1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10)
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4)
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.39.cfs8
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: don't use GFP_* in generic Lustre code.
18 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
22 Description: obdclass is limited by single OBD_ALLOC(idarray)
23 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
27 Description: warnings with build patchless client with vanila 2.6.19 and up
28 Details : change old ctl_table style and replace ctl_table/ctl_table_header
29 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
33 Frequency : Only for SLES
34 Description: sanity_quota fail test_1
35 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
36 file, which makes TSTID[2] inunique.
40 Frequency : Only for SLES9
41 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
42 Details : Update target/ChangeLog/which_patch .
46 Frequency : Only for test_10 in sanity-quota.sh
47 Description: a bug in quota test code
48 Details : checking a test flag in a wrong place
52 Frequency : Only for test_13 in sanity-quota.sh
53 Description: a bug in quota test script
54 Details : shouldn't check overfull allocated quota
58 Description: jbd statistics
59 Details : Port older jbd statistics patch for sles10
61 Severity : enhancement
63 Description: Add printing inode info into message about error in writepage.
65 --------------------------------------------------------------------------------
67 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
69 * Support for kernels:
72 2.6.16.46-0.14 (SLES 10)
73 2.6.18.8 vanilla (kernel.org)
74 * Client support for unpatched kernels:
75 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
76 2.6.16 - 2.6.22 vanilla (kernel.org)
77 * Due to recently discovered recovery problems, we do not recommend
78 using patchless RHEL 4 clients with this or any earlier release.
79 * Recommended e2fsprogs version: 1.39.cfs8
80 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
81 * Starting with this release, the ldiskfs backing filesystem required
82 by Lustre is now in its own package, lustre-ldiskfs. This package
83 should be installed. It is versioned separately from Lustre and
84 may be released separately in future.
88 Description: block reactivating mgc import until all deactivates complete
89 Details : Fix race when failing back MDT/MGS to itself (testing)
91 Severity : enhancement
93 Description: add optional extra BUILD_VERSION info
94 Details : add a new environment variable (namely LUSTRE_VERS) which allows
95 to override the lustre version.
98 Frequency : 2.6.18 servers only
100 Description: ll_kern_mount() doesn't release the module reference
101 Details : The ldiskfs module reference count never drops down to 0
102 because ll_kern_mount() doesn't release the module reference.
107 Description: server LBUG when using old ost_num_threads parameter
108 Details : Accept the old ost_num_threads parameter but warn that it
109 is deprecated, and fix an off-by-one error that caused an LBUG.
114 Description: Transient SCSI error results in persistent IO issue
115 Details : iobuf->dr_error is not reinitialized to 0 between two
119 Frequency : sometimes when underlying device returns I/O errors
121 Description: OSTs not going read-only during write failures
122 Details : OSTs are not remounted read-only when the journal commit threads
123 get I/O errors because fsfilt_ext3 calls journal_start/stop()
124 instead of the ext3 wrappers.
128 Description: poor connect scaling with increasing client count
129 Details : Don't run filter_grant_sanity_check for more than 100 exports
130 to improve scaling for large numbers of clients.
133 Frequency : SLES10 only
135 Description: sanity-quota.sh quotacheck failed: rc = -22
136 Details : Quotas cannot be enabled on SLES10.
139 Frequency : liblustre clients only
141 Description: getdirentries does not give error when run on compute nodes
142 Details : getdirentries does not fail when the size specified as an argument
143 is too small to contain at least one entry
145 Severity : enhancement
147 Description: Add LNET router traceability for debug purposes
148 Details : If a checksum failure occurs with a router as part of the
149 IO path, the NID of the last router that forwarded the bulk data
150 is printed so it can be identified.
155 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
156 Details : Due to a race condition, liblustre clients were occasionally
159 Severity : enhancement
161 Description: lfs setstripe use optional parameters instead of postional
164 Severity : enhancement
166 Description: Nanosecond timestamp support for ldiskfs
167 Details : The on-disk ldiskfs filesystem has added support for nanosecond
168 resolution timestamps. There is not yet support for this at
169 the Lustre filesystem level.
172 Frequency : during server recovery
174 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
175 Details : request with rq_no_resend flag not awake l_wait_event if they get a
179 Frequency : nfs export on patchless client
181 Description: connectathon hang when test nfs export over patchless client
182 Details : Disconnected dentry cannot be found with lookup, so we do not need
183 to unhash it or make it invalid
186 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
187 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
188 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
193 Description: open req refcounting wrong on reconnect
194 Details : If reconnect happened between getting open reply from server and
195 call to mdc_set_replay_data in ll_file_open, we will schedule
196 replay for unreferenced request that we are about to free.
197 Subsequent close will crash in variety of ways.
198 Check that request is still eligible for replay in
199 mdc_set_replay_data().
204 Description: disable writes to filesystem when reading health_check file
205 Details : the default for reading the health_check proc file has changed
206 to NOT do a journal transaction and write to disk, because this
207 can cause reads of the /proc file to hang and block HA state
208 checking on a healthy but otherwise heavily loaded system. It
209 is possible to return to the previous behaviour during configure
210 with --enable-health-write.
212 Severity : enhancement
214 Description: 64-bit inode version
215 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
216 to the inode. This will be required for version-based recovery.
221 Description: MDS fails to start if a duplicate client export is detected
222 Details : in some rare cases it was possible for a client to connect to
223 an MDS multiple times. Upon recovery the MDS would detect this
224 and fail during startup. Handle this more gracefully.
226 Severity : enhancement
228 Description: Add -o localflock option to simulate old noflock
230 Details : This will achieve local-only flock/fcntl locks
236 Description: log_commit_thread vs filter_destroy race leads to crash
237 Details : Take import reference before releasing llog record semaphore
242 Description: Wrong request locking in request set processing
243 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
244 imp_delayed_list, in this place should be used imp_lock.
247 Frequency : when reconnection
249 Description: Grant Leak when osc reconnect to OST
250 Details : When osc reconnect ost, OST(filter) should check whether it
251 should grant more space to client by comparing fed_grant and
252 cl_avail_grant, and return the granted space to client instead
253 of "new granted" space, because client will call osc_init_grant
254 to update the client grant space info.
257 Frequency : when client reconnect to OST
259 Description: Grant Leak when osc do resend and replay bulk write
260 Details : When osc reconnect to OST, OST(filter)should clear grant info of
261 bulk write request, because the grant info while be sync between
262 OSC and OST when reconnect, and we should ignore the grant info
263 these of resend/replay write req.
268 Description: Grant space more than avaiable left space sometimes.
269 Details : When then OST is about to be full, if two bulk writing from
270 different clients came to OST. Accord the avaliable space of the
271 OST, the first req should be permitted, and the second one
272 should be denied by ENOSPC. But if the seconde arrived before
273 the first one is commited. The OST might wrongly permit second
274 writing, which will cause grant space > avaiable space.
277 Frequency : when client is evicted
279 Description: Grant might be wrongly erased when osc is evicted by OST
280 Details : when the import is evicted by server, it will fork another
281 thread ptlrpc_invalidate_import_thread to invalidate the
282 import, where the grant will be set to 0. While the original
283 thread will update the grant it got when connecting. So if
284 the former happened latter, the grant will be wrongly errased
285 because of this race.
290 Description: Checking Stale with correct fid
291 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
292 to check whether it is stale, because sometimes, we want the
293 enqueue happened anyway, and op_data.fid2 will not be initialized.
295 Severity : enhancement
297 Description: update patchless client
298 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
301 Frequency : only with 2.4 kernel
303 Description: random memory corruption
304 Details : size of struct ll_inode_info is to big for union inode.u and this
305 can be cause of random memory corruption.
310 Description: Memory leak in recovery
311 Details : Lov_mds_md was not free in an error handler in mds_create_object.
312 It should also check obd_fail before fsfilt_start, otherwise if
313 fsfilt_start return -EROFS,(failover mds during mds recovery).
314 then the req will return with repmsg->transno = 0 and rc = EROFS.
315 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
316 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
317 be freed no matter whether fsfilt_commit success or not.
320 Frequency : only with huge count clients
322 Description: Prevents from taking the superblock lock in llap_from_page for
324 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
325 for llap_from_page call in ll_removepage prevents from taking the
326 superblock lock for a soon died page.
331 Description: Not check open intent error before release open handle
332 Details : in some rare cases, the open intent error is not checked before
333 release open handle, which may cause
334 ASSERTION(open_req->rq_transno != 0), because it tries to release
335 the failed open handle.
340 Description: Set cat log bitmap only after create log success.
341 Details : in some rare cases, the cat log bitmap is set too early. and it
342 should be set only after create log success.
346 Description: Accessing a block bevice can re-enable I/O when Lustre is
347 tearing down a device.
348 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
353 Description: service threads may hog cpus when there are a lot of requests
355 Details : Insert cond_resched to give other threads a chance to use some of
361 Description: the cat log was not initialized in recovery
362 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
363 unlink log on mds will not be initialized until mds post
364 recovery. And also in mds post recovery, the unlink log will
365 initialization will be done asynchronausly, so there will be race
366 between add unlink log and unlink log initialization.
370 Description: brw_stats were being printed incorrectly
371 Details : brw_stats were being printed as log2 but all of them were not
372 recorded as log2. Also remove some code duplication arising from
373 filter_tally_{read,write}.
377 Frequency : rare, only in recovery.
378 Description: ASSERTION(req->rq_type != LI_POISON) failed
379 Details : imp_lock should be held while iterating over imp_sending_list for
380 prevent destroy request after get timeout in ptlrpc_queue_wait.
384 Description: replay-single.sh test 52 fails
385 Details : A lock's skiplist need to be cleanup when it being unlinked
386 from its resource list.
390 Description: Short directio read returns full requested size rather than
392 Details : Direct I/O operations should return actual amount of bytes
393 transferred rather than requested size.
395 Severity : enhancement
397 Description: metadata RPC reduction (e.g. for rm performance)
398 Details : decrease the amount of synchronous RPC between clients and servers
399 by canceling conflicing lock before the operation on the client side
400 and packing thier handles into the main operation RPC to server.
402 --------------------------------------------------------------------------------
404 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
409 Frequency : on some architectures
411 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
412 Details : Uninitialized flags sometimes cause configuration commands to
416 Frequency : patchless clients only
418 Description: missing __iget() symbol export
419 Details : The __iget() symbol export is missing. To avoid the need for
420 this on patchless clients the deathrow inode reaper is turned
421 off, and we depend on the VM to clean up old inodes. This
422 dependency was during via the fix for bug 12181.
424 --------------------------------------------------------------------------------
426 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
428 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
429 INTEROPERATE with older versions automatically. In many cases a
430 special upgrade step is needed. Please read the
431 user documentation before upgrading any part of a live system.
432 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
433 version will not interoperate with 1.6 betas before beta5 (1.5.95).
434 * WARNING: Lustre configuration and startup changes are required with
435 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
440 Severity : enhancement
442 Description: MountConf
443 Details : Lustre configuration is now managed via mkfs and mount
444 commands instead of lmc and lconf. New obd types (MGS, MGC)
445 are added for dynamic configuration management. See
446 https://mail.clusterfs.com/wikis/lustre/MountConf for
449 Severity : enhancement
451 Description: dynamic OST addition
452 Details : OSTs can now be added to a live filesystem
454 Severity : enhancement
456 Description: startup order invariance
457 Details : MDTs and OSTs can be started in any order. Clients only
458 require the MDT to complete startup.
460 Severity : enhancement
462 Description: parallel, asynchronous orphan cleanup
463 Details : orphan cleanup is now performed in separate threads for each
464 OST, allowing parallel non-blocking operation.
466 Severity : enhancement
468 Description: optimized stripe assignment
469 Details : stripe assignments are now made based on ost space available,
470 ost previous usage, and OSS previous usage, in order to try
471 to optimize storage space and networking resources.
473 Severity : enhancement
475 Description: Permanently set tunables
476 Details : All writable /proc/fs/lustre tunables can now be permanently
477 set on a per-server basis, at mkfs time or on a live
480 Severity : enhancement
482 Description: Lustre message v2
483 Details : Add lustre message format v2.
485 Severity : enhancement
487 Description: client OST exclusion list
488 Details : Clients can be started with a list of OSTs that should be
489 declared "inactive" for known non-responsive OSTs.
493 Description: SPEC SFS validation failure on NFS v2 over lustre.
494 Details : Changes the blocksize for regular files to be 2x RPC size,
495 and not depend on stripe size.
497 Severity : enhancement
499 Description: Multiple MD RPCs in flight.
500 Details : Further unserialise some read-only MDS RPCs - learn about intents.
501 To avoid overly-overloading MDS, introduce a limit on number of
502 MDS RPCs in flight for a single client and add /proc controls
503 to adjust this limit.
505 Severity : enhancement
507 Description: client read/write statistics
508 Details : Add client read/write call usage stats for performance
509 analysis of user processes.
510 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
511 file access. extents_stats shows chunk size distribution.
512 extents_stats_per_process show chunk size distribution per
515 Severity : enhancement
517 Description: mds statistics
518 Details : Add detailed mds operations statistics in
519 /proc/fs/lustre/mds/*/stats.
523 Description: Failure of copying files with lustre special EAs.
524 Details : Client side always return success for setxattr call for lustre
525 special xattr (currently only "trusted.lov").
530 Description: Refcount LNET uuids
531 Details : The global LNET uuid list grew linearly with every startup;
532 refcount repeated list entries instead of always adding to
535 Severity : enhancement
537 Description: Dynamic service threads
538 Details : Within a small range, start extra service threads
539 automatically when the request queue builds up.
542 Frequency : mixed-endian client/server environments
544 Description: mixed-endian crashes
545 Details : The new msg_v2 system had some failures in mixed-endian
548 Severity : enhancement
550 Description: Easy OST removal
551 Details : OSTs can be permanently deactivated with e.g. 'lctl
552 conf_param lustre-OST0001.osc.active=0'
554 Severity : enhancement
556 Description: MGS proc entries
557 Details : Added basic proc entries for the MGS showing what filesystems
560 Severity : enhancement
562 Description: provide MGS failover
563 Details : Added config lock reacquisition after MGS server failover.
565 Severity : enhancement
567 Description: add Linux 2.4 support
568 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
572 Description: a large application tries to do I/O to the same resource and dies
574 Details : Check the req->rq_arrival time after the call to
575 ost_brw_lock_get(), but before we do anything about
576 processing it & sending the BULK transfer request. This
577 should help move old stale pending locks off the queue as
578 quickly as obd_timeout.
581 Frequency : when an incorrect nid is specified during startup
583 Description: ptlrpc connect to non-existant node causes kernel crash
584 Details : LNET can't be re-entered from an event callback, which
585 happened when we expire a message after the export has been
586 cleaned up. Instead, hand the zombie cleanup off to another
589 Severity : enhancement
591 Description: plain/inodebits lock performance improvement
592 Details : Grouping plain/inodebits in granted list by their request modes
593 and bits policy, thus improving the performance of search through
597 Frequency : only if OST filesystem is corrupted
599 Description: client incorrectly hits assertion in ptlrpc_replay_req()
600 Details : for a short time RPCs with bulk IO are in the replay list,
601 but replay of bulk IOs is unimplemented. If the OST filesystem
602 is corrupted due to disk cache incoherency and then replay is
603 started it is possible to trip an assertion. Avoid putting
604 committed RPCs into the replay list at all to avoid this issue.
607 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
609 Description: System hang on startup
610 Details : This bug allowed the liblustre (e.g. catamount) client to
611 return to the app before handling all startup RPCs. This
612 could leave the node unresponsive to lustre network traffic
613 and manifested as a server ptllnd timeout.
615 Severity : enhancement
617 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
618 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
619 boolean to control whether to print peer debug info when a
620 client's RPC times out.
623 Frequency : only for kernels with patches from Lustre below 1.4.3
625 Description: Remove old rdonly API
626 Details : Remove old rdonly API which unused from at least lustre 1.4.3
629 Frequency : only for devices with external journals
631 Description: Set external device read-only also
632 Details : During a commanded failover stop, we set the disk device
633 read-only while the server shuts down. We now also set any
634 external journal device read-only at the same time.
637 Frequency : when upgrading from 1.4 while trying to change parameters
639 Description: The wrong (new) MDC name was used when setting parameters for
640 upgraded MDT's. Also allows changing of OSC (and MDC)
641 parameters if --writeconf is specified at tunefs upgrade time.
644 Frequency : when setting specific ost indicies
646 Description: QOS code breaks on skipped indicies
647 Details : Add checks for missing OST indicies in the QOS code, so OSTs
648 created with --index need not be sequential.
650 Severity : enhancement
652 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
653 Details : The uninit_groups feature works in conjunction with the kernel
654 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
655 up the pass1 processing of e2fsck. This is a read-only feature
656 in ldiskfs2 only, so older kernels and current ldiskfs cannot
657 mount filesystems that have had this feature enabled.
659 Severity : enhancement
661 Description: Improve multi-block allocation algorithm to avoid fragmentation
662 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
663 allocation locality and avoid filesystem fragmentation.
665 ------------------------------------------------------------------------------
667 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
669 * Support for kernels:
670 2.4.21-47.0.1.EL (RHEL 3)
672 2.6.9-42.0.10.EL (RHEL 4)
673 2.6.12.6 vanilla (kernel.org)
674 2.6.16.27-0.9 (SLES 10)
675 * Recommended e2fsprogs version: 1.39.cfs5
677 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
681 Frequency : occasional, depends on client load and configuration
682 Bugzilla : 12181, 12203
683 Description: data loss for recently-modified files
685 Details : In some cases it is possible that recently written or created
686 files may not be written to disk in a timely manner (this should
687 normally be within 30s unless client IO load is very high).
688 The problem appears as zero-length files or files that are a
689 multiple of 1MB in size after a client crash or client eviction
690 that are missing data at the end of the file.
692 This problem is more likely to be hit on clients where files are
693 repeatedly created and unlinked in the same directory, clients
694 have a large amount of RAM, have many CPUs, the filesystem has
695 many OSTs, the clients are rebooted frequently, and/or the files
696 are not accessed by other nodes after being written.
698 The presence of the problem can be detected by looking at
699 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
700 smaller than the second (nr_unused) then dirty files will not
701 be flushed automatically to disk. "sync; sleep 10" should be
702 run several times on the node before unmounting it to update
703 Lustre (this is also safe to run on nodes without this problem).
705 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
706 that can cause this same problem, so customers using that kernel
707 also need to update the kernel in addition to Lustre. In order
708 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
710 It is normal that files written just before a client crash (less
711 than 30s) may not yet have been flushed to disk, even for local
715 Frequency : frequent on thin XT3 nodes
717 Description: UUID collision on thin XT3 Linux nodes
718 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
719 randomly, since we relied on an insufficiently-seeded PRNG.
724 Description: OSS hangs after "All ost request buffers busy"
725 Details : A deadlock between quota and journal operations caused OSS
726 hangs after printing "All ost request buffers busy."
729 Frequency : always on liblustre builds
731 Description: Cleanup compiler warnings on liblustre
734 Frequency : always on liblustre builds on XT3
736 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
740 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
741 > 32000 subdirectories
742 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
743 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
744 subdirectory count crosses 32000. This will aid e2fsck to
745 correctly handle more than 32000 subdirectories.
748 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
750 Description: System hang on startup
751 Details : This bug allowed the liblustre (e.g. catamount) client to
752 return to the app before handling all startup RPCs. This
753 could leave the node unresponsive to lustre network traffic
754 and manifested as a server ptllnd timeout.
756 Severity : enhancement
758 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
759 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
760 boolean to control whether to print peer debug info when a
761 client's RPC times out.
766 Description: make O_SYNC working on 2.6 kernels
767 Details : 2.6 kernels use different method for mark pages for write,
768 so need add a code to lustre for O_SYNC work.
773 Description: Failure to close file and release space on NFS
774 Details : Put inode details into lock acquired in ll_intent_file_open.
775 Use mdc_intent_lock in ll_intent_open to properly
776 detect all kind of errors unhandled by mdc_enqueue.
781 Description: proc file read during shutdown sometimes raced obd removal,
783 Details : Add lock to prevent obd access after proc file removal.
786 Frequency : Only for files larger than 4GB on 32-bit clients.
788 Description: improperly doing page alignment of locks
789 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
790 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
795 Description: RPCs being resent when they shouldn't be
796 Details : Some RPCs that should not be resent are being resent. This
797 can cause inconsistencies in the RPC state machine. Do not
798 resend such requests.
801 Frequency : rare, only with NFS export
803 Description: Crash on NFS re-export node
804 Details : under very unusual load conditions an assertion is hit in
805 ll_intent_file_open()
808 Frequency : only if OST filesystem is corrupted
810 Description: client incorrectly hits assertion in ptlrpc_replay_req()
811 Details : for a short time RPCs with bulk IO are in the replay list,
812 but replay of bulk IOs is unimplemented. If the OST filesystem
813 is corrupted due to disk cache incoherency and then replay is
814 started it is possible to trip an assertion. Avoid putting
815 committed RPCs into the replay list at all to avoid this issue.
820 Description: large O_DIRECT requests fail under memory pressure/fragmentation
821 Details : Large single O_DIRECT read and write calls can fail to allocate
822 a sufficiently large buffer to process the request. In case of
823 allocation failure the allocation is retried with a smaller
824 buffer and broken into smaller requests.
826 Severity : enhancement
828 Description: Add -o localflock option to simulate old noflock behaviour.
829 Details : This will achieve local-only flock/fcntl locks coherentness.
834 Description: versioning check is incomplete
835 Details : Checking the version difference of client vs. server, report
836 error if the gap is too big.
841 Description: add support PG_writeback bit
842 Details : add support for PG_writeback bit for Lustre, for more carefull
843 work with page cache in 2.6 kernel. This also fix some deadlocks
844 and remove hack for work O_SYNC with 2.6 kernel.
846 Severity : enhancement
848 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
849 Details : The uninit_groups feature works in conjunction with the kernel
850 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
851 up the pass1 processing of e2fsck. This is a read-only feature
852 in ldiskfs2 only, so older kernels and current ldiskfs cannot
853 mount filesystems that have had this feature enabled.
855 Severity : enhancement
857 Description: Improve multi-block allocation algorithm to avoid fragmentation
858 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
859 allocation locality and avoid filesystem fragmentation.
861 ------------------------------------------------------------------------------
863 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
865 * Support for kernels:
866 2.6.9-42.0.3.EL (RHEL 4)
868 2.4.21-47.0.1.EL (RHEL 3)
869 2.6.12.6 vanilla (kernel.org)
870 2.6.16.21-0.8 (SLES10)
871 * Recommended e2fsprogs version: 1.39.cfs2-0
873 * The backwards-compatible /proc/sys/portals symlink has been removed
874 in this release. Before upgrading, please ensure that you change
875 any configuration scripts or /etc/sysctl.conf files that access
876 /proc/sys/portals/* or sysctl portals.* to use the corresponding
877 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
878 in advance of the upgrade on any system running Lustre 1.4.6 or
879 newer, since /proc/sys/lnet was added in that version.
880 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
884 Frequency : only when quota is used
886 Description: avoid scanning export list for quota master
887 Details : Change the algorithms to avoid scanning export list in order
888 to improve the efficiency.
891 Frequency : MDS failover only, very rarely
893 Description: "went back in time" messages on mds failover
894 Details : The greatest transno may be lost when the current operation
895 finishes with an error (transno==0) and the client's last_rcvd
896 record is over-written. Save the greatest transno in the
897 mds_last_transno for this case.
900 Frequency : always for specific kernels and striping counts
902 Description: client may get "Matching packet too big" without ACL support
903 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
904 when trying to access files in certain configurations. The
905 clients should in fact be denied when mounting because they do
909 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
911 Description: many clients connecting with IO in progress causes connect timeouts
912 Details : Avoid synchronous journal commits to avoid delays caused by many
913 clients connecting/disconnecting when bulk IO is in progress.
914 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
915 OST_IO_PORTAL to avoid delays behind potentially many pending
919 Frequency : occasionally with multiple writers to a single file
921 Description: shared writes to file may result in wrong size reported by stat()
922 Details : Allow growing of kms when extent lock is cancelled
925 Frequency : always with random mmap IO to multi-striped file
927 Description: mmap write might be lost if we are writing to a 'hole' in stripe
928 Details : Only if the hole is at the end of OST object so that kms is too
929 small. Fix is to increase kms accordingly in ll_nopage.
932 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
934 Description: writes to a missing object would leak memory on the OST
935 Details : If there is an inconsistency between the MDS and OST filesystems,
936 such that the MDS references an object that doesn't exist, writes
937 to that object will leak memory due to incorrect cleanup in the
938 error handling path, eventually running out of memory on the OST.
943 Description: Creating too long symlink causes lustre errors
944 Details : Check symlink and name lengths before sending requests to MDS.
947 Frequency : only if flock is enabled (not on by default)
949 Description: posix locks not released on fd closure on 2.6.9+
950 Details : We failed to add posix locks to list of inode locks on 2.6.9+
951 kernels, this caused such locks not to be released on fd close and
952 then assertions on fs unmount about still used locks.
955 Frequency : MDS failover only, very rarely
957 Description: clients may get ASSERTION(granted_lock != NULL)
958 Details : When request was taking a long time, and a client was resending
959 a getattr by name lock request. The were multiple lock requests
960 with the same client lock handle and
961 mds_getattr_name->fixup_handle_for_resent_request found one of the
962 lock handles but later failed with ASSERTION(granted_lock != NULL).
967 Description: handle->h_buffer_credits > 0, assertion failure
968 Details : h_buffer_credits is zero after truncate, causing assertion
969 failure. This patch extends the transaction or creates a new
973 Frequency : NFS re-export or patchless client
974 Bugzilla : 11179, 10796
975 Description: Crash on NFS re-export node (__d_move)
976 Details : We do not want to hash the dentry if we don't have a lock.
977 But if this dentry is later used in d_move, we'd hit uninitialised
978 list head d_hash, so we just do this to init d_hash field but
979 leave dentry unhashed.
982 Frequency : NFS re-export or patchless client
984 Description: NFS exports has problem with symbolic link
985 Details : lustre client didn't properly install dentry when re-exported
986 to NFS or running patchless client.
989 Frequency : NFS re-export or patchless client
991 Description: Various nfs/patchless fixes.
992 Details : fixes reuse disconected alias for lookup process - this fixes
993 warning "find_exported_dentry: npd != pd",
994 fix permission error with open files at nfs.
995 fix apply umask when do revalidate.
998 Frequency : occasional
1000 Description: Crash on NFS re-export node
1001 Details : calling clear_page() on the wrong pointer triggered oops in
1002 generic_mapping_read().
1005 Frequency : rarely, using O_DIRECT IO
1007 Description: unaligned directio crashes client with LASSERT
1008 Details : check for unaligned buffers before trying any requests.
1011 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1013 Description: stale data returned from RAID cache
1014 Details : If only a small amount of IO is done to the RAID device before
1015 reading it again it is possible to get stale data from the RAID
1016 cache instead of reading it from disk.
1019 Frequency : always for sles10 kernel
1021 Description: sles10 support
1022 Details : ll_follow_link: compile fixes and using of nd_set_link
1023 under newer kernels.
1026 Frequency : depends on arch, kernel and compiler version, always on sles10
1029 Description: recursive or deep enough symlinks cause stack overflow
1030 Details : getting rid of large stack-allocated variable in
1034 Frequency : depends on hardware
1036 Description: lustre write performance loss in the SLES10 kernel
1037 Details : the performance loss is caused by using of write barriers in the
1038 ext3 code. The SLES10 kernel turns barrier support on by
1039 default. The fix is to undo that change for ldiskfs.
1041 ------------------------------------------------------------------------------
1043 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1045 * Support for kernels:
1046 2.6.9-42.0.3EL (RHEL 4)
1047 2.6.5-7.276 (SLES 9)
1048 2.4.21-47.0.1.EL (RHEL 3)
1049 2.6.12.6 vanilla (kernel.org)
1053 Frequency : quota enabled and large files being deleted
1055 Description: releasing more than 4GB of quota at once hangs OST
1056 Details : If a user deletes more than 4GB of files on a single OST it
1057 will cause the OST to spin in an infinite loop. Release
1058 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1063 Description: statfs data retrieved from /proc may be stale or zero
1064 Details : When reading per-device statfs data from /proc, in the
1065 {kbytes,files}_{total,free,avail} files, it may appear
1066 as zero or be out of date.
1069 Frequency : systems with MD RAID1 external journal devices
1071 Description: lconf's call to blkid is confused by RAID1 journal devices
1072 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1073 of returning all block devices that match the journal UUID.
1076 Frequency : always, for aggregate stripe size over 4GB
1078 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1079 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1080 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1083 Frequency : always if "lfs find" used on a local file/directory
1085 Description: "lfs find" segfaults if used on a local file/directory
1086 Details : The case where a directory component was not specified wasn't
1087 handled correctly. Handle this properly.
1090 Frequency : always on ppc64
1092 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1093 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1094 perform endianness conversion.
1097 Frequency : rarely (truncate to non-zero file size after write under load)
1098 Bugzilla : 10730, 10687
1099 Description: Files padded with zeros to next 4K multiple
1100 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1101 it is possible that files that are truncated to a non-zero size
1102 immediately after being written are filled with zero bytes beyond
1103 the truncated size. No file data is lost.
1105 Severity : enhancement
1107 Description: Allow recovery/failover for liblustre clients.
1108 Details : liblustre clients were unaware of failover configurations until
1111 Severity : enhancement
1113 Description: user file locks should fail when not mounting with flock option
1114 Details : Set up an error-returning stub in ll_file_operations.lock field
1115 to prevent incorrect behaviour when client is mounted without
1116 flock option. Also, set up properly f_op->flock field for
1120 Frequency : always on ia64
1122 Description: "lfs df" loops on printing out MDS statfs information
1123 Details : The obd_ioctl_data was not initialized and in some systems
1124 this caused a failure during the ioctl that did not return
1125 an error. Initialize the struct and return an error on failure.
1128 Frequency : SLES 9 only
1130 Description: Error of copying files with lustre special EAs as root
1131 Details : Client side always return success for setxattr call for lustre
1132 special xattr (currently only "trusted.lov").
1135 Frequency : rarely on clusters with both ia64+i386 clients
1137 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1138 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1139 clients are doing concurrent writes to the same file it is
1140 possible that the ia64 clients may LASSERT because the OST
1141 extent locks are not PAGE_SIZE aligned. Ensure that grown
1142 locks are always aligned on the request boundary.
1145 Frequency : specific use, occasional
1147 Description: Overwriting in use executable truncates on-disk binary image
1148 Details : If one node attempts to overwrite an executable in use by
1149 another node, we now correctly return ETXTBSY instead of
1150 truncating the file.
1152 Severity : enhancement
1154 Description: Async OSC create to avoid the blocking unnecessarily.
1155 Details : If a OST has no remain object, system will block on the creating
1156 when need to create a new object on this OST. Now, ways use
1157 pre-created objects when available, instead of blocking on an
1158 empty osc while others are not empty. If we must block, we block
1159 for the shortest possible period of time.
1164 Description: chmod on Lustre root is propagated to other clients
1165 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1166 invalid by the follow_mount time.
1169 Frequency : liblustre clients only
1171 Description: Race in 'instant cancel' lock handling could lead to such locks
1172 never to be granted in case of SMP MDS
1173 Details : Do not destroy not yet granted but cbpending locks in
1177 Frequency : replay/resend of open
1179 Description: non null lock assetion failure in mds_intent_policy
1180 Details : Trying to replay/resend lockless open requests resulted in
1181 mds_open() returning 0 with no lock. Now it sets a flag if
1182 a lock is going to be returned.
1184 Severity : enhancement
1186 Description: Checksum enhancements
1187 Details : New checksum enhancements allow for resending RPCs that failed
1190 Severity : enhancement
1192 Description: Tunables on number of dirty pages in cacche
1193 Details : Allow to set limit on number of dirty pages cached.
1198 Description: client crash on unmount - lock still has references
1199 Details : In some error handling cases it was possible to leak a lock
1200 reference on a client while accessing a file. This was not
1201 harmful to the client during operation, but would cause the
1202 client to crash when the filesystem is unmounted.
1205 Frequency : specific case, rare
1207 Description: ETXTBSY on mds though file not in use by client
1208 Details : ETXTBSY is no longer incorrectly returned when attempting to
1209 chmod or chown a directory that the user previously tried to
1210 execute or a currently-executing binary.
1213 Frequency : extremely rare except on liblustre-based clients
1215 Description: Lustre space not freed when files are deleted
1216 Details : Clean up open-unlinked files after client eviction. Previously
1217 the unlink was skipped and the files remained as orphans.
1222 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1223 Details : In some cases it was possible to send a blocking callback to
1224 a client doing a glimpse, even though that client didn't get
1225 a lock granted. When the glimpse lock is cancelled on the OST
1226 the freed lock is left on the waiting list and corrupted the list.
1229 Frequency : all core dumps
1231 Description: Broke core dumps to lustre
1232 Details : Negative dentry may be unhashed if parent does not have UPDATE
1233 lock, but some callers, e.g. do_coredump, expect dentry to be
1234 hashed after successful create, hash it in ll_create_it.
1236 ------------------------------------------------------------------------------
1238 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1240 * Support for kernels:
1241 2.6.9-42.0.2.EL (RHEL 4)
1242 2.6.5-7.276 (SLES 9)
1243 2.4.21-40.EL (RHEL 3)
1244 2.6.12.6 vanilla (kernel.org)
1248 Frequency : always on RHEL 3
1250 Description: Number of open files grows over time
1251 Details : The number of open files grows over time, whether or not
1252 Lustre is started. This was due to a filp leak introduced
1253 by one of our kernel patches.
1255 ------------------------------------------------------------------------------
1257 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1259 * Support for kernels:
1260 2.6.9-42.EL (RHEL 4)
1261 2.6.5-7.276 (SLES 9)
1262 2.4.21-40.EL (RHEL 3)
1263 2.6.12.6 vanilla (kernel.org)
1268 Bugzilla : 5719, 9635, 9792, 9684
1269 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1270 Details : If a server is under heavy load and cannot reply to new
1271 connection requests before the client resends the (re)connect,
1272 the connection handling code can behave badly if two service
1273 threads are concurrently handing separate (re)connections from
1274 the same client. Add better locking to the connection handling
1275 code, and ensure that only a single connection will be processed
1276 for a given client UUID, even if the lock is dropped.
1278 Severity : enhancement
1280 Description: add TCP zero-copy support to kernel
1281 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1282 sends if the hardware supports scatter-gather and checksumming.
1283 This allows socklnd to do client-write and server-read more
1284 efficiently and reduce CPU utilization from skbuf copying.
1287 Frequency : only if NFS exporting from client
1289 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1290 Details : If setting attributes on a file created under NFS that had
1291 never been opened it would be possible to oops the client
1292 if the file had no objects.
1295 Frequency : always for liblustre
1297 Description: liblustre client does MDS+OSTs setattr RPC for each write
1298 Details : When doing a write from a liblustre client, the client
1299 incorrectly issued an RPC to the MDS and each OST the file was
1300 striped over in order to update the timestamps. When writing
1301 with small chunks and many clients this could overwhelm the MDS
1302 with RPCs. In all cases it would slow down the write because
1303 these RPCs are unnecessary.
1305 Severity : enhancement
1307 Description: allow number of MDS service threads to be changed at module load
1308 Details : It is now possible to change the number of MDS service threads
1309 running. Adding "options mds mds_num_threads={N}" to the MDS's
1310 /etc/modprobe.conf will set the number of threads for the next
1311 time Lustre is restarted (assuming the "mds" module is also
1312 reloaded at that time). The default number of threads will
1313 stay the same, 32 for most systems.
1318 Description: OST crash if filesystem is unformatted or corrupt
1319 Details : If an OST is started on a device that has never been formatted
1320 or if the filesystem is corrupt and cannot even mount then the
1321 error handling cleanup routines would dereference a NULL pointer.
1326 Description: NULL pointer deref in llap_from_page.
1327 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1328 incorrect) mapping if the page was truncated/reclaimed while it was
1329 searched for. Check for this condition and skip such pages when
1330 doing readahead. Introduce extra check to llap_from_page() to
1331 verify page->mapping->host is non-NULL (so page is not anonymous).
1334 Frequency : Sometimes when using sys_sendfile
1336 Description: "page not covered by a lock" warnings from ll_readpage
1337 Details : sendfile called ll_readpage without right page locks present.
1338 Now we introduced ll_file_sendfile that does necessary locking
1339 around call to generic_file_sendfile() much like we do in
1343 Frequency : with certain MDS communication failures at client mount time
1345 Description: NULL pointer deref after failed client mount
1346 Details : a client connection request may delayed by the network layer
1347 and not be sent until after the PTLRPC layer has timed out the
1348 request. If the client fails the mount immediately it will try
1349 to clean up before the network times out the request. Add a
1350 reference from the request import to the obd device and delay
1351 the cleanup until the network drops the request.
1354 Frequency : occasionally during client (re)connect
1356 Description: assertion failure during client (re)connect
1357 Details : processing a client connection request may be delayed by the
1358 client or server longer than the client connect timeout. This
1359 causes the client to resend the connection request. If the
1360 original connection request is replied in this interval, the
1361 client may trip an assertion failure in ptlrpc_connect_interpret()
1362 which thought it would be the only running connect process.
1365 Frequency : only with obd_echo servers and clients that are rebooted
1367 Description: kernel BUG accessing uninitialized data structure
1368 Details : When running an obd_echo server it did not start the ping_evictor
1369 thread, and when a client was evicted an uninitialized data
1370 structure was accessed. Start the ping_evictor in the RPC
1371 service startup instead of the OBD startup.
1373 Severity : enhancement
1374 Bugzilla : 10193 (patchless)
1375 Description: Remove dependency on various unexported kernel interfaces.
1376 Details : No longer need reparent_to_init, exit_mm, exit_files,
1377 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1380 Frequency : rare (only users of deprecated and unsupported LDAP config)
1382 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1383 Details : LDAP apparently contains 'lustreName' attributes instead of
1384 'name'. A simple remapping of the name is sufficient.
1387 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1389 Description: waiting_locks_callback trips kernel BUG if client is evicted
1390 Details : Running with the dump_on_timeout debug flag turned on makes
1391 it possible that the waiting_locks_callback() can try to dump
1392 the Lustre kernel debug logs from an interrupt handler. Defer
1393 this log dumping to the expired_lock_main() thread.
1395 Severity : enhancement
1397 Description: Support NFS exporting on 2.6 kernels.
1398 Details : Implement non-rawops metadata methods for NFS server to use without
1399 changing NFS server code.
1402 Frequency : very rare (synthetic metadata workload only)
1404 Description: two racing renames might cause an MDS thread to deadlock
1405 Details : Running the "racer" program may cause one MDS thread to rename
1406 a file from being the source of a rename to being the target of
1407 a rename at exactly the same time that another thread is doing
1408 so, and the second thread has already enqueued these locks after
1409 doing a lookup of the target and is trying to relock them in
1410 order. Ensure that we don't try to re-lock the same resource.
1413 Frequency : only very large systems with liblustre clients
1415 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1416 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1417 all OSTs in parallel. This allows the eviction of stale liblustre
1418 clients to proceed much faster than if they were done in series,
1419 and also offers similar improvements for other set_info RPCs.
1424 Description: excessive CPU usage during initial read phase on client
1425 Details : During the initial read phase on a client, it would agressively
1426 retry readahead on the file, consuming too much CPU and impacting
1427 performance (since 1.4.5.8). Improve the readahead algorithm
1428 to avoid this, and also improve some other common cases (read
1429 of small files in particular, where "small" is files smaller than
1430 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1435 Description: MDS crash when receiving packet with unknown intent.
1436 Details : Do not LBUG in unknown intent case, just return -EFAULT
1438 Severity : enhancement
1439 Bugzilla : 9293, 9385
1440 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1441 Details : Do not serialize getattr (non-intent version) and statfs.
1444 Frequency : occasional, when OST network is overloaded/intermittent
1446 Description: client evicted by OST after bulk IO timeout
1447 Details : If a client sends a bulk IO request (read or write) the OST
1448 may evict the client if it is unresposive to its data GET/PUT
1449 request. This is incorrect if the network is overloaded (takes
1450 too long to transfer the RPC data) or dropped the OST GET/PUT
1451 request. There is no need to evict the client at all, since
1452 the pinger and/or lock callbacks will handle this, and the
1453 client can restart the bulk request.
1456 Frequency : Always when mmapping file with no objects
1458 Description: client crashes when mmapping file with no objects
1459 Details : Check that we actually have objects in a file before doing any
1460 operations on objects in ll_vm_open, ll_vm_close and
1466 Description: Request leak when working with deleted CWD
1467 Details : Introduce advanced request refcount tracking for requests
1468 referenced from lustre intent.
1470 Severity : Enhancement
1472 Description: Cache open file handles on client.
1473 Details : MDS now will return special lock along with openhandle, if
1474 requested and client is allowed to hold openhandle, even if unused,
1475 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1476 closing files for every read/write openration.
1478 Severity : Enhancement
1480 Description: Cache open negative dentries on client when possible.
1481 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1482 negative dentries on lock revocation.
1487 Description: Remounting a client read-only wasn't possible with a zconf mount
1488 Details : It wasn't possible to remount a client read-only with llmount.
1490 Severity : enhancement
1491 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1492 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1493 MPICH is included to add Lustre-specific ADIO interfaces.
1494 This is based closely on the UFS ADIO layer and only differs
1495 in file creation, in order to allow the OST striping to be set.
1496 This is user-contributed code and not supported by CFS.
1501 Description: extended inode attributes (immutable, append-only) work improperly
1502 when 2.4 and 2.6 kernels are used on client/server or vice versa
1503 Details : Introduce kernel-independent values for these flags.
1505 Severity : enhancement
1508 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1509 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1510 granularity. Now, Fractional MB granularity is be supported,
1511 this is very useful for low memory system.
1513 Severity : enhancement
1515 Description: Getattr by fid
1516 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1517 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1518 have locks and avoids getattr rpc requests for it on every stat.
1521 Frequency : Always, for filesystems larger than 2TB
1523 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1524 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1525 filesystems larger than 2TB to be mounted with mballoc enabled.
1528 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1530 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1531 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1532 filesystems larger than 2TB is mounted then the kernel will
1533 silently corrupt the start of the filesystem. CONFIG_LBD is
1534 enabled for all CFS-supported kernels, but the possibility of
1535 this happening with a modified kernel config exists.
1537 Severity : enhancement
1539 Description: add client O_DIRECT support for 2.6 kernels
1540 Details : It is now possible to do O_DIRECT reads and writes to files
1541 in the Lustre client mountpoint on 2.6 kernel clients.
1543 Severity : enhancement
1545 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1546 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1547 parallel, not waiting for response from every OST before sending
1548 a rpc to the next OST.
1553 Description: setattr vs write race when updating file timestamps
1554 Details : Client processes that update a file timestamp into the past
1555 right after writing to the file (e.g. tar) it is possible that
1556 the updated file modification time can be reset to the current
1557 time due to a race between processing the setattr and write RPC.
1559 Severity : enhancement
1561 Description: Bring 'lfs find' closer in line with regular Linux find.
1562 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1563 -print0 options and obtains all the needed info through the lustre
1566 Severity : enhancement
1568 Description: support up to 1024 configured devices on one node
1569 Details : change obd_dev array from statically allocated to dynamically
1570 allocated structs as they are first used to reduce memory usage
1575 Description: Flush dirty partially truncated pages during truncate
1576 Details : Immediatelly flush partially truncated pages in filter_setattr,
1577 this way we completely avoid having any pages in page cache on OST
1578 and can retire ugly workarounds during writes to flush such pages.
1583 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1584 Details : protect inode from truncation within vfs_unlink() context
1585 just take a reference before calling vfs_unlink() and release it
1586 when parent's i_sem is free.
1591 Description: last_id value checked outside lock on OST caused LASSERT failure
1592 Details : If there were multiple MDS->OST object precreate requests in
1593 flight, it was possible that the OST's last object id was checked
1594 outside a lock and incorrectly tripped an assertion. Move checks
1595 inside locks, and discard old precreate requests.
1598 Frequency : always, if extents are used on OSTs
1600 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1601 Details : OSTs using the extents format would not zero the high 16 bits of
1602 the index physical block number. This is not a problem for any
1603 OST filesystems smaller than 16TB, and no kernels support ext3
1604 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1605 new/modified files) and can be fixed for existing filesystems
1606 with e2fsprogs-1.39-cfs1.
1611 Description: import connection selection may be incorrect if timer wraps
1612 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1613 import connections to be ignored if the 32-bit jiffies counter
1614 wraps. Use a 64-bit jiffies counter.
1617 Frequency : very large clusters immediately after boot
1619 Description: LNET request buffers exhausted under heavy short-term load
1620 Details : If a large number of client requests are generated on a service
1621 that has previously never seen so many requests it is possible
1622 that the request buffer growth cannot keep up with the spike in
1623 demand. Instead of dropping incoming requests, they are held in
1624 the LND until the RPC service can accept more requests.
1627 Frequency : Sometimes during replay
1629 Description: Assertion failure in ll_local_open after replay.
1630 Details : If replay happened on an open request reply before we were able
1631 to set replay handler, reply will become not swabbed tripping the
1632 assertion in ll_local_open. Now we set the handler right after
1633 recognising of open request
1636 Frequency : very rare
1638 Description: kernel reports "badness in vsnprintf"
1639 Details : Reading from the "recovery_status" /proc file in small chunks
1640 may cause a negative length in lprocfs_obd_rd_recovery_status()
1641 call to vsnprintf() (which is otherwise harmless). Exit early
1642 if there is no more space in the output buffer.
1644 Severity : enhancement
1646 Description: clear OBD RPC statistics by writing to them
1647 Details : It is now possible to clear the OBD RPC statistics by writing
1648 to the "stats" file.
1653 Description: Client mtime is not the same on different clients after utimes
1654 Details : In some cases, the client was using the utimes() syscall on
1655 a file cached on another node. The clients now validate the
1656 ctime from the MDS + OSTs to determine which one is right.
1661 Description: Inability to activate failout mode
1662 Details : lconf script incorrectly assumed that in pythong string's numeric
1663 value is used in comparisons.
1666 Frequency : always with multiple stripes per file
1668 Description: Inefficient object allocation for mutli-stripe files
1669 Details : When selecting which OSTs to stripe files over, for files with
1670 a stripe count that divides evenly into the number of OSTs,
1671 the MDS is always picking the same starting OST for each file.
1672 Return the OST selection heuristic to the original design.
1677 Description: mount failures may take full timeout to return an error
1678 Details : Under some heavy load conditions it is possible that a
1679 failed mount can wait for the full obd_timeout interval,
1680 possibly several minutes, before reporting an error.
1681 Instead return an error as soon as the status is known.
1683 Frequency : quota enabled and large files being deleted
1685 Description: releasing more than 4GB of quota at once hangs OST
1686 Details : If a user deletes more than 4GB of files on a single OST it
1687 will cause the OST to spin in an infinite loop. Release
1688 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1693 Description: statfs data retrieved from /proc may be stale or zero
1694 Details : When reading per-device statfs data from /proc, in the
1695 {kbytes,files}_{total,free,avail} files, it may appear
1696 as zero or be out of date.
1699 Frequency : systems with MD RAID1 external journal devices
1701 Description: lconf's call to blkid is confused by RAID1 journal devices
1702 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1703 of returning all block devices that match the journal UUID.
1706 Frequency : always, for aggregate stripe size over 4GB
1708 Description: assertion fails when trying to use 4GB stripe size
1709 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1710 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1713 Frequency : always on ppc64
1715 Description: the first write on an ext3 filesystem with mballoc got stuck
1716 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1717 perform endianness conversion.
1719 ------------------------------------------------------------------------------
1721 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1723 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1724 INTEROPERATE with older versions automatically. Please read the
1725 user documentation before upgrading any part of a live system.
1726 * WARNING: Lustre networking configuration changes are required with
1727 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1730 * Support for kernels:
1731 2.6.9-22.0.2.EL (RHEL 4)
1732 2.6.5-7.244 (SLES 9)
1733 2.6.12.6 vanilla (kernel.org)
1736 Severity : enhancement
1737 Bugzilla : 7981/8208
1738 Description: Introduced Lustre Networking (LNET)
1739 Details : LNET is new networking infrastructure for Lustre, it includes
1740 a reorganized network configuration mode (see the user
1741 documentation for full details) as well as support for routing
1742 between different network fabrics. Lustre Networking Devices
1743 (LNDs) for the supported network fabrics have also been
1744 created for this new infrastructure.
1746 Severity : enhancement
1747 Description: Introduced Access control lists
1748 Details : clients can set ACLs on files and directories in order to have
1749 more fine-grained permissions than the standard Unix UGO+RWX.
1750 The MDS must be started with the "-o acl" mount option.
1752 Severity : enhancement
1753 Description: Introduced filesystem quotas
1754 Details : Administrators may now establish per-user quotas on the
1757 Severity : enhancement
1759 Description: Configuration change for the XT3
1760 The PTLLND is now used to run Lustre over Portals on the XT3
1761 The configure option(s) --with-cray-portals are no longer used.
1762 Rather --with-portals=<path-to-portals-includes> is used to
1763 enable building on the XT3. In addition to enable XT3 specific
1764 features the option --enable-cray-xt3 must be used.
1769 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1770 Details : A race between allocating a new llcd and re-getting the llcd_lock
1771 allowed another thread to grab newly-allocated llcd.
1773 Severity : enhancement
1775 Description: 2.6 OST async journal commit and locking fix to improve performance
1776 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1777 2.6 kernels are now async as they already were in 2.4 kernels so
1778 that they can commit concurrently with the network bulk transfer.
1779 For block-allocated files the filter allocation semaphore is held
1780 to avoid filesystem fragmentation during allocation. BKL lock
1781 removed for 2.6 xattr operations where it is no longer needed.
1786 Description: lconf incorrectly determined whether two IP networks could talk
1787 Details : In some more complicated routing and multiple-network
1788 configurations, lconf will avoid trying to make a network
1789 connection to a disjoint part of the IP space. It was doing the
1790 math incorrectly for one set of cases.
1795 Description: Fix for potential infinite loop processing records in an llog.
1796 Details : If an llog record is corrupted/zeroed, it is possible to loop
1797 forever in llog_process(). Validate the llog record length
1798 and skip the remainder of the block on error.
1801 Frequency : occasional (liblustre only)
1803 Description: liblustre could not open files whose last component is a symlink
1804 Details : sysio_path_walk() would incorrectly pass the open intent to
1805 intermediate path components.
1808 Frequency : rare (liblustre only with non-standard tuning)
1809 Bugzilla : 7201 (7350)
1810 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1811 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1812 after it has been released, only do this for referenced locks
1814 Severity : enhancement
1816 Description: specifying an (invalid) directory default stripe_size of -1
1817 would reset the directory default striping
1818 Details : stripe_size -1 was used internally to signal directory stripe
1819 removal, now use "all default" to signal dir stripe removal
1820 as a directory striping of "all default" is not useful
1823 Frequency : common for large clusters running liblustre clients
1825 Description: doing an ls when liblustre clients are running is slow
1826 Details : sending a glimpse AST to a liblustre client waits for every AST
1827 to time out, as liblustre clients will not respond. Since they
1828 cannot cache data we refresh the OST lock LVB from disk instead.
1830 Severity : enhancement
1832 Description: doing an ls at the same time as file IO can be slow
1833 Details : enqueue and other "small" requests can be blocked behind many
1834 large IO requests. Create a new OST IO portal for non-IO
1835 requests so they can be processed faster.
1838 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1840 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1841 created file with mode 000
1842 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1843 allow an ftruncate() as a non-root user to a file with mode 000.
1844 Lustre now respects this flag to disable mode checks when
1845 truncating a file owned by the user
1848 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1851 Description: Revoking locks from clients that went dead or catatonic might take
1853 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1854 cancellation of such locks instant on servers without waiting for
1855 any reply from clients. Clients drops these locks when cancel
1856 notification from server is received without replying.
1859 Frequency : liblustre-only, when liblustre client dies or becomes busy
1861 Description: Doing ls on Linux clients can take a long time with active
1863 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1864 granting such locks to it in the first place if possible. Locks
1865 are taken by proxy on the OST during the read or write and
1866 dropped immediately afterward. Add connect flags handling, do
1867 not grant locks to liblustre clients for glimpse ASTs.
1869 Severity : enhancement
1871 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1872 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1873 sequential reads and behaves badly with applications doing random
1874 reads. Improve it by reading ahead at least read region, and
1875 avoiding excessive large RPC for small reads.
1877 Severity : enhancement
1879 Description: Creating more than 1000 files for a single job may cause a load
1880 imbalance on the OSTs if there are also a large number of OSTs.
1881 Details : qos_prep_create() uses an OST index reseed value that is an
1882 even multiple of the number of available OSTs so that if the
1883 reseed happens in the middle of the object allocation it will
1884 still utilize the OSTs as uniformly as possible.
1889 Description: OST or MDS may oops in ping_evictor_main()
1890 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1891 but doesn't restart at beginning of obd_exports_timed list
1894 Severity : enhancement
1896 Description: improve by-nid export eviction on the MDS and OST
1897 Details : allow multiple exports with the same NID to be evicted at one
1898 time without re-searching the exports list.
1901 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1903 Description: MDS may oops in groups_free()
1904 Details : in rare race conditions a newly allocated group_info struct is
1905 freed again, and this can be NULL. The 2.4 compatibility code
1906 for groups_free() checked for a NULL pointer, but 2.6 did not.
1909 Frequency : common for liblustre clients doing little filesystem IO
1910 Bugzilla : 9352, 7313
1911 Description: server may evict liblustre clients accessing contended locks
1912 Details : if a client is granted a lock or receives a completion AST
1913 with a blocking AST already set it would not reply to the AST
1914 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1917 Frequency : lfs setstripe, only systems with more than 160 OSTs
1919 Description: unable to set striping with a starting offset beyond OST 160
1920 Details : llapi_create_file() incorrectly limited the starting stripe
1921 index to the maximum single-file stripe count.
1924 Frequency : LDAP users only
1926 Description: lconf did not handle in-kernel recovery with LDAP properly
1927 Details : lconf/LustreDB get_refs() is searching the wrong namespace
1929 Severity : enhancement
1931 Description: bind OST threads to NUMA nodes to improve performance
1932 Details : all OST threads are uniformly bound to CPUs on a single NUMA
1933 node and do their allocations there to localize memory access
1935 Severity : enhancement
1937 Description: llmount can determine client NID directly from Myrinet (GM)
1938 Details : the client NID code from gmnalnid was moved directly into
1939 llmount, removing the need to use this or specifying the
1940 client NID explicitly when mounting GM clients with zeroconf
1943 Frequency : if client is started with down MDS
1945 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
1946 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
1947 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
1948 ptlrpc_import_delayed_req() because we didn't check if the
1949 request was interrupted, and we also didn't break out of the
1950 event loop if there was no timeout
1955 Description: data loss during non-page-aligned writes to a single file from
1956 both multiple nodes and multiple threads on one node at same time
1957 Details : updates to KMS and lsm weren't protected by common lock. Resulting
1958 inconsistency led to false short-reads, that were cached and later
1959 used by ->prepare_write() to fill in partially written page,
1960 leading to data loss.
1963 Frequency : always, if lconf --abort_recovery used
1965 Description: lconf --abort_recovery fails with 'Operation not supported'
1966 Details : lconf was attempting to abort recovery on the MDT device and not
1969 Severity : enhancement
1971 Description: remove cleanup logs
1972 Details : replace lconf-generated cleanup logs with lustre internal
1973 cleanup routines. Eliminates the need for client-cleanup and
1976 Severity : enhancement
1978 Description: add support for EAs (user and system) on lustre filesystems
1979 Details : it is now possible to store extended attributes in the Lustre
1980 client filesystem, and with the user_xattr mount option it
1981 is possible to allow users to store EAs on their files also
1983 Severity : enhancement
1985 Description: Add possibility (config option) to show minimal available OST free
1987 Details : When compiled with --enable-mindf configure option, statfs(2)
1988 (and so, df) will return least minimal free space available from
1989 all OSTs as amount of free space on FS, instead of summary of
1990 free spaces of all OSTs.
1992 Severity : enhancement
1994 Description: do not expand extent locks acquired on OST-side
1995 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
1996 by server: they are not cached anyway.
1999 Frequency : when mmap is used/binaries executed from Lustre
2001 Description: Unmmap pages before throwing them away from read cache.
2002 Details : llap_shrink cache now attempts to unmap pages before discarding
2003 them (if unmapping failed - do not discard). SLES9 kernel has
2004 extra checks that trigger if this unmapping is not done first.
2009 Description: lconf didn't resolve symlinks before checking to see whether a
2010 given mountpoint was already in use
2013 Frequency : when migrating failover services
2014 Bugzilla : 6395, 9514
2015 Description: When migrating a subset of services from a node (e.g. failback
2016 from a failover service node) the remaining services would
2017 time out and evict clients.
2018 Details : lconf --force (implied by --failover) sets the global obd_timeout
2019 to 5 seconds in order to quickly disconnect, but this caused
2020 other RPCs to time out too quickly. Do not change the global
2021 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2023 Severity : enhancement
2024 Frequency : if MDS is started with down OST
2025 Bugzilla : 9439,5706
2026 Description: Allow startup/shutdown of an MDS without depending on the
2027 availability of the OSTs.
2028 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2029 Add appropriate locking and lov-osc refcounts for safe
2030 cleaning. Add osc abort_inflight calls in case the
2031 synchronize never started.
2034 Frequency : occasional (Cray XT3 only)
2036 Description: root not authorized to access files in CRAY_PORTALS environment
2037 Details : The client process capabilities were not honoured on the MDS in
2038 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2039 been accessed by an authorized user then root was able to access
2040 the file on the local client also. The root user capabilities
2041 are now allowed on the MDS, as this environment has secure UID.
2044 Frequency : occasional
2046 Description: ldiskfs "too long searching" message happens too often
2047 Details : A debugging message (otherwise harmless) prints too often on
2048 the OST console. This has been reduced to only happen when
2049 there are fragmentation problems on the filesystem.
2054 Description: Division by zero in statfs when all OSCs are inactive
2055 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2056 lov_statfs divides by value returned by lov_get_stripecnt().
2060 Bugzilla : 9489, 3273
2061 Description: First write from each client to each OST was only 4kB in size,
2062 to initialize client writeback cache, which caused sub-optimal
2063 RPCs and poor layout on disk for the first writen file.
2064 Details : Clients now request an initial cache grant at (re)connect time
2065 and so that they can start streaming writes to the cache right
2066 away and always do full-sized RPCs if there is enough data.
2067 If the OST is rebooted the client also re-establishes its grant
2068 so that client cached writes will be honoured under the grant.
2073 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2074 Details : Now I/O RPCs go to different portal number and (presumably) fast
2075 lock requests (and glimses) and other RPCs get their own service
2076 threads pool that should be able to service those RPCs
2079 Severity : enhancement
2081 Description: Ability to exchange lustre version between client and servers and
2082 issue warnings at client side if client is too old. Also for
2083 liblustre clients there is ability to refuse connection of too old
2085 Details : New 'version' field is added to connect data structure that is
2086 filled with version info. That info is later checked by server and
2090 Frequency : rare, liblustre only.
2091 Bugzilla : 9296, 9581
2092 Description: Two simultaneous writes from liblustre at offset within same page
2093 might proceed at the same time overwriting eachother with stale
2095 Details : I/O lock withing llu_file_prwv was released too early, before data
2096 actually was hitting the wire. Extended lock-holding time until
2097 server acknowledges receiving data.
2100 Frequency : extremely rare. Never observed in practice.
2102 Description: avoid generating lustre_handle cookie of 0.
2103 Details : class_handle_hash() generates handle cookies by incrementing
2104 global counter, and can hit 0 occasionaly (this is unlikely, but
2105 not impossible, because initial value of cookie counter is
2106 selected randonly). Value of 0 is used as a sentinel meaning
2107 "unassigned handle" --- avoid it. Also coalesce two critical
2108 sections in this function into one.
2110 Severity : enhancement
2112 Description: allow liblustre clients to delegate truncate locking to OST
2113 Details : To avoid overhead of locking, liblustre client instructs OST to
2114 take extent lock in ost_punch() on client's behalf. New connection
2115 flag is added to handle backward compatibility.
2117 Severity : enhancement
2118 Bugzilla : 4928, 7341, 9758
2119 Description: allow number of OST service threads to be specified
2120 Details : a module parameter allows the number of OST service threads
2121 to be specified via "options ost ost_num_threads={N}" in the
2122 OSS's /etc/modules.conf or /etc/modprobe.conf.
2126 Bugzilla : 6146, 9635, 9895
2127 Description: servers crash with bad pointer in target_handle_connect()
2128 Details : In rare cases when a client is reconnecting it was possible that
2129 the connection request was the last reference for that export.
2130 We would temporarily drop the export reference and get a new
2131 one, but this may have been the last reference and the export
2132 was just destroyed. Get new reference before dropping old one.
2134 Severity : enhancement
2135 Frequency : if client is started with failover MDS
2137 Description: Allow multiple MDS hostnames in the mount command
2138 Details : Try to read the configuration from all specified MDS
2139 hostnames during a client mount in case the "primary"
2142 Severity : enhancement
2144 Description: Stop sending data to evicted clients as soon as possible.
2145 Details : Check if the client we are about to send or are sending data to
2146 was evicted already. (Check is done every second of waiting,
2147 for which l_wait_event interface was extended to allow checking
2148 of exit condition at specified intervals).
2151 Frequency : rare, normally only when NFS exporting is done from client
2153 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2155 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2158 Frequency : always, if srand() is called before liblustre initialization
2160 Description: Liblustre uses system PRNG disturbing its usage by user application
2161 Details : Introduce internal to lustre fast and high-quality PRNG for
2162 lustre usage and make liblustre and some other places in generic
2163 lustre code to use it.
2165 Severity : enhancement
2166 Bugzilla : 9477, 9557, 9870
2167 Description: Verify that the MDS configuration logs are updated when xml is
2168 Details : Check if the .xml configuration logs are newer than the config
2169 logs stored on the MDS and report an error if this is the case.
2170 Request --write-conf, or allow starting with --old_conf.
2172 Severity : enhancement
2174 Description: Handle symlinks in the path when checking if Lustre is mounted.
2175 Details : Resolve intermediate symlinks when checking if a client has
2176 mounted a filesystem to avoid duplicate client mounts.
2181 Description: lconf can hit an error exception but still return success.
2182 Details : The lconf command catches the Command error exception at the top
2183 level script context and will exit with the associated exit
2184 status, but doesn't ensure that this exit status is non-zero.
2189 Description: failure of ptlrpc thread startup can cause oops
2190 Details : Starting a ptlrpc service thread can fail if there are a large
2191 number of threads or the server memory is very fragmented.
2192 Handle this without oopsing.
2195 Frequency : always, only if liblustre and non-default acceptor port was used
2197 Description: liblustre cannot connect to servers with non-default acceptor port
2198 Details : tcpnal_set_default_params() was not called and was therefore
2199 ignoring the environment varaible TCPNAL_PORT, as well as other
2200 TCPNAL_ environment variables
2205 Description: two objects could be created on the same OST for a single file
2206 Details : If an OST is down, in some cases it was possible to create two
2207 objects on a single OST for a single file. No problems other
2208 than potential performance impact and spurious error messages.
2212 Bugzilla : 5681, 9562
2213 Description: Client may oops in ll_unhash_aliases
2214 Details : Client dcache may become inconsistent in race condition.
2215 In some cases "getcwd" can fail if the current directory is
2221 Description: Inode refcounting problems in NFS export code
2222 Details : link_raw functions used to call d_instantiate without obtaining
2223 extra inode reference first.
2227 Bugzilla : 9942, 9903
2228 Description: Referencing freed requests leading to crash, memleaks with NFS.
2229 Details : We used to require that call to ll_revalidate_it was always
2230 followed by ll_lookup_it. Also with revalidate_special() it is
2231 possible to call ll_revalidate_it() twice for the same dentry
2232 even if first occurence returned success. This fix changes semantic
2233 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2234 reference on a request referred from the intent.
2235 ll_intent_release() then releases such a request.
2238 Frequency : rare, normally benchmark loads only
2240 Description: unlinked inodes were kept in memory on the client
2241 Details : If a client is repeatedly creating and unlinking files it
2242 can accumulate a lot of stale inodes in the inode slab cache.
2243 If there is no other client load running this can cause the
2244 client node to run out of memory. Instead flush old inodes
2245 from client cache that have the same inode number as a new inode.
2248 Frequency : SLES9 2.6.5 kernel and long filenames only
2249 Bugzilla : 9969, 10379
2250 Description: utime reports stale NFS file handle
2251 Details : SLES9 uses out-of-dentry names in some cases, which confused
2252 the lustre dentry revalidation. Change it to always use the
2256 Frequency : rare, unless heavy write-truncate concurrency is continuous
2257 Bugzilla : 4180, 6984, 7171, 9963, 9331
2258 Description: OST becomes very slow and/or deadlocked during object unlink
2259 Details : filter_destroy() was holding onto the parent directory lock
2260 while truncating+unlinking objects. For very large objects this
2261 may block other threads for a long time and slow overall OST
2262 responsiveness. It may also be possible to get a lock ordering
2263 deadlock in this case, or run out of journal credits because of
2264 the combined truncate+unlink. Solution is to do object truncate
2265 first in one transaction without parent lock, and then do the
2266 final unlink in a new transaction with the parent lock. This
2267 reduces the lock hold time dramatically.
2270 Frequency : rare, 2.4 kernels only
2272 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2273 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2274 before unmounting their filesystem in order to avoid deadlock.
2275 The kernel_locked() function in 2.4 kernels only checks whether
2276 the kernel lock is held, not whether it is this process that is
2277 holding it as 2.6 kernels do.
2282 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2283 Details : The client ptlrpc code may be trying to reconnect to a down
2284 server before a previous connection attempt has timed out.
2285 Increase the reconnect interval to be longer than the connection
2286 timeout interval to avoid sending duplicate connections to
2290 Frequency : echo_client brw_test command
2292 Description: fix echo_client to work with OST preallocated code
2293 Details : OST preallocation code (5137) didn't take echo_client IO path
2294 into account: echo_client calls filter methods outside of any
2295 OST thread and, hence, there is no per-thread preallocated
2296 pages and buffers to use. Solution: hijack pga pages for IO. As
2297 a byproduct, this avoids unnecessary data copying.
2301 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2302 Description: Client can oops in mdc_commit_close() after open replay
2303 Details : It was possible for the MDS to return an open request with no
2304 transaction number in mds_finish_transno() if the client was
2305 evicted, but without actually returning an error. Clients
2306 would later try to replay that open and may trip an assertion
2307 Simplify the client close codepath, and always return an error
2308 from the MDS in case the open is not successful.
2311 Frequency : rare, 2.6 OSTs only
2313 Description: OST may deadlock under high load on fragmented files
2314 Details : If there was a heavy load and highly-fragmented OST filesystems
2315 it was possible to have all the OST threads deadlock waiting on
2316 allocation of biovecs, because the biovecs were not released
2317 until the entire RPC IO was completed. Instead, release biovecs
2318 as soon as they are complete to ensure forward IO progress.
2320 Severity : enhancement
2322 Description: Support for specifying external journal device at mount
2323 Details : If an OST or MDS device is formatted with an external journal
2324 device, this device major/minor is stored in the ext3 superblock
2325 and may not be valid for failover. Allow detecting and
2326 specifying the external journal at mount time.
2331 Description: Mounting an MDS with pending unlinked files may cause oops
2332 Details : target_finish_recovery() calls mds_postrecov() which returned
2333 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2334 considers this an error and immediately begins cleaning up the
2335 lov, just after starting the mds_lov process
2337 Severity : enhancement
2339 Description: Implement 'lfs df' to report actual free space on per-OST basis
2340 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2341 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2342 usage of objects; '-h' to report in human readable format.
2344 ------------------------------------------------------------------------------
2346 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2353 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2354 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2355 at startup during orphan unlinking. Instead, skip buddy update
2356 before setup, as it will be regenerated after recovery is complete.
2361 Description: If an OST is inactive, its locks might reference stale inodes.
2362 Details : lov_change_cbdata() must iterate over all namespaces, even if
2363 they are inactive to clear inode references from the lock.
2365 Severity : enhancement
2366 Frequency : occasional, if non-standard max_dirty_mb used
2368 Description: Client will block write RPCs if not enough grant
2369 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2370 then the client will block writes while waiting for another RPC
2371 to complete instead of consuming its dirty limit. With change
2372 we get improved performance when max_dirty_mb is small.
2374 Severity : enhancement
2375 Bugzilla : 3389, 6253
2376 Description: Add support for supplementary groups on the MDS.
2377 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2378 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2379 lookups for user supplementary groups into a cache.
2383 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2384 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2385 for liblustre O_EXCL is enforced by MDS.
2389 Description: readdir never returns NULL in liblustre.
2390 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2391 dentry in struct dirent.
2395 Frequency : liblustre only, depends on application IO pattern
2396 Description: liblustre clients evicted if not contacting servers
2397 Details : Don't put liblustre clients into the ping_evictor list, so
2398 they will not be evicted by the pinger ever.
2400 Severity : enhancement
2402 Description: Add ability to evict clients by NID from MDS.
2403 Details : By echoing "nid:$NID" string into
2404 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2405 $NID would be instantly evicted from this MDS and from all active
2406 OSTs connected to it.
2410 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2411 Details : lookup_it_finish() used to query file size from OSTs that was not
2416 Description: service threads change working directory to that of init
2417 Details : Starting lustre service threads may pin the working directory
2418 of the parent thread, making that filesystem busy. Threads
2419 now change to the working directory of init to avoid this.
2423 Frequency : during shutdown only
2424 Description: shutdown with a failed MDS or OST can cause unmount to hang
2425 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2430 Frequency : 2.6 only
2431 Description: chmod/chown may include an extra supplementary group
2432 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2433 supplementary group and if none is specified this is used.
2436 Bugzilla : 5479 (6816)
2438 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2439 Details : If lookup is in progress on a file that is unlinked we might try
2440 to revalidate the inode and fail in revalidate after lookup is
2441 complete and ll_file_open() enqueues the open again but
2442 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2445 Frequency : always, if lconf --abort_recovery used
2447 Description: lconf --abort_recovery fails with 'Operation not supported'
2448 Details : lconf was attempting to abort recovery on the MDT device and not
2451 ------------------------------------------------------------------------------
2453 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2458 Frequency : rare (only unsupported configurations with a node running as an
2460 Bugzilla : 6514, 5137
2461 Description: Mounting a Lustre file system on a node running as an OST could
2463 Details : OSTs now preallocates memory needed to write out data at
2464 startup, instead of when needed, to avoid having to
2465 allocate memory in possibly low memory situations.
2466 Specifically, if the file system is mounted on on OST,
2467 memory pressure could force it to try to write out data,
2468 which it needed to allocate memory to do. Due to the low
2469 memory, it would be unable to do so and the node would
2470 become unresponsive.
2472 Severity : enhancement
2474 Description: Addition of lconf --service command line option
2475 Details : lconf now accepts a '--service <arg>' option, which is
2476 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2478 Severity : enhancement
2480 Description: Failover mode is now the default for OSTs.
2481 Details : By default, OSTs will now run in failover mode. To return to
2482 the old behaviour, add '--failout' to the lmc line for OSTs.
2484 Severity : enhancement
2486 Description: Health checks are now provided for MDS and OSTs
2487 Details : Additional detailed health check information on MSD and OSTs
2488 is now provided through the procfs health_check value.
2491 Frequency : occasional, depends on IO load
2493 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2494 after numerous create/delete cycles
2495 Details : The ext3 inode allocation policy would not allocate new inodes
2496 very well on the OSTs because there are no new directories
2497 being created. Instead we look for groups with free space if
2498 the parent directories are nearly full.
2503 Description: Network or server problems during mount may cause partially
2504 mounted clients instead of returning an error.
2505 Details : The config llog parsing code may overwrite the error return
2506 code during mount error handling, returning success instead
2512 Description: MDS can fail to allocate large reply buffers
2513 Details : After long uptimes the MDS can fail to allocate large reply
2514 buffers (e.g. zconf client mount config records) due to memory
2515 fragmentation or consumption by the buffer cache. Preallocate
2516 some large reply buffers so that these replies can be sent even
2517 under memory pressure.
2521 Frequency : rare (liblustre)
2522 Description: fsx running with liblustre complained that using truncate() to
2523 extend the file doesn't work. This patch corrects that issue.
2524 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2525 ATTR_SIZE and lsm use in llu_setattr_raw.
2529 Frequency : rare, only 2.6 kernels
2530 Description: Unusual file access patterns on the MDS may result in inode
2531 data being lost in very rare circumstances.
2532 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2533 code in the 2.6 kernel caused the inode and/or EA data not to
2534 be read from disk, causing single-file corruption.
2538 Frequency : rare, only 2.6 filesystems using extents
2539 Description: Heavy concurrent write and delete load may cause data corruption.
2540 Details : It was possible under high-load situations to have an extent
2541 metadata block in the block device cache from a just-unlinked
2542 file overwrite a newly-allocated data block. We now unmap any
2543 metadata buffers that alias just-allocated data blocks.
2547 Frequency : filesystems with default stripe_count larger than 77
2548 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2549 Details : lconf specifies an inode size of 4096 bytes when the default
2550 stripe_count is larger than 77. This conflicts with the default
2551 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2552 this case to avoid pinning too much memory for large EAs.
2554 ------------------------------------------------------------------------------
2556 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2561 Frequency : rare (extremely heavy IO load with hundreds of clients)
2563 Description: Client is evicted, gets IO error writing to file
2564 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2565 caused clients to be evicted for AST timeouts. The fixes in
2566 bug 5192 mean we no longer need to have such short AST timeouts
2567 so ldlm_timeout has been increased.
2570 Frequency : occasional during --force or --failover shutdown under load
2571 Bugzilla : 5949, 4834
2572 Description: Server oops/LBUG if stopped with --force or --failover under load
2573 Details : a collection of import/export refcount and cleanup ordering
2574 issues fixed for safer force cleanup
2577 Frequency : only filesystems larger than 120 OSTs
2578 Bugzilla : 5990, 6223
2579 Description: lfs getstripe would oops on a very large filesystem
2580 Details : lov_getconfig used kfree on vmalloc'd memory
2583 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2584 Bugzilla : 6242, 6243
2585 Description: reading from files that had been truncated to a non-zero size
2586 but never opened returned no data
2587 Details : ll_file_read() reads zeros from no-object files to EOF
2592 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2593 Details : There's little harm in aborting MDS/OSS recovery and letting it
2594 try again, so I removed the LASSERT and return an error instead.
2596 Severity : enhancement
2598 Description: New debugging infrastructure for tracking down data corruption
2599 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2600 (b) cover more of the client-side code path, and (c) try to narrow
2601 down where problems occurred
2605 Bugzilla : 3819, 4364, 4397, 6313
2606 Description: Racing close and eviction MDS could cause assertion in mds_close
2607 Details : It was possible to get multiple mfd references during close and
2608 client eviction, leading to one thread referencing a freed mfd.
2610 Severity: : enhancement
2611 Bugzilla : 3262, 6359
2612 Description: Attempts to reconnect to servers are now more aggressive.
2613 Details : This builds on the enhanced upcall-less recovery that was added
2614 in 1.4.2. When trying to reconnect to servers, clients will
2615 now try each server in the failover group every 10 seconds. By
2616 default, clients would previously try one server every 25 seconds.
2621 Description: After recovery, certain operations trigger a failed
2622 assertion on a client.
2623 Details : Failing over an mds, using lconf -d --failover, while a
2624 client was doing a readdir() call would cause the client to
2625 LBUG after recovery completed and the readdir() was resent.
2627 Severity : enhancement
2629 Description: Default groups are now added by lconf
2630 Details : You can now run lconf --group <servicename> without having to
2631 manually add groups with lmc.
2634 Frequency : occasional
2636 Description: Nodes with an elan id of 0 trigger a failed assertion
2639 Frequency : always when accessing e.g. tty/console device nodes
2641 Description: tty and some other devices nodes cannot be used on lustre
2642 Details : file's private_data field is used by device data and lustre
2643 values in there got lost. New field was added to struct file to
2644 store fs-specific private data.
2647 Frequency : when exporting Lustre via NFS
2649 Description: NFSD failed occasionally when looking up a path component
2650 Details : NFSD is looking up ".." which was broken in ext3 directories
2651 that had grown large enough to become hashed.
2654 Frequency : Clusters with multiple interfaces not on the same subnet
2656 Description: Nodes will repeatedly try to reconnect to an interface which it
2657 cannot reach and report an error to the log.
2658 Details : Extra peer list entries will be created by lconf with some peers
2659 unreachable. lconf now validates the peer before adding it.
2662 Frequency : Only if a default stripe is set on the filesystem root.
2664 Description: Setting a default stripe on the filesystem root prevented the
2665 filesystem from being remounted.
2666 Details : The client was sending extra request flags in the root getattr
2667 request and did not allocate a reply buffer for the dir EA.
2670 Frequency : occasional, higher if lots of files are accessed by one client
2671 Bugzilla : 6159, 6097
2672 Description: Client trips assertion regarding lsm mismatch/magic
2673 Details : While revalidating inodes the VFS looks up inodes with ifind()
2674 and in rare cases can find an inode that is being freed.
2675 The ll_test_inode() code will free the lsm during ifind()
2676 when it finds an existing inode and then the VFS later attaches
2677 this free lsm to a new inode.
2681 Bugzilla : 6422, 7030
2682 Description: MDS deadlock between mkdir and client eviction
2683 Details : Creating a new file via mkdir or mknod (starting a transaction
2684 and getting the ns lock) can deadlock with client eviction
2685 (gets ns lock and trying to finish a synchronous transaction).
2688 Frequency : occasional
2689 Description: While starting a server, the fsfilt_ext3 module could not be
2691 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2692 kernels. Previously, lconf would still use the ext3 name
2693 when trying to load modules. Now, it will correctly use
2694 ext3 on 2.4 and ldiskfs on 2.6.
2696 Severity : enhancement
2697 Description: The default stripe count has been changed to 1
2698 Details : The interpretation of the default stripe count (0, to lfs
2699 or lmc) has been changed to mean striping across a single
2700 OST, rather than all available. For general usage we have
2701 found a stripe count of 1 or 2 works best.
2703 Severity : enhancement
2704 Description: Add support for compiling against Cray portals.
2705 Details : Conditional compiling for some areas that are different
2709 Frequency : occasional
2710 Bugzilla : 6409, 6834
2711 Description: Creating files with an explicit stripe count may lead to
2712 a failed assertion on the MDS
2713 Details : If some OSTs are full or unavailable, creating files may
2714 trigger a failed assertion on the MDS. Now, Lustre will
2715 try to use other servers or return an error to the
2719 Frequency : occasional
2721 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2722 caused lock timeout during readahead (since 1.4.2).
2723 Details : Processes doing readahead might match a lock that hasn't been
2724 granted yet if there are overlapping and conflicting lock
2725 requests. The readahead process waits on ungranted lock
2726 (original lock is CBPENDING), while OST waits for that process
2727 to cancel CBPENDING read lock and eventually evicts client.
2729 Severity : enhancement
2731 Description: Initial enabling of flock support for clients
2732 Details : Implements fcntl advisory locking and file status functions.
2733 This feature is provided as an optional mount flag (default
2734 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2735 locking are implemented yet, and those that are are not guaranteed
2736 to be completely correct in production environments.
2737 mount -t lustre -o [flock|noflock] ...
2740 Frequency : occasional
2742 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2743 assertion in the ext3 JBD (journaling) layer.
2744 Details : The b_committed_data struct is protected by the big kernel lock
2745 in 2.4 kernels, serializing journal_commit_transaction() and
2746 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2747 access to this struct. In 2.6 kernels there is finer grained
2748 locking to improve SMP performance of the JBD layer.
2752 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2754 -----------------------------------------------------------------------------
2756 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2758 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2759 versions of Lustre. Please update all servers and clients to
2760 version 1.4.2 or later at the same time. You must also run
2761 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2762 to update the configuration logs.
2764 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2765 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2766 - swab llog records of type '0' so we get proper header size/idx (5861)
2767 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2768 - fix rename of one directory over another leaking an inode (5953)
2769 - avoid SetPageDirty on 2.6 (5981)
2770 - don't re-add just-being-destroyed locks to the waiting list (5653)
2771 - when creating new directories, inherit the parent's custom
2772 striping settings if present parent (3048)
2773 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2774 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2775 - don't hold client locks on temporary worklist from l_lru (5666)
2776 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2777 - automatically evict dead clients (5921)
2778 - Update file size properly in create+truncate+fstat case (6196)
2779 - Do not unhash mountpoint dentries, do not allow removal of
2781 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2782 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2783 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2784 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2785 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2786 - Correctly handle reads of files with no objects (6243)
2787 - lctl recover will also mark a device active if deactivate used (5933)
2789 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2790 - allow --write-conf on an MDS with different nettype than client (5619)
2791 - don't write config llogs to MDS for mounts not from that MDS (5617)
2792 - lconf should create multiple TCP connections from a client (5201)
2793 - init scripts are now turned off by default; run chkconfig --on
2794 lustre and chkconfig --on lustrefs to use them
2795 - upcalls are no longer needed for clients to recover to failover
2797 - add --abort-recovery option to lconf to abort recovery on device
2799 - add support for an arbitrary number of OSTs (3026)
2800 - Quota support protocol changes.
2801 - forward compatibility changes to wire structs (6007)
2802 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2803 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2804 - improved "lustre status" script
2805 - initialize blocksize for non-regular files (6062)
2806 - added --disable-server and --disable-client configure options (5782)
2807 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2808 - Vanilla 2.4.29 support
2809 - increase maximum number of obd devices to 520 (6242)
2810 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2811 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2812 SLES 9 2.6.5 kernels
2813 - limit stripes per file to 160 (the maximum EA size) (6093)
2815 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2818 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2819 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2820 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2821 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2822 - small ext3 extents cleanups and fixes (5733)
2823 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2824 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2825 - avoid needless client->OST connect, fix handle mismatch (5317)
2826 - fix DLM error path that led to out-of-sync client, long delays (5779)
2827 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2828 - fix several locking issues related to i_size (5492,5624,5654,5672)
2829 - don't move pending lock onto export if it is already evicted (5683)
2830 - fix kernel oops when creating .foo in unlinked directory (5548)
2831 - fix deadlock in obdfilter statistics vs. object create (5811)
2832 - use time_{before,after} to avoid timer jiffies wrap (5882)
2833 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2834 - Do not leak request if server was not able to process it (5154)
2835 - If mds_open unable to find parent dir, make that negative lookup(5154)
2836 - don't create new directories with extent-mapping (5909, 5936)
2838 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2839 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2840 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2841 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2842 - support for RHEL4 kernel on i686 (5773)
2843 - provide error messages when incompatible logs are encountered (5898)
2845 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2846 * version 1.4.0.10 (1.4.1 release candidate 1)
2848 - don't keep a lock reference when lock is not granted (4238)
2849 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2850 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2851 - drop import inflight refcount on signal_completed_replay error (5255)
2852 - unlock page after async write error during send (3677)
2853 - handle missing objects in filter_preprw_read properly (5265)
2854 - no transno return for symlink open, don't save no-trasno open (3440)
2855 - don't try to complete elan receive that already failed (4012)
2856 - free RPC server reply state on error (5406)
2857 - clean up thread from ptlrpc_start_thread() on error (5160)
2858 - readahead could read extra page into cache that wasn't ejected (5388)
2859 - prevent races in class_attach/setup/cleanup/detach (5260)
2860 - don't dereference de->d_inode after l_dput of de (5458)
2861 - use "int" for stripe value returned from lock_to_stripe (5544)
2862 - mballoc allocation and error-checking fixes in 2.6 (5504)
2863 - block device patches to fix I/O request sizes in 2.6 (5482)
2864 - look up hostnames for IB nals (5602)
2865 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2866 - don't start multiple acceptors for the same port (5277)
2867 - fix incorrect LASSERT in mds_getattr_name (5635)
2868 - export a proc file for general "ping" checking (5628)
2869 - fix "lfs check" to not block when the MDS is down (5628)
2871 - service request history (4965)
2872 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2873 - create an "evict_client" /proc entry on OSTs, like the MDS has
2874 - fix mount usage message, return errors per mount(8) (5168)
2875 - change grep [] to grep "[]" in tests so they work in more UMLs
2876 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2877 - remove ancient LOV_MAGIC_V0 EA support (5047)
2878 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2879 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2880 - allow readahead allocations to fail when low on memory (5383)
2881 - mmap locking landed again, after considerable improvement (2828)
2882 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2884 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2887 - send OST transaction number in read/write reply to free req (4966)
2888 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2889 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2890 - the watchdog thread now runs as interruptible (5246)
2891 - flock/lockf fixes (but it's still disabled, pending 5135)
2892 - don't use EXT3 constants in llite code (5094)
2893 - memory shortage at startup could cause assertion (5176)
2895 - reorganization of lov code
2896 - single portals codebase
2898 - add extents/mballoc support (5025)
2899 - direct I/O reads in the obdfilter (4048)
2900 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2902 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2905 - send OST transaction number in read/write reply to free req (4966)
2906 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2907 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2908 - the watchdog thread now runs as interruptible (5246)
2909 - handle missing objects in filter_preprw_read properly (5265)
2910 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2911 - drop import inflight refcount on signal_completed_replay error (5255)
2912 - unlock page after async write error during send (3677)
2913 - return original error code on reconstructed replies (3761)
2914 - no transno return for symlink open, don't save no-trasno open (3440)
2916 - add pid to ldlm debugging output (4922)
2917 - bump the watchdog timeouts -- we can't handle 30sec yet
2918 - extra debugging for orphan dentry/inode bug (5259)
2920 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
2923 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
2924 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
2925 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
2926 - fix NULL dereference in /proc/sys/portals/routes (4827)
2927 - allow failed mdc_close() operations to be interrupted (4561)
2928 - stop precreate on OST before MDS would time out on it (4778)
2929 - don't send partial-page writes before EOF from client (4410)
2930 - discard client grant for sub-page writes on large-page clients (4520)
2931 - don't free dentries not owned by NFS code, check generation (4806)
2932 - fix lsm leak if mds_create_objects() fails (4801)
2933 - limit debug_daemon file size, always print CERROR messages (4789)
2934 - use transno after validating reply (3892)
2935 - process timed out requests if import state changes (3754)
2936 - update mtime on OST during writes, return in glimpse (4829)
2937 - add mkfsoptions to LDAP (4679)
2938 - use ->max_readahead method instead of zapping global ra (5039)
2939 - don't interrupt __l_wait_event() during strace
2941 - add software watchdogs to catch hung threads quickly (4941)
2942 - make lustrefs init script start after nfs is mounted
2943 - fix CWARN/ERROR duplication (4930)
2944 - return async write errors to application if possible (2248)
2945 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
2946 - print NAL number in %x format (4645)
2947 - update barely-supported suse-2.4.21-171 series (4842)
2948 - support for sles 9 %post scripts
2949 - support for building 2.6 kernel-source packages
2950 - support for sles km_* packages
2952 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
2955 - ignore -ENOENT errors in osc_destroy (3639)
2956 - notify osc create thread that OSC is being cleaned up (4600)
2957 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
2958 - reconstruct ost_handle() like mds_handle() (4657)
2959 - create a new thread to do import eviction to avoid deadlock (3969)
2960 - let lconf resolve symlinked-to devices (4629)
2961 - don't unlink "objects" from directory with default EA (4554)
2962 - hold socknal file ref over connect in case target is down (4394)
2963 - allow more than 32000 subdirectories in a single directory (3244)
2964 - fix blocks count for O_DIRECT writes (3751)
2965 - OST returns ENOSPC from object create when no space left (4539)
2966 - don't send truncate RPC if file size isn't changing (4410)
2967 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
2968 - bind to privileged port in socknal and tcpnal (3689)
2970 - rate limit CERROR/CWARN console message to avoid overload (4519)
2971 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
2972 - basic mmap support (3918)
2973 - kernel patch series update from b1_4 (4711)
2975 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
2978 - avoid crash during MDS cleanup with OST shut down (2775)
2979 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
2980 - don't use bad inodes on the MDS (3744)
2981 - dynamic object preallocation to improve recovery speed (4236)
2982 - don't hold spinlock over lock dumping or change debug flags (4401)
2983 - don't zero obd_dev when it is force cleaned (3651)
2984 - print grants to console if they go negative (4431)
2985 - "lctl deactivate" will stop automatic recovery attempts (3406)
2986 - look for existing locks in ldlm_handle_enqueue() (3764)
2987 - don't resolve lock handle twice in recovery avoiding race (4401)
2988 - revalidate should check working dir is a directory (4134)
2990 - don't always mark "slow" obdfilter messages as errors (4418)
2992 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
2995 - don't close LustreDB during write_conf until it is done (3860)
2996 - fix typo in lconf for_each_profile (3821)
2997 - allow dumping logs from multiple threads at one time (3820)
2998 - don't allow multiple threads in OSC recovery (3812)
2999 - fix debug_size parameters (3864)
3000 - fix mds_postrecov to initialize import for llog ctxt (3121)
3001 - replace config semaphore with spinlock (3306)
3002 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3003 - don't allow enqueue to complete on a destroyed export (3822)
3004 - down write_lock before checking llog header bitmap (3825)
3005 - recover from lock replay timeout (3764)
3006 - up llog sem before sending rpc (3652)
3007 - reduce ns lock hold times when setting kms (3267)
3008 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3009 - fix NULL deref and obd_dev leak on setup error (3312)
3010 - replace some LBUG about llog ops with error handling (3841)
3011 - don't match INVALID dentries from d_lookup and spin (3784)
3012 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3013 - fix invalid assertion in ptlrpc_set_wait (3880)
3015 - add libwrap support for the TCP acceptor (3996)
3016 - add /proc/sys/portals/routes for non-root route listing (3994)
3017 - allow setting MDS UUID in .xml (2580)
3018 - print the stack of a process that LBUGs (4228)
3020 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3023 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3024 - make sure to unset replay flag from failed open requests (3440)
3025 - if default stripe count is 0, use OST count for inode size (3636)
3026 - update parent mtime/ctime on client for create/unlink (2611)
3027 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3028 - free recovery state on server during a forced cleanup (3571)
3029 - unregister_reply for resent reqs (3063)
3030 - loop back devices mounting and status check on 2.6 (3563)
3031 - fix resource-creation race that can provoke i_size == 0 (3513)
3032 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3033 - more debugging for page-accounting assertion (3746)
3034 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3035 - avoid deadlock after precreation failure (3758)
3036 - fix race and lock order deadlock in orphan handling (3450, 3750)
3037 - add validity checks when grabbing inodes from l_ast_data (3599)
3039 - add /proc/.../recovery_status to obdfilter (3428)
3040 - lightweight CDEBUG infrastructure, debug daemon (3668)
3041 - change default OSC RPC parameters to be better on small clusters
3042 - turn off OST read cache for files smaller than 32MB
3043 - install man pages and include them in rpms (3100)
3044 - add new init script for (un)mounting lustre filesystems (2593)
3045 - run chkconfig in %post for init scripts (3701)
3046 - drop scimac NAL (unmaintained)
3048 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3051 - clean kiobufs before and after use (3485)
3052 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3053 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3054 - change init script to fail more clearly if not run as root (1528)
3055 - allow clients to reconnect during replay (1742)
3056 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3057 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3058 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3059 - fix cond_resched (3554)
3060 - only evict unfinished clients after recovery (3515)
3061 - allow bulk resend, prevent data loss (3570)
3062 - dynamic ptlrpc request buffer allocation (2102)
3063 - don't allow unlinking open directory if it isn't empty (2904)
3064 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3065 - remove extraneous obd dereference causing LASSERT failure (3334)
3066 - don't use get_cycles() when creating temp. files on the mds (3156)
3067 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3068 - handle EEXIST for set-stripe, set proper directory name (3336)
3070 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3071 - fix ksocknal_fmb_callback() error messages (2918)
3073 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3076 - don't copy lvb into (possibly NULL) reply on error (2983)
3077 - don't deref dentry after dput, don't free lvb on error (2922)
3078 - use the kms to determine writeback rpc length (2947)
3079 - increment oti_logcookies when osc is inactive (2948)
3080 - update client's i_blocks count via lvb messages (2543)
3081 - handle intent open/close of special files properly (1557)
3082 - mount MDS with errors=remount-ro, like obdfilter (2009)
3083 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3084 - don't use cancelling-locks' kms values (2947)
3085 - use highest lock extent for kms, not last one (2925)
3086 - don't dereference ERR_PTR() dentry in error handling path (3107)
3087 - fix thread race in portals_debug_dumplog() (3122)
3088 - create lprocfs device entries at setup instead of at attach (1519)
3089 - common AST error handler, don't evict client on completion race (3145)
3090 - zero nameidata in detach_mnt in 2.6 (3118)
3091 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3092 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3093 - initialize RPC timeout timer earlier for 2.6 (3219)
3094 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3095 - print nal/nid for unknown nid (3258)
3096 - additional checks for oscc recovery before doing precreate (3284)
3097 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3098 - don't crash in mdc_close for bad permissions on open (3285)
3099 - zero i_rdev for non-device files (3147)
3100 - clear page->private before handing to FS, better assertion (3119)
3101 - tune the read pipeline (3236)
3102 - fix incorrect decref of invalidated dentry (2350)
3103 - provide read-ahead stats and refine rpc in flight stats (3328)
3104 - don't hold journal transaction open across create RPC (3313)
3105 - update atime on MDS at close time (3265)
3106 - close LDAP connection when recovering to avoid server load (3315)
3107 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3108 - don't leak open file on MDS after open resend (3325)
3109 - serialize filter_precreate and filter_destroy_precreated (3329)
3110 - loop device shouldn't call sync_dev() for nul device (3092)
3111 - clear page cache after eviction (2766)
3112 - resynchronize MDS->OST in background (2824)
3113 - refuse to mount the same filesystem twice on same mountpoint (3394)
3114 - allow llmount to create routes for mounting behind routers (3320)
3115 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3116 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3117 - fix rare problem with rename on htree directories (3417)
3119 - allow default OST striping configuration per directory (1414)
3120 - fix compilation for qswnal for 2.6 kernels (3125)
3121 - increase maximum number of MDS request buffers for large systems
3122 - change liblustreapi to be useful for external progs like lfsck (3098)
3123 - increase local configuration timeout for slow disks (3353)
3124 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3126 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3129 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3130 - don't overwrite extent policy data in reply if lock was blocked (2901)
3131 - drop filter export grants atomically with removal from device (2663)
3132 - del obd_self_export from work_list in class_disconnect_exports (2908)
3133 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3134 - swab reply message in mdc_close, other PPC fixes (2464)
3135 - fix destroying of named logs (2325)
3136 - overwrite old logs when running lconf --write_conf (2264)
3137 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3138 - fix race in target_handle_connect (2898)
3139 - mds_reint_create() should take same inode create lock (2926)
3140 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3141 - don't close files for self_export to avoid uninitialized obd (2936)
3142 - allow MDS with the same name as client node (2939)
3143 - hold dentry reference for closed log files for unlink (2325)
3144 - reserve space for all logs during transactions (2059)
3145 - don't evict page beyond end of stripe extent (2925)
3146 - don't oops on a deleted current working directory (2399)
3147 - handle hard links to targets without a parent properly (2517)
3148 - don't dereference NULL lock when racing during eviction (2867)
3149 - don't grow lock extents when lots of conflicting locks (2919)
3151 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3154 - account for cache space usage on clients to avoid data loss (974)
3155 - lfsck support in lustre kernel code (2349)
3156 - reduce journal credits needed for BRW writes (2370)
3157 - orphan handling to avoid losing space on client/server crashes
3158 - ptlrpcd can be blocked, stopping ALL progress (2477)
3159 - use lock value blocks to assist in proper KMS, faster stat (1021)
3160 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3161 - recovery for initial connections (2355)
3162 - fixes for mds_cleanup_orphans (1934)
3163 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3164 - block all file creations until orphan recovery completes (1901)
3165 - client remove rq_connection from request struct (2423)
3166 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3167 - recovery timer race (2670)
3168 - mdc_close recovey bug (2532)
3169 - ptlrpc cleanup bug (2710)
3170 - mds timeout on local locks (2588)
3171 - namespace lock held during RPCs (2431)
3172 - handle interrupted sync write properly (2503)
3173 - don't try to handle a message that hasn't been replied to (2699)
3174 - client assert failure during cleanup after abort recovery (2701)
3175 - leak mdc device after failed mount (2712)
3176 - ptlrpc_check_set allows timedout requests to complete (2714)
3177 - wait for inflight reqs when ptlrpcd finishes (2710)
3178 - make sure unregistered services are removed from the srv_list
3179 - reset bulk XID's when resending them (caught by 1138 test)
3180 - unregister_bulk after timeout
3181 - fix lconf error (2694)
3182 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3183 - readahead locks pages, leaves pending causing memory pressure (2673)
3184 - increase OST request buffers to 4096 on large machines (2729)
3185 - fix up permission of existing directories in simple_mkdir (2661)
3186 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3187 - don't assign transno to errored transactions (2742)
3188 - don't delete objects on OST if given a bogus objid from MDS (2751)
3189 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3190 - if rq_no_resend, then timeout request after recovery (2432)
3191 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3192 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3193 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3194 - MDS recovery shouldn't delete objects that it has given out (2730)
3195 - if enqueue arrives after completion, don't clobber LVB (2819)
3196 - don't unlock pages twice when trigger_group_io returns error (2814)
3197 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3198 - don't write pages to disk if there was an error (1450)
3199 - don't ping imports that have recovery disabled (2676)
3200 - take buffered bytes into account when balancing socknal conn (2817)
3201 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3202 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3203 - remove little-endian swabbing of llog records (1987)
3204 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3205 - retry reposting request buffers if they fail (1191)
3206 - grow extent at grant time to avoid granting a revoked lock (2809)
3207 - lock revoke doesn't evict page if covered by a second lock (2765)
3208 - disable VM readahead to avoid reading outside lock extents (2805)
3210 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3211 - updated kernel patches for hp-2.4.20 kernel (2681)
3213 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3216 - fix truncated write corruption (2366)
3217 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3219 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3220 - ensure len(uuid) < 37 in lmc (1171)
3221 - fix ia64 OOPS in llog_test (2255)
3222 - zero end of page at obdfilter for partial page writes (2648)
3223 - don't leave stale dentries around after renames (bug 2428)
3224 - fix timeouts when evicting a client with a single lock held (2642)
3225 - set deadline for the initial HELLO message to drain (2634)
3226 - print out dotted-quad IP addresses in the socknal (2302)
3228 - additional debugging for MDS client eviction problem (2443)
3229 - fix mkfsoptions support for osts (2603, 2604)
3231 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3234 - add series for the vanilla 2.6.0 kernel
3235 - add series for the vanilla 2.4.24 kernel
3236 - add series for a cray x86/64 UL kernel drop
3237 - fix xattr patches for the vanilla 2.4.19 series
3239 - generate true UUIDs in lmc (1171)
3240 - have portals stack dumping break in UML (2466)
3241 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3242 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3243 - wake callback waiting threads on client eviction (2460)
3244 - Add --ptldebug and --subsystem to lmc (1719)
3245 - update assertion to allow safe interrupt allocation
3246 - set rq_no_resend for cancel requests (2432)
3247 - recalculate ptlrpcd timeout after resend (2494)
3248 - call vfs_rmdir when removing pending directories (2368)
3249 - fix renaming a file to itself (2429)
3250 - lmc creates a default one-stripe lov (2454)
3251 - expand procfs space to handle large clusters (2326)
3252 - increase UML stack to avoid overflow
3253 - update lconf's list of debug and subsystem masks
3254 - fix lfs find --obd (2510)
3255 - /proc tunable for disabling filter read caching (2591)
3256 - stop rpm packages from altering slapd.conf (2301)
3257 - disable nagle in the socknal under 0conf (2578)
3258 - choose mds inode size based on stripe count (2572)
3259 - fix kernel-source rpm problems (2516)
3261 - add --disable-doc to avoid pdf generation (2421)
3262 - update documentation, tests, type-os, comments
3263 - avoid format warnings on ia64
3264 - remove the TOE NAL
3265 - tiny code cleanups by removing unused fields
3267 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3270 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3271 - remove the most blatant lies from BUILDING file (2371)
3272 - change default debug level to reasonable production setting
3273 - reduce client side cache size to reduce cache flush time
3274 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3275 - make TCP zerocopy and pinger support enabled by default (2476)
3276 - sync writes completed after process exits caused crashes (2319)
3277 - maintain correct mount count on the MDS (2356)
3278 - backout 1557, because 2316 wasn't really fixed
3279 - better file I/O statistics gathering in /proc
3280 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3281 - another kernel patch to fix zero-copy TCP function export
3282 - don't take duplicate lock when processing re-sent getattr (2420)
3283 - lctl uses obd_self_export instead of creating new conn (2353)
3284 - MDS/OST recovery case which requires object creation asserted (2425)
3285 - move lfs from /usr/sbin to /usr/bin in packages
3286 - fix race between mds_client_add and mds_client_free (2417)
3287 - use kmalloc instead of slabs in portals (2430)
3288 - don't create duplicate records when a failover MDS is present (2442)
3289 - remove unnecessary mount age check (2332)
3290 - don't remove directory inodes from locks prematurely (2451)
3291 - don't break if MDS service name is the same as hostname (2103)
3292 - fix races in client write RPC generation when cache full (2482)
3294 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3297 - remove now-unused request->rq_obd (278)
3298 - if an allocation fails, print out how much memory we've used (1933)
3299 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3300 - add the "configurable stack size" patch to most series files (1256)
3301 - ability to write large log records, for 100+ OST configs (2306)
3302 - fix NULL deref when filter_prep fails (2314)
3303 - fix operator precedence error in filter_sync
3304 - dynamic allocation of socknal TX descriptors (2315)
3305 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3306 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3307 - ensure that CWARN messages in Portals always get to the syslog
3308 - __init/__exit are not for prototype decls (ldlm_init/exit)
3309 - x86-64 compile warning fixes
3310 - fix gateway LMC keyword conflict (2318)
3311 - fix MDS lock inversions in getattr/reint paths (1844)
3312 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3313 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3314 - DLM race condition prevented some lock evictions (2328)
3315 - ENOMEM detection and retry on socknal sends (2230)
3316 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3317 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3318 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3319 - better error messages when a client is rejected during recovery (1505)
3320 - avoid cancelling locks which were never granted, after failure (2330)
3321 - fix i_sem/journal inversion in mds_client_add (2333)
3322 - fix truncate/getattr lock cycle deadlock (2334)
3323 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3324 - fix two rare exit paths which could leak an l_lock() ref (2321)
3325 - fencepost error in MDS/OST orphan recovery (2226)
3326 - make log record alignment 8 bytes (1988)
3327 - lstripe now fails when requested offset > ost_count (2237)
3328 - ensure that all kernel series have a complete list.h (1607)
3329 - fix crashes in special-file operations (2316)
3330 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3332 - allow configurable automake binary, for testing new versions
3333 - small update to the lfs documentation
3335 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3337 * fix negative export reference count in fsfilt_sync (2312)
3339 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3340 * release candidate 0.9.1
3342 - orphans are moved into the PENDING directory for possible recovery
3343 - replayed opens now open by fid for orphan/rename safety (1042)
3344 - last close of an orphan inode generates a transno (683)
3345 - chdir() and mount() now pin the directory entry (1020)
3346 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3347 - discard very old requests without processing them (1502)
3348 - remove some common, well-understood CERRORs (1505)
3349 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3350 - clear "grant" flags in OST replies until OST grant code lands (1644)
3351 - fix read performance by not clobbering i_blksize on client (1598)
3352 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3353 - make LRU size a /proc tunable, clears locks when reduced (707)
3354 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3355 - server locks take references on exports now (1558)
3356 - build fixes for 2.4.20-rh trees (1663)
3357 - return an error from lov_create if all OSCs are inactive (1751)
3358 - fix import levels when a reconnect happens without a timeout (1597)
3359 - exit early from mds_open if we get a lookup error (1749)
3360 - partial page read at EOF wouldn't wait for disk before sending (1642)
3361 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3362 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3363 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3364 - write/truncate lock inversion (1639)
3365 - Don't auto-load obdclass, portals modules during cleanup (1495)
3366 - fix timestamps from jumping to "now" (1763)
3367 - extra journal assertions (1648)
3368 - add an extra multiunlink test (1771)
3369 - fix read_record/write_record API (1776)
3370 - fix leak of offset_extent, possible incorrect i_size later (1772)
3371 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3372 - Debugging for the kqswnal_get_idle_tx problems (1820)
3373 - Allow recovery to be attempted multiple times (1536)
3374 - Write out MDS last_rcvd file after it is first created (1600)
3375 - Fix tx_descriptor leak in failed transmit situations (1827)
3376 - ext3 journaling fixes for assertion failure after IO error (1871)
3377 - class_export_put() on freed export after completion AST error (1896)
3378 - Fix revalidate looping in VFS (1322)
3379 - Don't access a freed export during MDS_REINT timeout (1521)
3380 - Add open-unlink recovery support on the MDS (1673,1764)
3381 - Return an error if no MDS data was read from last_rcvd (1946)
3382 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3383 - Don't setup a disk device that doesn't match exported UUID (317)
3384 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3385 - avoid committing NULL handle in force close
3386 - local.sh is now a one-stripe LOV configuration
3387 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3388 - fix ext3 htree duplicate directory entry corruption (1516)
3389 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3390 - update to new LOV EA format (2097)
3391 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3393 2003-06-15 Phil Schwan <phil@clusterfs.com>
3396 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3397 - per-import/export recovery handling (958, 931, 959)
3398 - multiple last-rcvd slots, for serving multiple FSes (949)
3399 - connections are again shared between multiple imp/exports (963, 964)
3400 - "umount -f" would hang if any requests needed to be sent (393, 978)
3401 - avoid pinning large req buffer by copying for queued messages (989)
3402 - add "uuid" to "lctl device" command to help upcalls (991)
3403 - "open" RPCs with transnos would confuse recovery counters (1037)
3404 - do proper endian conversion of all wire messages (288, 340, 891)
3405 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3406 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3407 - fix ptlrpc_connection leak in target_handle_connect (1174)
3408 - fix import refcounting bug in OST and MDS cleanup (1134)
3409 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3410 - fix very unlikely obd_types race condition (501)
3411 - remove osc_open hack for echo_client (1187)
3412 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3413 - a failure in read_inode2 leads to deadlock (1139)
3414 - cancel ack-locks as soon as transaction is committed (1072)
3415 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3416 - make sure to commitrw after any preprw to avoid deadlock (1162)
3417 - failing to execute a file in a lustre FS would lock inode (1203)
3418 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3419 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3420 - obd_uuid2tgt would walk off the end of the list (1255)
3421 - on IA64 the osc would give portals incorrect bulk size (1258)
3422 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3423 - fix lock inversion caused by new llite matching code (1282)
3424 - limit the number of dirty pages on a client to 10MB (1286)
3425 - timed out locks were not being corrected cancelled (1289)
3426 - fix O_DIRECT above 4GB on IA-32 (1292)
3427 * major user-visible changes
3428 - fail out/fail over policy now controlled by the upcall (993)
3430 - add OBD_PING to check server availability and failure (954)
3431 - lustre messages are now sent in sending host order (288, 340, 891)
3432 - add eadatalen to MDS getattr reply (340)
3433 - OST read replies may contain second buffer, with per-page status (593)
3435 2003-03-11 Phil Schwan <phil@clusterfs.com>
3438 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3439 - failed open()s could cause deadlock; fixed (867, 869)
3440 - stop cancelling OST locks when files are closed (481)
3441 - overlapping XID spaces caused network corruption (851, 853)
3442 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3443 - setattr_raw added, to do single-RPC, server-side setattrs
3444 - lmc/lconf syntax change for OST UUIDs
3445 - fix crashy race condition between ptlrpc_free_req and osc_close
3446 - don't use request in mdc_enqueue if we hit a timeout (889)
3447 - don't set the inode i_size for regular files from the MDS (896)
3448 - handle out of order completion AST (842)
3449 - don't LBUG if a lock request times out after receiving AST (913)
3450 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3451 - if a bad lock AST arrives, send an error instead of dropping entirely
3452 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3453 - fix leak in bulk IO when only partially completed (899, 900, 926)
3454 - fix O_DIRECT for ia64 (55)
3455 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3456 - C-z after timeout could hang a process forever; fixed (977)
3458 - client-side I/O cache (678, 924, 929, 941, 970)
3460 - READPAGE and SETATTRs which don't take server-side locks get
3463 2003-02-11 Phil Schwan <phil@clusterfs.com>
3466 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3467 - Fix fsfilt_extN_readpage() to read a full page of directory
3468 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3469 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3470 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3471 - Most intent operations were removed, in favour of a new RPC mode
3472 that does a single RPC to the server and bypasses most of the VFS
3473 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3474 - Aggressively cancel local locks on DLM servers
3475 - mds_reint_unlink sends EA to the client if it's the last nlink.
3476 client uses that EA to unlink OST objects.
3477 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3478 - recursive symlinks were fixed (439)
3479 - fixed NULL deref in DEBUG_REQ
3480 - filter_update_lastobjid no longer calls sync, which annoyed extN
3481 - fixed multi-client small-writes to a single file problem (445)
3482 - fixed mtime updates during file writes (607)
3483 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3484 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3485 - fixed Linux OST size reporting problem (444, 656)
3486 - OST now updates object mtime with writes or setattr (607, 619)
3487 - client verifies file size before zeroing page past EOF (445)
3488 - OST now writes last allocated objid to disk with allocation (108)
3489 - LOV on echo now works (409)
3491 - mds_reint_unlink sends a new buffer, with the EA included. this
3492 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3493 set if a regular file was being unlinked, and it was the last link
3494 - use PtlGet from the target for bulk writes (315)
3495 - OST now updates object mtime with writes or setattr (607, 619)
3496 - LDLM now has a grant-time callback to revalidate locked items, if
3498 - Many MDS operations were reorganized to combat race conditions
3500 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3501 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3503 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3506 - Fully reactivate OST imports after reconnection (512, others)
3507 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3508 - More graceful error handling for truncating on dead OST (515)
3509 - Don't error out unless we're actually accessing dead stripes (474)
3510 - Fix garbage sizes when stripes are missing (410)
3511 - LRU counters were broken, causing constant lock purge (433, 432)
3512 - garbage on read from stripes with failed OSTs (441)
3513 - mark OSCs as active before reconnecting during recovery (438)
3514 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3515 - lfind did not preserve OST order in output (443)
3516 - symlinks cause hung clients, incorrect data (439)
3517 - stop dereferencing request after dropping refcount (457)
3518 - don't LASSERT(spin_is_locked) on non-SMP (455)
3519 - fixes for many rename() bugs
3520 - fstat didn't correctly synchronize attributes (399)
3521 - server must handle lock cancellation during blocking AST prep (487)
3522 - bulk descriptors were free()d too soon (511)
3523 - fix paths in lconf, which would load incorrect modules (451, 507)
3524 - fix confusing lconf 'host not found' error message (386)
3525 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3526 - fix race condition in mdc_blocking_ast() for inode access (526)
3527 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3528 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3529 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3530 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3531 - fix lstripe in several ways (kernel side) (527)
3532 - fix request leak in ldlm_cli_enqueue (262)
3533 - incorrect OSC was marked inactive after OST failure
3534 - call mds_fs_cleanup before unmounting filesystem (524)
3535 - fix races between taking ns_lock and ldlm_lock_change_resource
3536 - fix races updating LOV export open file list
3537 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3538 - fix recovery NULL deref in ldlm_cli_cancel_unused
3539 - fix some DLM races by using new hash table for lock handles (419)
3540 - permit the client to specify desired inodes, at replay
3541 - duplicate requests when we queue them for replay reintegration
3542 - fix last_rcvd offset calculation
3543 - sync after each recovered transaction, so we always make progress
3544 - never, not always, ERESTART requests without transnos
3545 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3547 - skip replay if the MDS says that the client is already connected
3548 - don't check for a recovery-enabled export to match lctl's UUID
3549 - don't INC_USE_COUNT for phantom exports
3550 - don't crash when cleaning up phantom exports (567)
3551 - don't double-finish or set replay data for errored mdc_open requests
3552 - abort requests when they time out, so we don't get old replies
3553 - send/receive replies for AST messages again
3554 - if the client says that it doesn't have the lock, cancel it on the
3556 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3557 mark it as destroyed, it will all work out when decref is called
3558 - fix module use counts (22, 581)
3560 - ASTs now expect a reply (server cancels lock on error reply)
3562 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3565 - fix many simultaneous client startup (392)
3566 - fix dentry->d_it clobbering
3567 - credentials weren't being shipped for readdir/getattr operations
3568 - remove invalid assertions triggered during some concurrent MD
3570 - proper Lustre versions added (336, 389)
3571 - fix memory leak for create error case (398)
3572 - fix LOV locking bug that would get cli/srv out of sync
3573 - fix echo client over LOV (409)
3574 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3575 - fix double-O_EXCL intent crash (424)
3576 - avoid sending multiple lock CANCELs (352)
3578 - MDS can do multi-client recovery (modulo bugs in new code)
3580 - many updates, edits, cleanups
3582 2002-11-18 Phil Schwan <phil@clusterfs.com>
3585 - fix null d_it dereference (346)
3586 - fix full OST/dbench hang (333)
3587 - fix permission problem with file removal (286)
3588 - fix removal of OSCs from LOV when they fail
3589 - fix NULL deref during bulk timeout (214)
3590 - fix problems related to multiple filesystems on one MDS (241)
3591 - fixed serious subtle metadata locking bugs
3592 - free locks on clients when inodes are removed due to memory
3594 - fix inode pointer in lock data (285)
3595 - partial support for multiple MDS on a single host (241)
3596 - data locks weren't cancelled at clear_inode time (290, 311)
3597 - intent locks could lead to unbounded lock growth (205)
3598 - added a maximum lock count, an LRU list, and a flusher
3599 - fix multiple rename (365)
3600 - properly abstracted the echo client
3601 - OSC locked 1 byte too many; fixed
3602 - rewrote brw callback code:
3603 - fixed recovery bugs related to LOVs (306)
3604 - fixed too-many-pages-in-one-write crash (191)
3605 - fixed (again) crash in sync_io_timeout (214)
3606 - probably fixed callback-related race (385)
3608 - Add capability to MDS protocol
3609 - LDLM cancellations and callbacks on different portals
3611 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3614 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3615 - defer open object destruction to close time (601981, 138)
3616 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3617 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3618 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3619 - don't call obd_setattr in ll_file_release for destroyed objects
3620 * protocol change to lustre_msg: move |version| and add |flags|
3621 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3622 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3623 * added replay of create, unlink, link and rename operations during
3624 MDS failover; recovery should be much more robust now
3625 * remove failed OSCs from LOVs (only lov_create uses this so far)
3626 * the lustre-HOWTO was brought (more) up to date (582544)
3628 2002-10-23 Phil Schwan <phil@clusterfs.com>
3631 - in-use dentries weren't being reused properly (617851)
3632 - prevent multiple LDLM setup (599178)
3633 - fix LOV size calculations for truncate (617853)
3634 - fix client handling of MDS intent errors (POSIX)
3635 - fix permission bug in lovstripe.c test (624321)
3636 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3637 - truncate past end of file could corrupt data
3638 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3639 - a race in recovery could return ETIMEDOUT to apps (623947)
3640 - building outside the source directory was fixed
3641 * the lustre-HOWTO was brought (more) up to date (582544)
3642 * major progress was made on recovery functionality
3644 2002-10-10 Phil Schwan <phil@clusterfs.com>
3647 - recovery deadlock fix
3648 - rm -rf causes LBUG fix (617817)
3649 - file open by multiple tasks fix (618962)
3650 - directory permissions bugs (602707 and 620007)
3651 - journal_stop fixed with locking (611313)
3652 - O_APPEND failures resolved (618273, perhaps 614459)
3653 - lconf PATH fix (619770)
3654 - IA64 build fix (621450)
3655 - RPC buffer sizes scale with amount of memory
3657 2002-10-01 Phil Schwan <phil@clusterfs.com>
3660 - locks would be cancelled without throwing away data pages,
3661 resulting in inconsistent data (605627)
3662 - inode attributes were not always being refreshed (605627, 612449)
3663 - lconf now continues to cleanup after lctl reports an error
3664 - MDS now enforces user permissions (602707)
3665 - lprocfs cleanup fixed, but not yet enabled (614157)
3666 - fixed infinite server hang, should a client not respond to an AST
3667 - avoid going into recovery if user calls readlink() with a buffer
3668 that's too small (613941)
3669 - AST RPCs no longer require replies (614867) -- this may be changed
3670 - don't crash server if client sends an IOV that's too big (611336)
3671 - fixed lock conversion deadlock (611892)
3672 - fixed the following of symlinks (614622)
3673 * recovery: the server can remove locks from a client that dies, other
3674 clients can make progress
3675 * more extN patch fixes
3676 * compile-time configurable ptlrpc buffer allocations
3678 - collaborative read cache document
3679 - Lustre Lite Performance CDR document-in-progress
3681 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3684 - fix typo in patch-2.4.18
3686 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3689 - clear ptlrpc request each time in handle_incoming_request()
3690 - unlink of files now destroys the object on the OST
3692 2002-09-19 Peter Braam <braam@clusterfs.com>
3694 * add hard link support
3695 * change obdfile creation method
3696 * kernel patch changed
3698 2002-09-19 Peter Braam <braam@clusterfs.com>
3701 - stack overflow bug in extN fixed
3703 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3705 * documentation updates
3706 - add man pages for config tools
3707 - update tests/README to describe testing with new config tools
3708 - finish metadata API descriptions
3709 * bug fixes and cleanups
3710 - statfs workaround for 16TB limit
3711 - LOV stripe allocation improved, can stripe on subset of OSTs
3712 - LOV file size/IO offset was wrong for files > 4GB in size
3713 - object EA data was being dropped, caused files to be unreadable
3714 - memory overflow with non-LOV OST caused memory corruption
3715 - fixed regression tests to work with new config tools, obdfilter
3716 - fixed bug when directory size became larger than 1 block
3717 - fixed bug (for single client case) when PWD was deleted
3718 - invalidate local directory pages when doing intent-based ops
3719 - avoid LDLM oops when lock callback contained bad data
3721 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3723 * documentation updates
3724 * bug fixes and cleanups
3725 - configuration tools
3728 - 64-bit compile warnings
3729 - 64-bit internal statfs data
3731 * test_brw on persistent OST devices
3733 * lprocfs (disabled)
3735 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3737 * documentation updates
3738 * bug fixes and cleanups
3739 * configuration tools
3741 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3744 * many small fixes to 0_5_4
3745 * io/network handling
3746 * thinkos in MDS operations
3748 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3751 * crucial basic fixes to 0.5.3
3752 * IOR, Iozone work over Elan
3755 2002-08-07 Phil Schwan <phil@clusterfs.com>
3756 * version 0_5_3, our first alpha
3757 * we use the new Portals iovs
3758 * documentation updates
3759 * bug fixes and cleanups
3760 * small changes in the DLM wire protocol
3762 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3763 * version 0_5_1 with some initial stability,
3764 * locking on MD and file I/O.
3765 * documentation updates
3766 * several bug fixes since 0.5.0
3767 * small changes in wire protocol
3769 2002-07-18 Phil Schwan <phil@clusterfs.com>
3771 * delivered as Lustre Light Alpha
3772 * fixed a crash after handling invalid MDS requests
3773 * fixed directory pages for architectures with non-4k pages sizes
3775 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3776 * release version v0_4_4
3777 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3779 * Fixes a number of interruption problems with OST operations.
3780 * Update documentation for portals header changes
3781 * Move all wire protocol structs/defines to lustre_idl.h
3782 * Fixes symlink length bug.
3783 * Add tcpdump to repository.
3785 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3786 * release version v0_4_3
3787 * Fixes statfs for inodes on extN.
3788 * Fixes bug in runtests which would delete /etc/hosts.
3789 * Use 64-bit object IDs wherever possible (not into VFS though)
3790 Remove ost_get_info, which is unused by lustre, and out of date.
3792 2002-07-03 Peter Braam <braam@clusterfs.com>
3793 * release version v0_4_2 Fixes a lookup error (type not passed)
3794 * move forward to head of Portals
3795 * move forward to latest Lustre kernel
3797 2002-06-25 Peter Braam <braam@clusterfs.com>
3798 * release version v0_4_1. Hopefully stable on single node use.