTs connected to it.
2406 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2407 Details : lookup_it_finish() used to query file size from OSTs that was not
2412 Description: service threads change working directory to that of init
2413 Details : Starting lustre service threads may pin the working directory
2414 of the parent thread, making that filesystem busy. Threads
2415 now change to the working directory of init to avoid this.
2419 Frequency : during shutdown only
2420 Description: shutdown with a failed MDS or OST can cause unmount to hang
2421 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2426 Frequency : 2.6 only
2427 Description: chmod/chown may include an extra supplementary group
2428 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2429 supplementary group and if none is specified this is used.
2432 Bugzilla : 5479 (6816)
2434 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2435 Details : If lookup is in progress on a file that is unlinked we might try
2436 to revalidate the inode and fail in revalidate after lookup is
2437 complete and ll_file_open() enqueues the open again but
2438 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2441 Frequency : always, if lconf --abort_recovery used
2443 Description: lconf --abort_recovery fails with 'Operation not supported'
2444 Details : lconf was attempting to abort recovery on the MDT device and not
2447 ------------------------------------------------------------------------------
2449 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2454 Frequency : rare (only unsupported configurations with a node running as an
2456 Bugzilla : 6514, 5137
2457 Description: Mounting a Lustre file system on a node running as an OST could
2459 Details : OSTs now preallocates memory needed to write out data at
2460 startup, instead of when needed, to avoid having to
2461 allocate memory in possibly low memory situations.
2462 Specifically, if the file system is mounted on on OST,
2463 memory pressure could force it to try to write out data,
2464 which it needed to allocate memory to do. Due to the low
2465 memory, it would be unable to do so and the node would
2466 become unresponsive.
2468 Severity : enhancement
2470 Description: Addition of lconf --service command line option
2471 Details : lconf now accepts a '--service <arg>' option, which is
2472 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2474 Severity : enhancement
2476 Description: Failover mode is now the default for OSTs.
2477 Details : By default, OSTs will now run in failover mode. To return to
2478 the old behaviour, add '--failout' to the lmc line for OSTs.
2480 Severity : enhancement
2482 Description: Health checks are now provided for MDS and OSTs
2483 Details : Additional detailed health check information on MSD and OSTs
2484 is now provided through the procfs health_check value.
2487 Frequency : occasional, depends on IO load
2489 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2490 after numerous create/delete cycles
2491 Details : The ext3 inode allocation policy would not allocate new inodes
2492 very well on the OSTs because there are no new directories
2493 being created. Instead we look for groups with free space if
2494 the parent directories are nearly full.
2499 Description: Network or server problems during mount may cause partially
2500 mounted clients instead of returning an error.
2501 Details : The config llog parsing code may overwrite the error return
2502 code during mount error handling, returning success instead
2508 Description: MDS can fail to allocate large reply buffers
2509 Details : After long uptimes the MDS can fail to allocate large reply
2510 buffers (e.g. zconf client mount config records) due to memory
2511 fragmentation or consumption by the buffer cache. Preallocate
2512 some large reply buffers so that these replies can be sent even
2513 under memory pressure.
2517 Frequency : rare (liblustre)
2518 Description: fsx running with liblustre complained that using truncate() to
2519 extend the file doesn't work. This patch corrects that issue.
2520 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2521 ATTR_SIZE and lsm use in llu_setattr_raw.
2525 Frequency : rare, only 2.6 kernels
2526 Description: Unusual file access patterns on the MDS may result in inode
2527 data being lost in very rare circumstances.
2528 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2529 code in the 2.6 kernel caused the inode and/or EA data not to
2530 be read from disk, causing single-file corruption.
2534 Frequency : rare, only 2.6 filesystems using extents
2535 Description: Heavy concurrent write and delete load may cause data corruption.
2536 Details : It was possible under high-load situations to have an extent
2537 metadata block in the block device cache from a just-unlinked
2538 file overwrite a newly-allocated data block. We now unmap any
2539 metadata buffers that alias just-allocated data blocks.
2543 Frequency : filesystems with default stripe_count larger than 77
2544 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2545 Details : lconf specifies an inode size of 4096 bytes when the default
2546 stripe_count is larger than 77. This conflicts with the default
2547 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2548 this case to avoid pinning too much memory for large EAs.
2550 ------------------------------------------------------------------------------
2552 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2557 Frequency : rare (extremely heavy IO load with hundreds of clients)
2559 Description: Client is evicted, gets IO error writing to file
2560 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2561 caused clients to be evicted for AST timeouts. The fixes in
2562 bug 5192 mean we no longer need to have such short AST timeouts
2563 so ldlm_timeout has been increased.
2566 Frequency : occasional during --force or --failover shutdown under load
2567 Bugzilla : 5949, 4834
2568 Description: Server oops/LBUG if stopped with --force or --failover under load
2569 Details : a collection of import/export refcount and cleanup ordering
2570 issues fixed for safer force cleanup
2573 Frequency : only filesystems larger than 120 OSTs
2574 Bugzilla : 5990, 6223
2575 Description: lfs getstripe would oops on a very large filesystem
2576 Details : lov_getconfig used kfree on vmalloc'd memory
2579 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2580 Bugzilla : 6242, 6243
2581 Description: reading from files that had been truncated to a non-zero size
2582 but never opened returned no data
2583 Details : ll_file_read() reads zeros from no-object files to EOF
2588 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2589 Details : There's little harm in aborting MDS/OSS recovery and letting it
2590 try again, so I removed the LASSERT and return an error instead.
2592 Severity : enhancement
2594 Description: New debugging infrastructure for tracking down data corruption
2595 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2596 (b) cover more of the client-side code path, and (c) try to narrow
2597 down where problems occurred
2601 Bugzilla : 3819, 4364, 4397, 6313
2602 Description: Racing close and eviction MDS could cause assertion in mds_close
2603 Details : It was possible to get multiple mfd references during close and
2604 client eviction, leading to one thread referencing a freed mfd.
2606 Severity: : enhancement
2607 Bugzilla : 3262, 6359
2608 Description: Attempts to reconnect to servers are now more aggressive.
2609 Details : This builds on the enhanced upcall-less recovery that was added
2610 in 1.4.2. When trying to reconnect to servers, clients will
2611 now try each server in the failover group every 10 seconds. By
2612 default, clients would previously try one server every 25 seconds.
2617 Description: After recovery, certain operations trigger a failed
2618 assertion on a client.
2619 Details : Failing over an mds, using lconf -d --failover, while a
2620 client was doing a readdir() call would cause the client to
2621 LBUG after recovery completed and the readdir() was resent.
2623 Severity : enhancement
2625 Description: Default groups are now added by lconf
2626 Details : You can now run lconf --group <servicename> without having to
2627 manually add groups with lmc.
2630 Frequency : occasional
2632 Description: Nodes with an elan id of 0 trigger a failed assertion
2635 Frequency : always when accessing e.g. tty/console device nodes
2637 Description: tty and some other devices nodes cannot be used on lustre
2638 Details : file's private_data field is used by device data and lustre
2639 values in there got lost. New field was added to struct file to
2640 store fs-specific private data.
2643 Frequency : when exporting Lustre via NFS
2645 Description: NFSD failed occasionally when looking up a path component
2646 Details : NFSD is looking up ".." which was broken in ext3 directories
2647 that had grown large enough to become hashed.
2650 Frequency : Clusters with multiple interfaces not on the same subnet
2652 Description: Nodes will repeatedly try to reconnect to an interface which it
2653 cannot reach and report an error to the log.
2654 Details : Extra peer list entries will be created by lconf with some peers
2655 unreachable. lconf now validates the peer before adding it.
2658 Frequency : Only if a default stripe is set on the filesystem root.
2660 Description: Setting a default stripe on the filesystem root prevented the
2661 filesystem from being remounted.
2662 Details : The client was sending extra request flags in the root getattr
2663 request and did not allocate a reply buffer for the dir EA.
2666 Frequency : occasional, higher if lots of files are accessed by one client
2667 Bugzilla : 6159, 6097
2668 Description: Client trips assertion regarding lsm mismatch/magic
2669 Details : While revalidating inodes the VFS looks up inodes with ifind()
2670 and in rare cases can find an inode that is being freed.
2671 The ll_test_inode() code will free the lsm during ifind()
2672 when it finds an existing inode and then the VFS later attaches
2673 this free lsm to a new inode.
2677 Bugzilla : 6422, 7030
2678 Description: MDS deadlock between mkdir and client eviction
2679 Details : Creating a new file via mkdir or mknod (starting a transaction
2680 and getting the ns lock) can deadlock with client eviction
2681 (gets ns lock and trying to finish a synchronous transaction).
2684 Frequency : occasional
2685 Description: While starting a server, the fsfilt_ext3 module could not be
2687 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2688 kernels. Previously, lconf would still use the ext3 name
2689 when trying to load modules. Now, it will correctly use
2690 ext3 on 2.4 and ldiskfs on 2.6.
2692 Severity : enhancement
2693 Description: The default stripe count has been changed to 1
2694 Details : The interpretation of the default stripe count (0, to lfs
2695 or lmc) has been changed to mean striping across a single
2696 OST, rather than all available. For general usage we have
2697 found a stripe count of 1 or 2 works best.
2699 Severity : enhancement
2700 Description: Add support for compiling against Cray portals.
2701 Details : Conditional compiling for some areas that are different
2705 Frequency : occasional
2706 Bugzilla : 6409, 6834
2707 Description: Creating files with an explicit stripe count may lead to
2708 a failed assertion on the MDS
2709 Details : If some OSTs are full or unavailable, creating files may
2710 trigger a failed assertion on the MDS. Now, Lustre will
2711 try to use other servers or return an error to the
2715 Frequency : occasional
2717 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2718 caused lock timeout during readahead (since 1.4.2).
2719 Details : Processes doing readahead might match a lock that hasn't been
2720 granted yet if there are overlapping and conflicting lock
2721 requests. The readahead process waits on ungranted lock
2722 (original lock is CBPENDING), while OST waits for that process
2723 to cancel CBPENDING read lock and eventually evicts client.
2725 Severity : enhancement
2727 Description: Initial enabling of flock support for clients
2728 Details : Implements fcntl advisory locking and file status functions.
2729 This feature is provided as an optional mount flag (default
2730 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2731 locking are implemented yet, and those that are are not guaranteed
2732 to be completely correct in production environments.
2733 mount -t lustre -o [flock|noflock] ...
2736 Frequency : occasional
2738 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2739 assertion in the ext3 JBD (journaling) layer.
2740 Details : The b_committed_data struct is protected by the big kernel lock
2741 in 2.4 kernels, serializing journal_commit_transaction() and
2742 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2743 access to this struct. In 2.6 kernels there is finer grained
2744 locking to improve SMP performance of the JBD layer.
2748 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2750 -----------------------------------------------------------------------------
2752 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2754 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2755 versions of Lustre. Please update all servers and clients to
2756 version 1.4.2 or later at the same time. You must also run
2757 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2758 to update the configuration logs.
2760 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2761 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2762 - swab llog records of type '0' so we get proper header size/idx (5861)
2763 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2764 - fix rename of one directory over another leaking an inode (5953)
2765 - avoid SetPageDirty on 2.6 (5981)
2766 - don't re-add just-being-destroyed locks to the waiting list (5653)
2767 - when creating new directories, inherit the parent's custom
2768 striping settings if present parent (3048)
2769 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2770 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2771 - don't hold client locks on temporary worklist from l_lru (5666)
2772 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2773 - automatically evict dead clients (5921)
2774 - Update file size properly in create+truncate+fstat case (6196)
2775 - Do not unhash mountpoint dentries, do not allow removal of
2777 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2778 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2779 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2780 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2781 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2782 - Correctly handle reads of files with no objects (6243)
2783 - lctl recover will also mark a device active if deactivate used (5933)
2785 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2786 - allow --write-conf on an MDS with different nettype than client (5619)
2787 - don't write config llogs to MDS for mounts not from that MDS (5617)
2788 - lconf should create multiple TCP connections from a client (5201)
2789 - init scripts are now turned off by default; run chkconfig --on
2790 lustre and chkconfig --on lustrefs to use them
2791 - upcalls are no longer needed for clients to recover to failover
2793 - add --abort-recovery option to lconf to abort recovery on device
2795 - add support for an arbitrary number of OSTs (3026)
2796 - Quota support protocol changes.
2797 - forward compatibility changes to wire structs (6007)
2798 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2799 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2800 - improved "lustre status" script
2801 - initialize blocksize for non-regular files (6062)
2802 - added --disable-server and --disable-client configure options (5782)
2803 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2804 - Vanilla 2.4.29 support
2805 - increase maximum number of obd devices to 520 (6242)
2806 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2807 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2808 SLES 9 2.6.5 kernels
2809 - limit stripes per file to 160 (the maximum EA size) (6093)
2811 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2814 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2815 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2816 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2817 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2818 - small ext3 extents cleanups and fixes (5733)
2819 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2820 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2821 - avoid needless client->OST connect, fix handle mismatch (5317)
2822 - fix DLM error path that led to out-of-sync client, long delays (5779)
2823 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2824 - fix several locking issues related to i_size (5492,5624,5654,5672)
2825 - don't move pending lock onto export if it is already evicted (5683)
2826 - fix kernel oops when creating .foo in unlinked directory (5548)
2827 - fix deadlock in obdfilter statistics vs. object create (5811)
2828 - use time_{before,after} to avoid timer jiffies wrap (5882)
2829 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2830 - Do not leak request if server was not able to process it (5154)
2831 - If mds_open unable to find parent dir, make that negative lookup(5154)
2832 - don't create new directories with extent-mapping (5909, 5936)
2834 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2835 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2836 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2837 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2838 - support for RHEL4 kernel on i686 (5773)
2839 - provide error messages when incompatible logs are encountered (5898)
2841 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2842 * version 1.4.0.10 (1.4.1 release candidate 1)
2844 - don't keep a lock reference when lock is not granted (4238)
2845 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2846 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2847 - drop import inflight refcount on signal_completed_replay error (5255)
2848 - unlock page after async write error during send (3677)
2849 - handle missing objects in filter_preprw_read properly (5265)
2850 - no transno return for symlink open, don't save no-trasno open (3440)
2851 - don't try to complete elan receive that already failed (4012)
2852 - free RPC server reply state on error (5406)
2853 - clean up thread from ptlrpc_start_thread() on error (5160)
2854 - readahead could read extra page into cache that wasn't ejected (5388)
2855 - prevent races in class_attach/setup/cleanup/detach (5260)
2856 - don't dereference de->d_inode after l_dput of de (5458)
2857 - use "int" for stripe value returned from lock_to_stripe (5544)
2858 - mballoc allocation and error-checking fixes in 2.6 (5504)
2859 - block device patches to fix I/O request sizes in 2.6 (5482)
2860 - look up hostnames for IB nals (5602)
2861 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2862 - don't start multiple acceptors for the same port (5277)
2863 - fix incorrect LASSERT in mds_getattr_name (5635)
2864 - export a proc file for general "ping" checking (5628)
2865 - fix "lfs check" to not block when the MDS is down (5628)
2867 - service request history (4965)
2868 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2869 - create an "evict_client" /proc entry on OSTs, like the MDS has
2870 - fix mount usage message, return errors per mount(8) (5168)
2871 - change grep [] to grep "[]" in tests so they work in more UMLs
2872 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2873 - remove ancient LOV_MAGIC_V0 EA support (5047)
2874 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2875 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2876 - allow readahead allocations to fail when low on memory (5383)
2877 - mmap locking landed again, after considerable improvement (2828)
2878 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2880 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2883 - send OST transaction number in read/write reply to free req (4966)
2884 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2885 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2886 - the watchdog thread now runs as interruptible (5246)
2887 - flock/lockf fixes (but it's still disabled, pending 5135)
2888 - don't use EXT3 constants in llite code (5094)
2889 - memory shortage at startup could cause assertion (5176)
2891 - reorganization of lov code
2892 - single portals codebase
2894 - add extents/mballoc support (5025)
2895 - direct I/O reads in the obdfilter (4048)
2896 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2898 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2901 - send OST transaction number in read/write reply to free req (4966)
2902 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2903 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2904 - the watchdog thread now runs as interruptible (5246)
2905 - handle missing objects in filter_preprw_read properly (5265)
2906 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2907 - drop import inflight refcount on signal_completed_replay error (5255)
2908 - unlock page after async write error during send (3677)
2909 - return original error code on reconstructed replies (3761)
2910 - no transno return for symlink open, don't save no-trasno open (3440)
2912 - add pid to ldlm debugging output (4922)
2913 - bump the watchdog timeouts -- we can't handle 30sec yet
2914 - extra debugging for orphan dentry/inode bug (5259)
2916 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
2919 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
2920 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
2921 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
2922 - fix NULL dereference in /proc/sys/portals/routes (4827)
2923 - allow failed mdc_close() operations to be interrupted (4561)
2924 - stop precreate on OST before MDS would time out on it (4778)
2925 - don't send partial-page writes before EOF from client (4410)
2926 - discard client grant for sub-page writes on large-page clients (4520)
2927 - don't free dentries not owned by NFS code, check generation (4806)
2928 - fix lsm leak if mds_create_objects() fails (4801)
2929 - limit debug_daemon file size, always print CERROR messages (4789)
2930 - use transno after validating reply (3892)
2931 - process timed out requests if import state changes (3754)
2932 - update mtime on OST during writes, return in glimpse (4829)
2933 - add mkfsoptions to LDAP (4679)
2934 - use ->max_readahead method instead of zapping global ra (5039)
2935 - don't interrupt __l_wait_event() during strace
2937 - add software watchdogs to catch hung threads quickly (4941)
2938 - make lustrefs init script start after nfs is mounted
2939 - fix CWARN/ERROR duplication (4930)
2940 - return async write errors to application if possible (2248)
2941 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
2942 - print NAL number in %x format (4645)
2943 - update barely-supported suse-2.4.21-171 series (4842)
2944 - support for sles 9 %post scripts
2945 - support for building 2.6 kernel-source packages
2946 - support for sles km_* packages
2948 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
2951 - ignore -ENOENT errors in osc_destroy (3639)
2952 - notify osc create thread that OSC is being cleaned up (4600)
2953 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
2954 - reconstruct ost_handle() like mds_handle() (4657)
2955 - create a new thread to do import eviction to avoid deadlock (3969)
2956 - let lconf resolve symlinked-to devices (4629)
2957 - don't unlink "objects" from directory with default EA (4554)
2958 - hold socknal file ref over connect in case target is down (4394)
2959 - allow more than 32000 subdirectories in a single directory (3244)
2960 - fix blocks count for O_DIRECT writes (3751)
2961 - OST returns ENOSPC from object create when no space left (4539)
2962 - don't send truncate RPC if file size isn't changing (4410)
2963 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
2964 - bind to privileged port in socknal and tcpnal (3689)
2966 - rate limit CERROR/CWARN console message to avoid overload (4519)
2967 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
2968 - basic mmap support (3918)
2969 - kernel patch series update from b1_4 (4711)
2971 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
2974 - avoid crash during MDS cleanup with OST shut down (2775)
2975 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
2976 - don't use bad inodes on the MDS (3744)
2977 - dynamic object preallocation to improve recovery speed (4236)
2978 - don't hold spinlock over lock dumping or change debug flags (4401)
2979 - don't zero obd_dev when it is force cleaned (3651)
2980 - print grants to console if they go negative (4431)
2981 - "lctl deactivate" will stop automatic recovery attempts (3406)
2982 - look for existing locks in ldlm_handle_enqueue() (3764)
2983 - don't resolve lock handle twice in recovery avoiding race (4401)
2984 - revalidate should check working dir is a directory (4134)
2986 - don't always mark "slow" obdfilter messages as errors (4418)
2988 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
2991 - don't close LustreDB during write_conf until it is done (3860)
2992 - fix typo in lconf for_each_profile (3821)
2993 - allow dumping logs from multiple threads at one time (3820)
2994 - don't allow multiple threads in OSC recovery (3812)
2995 - fix debug_size parameters (3864)
2996 - fix mds_postrecov to initialize import for llog ctxt (3121)
2997 - replace config semaphore with spinlock (3306)
2998 - be sure to send a reply for a CANCEL rpc with bad export (3863)
2999 - don't allow enqueue to complete on a destroyed export (3822)
3000 - down write_lock before checking llog header bitmap (3825)
3001 - recover from lock replay timeout (3764)
3002 - up llog sem before sending rpc (3652)
3003 - reduce ns lock hold times when setting kms (3267)
3004 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3005 - fix NULL deref and obd_dev leak on setup error (3312)
3006 - replace some LBUG about llog ops with error handling (3841)
3007 - don't match INVALID dentries from d_lookup and spin (3784)
3008 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3009 - fix invalid assertion in ptlrpc_set_wait (3880)
3011 - add libwrap support for the TCP acceptor (3996)
3012 - add /proc/sys/portals/routes for non-root route listing (3994)
3013 - allow setting MDS UUID in .xml (2580)
3014 - print the stack of a process that LBUGs (4228)
3016 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3019 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3020 - make sure to unset replay flag from failed open requests (3440)
3021 - if default stripe count is 0, use OST count for inode size (3636)
3022 - update parent mtime/ctime on client for create/unlink (2611)
3023 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3024 - free recovery state on server during a forced cleanup (3571)
3025 - unregister_reply for resent reqs (3063)
3026 - loop back devices mounting and status check on 2.6 (3563)
3027 - fix resource-creation race that can provoke i_size == 0 (3513)
3028 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3029 - more debugging for page-accounting assertion (3746)
3030 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3031 - avoid deadlock after precreation failure (3758)
3032 - fix race and lock order deadlock in orphan handling (3450, 3750)
3033 - add validity checks when grabbing inodes from l_ast_data (3599)
3035 - add /proc/.../recovery_status to obdfilter (3428)
3036 - lightweight CDEBUG infrastructure, debug daemon (3668)
3037 - change default OSC RPC parameters to be better on small clusters
3038 - turn off OST read cache for files smaller than 32MB
3039 - install man pages and include them in rpms (3100)
3040 - add new init script for (un)mounting lustre filesystems (2593)
3041 - run chkconfig in %post for init scripts (3701)
3042 - drop scimac NAL (unmaintained)
3044 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3047 - clean kiobufs before and after use (3485)
3048 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3049 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3050 - change init script to fail more clearly if not run as root (1528)
3051 - allow clients to reconnect during replay (1742)
3052 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3053 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3054 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3055 - fix cond_resched (3554)
3056 - only evict unfinished clients after recovery (3515)
3057 - allow bulk resend, prevent data loss (3570)
3058 - dynamic ptlrpc request buffer allocation (2102)
3059 - don't allow unlinking open directory if it isn't empty (2904)
3060 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3061 - remove extraneous obd dereference causing LASSERT failure (3334)
3062 - don't use get_cycles() when creating temp. files on the mds (3156)
3063 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3064 - handle EEXIST for set-stripe, set proper directory name (3336)
3066 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3067 - fix ksocknal_fmb_callback() error messages (2918)
3069 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3072 - don't copy lvb into (possibly NULL) reply on error (2983)
3073 - don't deref dentry after dput, don't free lvb on error (2922)
3074 - use the kms to determine writeback rpc length (2947)
3075 - increment oti_logcookies when osc is inactive (2948)
3076 - update client's i_blocks count via lvb messages (2543)
3077 - handle intent open/close of special files properly (1557)
3078 - mount MDS with errors=remount-ro, like obdfilter (2009)
3079 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3080 - don't use cancelling-locks' kms values (2947)
3081 - use highest lock extent for kms, not last one (2925)
3082 - don't dereference ERR_PTR() dentry in error handling path (3107)
3083 - fix thread race in portals_debug_dumplog() (3122)
3084 - create lprocfs device entries at setup instead of at attach (1519)
3085 - common AST error handler, don't evict client on completion race (3145)
3086 - zero nameidata in detach_mnt in 2.6 (3118)
3087 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3088 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3089 - initialize RPC timeout timer earlier for 2.6 (3219)
3090 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3091 - print nal/nid for unknown nid (3258)
3092 - additional checks for oscc recovery before doing precreate (3284)
3093 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3094 - don't crash in mdc_close for bad permissions on open (3285)
3095 - zero i_rdev for non-device files (3147)
3096 - clear page->private before handing to FS, better assertion (3119)
3097 - tune the read pipeline (3236)
3098 - fix incorrect decref of invalidated dentry (2350)
3099 - provide read-ahead stats and refine rpc in flight stats (3328)
3100 - don't hold journal transaction open across create RPC (3313)
3101 - update atime on MDS at close time (3265)
3102 - close LDAP connection when recovering to avoid server load (3315)
3103 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3104 - don't leak open file on MDS after open resend (3325)
3105 - serialize filter_precreate and filter_destroy_precreated (3329)
3106 - loop device shouldn't call sync_dev() for nul device (3092)
3107 - clear page cache after eviction (2766)
3108 - resynchronize MDS->OST in background (2824)
3109 - refuse to mount the same filesystem twice on same mountpoint (3394)
3110 - allow llmount to create routes for mounting behind routers (3320)
3111 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3112 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3113 - fix rare problem with rename on htree directories (3417)
3115 - allow default OST striping configuration per directory (1414)
3116 - fix compilation for qswnal for 2.6 kernels (3125)
3117 - increase maximum number of MDS request buffers for large systems
3118 - change liblustreapi to be useful for external progs like lfsck (3098)
3119 - increase local configuration timeout for slow disks (3353)
3120 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3122 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3125 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3126 - don't overwrite extent policy data in reply if lock was blocked (2901)
3127 - drop filter export grants atomically with removal from device (2663)
3128 - del obd_self_export from work_list in class_disconnect_exports (2908)
3129 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3130 - swab reply message in mdc_close, other PPC fixes (2464)
3131 - fix destroying of named logs (2325)
3132 - overwrite old logs when running lconf --write_conf (2264)
3133 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3134 - fix race in target_handle_connect (2898)
3135 - mds_reint_create() should take same inode create lock (2926)
3136 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3137 - don't close files for self_export to avoid uninitialized obd (2936)
3138 - allow MDS with the same name as client node (2939)
3139 - hold dentry reference for closed log files for unlink (2325)
3140 - reserve space for all logs during transactions (2059)
3141 - don't evict page beyond end of stripe extent (2925)
3142 - don't oops on a deleted current working directory (2399)
3143 - handle hard links to targets without a parent properly (2517)
3144 - don't dereference NULL lock when racing during eviction (2867)
3145 - don't grow lock extents when lots of conflicting locks (2919)
3147 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3150 - account for cache space usage on clients to avoid data loss (974)
3151 - lfsck support in lustre kernel code (2349)
3152 - reduce journal credits needed for BRW writes (2370)
3153 - orphan handling to avoid losing space on client/server crashes
3154 - ptlrpcd can be blocked, stopping ALL progress (2477)
3155 - use lock value blocks to assist in proper KMS, faster stat (1021)
3156 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3157 - recovery for initial connections (2355)
3158 - fixes for mds_cleanup_orphans (1934)
3159 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3160 - block all file creations until orphan recovery completes (1901)
3161 - client remove rq_connection from request struct (2423)
3162 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3163 - recovery timer race (2670)
3164 - mdc_close recovey bug (2532)
3165 - ptlrpc cleanup bug (2710)
3166 - mds timeout on local locks (2588)
3167 - namespace lock held during RPCs (2431)
3168 - handle interrupted sync write properly (2503)
3169 - don't try to handle a message that hasn't been replied to (2699)
3170 - client assert failure during cleanup after abort recovery (2701)
3171 - leak mdc device after failed mount (2712)
3172 - ptlrpc_check_set allows timedout requests to complete (2714)
3173 - wait for inflight reqs when ptlrpcd finishes (2710)
3174 - make sure unregistered services are removed from the srv_list
3175 - reset bulk XID's when resending them (caught by 1138 test)
3176 - unregister_bulk after timeout
3177 - fix lconf error (2694)
3178 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3179 - readahead locks pages, leaves pending causing memory pressure (2673)
3180 - increase OST request buffers to 4096 on large machines (2729)
3181 - fix up permission of existing directories in simple_mkdir (2661)
3182 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3183 - don't assign transno to errored transactions (2742)
3184 - don't delete objects on OST if given a bogus objid from MDS (2751)
3185 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3186 - if rq_no_resend, then timeout request after recovery (2432)
3187 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3188 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3189 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3190 - MDS recovery shouldn't delete objects that it has given out (2730)
3191 - if enqueue arrives after completion, don't clobber LVB (2819)
3192 - don't unlock pages twice when trigger_group_io returns error (2814)
3193 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3194 - don't write pages to disk if there was an error (1450)
3195 - don't ping imports that have recovery disabled (2676)
3196 - take buffered bytes into account when balancing socknal conn (2817)
3197 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3198 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3199 - remove little-endian swabbing of llog records (1987)
3200 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3201 - retry reposting request buffers if they fail (1191)
3202 - grow extent at grant time to avoid granting a revoked lock (2809)
3203 - lock revoke doesn't evict page if covered by a second lock (2765)
3204 - disable VM readahead to avoid reading outside lock extents (2805)
3206 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3207 - updated kernel patches for hp-2.4.20 kernel (2681)
3209 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3212 - fix truncated write corruption (2366)
3213 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3215 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3216 - ensure len(uuid) < 37 in lmc (1171)
3217 - fix ia64 OOPS in llog_test (2255)
3218 - zero end of page at obdfilter for partial page writes (2648)
3219 - don't leave stale dentries around after renames (bug 2428)
3220 - fix timeouts when evicting a client with a single lock held (2642)
3221 - set deadline for the initial HELLO message to drain (2634)
3222 - print out dotted-quad IP addresses in the socknal (2302)
3224 - additional debugging for MDS client eviction problem (2443)
3225 - fix mkfsoptions support for osts (2603, 2604)
3227 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3230 - add series for the vanilla 2.6.0 kernel
3231 - add series for the vanilla 2.4.24 kernel
3232 - add series for a cray x86/64 UL kernel drop
3233 - fix xattr patches for the vanilla 2.4.19 series
3235 - generate true UUIDs in lmc (1171)
3236 - have portals stack dumping break in UML (2466)
3237 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3238 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3239 - wake callback waiting threads on client eviction (2460)
3240 - Add --ptldebug and --subsystem to lmc (1719)
3241 - update assertion to allow safe interrupt allocation
3242 - set rq_no_resend for cancel requests (2432)
3243 - recalculate ptlrpcd timeout after resend (2494)
3244 - call vfs_rmdir when removing pending directories (2368)
3245 - fix renaming a file to itself (2429)
3246 - lmc creates a default one-stripe lov (2454)
3247 - expand procfs space to handle large clusters (2326)
3248 - increase UML stack to avoid overflow
3249 - update lconf's list of debug and subsystem masks
3250 - fix lfs find --obd (2510)
3251 - /proc tunable for disabling filter read caching (2591)
3252 - stop rpm packages from altering slapd.conf (2301)
3253 - disable nagle in the socknal under 0conf (2578)
3254 - choose mds inode size based on stripe count (2572)
3255 - fix kernel-source rpm problems (2516)
3257 - add --disable-doc to avoid pdf generation (2421)
3258 - update documentation, tests, type-os, comments
3259 - avoid format warnings on ia64
3260 - remove the TOE NAL
3261 - tiny code cleanups by removing unused fields
3263 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3266 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3267 - remove the most blatant lies from BUILDING file (2371)
3268 - change default debug level to reasonable production setting
3269 - reduce client side cache size to reduce cache flush time
3270 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3271 - make TCP zerocopy and pinger support enabled by default (2476)
3272 - sync writes completed after process exits caused crashes (2319)
3273 - maintain correct mount count on the MDS (2356)
3274 - backout 1557, because 2316 wasn't really fixed
3275 - better file I/O statistics gathering in /proc
3276 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3277 - another kernel patch to fix zero-copy TCP function export
3278 - don't take duplicate lock when processing re-sent getattr (2420)
3279 - lctl uses obd_self_export instead of creating new conn (2353)
3280 - MDS/OST recovery case which requires object creation asserted (2425)
3281 - move lfs from /usr/sbin to /usr/bin in packages
3282 - fix race between mds_client_add and mds_client_free (2417)
3283 - use kmalloc instead of slabs in portals (2430)
3284 - don't create duplicate records when a failover MDS is present (2442)
3285 - remove unnecessary mount age check (2332)
3286 - don't remove directory inodes from locks prematurely (2451)
3287 - don't break if MDS service name is the same as hostname (2103)
3288 - fix races in client write RPC generation when cache full (2482)
3290 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3293 - remove now-unused request->rq_obd (278)
3294 - if an allocation fails, print out how much memory we've used (1933)
3295 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3296 - add the "configurable stack size" patch to most series files (1256)
3297 - ability to write large log records, for 100+ OST configs (2306)
3298 - fix NULL deref when filter_prep fails (2314)
3299 - fix operator precedence error in filter_sync
3300 - dynamic allocation of socknal TX descriptors (2315)
3301 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3302 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3303 - ensure that CWARN messages in Portals always get to the syslog
3304 - __init/__exit are not for prototype decls (ldlm_init/exit)
3305 - x86-64 compile warning fixes
3306 - fix gateway LMC keyword conflict (2318)
3307 - fix MDS lock inversions in getattr/reint paths (1844)
3308 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3309 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3310 - DLM race condition prevented some lock evictions (2328)
3311 - ENOMEM detection and retry on socknal sends (2230)
3312 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3313 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3314 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3315 - better error messages when a client is rejected during recovery (1505)
3316 - avoid cancelling locks which were never granted, after failure (2330)
3317 - fix i_sem/journal inversion in mds_client_add (2333)
3318 - fix truncate/getattr lock cycle deadlock (2334)
3319 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3320 - fix two rare exit paths which could leak an l_lock() ref (2321)
3321 - fencepost error in MDS/OST orphan recovery (2226)
3322 - make log record alignment 8 bytes (1988)
3323 - lstripe now fails when requested offset > ost_count (2237)
3324 - ensure that all kernel series have a complete list.h (1607)
3325 - fix crashes in special-file operations (2316)
3326 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3328 - allow configurable automake binary, for testing new versions
3329 - small update to the lfs documentation
3331 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3333 * fix negative export reference count in fsfilt_sync (2312)
3335 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3336 * release candidate 0.9.1
3338 - orphans are moved into the PENDING directory for possible recovery
3339 - replayed opens now open by fid for orphan/rename safety (1042)
3340 - last close of an orphan inode generates a transno (683)
3341 - chdir() and mount() now pin the directory entry (1020)
3342 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3343 - discard very old requests without processing them (1502)
3344 - remove some common, well-understood CERRORs (1505)
3345 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3346 - clear "grant" flags in OST replies until OST grant code lands (1644)
3347 - fix read performance by not clobbering i_blksize on client (1598)
3348 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3349 - make LRU size a /proc tunable, clears locks when reduced (707)
3350 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3351 - server locks take references on exports now (1558)
3352 - build fixes for 2.4.20-rh trees (1663)
3353 - return an error from lov_create if all OSCs are inactive (1751)
3354 - fix import levels when a reconnect happens without a timeout (1597)
3355 - exit early from mds_open if we get a lookup error (1749)
3356 - partial page read at EOF wouldn't wait for disk before sending (1642)
3357 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3358 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3359 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3360 - write/truncate lock inversion (1639)
3361 - Don't auto-load obdclass, portals modules during cleanup (1495)
3362 - fix timestamps from jumping to "now" (1763)
3363 - extra journal assertions (1648)
3364 - add an extra multiunlink test (1771)
3365 - fix read_record/write_record API (1776)
3366 - fix leak of offset_extent, possible incorrect i_size later (1772)
3367 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3368 - Debugging for the kqswnal_get_idle_tx problems (1820)
3369 - Allow recovery to be attempted multiple times (1536)
3370 - Write out MDS last_rcvd file after it is first created (1600)
3371 - Fix tx_descriptor leak in failed transmit situations (1827)
3372 - ext3 journaling fixes for assertion failure after IO error (1871)
3373 - class_export_put() on freed export after completion AST error (1896)
3374 - Fix revalidate looping in VFS (1322)
3375 - Don't access a freed export during MDS_REINT timeout (1521)
3376 - Add open-unlink recovery support on the MDS (1673,1764)
3377 - Return an error if no MDS data was read from last_rcvd (1946)
3378 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3379 - Don't setup a disk device that doesn't match exported UUID (317)
3380 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3381 - avoid committing NULL handle in force close
3382 - local.sh is now a one-stripe LOV configuration
3383 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3384 - fix ext3 htree duplicate directory entry corruption (1516)
3385 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3386 - update to new LOV EA format (2097)
3387 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3389 2003-06-15 Phil Schwan <phil@clusterfs.com>
3392 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3393 - per-import/export recovery handling (958, 931, 959)
3394 - multiple last-rcvd slots, for serving multiple FSes (949)
3395 - connections are again shared between multiple imp/exports (963, 964)
3396 - "umount -f" would hang if any requests needed to be sent (393, 978)
3397 - avoid pinning large req buffer by copying for queued messages (989)
3398 - add "uuid" to "lctl device" command to help upcalls (991)
3399 - "open" RPCs with transnos would confuse recovery counters (1037)
3400 - do proper endian conversion of all wire messages (288, 340, 891)
3401 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3402 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3403 - fix ptlrpc_connection leak in target_handle_connect (1174)
3404 - fix import refcounting bug in OST and MDS cleanup (1134)
3405 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3406 - fix very unlikely obd_types race condition (501)
3407 - remove osc_open hack for echo_client (1187)
3408 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3409 - a failure in read_inode2 leads to deadlock (1139)
3410 - cancel ack-locks as soon as transaction is committed (1072)
3411 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3412 - make sure to commitrw after any preprw to avoid deadlock (1162)
3413 - failing to execute a file in a lustre FS would lock inode (1203)
3414 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3415 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3416 - obd_uuid2tgt would walk off the end of the list (1255)
3417 - on IA64 the osc would give portals incorrect bulk size (1258)
3418 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3419 - fix lock inversion caused by new llite matching code (1282)
3420 - limit the number of dirty pages on a client to 10MB (1286)
3421 - timed out locks were not being corrected cancelled (1289)
3422 - fix O_DIRECT above 4GB on IA-32 (1292)
3423 * major user-visible changes
3424 - fail out/fail over policy now controlled by the upcall (993)
3426 - add OBD_PING to check server availability and failure (954)
3427 - lustre messages are now sent in sending host order (288, 340, 891)
3428 - add eadatalen to MDS getattr reply (340)
3429 - OST read replies may contain second buffer, with per-page status (593)
3431 2003-03-11 Phil Schwan <phil@clusterfs.com>
3434 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3435 - failed open()s could cause deadlock; fixed (867, 869)
3436 - stop cancelling OST locks when files are closed (481)
3437 - overlapping XID spaces caused network corruption (851, 853)
3438 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3439 - setattr_raw added, to do single-RPC, server-side setattrs
3440 - lmc/lconf syntax change for OST UUIDs
3441 - fix crashy race condition between ptlrpc_free_req and osc_close
3442 - don't use request in mdc_enqueue if we hit a timeout (889)
3443 - don't set the inode i_size for regular files from the MDS (896)
3444 - handle out of order completion AST (842)
3445 - don't LBUG if a lock request times out after receiving AST (913)
3446 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3447 - if a bad lock AST arrives, send an error instead of dropping entirely
3448 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3449 - fix leak in bulk IO when only partially completed (899, 900, 926)
3450 - fix O_DIRECT for ia64 (55)
3451 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3452 - C-z after timeout could hang a process forever; fixed (977)
3454 - client-side I/O cache (678, 924, 929, 941, 970)
3456 - READPAGE and SETATTRs which don't take server-side locks get
3459 2003-02-11 Phil Schwan <phil@clusterfs.com>
3462 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3463 - Fix fsfilt_extN_readpage() to read a full page of directory
3464 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3465 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3466 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3467 - Most intent operations were removed, in favour of a new RPC mode
3468 that does a single RPC to the server and bypasses most of the VFS
3469 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3470 - Aggressively cancel local locks on DLM servers
3471 - mds_reint_unlink sends EA to the client if it's the last nlink.
3472 client uses that EA to unlink OST objects.
3473 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3474 - recursive symlinks were fixed (439)
3475 - fixed NULL deref in DEBUG_REQ
3476 - filter_update_lastobjid no longer calls sync, which annoyed extN
3477 - fixed multi-client small-writes to a single file problem (445)
3478 - fixed mtime updates during file writes (607)
3479 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3480 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3481 - fixed Linux OST size reporting problem (444, 656)
3482 - OST now updates object mtime with writes or setattr (607, 619)
3483 - client verifies file size before zeroing page past EOF (445)
3484 - OST now writes last allocated objid to disk with allocation (108)
3485 - LOV on echo now works (409)
3487 - mds_reint_unlink sends a new buffer, with the EA included. this
3488 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3489 set if a regular file was being unlinked, and it was the last link
3490 - use PtlGet from the target for bulk writes (315)
3491 - OST now updates object mtime with writes or setattr (607, 619)
3492 - LDLM now has a grant-time callback to revalidate locked items, if
3494 - Many MDS operations were reorganized to combat race conditions
3496 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3497 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3499 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3502 - Fully reactivate OST imports after reconnection (512, others)
3503 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3504 - More graceful error handling for truncating on dead OST (515)
3505 - Don't error out unless we're actually accessing dead stripes (474)
3506 - Fix garbage sizes when stripes are missing (410)
3507 - LRU counters were broken, causing constant lock purge (433, 432)
3508 - garbage on read from stripes with failed OSTs (441)
3509 - mark OSCs as active before reconnecting during recovery (438)
3510 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3511 - lfind did not preserve OST order in output (443)
3512 - symlinks cause hung clients, incorrect data (439)
3513 - stop dereferencing request after dropping refcount (457)
3514 - don't LASSERT(spin_is_locked) on non-SMP (455)
3515 - fixes for many rename() bugs
3516 - fstat didn't correctly synchronize attributes (399)
3517 - server must handle lock cancellation during blocking AST prep (487)
3518 - bulk descriptors were free()d too soon (511)
3519 - fix paths in lconf, which would load incorrect modules (451, 507)
3520 - fix confusing lconf 'host not found' error message (386)
3521 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3522 - fix race condition in mdc_blocking_ast() for inode access (526)
3523 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3524 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3525 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3526 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3527 - fix lstripe in several ways (kernel side) (527)
3528 - fix request leak in ldlm_cli_enqueue (262)
3529 - incorrect OSC was marked inactive after OST failure
3530 - call mds_fs_cleanup before unmounting filesystem (524)
3531 - fix races between taking ns_lock and ldlm_lock_change_resource
3532 - fix races updating LOV export open file list
3533 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3534 - fix recovery NULL deref in ldlm_cli_cancel_unused
3535 - fix some DLM races by using new hash table for lock handles (419)
3536 - permit the client to specify desired inodes, at replay
3537 - duplicate requests when we queue them for replay reintegration
3538 - fix last_rcvd offset calculation
3539 - sync after each recovered transaction, so we always make progress
3540 - never, not always, ERESTART requests without transnos
3541 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3543 - skip replay if the MDS says that the client is already connected
3544 - don't check for a recovery-enabled export to match lctl's UUID
3545 - don't INC_USE_COUNT for phantom exports
3546 - don't crash when cleaning up phantom exports (567)
3547 - don't double-finish or set replay data for errored mdc_open requests
3548 - abort requests when they time out, so we don't get old replies
3549 - send/receive replies for AST messages again
3550 - if the client says that it doesn't have the lock, cancel it on the
3552 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3553 mark it as destroyed, it will all work out when decref is called
3554 - fix module use counts (22, 581)
3556 - ASTs now expect a reply (server cancels lock on error reply)
3558 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3561 - fix many simultaneous client startup (392)
3562 - fix dentry->d_it clobbering
3563 - credentials weren't being shipped for readdir/getattr operations
3564 - remove invalid assertions triggered during some concurrent MD
3566 - proper Lustre versions added (336, 389)
3567 - fix memory leak for create error case (398)
3568 - fix LOV locking bug that would get cli/srv out of sync
3569 - fix echo client over LOV (409)
3570 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3571 - fix double-O_EXCL intent crash (424)
3572 - avoid sending multiple lock CANCELs (352)
3574 - MDS can do multi-client recovery (modulo bugs in new code)
3576 - many updates, edits, cleanups
3578 2002-11-18 Phil Schwan <phil@clusterfs.com>
3581 - fix null d_it dereference (346)
3582 - fix full OST/dbench hang (333)
3583 - fix permission problem with file removal (286)
3584 - fix removal of OSCs from LOV when they fail
3585 - fix NULL deref during bulk timeout (214)
3586 - fix problems related to multiple filesystems on one MDS (241)
3587 - fixed serious subtle metadata locking bugs
3588 - free locks on clients when inodes are removed due to memory
3590 - fix inode pointer in lock data (285)
3591 - partial support for multiple MDS on a single host (241)
3592 - data locks weren't cancelled at clear_inode time (290, 311)
3593 - intent locks could lead to unbounded lock growth (205)
3594 - added a maximum lock count, an LRU list, and a flusher
3595 - fix multiple rename (365)
3596 - properly abstracted the echo client
3597 - OSC locked 1 byte too many; fixed
3598 - rewrote brw callback code:
3599 - fixed recovery bugs related to LOVs (306)
3600 - fixed too-many-pages-in-one-write crash (191)
3601 - fixed (again) crash in sync_io_timeout (214)
3602 - probably fixed callback-related race (385)
3604 - Add capability to MDS protocol
3605 - LDLM cancellations and callbacks on different portals
3607 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3610 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3611 - defer open object destruction to close time (601981, 138)
3612 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3613 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3614 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3615 - don't call obd_setattr in ll_file_release for destroyed objects
3616 * protocol change to lustre_msg: move |version| and add |flags|
3617 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3618 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3619 * added replay of create, unlink, link and rename operations during
3620 MDS failover; recovery should be much more robust now
3621 * remove failed OSCs from LOVs (only lov_create uses this so far)
3622 * the lustre-HOWTO was brought (more) up to date (582544)
3624 2002-10-23 Phil Schwan <phil@clusterfs.com>
3627 - in-use dentries weren't being reused properly (617851)
3628 - prevent multiple LDLM setup (599178)
3629 - fix LOV size calculations for truncate (617853)
3630 - fix client handling of MDS intent errors (POSIX)
3631 - fix permission bug in lovstripe.c test (624321)
3632 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3633 - truncate past end of file could corrupt data
3634 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3635 - a race in recovery could return ETIMEDOUT to apps (623947)
3636 - building outside the source directory was fixed
3637 * the lustre-HOWTO was brought (more) up to date (582544)
3638 * major progress was made on recovery functionality
3640 2002-10-10 Phil Schwan <phil@clusterfs.com>
3643 - recovery deadlock fix
3644 - rm -rf causes LBUG fix (617817)
3645 - file open by multiple tasks fix (618962)
3646 - directory permissions bugs (602707 and 620007)
3647 - journal_stop fixed with locking (611313)
3648 - O_APPEND failures resolved (618273, perhaps 614459)
3649 - lconf PATH fix (619770)
3650 - IA64 build fix (621450)
3651 - RPC buffer sizes scale with amount of memory
3653 2002-10-01 Phil Schwan <phil@clusterfs.com>
3656 - locks would be cancelled without throwing away data pages,
3657 resulting in inconsistent data (605627)
3658 - inode attributes were not always being refreshed (605627, 612449)
3659 - lconf now continues to cleanup after lctl reports an error
3660 - MDS now enforces user permissions (602707)
3661 - lprocfs cleanup fixed, but not yet enabled (614157)
3662 - fixed infinite server hang, should a client not respond to an AST
3663 - avoid going into recovery if user calls readlink() with a buffer
3664 that's too small (613941)
3665 - AST RPCs no longer require replies (614867) -- this may be changed
3666 - don't crash server if client sends an IOV that's too big (611336)
3667 - fixed lock conversion deadlock (611892)
3668 - fixed the following of symlinks (614622)
3669 * recovery: the server can remove locks from a client that dies, other
3670 clients can make progress
3671 * more extN patch fixes
3672 * compile-time configurable ptlrpc buffer allocations
3674 - collaborative read cache document
3675 - Lustre Lite Performance CDR document-in-progress
3677 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3680 - fix typo in patch-2.4.18
3682 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3685 - clear ptlrpc request each time in handle_incoming_request()
3686 - unlink of files now destroys the object on the OST
3688 2002-09-19 Peter Braam <braam@clusterfs.com>
3690 * add hard link support
3691 * change obdfile creation method
3692 * kernel patch changed
3694 2002-09-19 Peter Braam <braam@clusterfs.com>
3697 - stack overflow bug in extN fixed
3699 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3701 * documentation updates
3702 - add man pages for config tools
3703 - update tests/README to describe testing with new config tools
3704 - finish metadata API descriptions
3705 * bug fixes and cleanups
3706 - statfs workaround for 16TB limit
3707 - LOV stripe allocation improved, can stripe on subset of OSTs
3708 - LOV file size/IO offset was wrong for files > 4GB in size
3709 - object EA data was being dropped, caused files to be unreadable
3710 - memory overflow with non-LOV OST caused memory corruption
3711 - fixed regression tests to work with new config tools, obdfilter
3712 - fixed bug when directory size became larger than 1 block
3713 - fixed bug (for single client case) when PWD was deleted
3714 - invalidate local directory pages when doing intent-based ops
3715 - avoid LDLM oops when lock callback contained bad data
3717 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3719 * documentation updates
3720 * bug fixes and cleanups
3721 - configuration tools
3724 - 64-bit compile warnings
3725 - 64-bit internal statfs data
3727 * test_brw on persistent OST devices
3729 * lprocfs (disabled)
3731 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3733 * documentation updates
3734 * bug fixes and cleanups
3735 * configuration tools
3737 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3740 * many small fixes to 0_5_4
3741 * io/network handling
3742 * thinkos in MDS operations
3744 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3747 * crucial basic fixes to 0.5.3
3748 * IOR, Iozone work over Elan
3751 2002-08-07 Phil Schwan <phil@clusterfs.com>
3752 * version 0_5_3, our first alpha
3753 * we use the new Portals iovs
3754 * documentation updates
3755 * bug fixes and cleanups
3756 * small changes in the DLM wire protocol
3758 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3759 * version 0_5_1 with some initial stability,
3760 * locking on MD and file I/O.
3761 * documentation updates
3762 * several bug fixes since 0.5.0
3763 * small changes in wire protocol
3765 2002-07-18 Phil Schwan <phil@clusterfs.com>
3767 * delivered as Lustre Light Alpha
3768 * fixed a crash after handling invalid MDS requests
3769 * fixed directory pages for architectures with non-4k pages sizes
3771 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3772 * release version v0_4_4
3773 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3775 * Fixes a number of interruption problems with OST operations.
3776 * Update documentation for portals header changes
3777 * Move all wire protocol structs/defines to lustre_idl.h
3778 * Fixes symlink length bug.
3779 * Add tcpdump to repository.
3781 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3782 * release version v0_4_3
3783 * Fixes statfs for inodes on extN.
3784 * Fixes bug in runtests which would delete /etc/hosts.
3785 * Use 64-bit object IDs wherever possible (not into VFS though)
3786 Remove ost_get_info, which is unused by lustre, and out of date.
3788 2002-07-03 Peter Braam <braam@clusterfs.com>
3789 * release version v0_4_2 Fixes a lookup error (type not passed)
3790 * move forward to head of Portals
3791 * move forward to latest Lustre kernel
3793 2002-06-25 Peter Braam <braam@clusterfs.com>
3794 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob