1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10)
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4)
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.39.cfs8
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: don't use GFP_* in generic Lustre code.
18 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
22 Description: obdclass is limited by single OBD_ALLOC(idarray)
23 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
27 Description: warnings with build patchless client with vanila 2.6.19 and up
28 Details : change old ctl_table style and replace ctl_table/ctl_table_header
29 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
33 Frequency : Only for SLES
34 Description: sanity_quota fail test_1
35 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
36 file, which makes TSTID[2] inunique.
40 Frequency : Only for SLES9
41 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
42 Details : Update target/ChangeLog/which_patch .
46 Frequency : Only for test_10 in sanity-quota.sh
47 Description: a bug in quota test code
48 Details : checking a test flag in a wrong place
52 Frequency : Only for test_13 in sanity-quota.sh
53 Description: a bug in quota test script
54 Details : shouldn't check overfull allocated quota
58 Description: jbd statistics
59 Details : Port older jbd statistics patch for sles10
61 --------------------------------------------------------------------------------
63 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
65 * Support for kernels:
68 2.6.16.46-0.14 (SLES 10)
69 2.6.18.8 vanilla (kernel.org)
70 * Client support for unpatched kernels:
71 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
72 2.6.16 - 2.6.22 vanilla (kernel.org)
73 * Due to recently discovered recovery problems, we do not recommend
74 using patchless RHEL 4 clients with this or any earlier release.
75 * Recommended e2fsprogs version: 1.39.cfs8
76 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
77 * Starting with this release, the ldiskfs backing filesystem required
78 by Lustre is now in its own package, lustre-ldiskfs. This package
79 should be installed. It is versioned separately from Lustre and
80 may be released separately in future.
82 Severity : enhancement
84 Description: add optional extra BUILD_VERSION info
85 Details : add a new environment variable (namely LUSTRE_VERS) which allows
86 to override the lustre version.
89 Frequency : 2.6.18 servers only
91 Description: ll_kern_mount() doesn't release the module reference
92 Details : The ldiskfs module reference count never drops down to 0
93 because ll_kern_mount() doesn't release the module reference.
98 Description: server LBUG when using old ost_num_threads parameter
99 Details : Accept the old ost_num_threads parameter but warn that it
100 is deprecated, and fix an off-by-one error that caused an LBUG.
105 Description: Transient SCSI error results in persistent IO issue
106 Details : iobuf->dr_error is not reinitialized to 0 between two
110 Frequency : sometimes when underlying device returns I/O errors
112 Description: OSTs not going read-only during write failures
113 Details : OSTs are not remounted read-only when the journal commit threads
114 get I/O errors because fsfilt_ext3 calls journal_start/stop()
115 instead of the ext3 wrappers.
119 Description: poor connect scaling with increasing client count
120 Details : Don't run filter_grant_sanity_check for more than 100 exports
121 to improve scaling for large numbers of clients.
124 Frequency : SLES10 only
126 Description: sanity-quota.sh quotacheck failed: rc = -22
127 Details : Quotas cannot be enabled on SLES10.
130 Frequency : liblustre clients only
132 Description: getdirentries does not give error when run on compute nodes
133 Details : getdirentries does not fail when the size specified as an argument
134 is too small to contain at least one entry
136 Severity : enhancement
138 Description: Add LNET router traceability for debug purposes
139 Details : If a checksum failure occurs with a router as part of the
140 IO path, the NID of the last router that forwarded the bulk data
141 is printed so it can be identified.
146 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
147 Details : Due to a race condition, liblustre clients were occasionally
150 Severity : enhancement
152 Description: lfs setstripe use optional parameters instead of postional
155 Severity : enhancement
157 Description: Nanosecond timestamp support for ldiskfs
158 Details : The on-disk ldiskfs filesystem has added support for nanosecond
159 resolution timestamps. There is not yet support for this at
160 the Lustre filesystem level.
163 Frequency : during server recovery
165 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
166 Details : request with rq_no_resend flag not awake l_wait_event if they get a
170 Frequency : nfs export on patchless client
172 Description: connectathon hang when test nfs export over patchless client
173 Details : Disconnected dentry cannot be found with lookup, so we do not need
174 to unhash it or make it invalid
177 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
178 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
179 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
184 Description: open req refcounting wrong on reconnect
185 Details : If reconnect happened between getting open reply from server and
186 call to mdc_set_replay_data in ll_file_open, we will schedule
187 replay for unreferenced request that we are about to free.
188 Subsequent close will crash in variety of ways.
189 Check that request is still eligible for replay in
190 mdc_set_replay_data().
195 Description: disable writes to filesystem when reading health_check file
196 Details : the default for reading the health_check proc file has changed
197 to NOT do a journal transaction and write to disk, because this
198 can cause reads of the /proc file to hang and block HA state
199 checking on a healthy but otherwise heavily loaded system. It
200 is possible to return to the previous behaviour during configure
201 with --enable-health-write.
203 Severity : enhancement
205 Description: 64-bit inode version
206 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
207 to the inode. This will be required for version-based recovery.
212 Description: MDS fails to start if a duplicate client export is detected
213 Details : in some rare cases it was possible for a client to connect to
214 an MDS multiple times. Upon recovery the MDS would detect this
215 and fail during startup. Handle this more gracefully.
217 Severity : enhancement
219 Description: Add -o localflock option to simulate old noflock
221 Details : This will achieve local-only flock/fcntl locks
227 Description: log_commit_thread vs filter_destroy race leads to crash
228 Details : Take import reference before releasing llog record semaphore
233 Description: Wrong request locking in request set processing
234 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
235 imp_delayed_list, in this place should be used imp_lock.
238 Frequency : when reconnection
240 Description: Grant Leak when osc reconnect to OST
241 Details : When osc reconnect ost, OST(filter) should check whether it
242 should grant more space to client by comparing fed_grant and
243 cl_avail_grant, and return the granted space to client instead
244 of "new granted" space, because client will call osc_init_grant
245 to update the client grant space info.
248 Frequency : when client reconnect to OST
250 Description: Grant Leak when osc do resend and replay bulk write
251 Details : When osc reconnect to OST, OST(filter)should clear grant info of
252 bulk write request, because the grant info while be sync between
253 OSC and OST when reconnect, and we should ignore the grant info
254 these of resend/replay write req.
259 Description: Grant space more than avaiable left space sometimes.
260 Details : When then OST is about to be full, if two bulk writing from
261 different clients came to OST. Accord the avaliable space of the
262 OST, the first req should be permitted, and the second one
263 should be denied by ENOSPC. But if the seconde arrived before
264 the first one is commited. The OST might wrongly permit second
265 writing, which will cause grant space > avaiable space.
268 Frequency : when client is evicted
270 Description: Grant might be wrongly erased when osc is evicted by OST
271 Details : when the import is evicted by server, it will fork another
272 thread ptlrpc_invalidate_import_thread to invalidate the
273 import, where the grant will be set to 0. While the original
274 thread will update the grant it got when connecting. So if
275 the former happened latter, the grant will be wrongly errased
276 because of this race.
281 Description: Checking Stale with correct fid
282 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
283 to check whether it is stale, because sometimes, we want the
284 enqueue happened anyway, and op_data.fid2 will not be initialized.
286 Severity : enhancement
288 Description: update patchless client
289 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
292 Frequency : only with 2.4 kernel
294 Description: random memory corruption
295 Details : size of struct ll_inode_info is to big for union inode.u and this
296 can be cause of random memory corruption.
301 Description: Memory leak in recovery
302 Details : Lov_mds_md was not free in an error handler in mds_create_object.
303 It should also check obd_fail before fsfilt_start, otherwise if
304 fsfilt_start return -EROFS,(failover mds during mds recovery).
305 then the req will return with repmsg->transno = 0 and rc = EROFS.
306 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
307 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
308 be freed no matter whether fsfilt_commit success or not.
311 Frequency : only with huge count clients
313 Description: Prevents from taking the superblock lock in llap_from_page for
315 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
316 for llap_from_page call in ll_removepage prevents from taking the
317 superblock lock for a soon died page.
322 Description: Not check open intent error before release open handle
323 Details : in some rare cases, the open intent error is not checked before
324 release open handle, which may cause
325 ASSERTION(open_req->rq_transno != 0), because it tries to release
326 the failed open handle.
331 Description: Set cat log bitmap only after create log success.
332 Details : in some rare cases, the cat log bitmap is set too early. and it
333 should be set only after create log success.
337 Description: Accessing a block bevice can re-enable I/O when Lustre is
338 tearing down a device.
339 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
344 Description: service threads may hog cpus when there are a lot of requests
346 Details : Insert cond_resched to give other threads a chance to use some of
352 Description: the cat log was not initialized in recovery
353 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
354 unlink log on mds will not be initialized until mds post
355 recovery. And also in mds post recovery, the unlink log will
356 initialization will be done asynchronausly, so there will be race
357 between add unlink log and unlink log initialization.
361 Description: brw_stats were being printed incorrectly
362 Details : brw_stats were being printed as log2 but all of them were not
363 recorded as log2. Also remove some code duplication arising from
364 filter_tally_{read,write}.
368 Frequency : rare, only in recovery.
369 Description: ASSERTION(req->rq_type != LI_POISON) failed
370 Details : imp_lock should be held while iterating over imp_sending_list for
371 prevent destroy request after get timeout in ptlrpc_queue_wait.
375 Description: replay-single.sh test 52 fails
376 Details : A lock's skiplist need to be cleanup when it being unlinked
377 from its resource list.
381 Description: Short directio read returns full requested size rather than
383 Details : Direct I/O operations should return actual amount of bytes
384 transferred rather than requested size.
386 Severity : enhancement
388 Description: metadata RPC reduction (e.g. for rm performance)
389 Details : decrease the amount of synchronous RPC between clients and servers
390 by canceling conflicing lock before the operation on the client side
391 and packing thier handles into the main operation RPC to server.
393 --------------------------------------------------------------------------------
395 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
400 Frequency : on some architectures
402 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
403 Details : Uninitialized flags sometimes cause configuration commands to
407 Frequency : patchless clients only
409 Description: missing __iget() symbol export
410 Details : The __iget() symbol export is missing. To avoid the need for
411 this on patchless clients the deathrow inode reaper is turned
412 off, and we depend on the VM to clean up old inodes. This
413 dependency was during via the fix for bug 12181.
415 --------------------------------------------------------------------------------
417 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
419 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
420 INTEROPERATE with older versions automatically. In many cases a
421 special upgrade step is needed. Please read the
422 user documentation before upgrading any part of a live system.
423 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
424 version will not interoperate with 1.6 betas before beta5 (1.5.95).
425 * WARNING: Lustre configuration and startup changes are required with
426 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
431 Severity : enhancement
433 Description: MountConf
434 Details : Lustre configuration is now managed via mkfs and mount
435 commands instead of lmc and lconf. New obd types (MGS, MGC)
436 are added for dynamic configuration management. See
437 https://mail.clusterfs.com/wikis/lustre/MountConf for
440 Severity : enhancement
442 Description: dynamic OST addition
443 Details : OSTs can now be added to a live filesystem
445 Severity : enhancement
447 Description: startup order invariance
448 Details : MDTs and OSTs can be started in any order. Clients only
449 require the MDT to complete startup.
451 Severity : enhancement
453 Description: parallel, asynchronous orphan cleanup
454 Details : orphan cleanup is now performed in separate threads for each
455 OST, allowing parallel non-blocking operation.
457 Severity : enhancement
459 Description: optimized stripe assignment
460 Details : stripe assignments are now made based on ost space available,
461 ost previous usage, and OSS previous usage, in order to try
462 to optimize storage space and networking resources.
464 Severity : enhancement
466 Description: Permanently set tunables
467 Details : All writable /proc/fs/lustre tunables can now be permanently
468 set on a per-server basis, at mkfs time or on a live
471 Severity : enhancement
473 Description: Lustre message v2
474 Details : Add lustre message format v2.
476 Severity : enhancement
478 Description: client OST exclusion list
479 Details : Clients can be started with a list of OSTs that should be
480 declared "inactive" for known non-responsive OSTs.
484 Description: SPEC SFS validation failure on NFS v2 over lustre.
485 Details : Changes the blocksize for regular files to be 2x RPC size,
486 and not depend on stripe size.
488 Severity : enhancement
490 Description: Multiple MD RPCs in flight.
491 Details : Further unserialise some read-only MDS RPCs - learn about intents.
492 To avoid overly-overloading MDS, introduce a limit on number of
493 MDS RPCs in flight for a single client and add /proc controls
494 to adjust this limit.
496 Severity : enhancement
498 Description: client read/write statistics
499 Details : Add client read/write call usage stats for performance
500 analysis of user processes.
501 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
502 file access. extents_stats shows chunk size distribution.
503 extents_stats_per_process show chunk size distribution per
506 Severity : enhancement
508 Description: mds statistics
509 Details : Add detailed mds operations statistics in
510 /proc/fs/lustre/mds/*/stats.
514 Description: Failure of copying files with lustre special EAs.
515 Details : Client side always return success for setxattr call for lustre
516 special xattr (currently only "trusted.lov").
521 Description: Refcount LNET uuids
522 Details : The global LNET uuid list grew linearly with every startup;
523 refcount repeated list entries instead of always adding to
526 Severity : enhancement
528 Description: Dynamic service threads
529 Details : Within a small range, start extra service threads
530 automatically when the request queue builds up.
533 Frequency : mixed-endian client/server environments
535 Description: mixed-endian crashes
536 Details : The new msg_v2 system had some failures in mixed-endian
539 Severity : enhancement
541 Description: Easy OST removal
542 Details : OSTs can be permanently deactivated with e.g. 'lctl
543 conf_param lustre-OST0001.osc.active=0'
545 Severity : enhancement
547 Description: MGS proc entries
548 Details : Added basic proc entries for the MGS showing what filesystems
551 Severity : enhancement
553 Description: provide MGS failover
554 Details : Added config lock reacquisition after MGS server failover.
556 Severity : enhancement
558 Description: add Linux 2.4 support
559 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
563 Description: a large application tries to do I/O to the same resource and dies
565 Details : Check the req->rq_arrival time after the call to
566 ost_brw_lock_get(), but before we do anything about
567 processing it & sending the BULK transfer request. This
568 should help move old stale pending locks off the queue as
569 quickly as obd_timeout.
572 Frequency : when an incorrect nid is specified during startup
574 Description: ptlrpc connect to non-existant node causes kernel crash
575 Details : LNET can't be re-entered from an event callback, which
576 happened when we expire a message after the export has been
577 cleaned up. Instead, hand the zombie cleanup off to another
580 Severity : enhancement
582 Description: plain/inodebits lock performance improvement
583 Details : Grouping plain/inodebits in granted list by their request modes
584 and bits policy, thus improving the performance of search through
588 Frequency : only if OST filesystem is corrupted
590 Description: client incorrectly hits assertion in ptlrpc_replay_req()
591 Details : for a short time RPCs with bulk IO are in the replay list,
592 but replay of bulk IOs is unimplemented. If the OST filesystem
593 is corrupted due to disk cache incoherency and then replay is
594 started it is possible to trip an assertion. Avoid putting
595 committed RPCs into the replay list at all to avoid this issue.
598 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
600 Description: System hang on startup
601 Details : This bug allowed the liblustre (e.g. catamount) client to
602 return to the app before handling all startup RPCs. This
603 could leave the node unresponsive to lustre network traffic
604 and manifested as a server ptllnd timeout.
606 Severity : enhancement
608 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
609 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
610 boolean to control whether to print peer debug info when a
611 client's RPC times out.
614 Frequency : only for kernels with patches from Lustre below 1.4.3
616 Description: Remove old rdonly API
617 Details : Remove old rdonly API which unused from at least lustre 1.4.3
620 Frequency : only for devices with external journals
622 Description: Set external device read-only also
623 Details : During a commanded failover stop, we set the disk device
624 read-only while the server shuts down. We now also set any
625 external journal device read-only at the same time.
628 Frequency : when upgrading from 1.4 while trying to change parameters
630 Description: The wrong (new) MDC name was used when setting parameters for
631 upgraded MDT's. Also allows changing of OSC (and MDC)
632 parameters if --writeconf is specified at tunefs upgrade time.
635 Frequency : when setting specific ost indicies
637 Description: QOS code breaks on skipped indicies
638 Details : Add checks for missing OST indicies in the QOS code, so OSTs
639 created with --index need not be sequential.
641 Severity : enhancement
643 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
644 Details : The uninit_groups feature works in conjunction with the kernel
645 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
646 up the pass1 processing of e2fsck. This is a read-only feature
647 in ldiskfs2 only, so older kernels and current ldiskfs cannot
648 mount filesystems that have had this feature enabled.
650 Severity : enhancement
652 Description: Improve multi-block allocation algorithm to avoid fragmentation
653 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
654 allocation locality and avoid filesystem fragmentation.
656 ------------------------------------------------------------------------------
658 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
660 * Support for kernels:
661 2.4.21-47.0.1.EL (RHEL 3)
663 2.6.9-42.0.10.EL (RHEL 4)
664 2.6.12.6 vanilla (kernel.org)
665 2.6.16.27-0.9 (SLES 10)
666 * Recommended e2fsprogs version: 1.39.cfs5
668 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
672 Frequency : occasional, depends on client load and configuration
673 Bugzilla : 12181, 12203
674 Description: data loss for recently-modified files
676 Details : In some cases it is possible that recently written or created
677 files may not be written to disk in a timely manner (this should
678 normally be within 30s unless client IO load is very high).
679 The problem appears as zero-length files or files that are a
680 multiple of 1MB in size after a client crash or client eviction
681 that are missing data at the end of the file.
683 This problem is more likely to be hit on clients where files are
684 repeatedly created and unlinked in the same directory, clients
685 have a large amount of RAM, have many CPUs, the filesystem has
686 many OSTs, the clients are rebooted frequently, and/or the files
687 are not accessed by other nodes after being written.
689 The presence of the problem can be detected by looking at
690 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
691 smaller than the second (nr_unused) then dirty files will not
692 be flushed automatically to disk. "sync; sleep 10" should be
693 run several times on the node before unmounting it to update
694 Lustre (this is also safe to run on nodes without this problem).
696 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
697 that can cause this same problem, so customers using that kernel
698 also need to update the kernel in addition to Lustre. In order
699 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
701 It is normal that files written just before a client crash (less
702 than 30s) may not yet have been flushed to disk, even for local
706 Frequency : frequent on thin XT3 nodes
708 Description: UUID collision on thin XT3 Linux nodes
709 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
710 randomly, since we relied on an insufficiently-seeded PRNG.
715 Description: OSS hangs after "All ost request buffers busy"
716 Details : A deadlock between quota and journal operations caused OSS
717 hangs after printing "All ost request buffers busy."
720 Frequency : always on liblustre builds
722 Description: Cleanup compiler warnings on liblustre
725 Frequency : always on liblustre builds on XT3
727 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
731 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
732 > 32000 subdirectories
733 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
734 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
735 subdirectory count crosses 32000. This will aid e2fsck to
736 correctly handle more than 32000 subdirectories.
739 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
741 Description: System hang on startup
742 Details : This bug allowed the liblustre (e.g. catamount) client to
743 return to the app before handling all startup RPCs. This
744 could leave the node unresponsive to lustre network traffic
745 and manifested as a server ptllnd timeout.
747 Severity : enhancement
749 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
750 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
751 boolean to control whether to print peer debug info when a
752 client's RPC times out.
757 Description: make O_SYNC working on 2.6 kernels
758 Details : 2.6 kernels use different method for mark pages for write,
759 so need add a code to lustre for O_SYNC work.
764 Description: Failure to close file and release space on NFS
765 Details : Put inode details into lock acquired in ll_intent_file_open.
766 Use mdc_intent_lock in ll_intent_open to properly
767 detect all kind of errors unhandled by mdc_enqueue.
772 Description: proc file read during shutdown sometimes raced obd removal,
774 Details : Add lock to prevent obd access after proc file removal.
777 Frequency : Only for files larger than 4GB on 32-bit clients.
779 Description: improperly doing page alignment of locks
780 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
781 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
786 Description: RPCs being resent when they shouldn't be
787 Details : Some RPCs that should not be resent are being resent. This
788 can cause inconsistencies in the RPC state machine. Do not
789 resend such requests.
792 Frequency : rare, only with NFS export
794 Description: Crash on NFS re-export node
795 Details : under very unusual load conditions an assertion is hit in
796 ll_intent_file_open()
799 Frequency : only if OST filesystem is corrupted
801 Description: client incorrectly hits assertion in ptlrpc_replay_req()
802 Details : for a short time RPCs with bulk IO are in the replay list,
803 but replay of bulk IOs is unimplemented. If the OST filesystem
804 is corrupted due to disk cache incoherency and then replay is
805 started it is possible to trip an assertion. Avoid putting
806 committed RPCs into the replay list at all to avoid this issue.
811 Description: large O_DIRECT requests fail under memory pressure/fragmentation
812 Details : Large single O_DIRECT read and write calls can fail to allocate
813 a sufficiently large buffer to process the request. In case of
814 allocation failure the allocation is retried with a smaller
815 buffer and broken into smaller requests.
817 Severity : enhancement
819 Description: Add -o localflock option to simulate old noflock behaviour.
820 Details : This will achieve local-only flock/fcntl locks coherentness.
825 Description: versioning check is incomplete
826 Details : Checking the version difference of client vs. server, report
827 error if the gap is too big.
832 Description: add support PG_writeback bit
833 Details : add support for PG_writeback bit for Lustre, for more carefull
834 work with page cache in 2.6 kernel. This also fix some deadlocks
835 and remove hack for work O_SYNC with 2.6 kernel.
837 Severity : enhancement
839 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
840 Details : The uninit_groups feature works in conjunction with the kernel
841 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
842 up the pass1 processing of e2fsck. This is a read-only feature
843 in ldiskfs2 only, so older kernels and current ldiskfs cannot
844 mount filesystems that have had this feature enabled.
846 Severity : enhancement
848 Description: Improve multi-block allocation algorithm to avoid fragmentation
849 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
850 allocation locality and avoid filesystem fragmentation.
852 ------------------------------------------------------------------------------
854 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
856 * Support for kernels:
857 2.6.9-42.0.3.EL (RHEL 4)
859 2.4.21-47.0.1.EL (RHEL 3)
860 2.6.12.6 vanilla (kernel.org)
861 2.6.16.21-0.8 (SLES10)
862 * Recommended e2fsprogs version: 1.39.cfs2-0
864 * The backwards-compatible /proc/sys/portals symlink has been removed
865 in this release. Before upgrading, please ensure that you change
866 any configuration scripts or /etc/sysctl.conf files that access
867 /proc/sys/portals/* or sysctl portals.* to use the corresponding
868 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
869 in advance of the upgrade on any system running Lustre 1.4.6 or
870 newer, since /proc/sys/lnet was added in that version.
871 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
875 Frequency : only when quota is used
877 Description: avoid scanning export list for quota master
878 Details : Change the algorithms to avoid scanning export list in order
879 to improve the efficiency.
882 Frequency : MDS failover only, very rarely
884 Description: "went back in time" messages on mds failover
885 Details : The greatest transno may be lost when the current operation
886 finishes with an error (transno==0) and the client's last_rcvd
887 record is over-written. Save the greatest transno in the
888 mds_last_transno for this case.
891 Frequency : always for specific kernels and striping counts
893 Description: client may get "Matching packet too big" without ACL support
894 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
895 when trying to access files in certain configurations. The
896 clients should in fact be denied when mounting because they do
900 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
902 Description: many clients connecting with IO in progress causes connect timeouts
903 Details : Avoid synchronous journal commits to avoid delays caused by many
904 clients connecting/disconnecting when bulk IO is in progress.
905 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
906 OST_IO_PORTAL to avoid delays behind potentially many pending
910 Frequency : occasionally with multiple writers to a single file
912 Description: shared writes to file may result in wrong size reported by stat()
913 Details : Allow growing of kms when extent lock is cancelled
916 Frequency : always with random mmap IO to multi-striped file
918 Description: mmap write might be lost if we are writing to a 'hole' in stripe
919 Details : Only if the hole is at the end of OST object so that kms is too
920 small. Fix is to increase kms accordingly in ll_nopage.
923 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
925 Description: writes to a missing object would leak memory on the OST
926 Details : If there is an inconsistency between the MDS and OST filesystems,
927 such that the MDS references an object that doesn't exist, writes
928 to that object will leak memory due to incorrect cleanup in the
929 error handling path, eventually running out of memory on the OST.
934 Description: Creating too long symlink causes lustre errors
935 Details : Check symlink and name lengths before sending requests to MDS.
938 Frequency : only if flock is enabled (not on by default)
940 Description: posix locks not released on fd closure on 2.6.9+
941 Details : We failed to add posix locks to list of inode locks on 2.6.9+
942 kernels, this caused such locks not to be released on fd close and
943 then assertions on fs unmount about still used locks.
946 Frequency : MDS failover only, very rarely
948 Description: clients may get ASSERTION(granted_lock != NULL)
949 Details : When request was taking a long time, and a client was resending
950 a getattr by name lock request. The were multiple lock requests
951 with the same client lock handle and
952 mds_getattr_name->fixup_handle_for_resent_request found one of the
953 lock handles but later failed with ASSERTION(granted_lock != NULL).
958 Description: handle->h_buffer_credits > 0, assertion failure
959 Details : h_buffer_credits is zero after truncate, causing assertion
960 failure. This patch extends the transaction or creates a new
964 Frequency : NFS re-export or patchless client
965 Bugzilla : 11179, 10796
966 Description: Crash on NFS re-export node (__d_move)
967 Details : We do not want to hash the dentry if we don't have a lock.
968 But if this dentry is later used in d_move, we'd hit uninitialised
969 list head d_hash, so we just do this to init d_hash field but
970 leave dentry unhashed.
973 Frequency : NFS re-export or patchless client
975 Description: NFS exports has problem with symbolic link
976 Details : lustre client didn't properly install dentry when re-exported
977 to NFS or running patchless client.
980 Frequency : NFS re-export or patchless client
982 Description: Various nfs/patchless fixes.
983 Details : fixes reuse disconected alias for lookup process - this fixes
984 warning "find_exported_dentry: npd != pd",
985 fix permission error with open files at nfs.
986 fix apply umask when do revalidate.
989 Frequency : occasional
991 Description: Crash on NFS re-export node
992 Details : calling clear_page() on the wrong pointer triggered oops in
993 generic_mapping_read().
996 Frequency : rarely, using O_DIRECT IO
998 Description: unaligned directio crashes client with LASSERT
999 Details : check for unaligned buffers before trying any requests.
1002 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1004 Description: stale data returned from RAID cache
1005 Details : If only a small amount of IO is done to the RAID device before
1006 reading it again it is possible to get stale data from the RAID
1007 cache instead of reading it from disk.
1010 Frequency : always for sles10 kernel
1012 Description: sles10 support
1013 Details : ll_follow_link: compile fixes and using of nd_set_link
1014 under newer kernels.
1017 Frequency : depends on arch, kernel and compiler version, always on sles10
1020 Description: recursive or deep enough symlinks cause stack overflow
1021 Details : getting rid of large stack-allocated variable in
1025 Frequency : depends on hardware
1027 Description: lustre write performance loss in the SLES10 kernel
1028 Details : the performance loss is caused by using of write barriers in the
1029 ext3 code. The SLES10 kernel turns barrier support on by
1030 default. The fix is to undo that change for ldiskfs.
1032 ------------------------------------------------------------------------------
1034 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1036 * Support for kernels:
1037 2.6.9-42.0.3EL (RHEL 4)
1038 2.6.5-7.276 (SLES 9)
1039 2.4.21-47.0.1.EL (RHEL 3)
1040 2.6.12.6 vanilla (kernel.org)
1044 Frequency : quota enabled and large files being deleted
1046 Description: releasing more than 4GB of quota at once hangs OST
1047 Details : If a user deletes more than 4GB of files on a single OST it
1048 will cause the OST to spin in an infinite loop. Release
1049 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1054 Description: statfs data retrieved from /proc may be stale or zero
1055 Details : When reading per-device statfs data from /proc, in the
1056 {kbytes,files}_{total,free,avail} files, it may appear
1057 as zero or be out of date.
1060 Frequency : systems with MD RAID1 external journal devices
1062 Description: lconf's call to blkid is confused by RAID1 journal devices
1063 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1064 of returning all block devices that match the journal UUID.
1067 Frequency : always, for aggregate stripe size over 4GB
1069 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1070 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1071 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1074 Frequency : always if "lfs find" used on a local file/directory
1076 Description: "lfs find" segfaults if used on a local file/directory
1077 Details : The case where a directory component was not specified wasn't
1078 handled correctly. Handle this properly.
1081 Frequency : always on ppc64
1083 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1084 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1085 perform endianness conversion.
1088 Frequency : rarely (truncate to non-zero file size after write under load)
1089 Bugzilla : 10730, 10687
1090 Description: Files padded with zeros to next 4K multiple
1091 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1092 it is possible that files that are truncated to a non-zero size
1093 immediately after being written are filled with zero bytes beyond
1094 the truncated size. No file data is lost.
1096 Severity : enhancement
1098 Description: Allow recovery/failover for liblustre clients.
1099 Details : liblustre clients were unaware of failover configurations until
1102 Severity : enhancement
1104 Description: user file locks should fail when not mounting with flock option
1105 Details : Set up an error-returning stub in ll_file_operations.lock field
1106 to prevent incorrect behaviour when client is mounted without
1107 flock option. Also, set up properly f_op->flock field for
1111 Frequency : always on ia64
1113 Description: "lfs df" loops on printing out MDS statfs information
1114 Details : The obd_ioctl_data was not initialized and in some systems
1115 this caused a failure during the ioctl that did not return
1116 an error. Initialize the struct and return an error on failure.
1119 Frequency : SLES 9 only
1121 Description: Error of copying files with lustre special EAs as root
1122 Details : Client side always return success for setxattr call for lustre
1123 special xattr (currently only "trusted.lov").
1126 Frequency : rarely on clusters with both ia64+i386 clients
1128 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1129 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1130 clients are doing concurrent writes to the same file it is
1131 possible that the ia64 clients may LASSERT because the OST
1132 extent locks are not PAGE_SIZE aligned. Ensure that grown
1133 locks are always aligned on the request boundary.
1136 Frequency : specific use, occasional
1138 Description: Overwriting in use executable truncates on-disk binary image
1139 Details : If one node attempts to overwrite an executable in use by
1140 another node, we now correctly return ETXTBSY instead of
1141 truncating the file.
1143 Severity : enhancement
1145 Description: Async OSC create to avoid the blocking unnecessarily.
1146 Details : If a OST has no remain object, system will block on the creating
1147 when need to create a new object on this OST. Now, ways use
1148 pre-created objects when available, instead of blocking on an
1149 empty osc while others are not empty. If we must block, we block
1150 for the shortest possible period of time.
1155 Description: chmod on Lustre root is propagated to other clients
1156 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1157 invalid by the follow_mount time.
1160 Frequency : liblustre clients only
1162 Description: Race in 'instant cancel' lock handling could lead to such locks
1163 never to be granted in case of SMP MDS
1164 Details : Do not destroy not yet granted but cbpending locks in
1168 Frequency : replay/resend of open
1170 Description: non null lock assetion failure in mds_intent_policy
1171 Details : Trying to replay/resend lockless open requests resulted in
1172 mds_open() returning 0 with no lock. Now it sets a flag if
1173 a lock is going to be returned.
1175 Severity : enhancement
1177 Description: Checksum enhancements
1178 Details : New checksum enhancements allow for resending RPCs that failed
1181 Severity : enhancement
1183 Description: Tunables on number of dirty pages in cacche
1184 Details : Allow to set limit on number of dirty pages cached.
1189 Description: client crash on unmount - lock still has references
1190 Details : In some error handling cases it was possible to leak a lock
1191 reference on a client while accessing a file. This was not
1192 harmful to the client during operation, but would cause the
1193 client to crash when the filesystem is unmounted.
1196 Frequency : specific case, rare
1198 Description: ETXTBSY on mds though file not in use by client
1199 Details : ETXTBSY is no longer incorrectly returned when attempting to
1200 chmod or chown a directory that the user previously tried to
1201 execute or a currently-executing binary.
1204 Frequency : extremely rare except on liblustre-based clients
1206 Description: Lustre space not freed when files are deleted
1207 Details : Clean up open-unlinked files after client eviction. Previously
1208 the unlink was skipped and the files remained as orphans.
1213 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1214 Details : In some cases it was possible to send a blocking callback to
1215 a client doing a glimpse, even though that client didn't get
1216 a lock granted. When the glimpse lock is cancelled on the OST
1217 the freed lock is left on the waiting list and corrupted the list.
1220 Frequency : all core dumps
1222 Description: Broke core dumps to lustre
1223 Details : Negative dentry may be unhashed if parent does not have UPDATE
1224 lock, but some callers, e.g. do_coredump, expect dentry to be
1225 hashed after successful create, hash it in ll_create_it.
1227 ------------------------------------------------------------------------------
1229 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1231 * Support for kernels:
1232 2.6.9-42.0.2.EL (RHEL 4)
1233 2.6.5-7.276 (SLES 9)
1234 2.4.21-40.EL (RHEL 3)
1235 2.6.12.6 vanilla (kernel.org)
1239 Frequency : always on RHEL 3
1241 Description: Number of open files grows over time
1242 Details : The number of open files grows over time, whether or not
1243 Lustre is started. This was due to a filp leak introduced
1244 by one of our kernel patches.
1246 ------------------------------------------------------------------------------
1248 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1250 * Support for kernels:
1251 2.6.9-42.EL (RHEL 4)
1252 2.6.5-7.276 (SLES 9)
1253 2.4.21-40.EL (RHEL 3)
1254 2.6.12.6 vanilla (kernel.org)
1259 Bugzilla : 5719, 9635, 9792, 9684
1260 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1261 Details : If a server is under heavy load and cannot reply to new
1262 connection requests before the client resends the (re)connect,
1263 the connection handling code can behave badly if two service
1264 threads are concurrently handing separate (re)connections from
1265 the same client. Add better locking to the connection handling
1266 code, and ensure that only a single connection will be processed
1267 for a given client UUID, even if the lock is dropped.
1269 Severity : enhancement
1271 Description: add TCP zero-copy support to kernel
1272 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1273 sends if the hardware supports scatter-gather and checksumming.
1274 This allows socklnd to do client-write and server-read more
1275 efficiently and reduce CPU utilization from skbuf copying.
1278 Frequency : only if NFS exporting from client
1280 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1281 Details : If setting attributes on a file created under NFS that had
1282 never been opened it would be possible to oops the client
1283 if the file had no objects.
1286 Frequency : always for liblustre
1288 Description: liblustre client does MDS+OSTs setattr RPC for each write
1289 Details : When doing a write from a liblustre client, the client
1290 incorrectly issued an RPC to the MDS and each OST the file was
1291 striped over in order to update the timestamps. When writing
1292 with small chunks and many clients this could overwhelm the MDS
1293 with RPCs. In all cases it would slow down the write because
1294 these RPCs are unnecessary.
1296 Severity : enhancement
1298 Description: allow number of MDS service threads to be changed at module load
1299 Details : It is now possible to change the number of MDS service threads
1300 running. Adding "options mds mds_num_threads={N}" to the MDS's
1301 /etc/modprobe.conf will set the number of threads for the next
1302 time Lustre is restarted (assuming the "mds" module is also
1303 reloaded at that time). The default number of threads will
1304 stay the same, 32 for most systems.
1309 Description: OST crash if filesystem is unformatted or corrupt
1310 Details : If an OST is started on a device that has never been formatted
1311 or if the filesystem is corrupt and cannot even mount then the
1312 error handling cleanup routines would dereference a NULL pointer.
1317 Description: NULL pointer deref in llap_from_page.
1318 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1319 incorrect) mapping if the page was truncated/reclaimed while it was
1320 searched for. Check for this condition and skip such pages when
1321 doing readahead. Introduce extra check to llap_from_page() to
1322 verify page->mapping->host is non-NULL (so page is not anonymous).
1325 Frequency : Sometimes when using sys_sendfile
1327 Description: "page not covered by a lock" warnings from ll_readpage
1328 Details : sendfile called ll_readpage without right page locks present.
1329 Now we introduced ll_file_sendfile that does necessary locking
1330 around call to generic_file_sendfile() much like we do in
1334 Frequency : with certain MDS communication failures at client mount time
1336 Description: NULL pointer deref after failed client mount
1337 Details : a client connection request may delayed by the network layer
1338 and not be sent until after the PTLRPC layer has timed out the
1339 request. If the client fails the mount immediately it will try
1340 to clean up before the network times out the request. Add a
1341 reference from the request import to the obd device and delay
1342 the cleanup until the network drops the request.
1345 Frequency : occasionally during client (re)connect
1347 Description: assertion failure during client (re)connect
1348 Details : processing a client connection request may be delayed by the
1349 client or server longer than the client connect timeout. This
1350 causes the client to resend the connection request. If the
1351 original connection request is replied in this interval, the
1352 client may trip an assertion failure in ptlrpc_connect_interpret()
1353 which thought it would be the only running connect process.
1356 Frequency : only with obd_echo servers and clients that are rebooted
1358 Description: kernel BUG accessing uninitialized data structure
1359 Details : When running an obd_echo server it did not start the ping_evictor
1360 thread, and when a client was evicted an uninitialized data
1361 structure was accessed. Start the ping_evictor in the RPC
1362 service startup instead of the OBD startup.
1364 Severity : enhancement
1365 Bugzilla : 10193 (patchless)
1366 Description: Remove dependency on various unexported kernel interfaces.
1367 Details : No longer need reparent_to_init, exit_mm, exit_files,
1368 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1371 Frequency : rare (only users of deprecated and unsupported LDAP config)
1373 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1374 Details : LDAP apparently contains 'lustreName' attributes instead of
1375 'name'. A simple remapping of the name is sufficient.
1378 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1380 Description: waiting_locks_callback trips kernel BUG if client is evicted
1381 Details : Running with the dump_on_timeout debug flag turned on makes
1382 it possible that the waiting_locks_callback() can try to dump
1383 the Lustre kernel debug logs from an interrupt handler. Defer
1384 this log dumping to the expired_lock_main() thread.
1386 Severity : enhancement
1388 Description: Support NFS exporting on 2.6 kernels.
1389 Details : Implement non-rawops metadata methods for NFS server to use without
1390 changing NFS server code.
1393 Frequency : very rare (synthetic metadata workload only)
1395 Description: two racing renames might cause an MDS thread to deadlock
1396 Details : Running the "racer" program may cause one MDS thread to rename
1397 a file from being the source of a rename to being the target of
1398 a rename at exactly the same time that another thread is doing
1399 so, and the second thread has already enqueued these locks after
1400 doing a lookup of the target and is trying to relock them in
1401 order. Ensure that we don't try to re-lock the same resource.
1404 Frequency : only very large systems with liblustre clients
1406 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1407 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1408 all OSTs in parallel. This allows the eviction of stale liblustre
1409 clients to proceed much faster than if they were done in series,
1410 and also offers similar improvements for other set_info RPCs.
1415 Description: excessive CPU usage during initial read phase on client
1416 Details : During the initial read phase on a client, it would agressively
1417 retry readahead on the file, consuming too much CPU and impacting
1418 performance (since 1.4.5.8). Improve the readahead algorithm
1419 to avoid this, and also improve some other common cases (read
1420 of small files in particular, where "small" is files smaller than
1421 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1426 Description: MDS crash when receiving packet with unknown intent.
1427 Details : Do not LBUG in unknown intent case, just return -EFAULT
1429 Severity : enhancement
1430 Bugzilla : 9293, 9385
1431 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1432 Details : Do not serialize getattr (non-intent version) and statfs.
1435 Frequency : occasional, when OST network is overloaded/intermittent
1437 Description: client evicted by OST after bulk IO timeout
1438 Details : If a client sends a bulk IO request (read or write) the OST
1439 may evict the client if it is unresposive to its data GET/PUT
1440 request. This is incorrect if the network is overloaded (takes
1441 too long to transfer the RPC data) or dropped the OST GET/PUT
1442 request. There is no need to evict the client at all, since
1443 the pinger and/or lock callbacks will handle this, and the
1444 client can restart the bulk request.
1447 Frequency : Always when mmapping file with no objects
1449 Description: client crashes when mmapping file with no objects
1450 Details : Check that we actually have objects in a file before doing any
1451 operations on objects in ll_vm_open, ll_vm_close and
1457 Description: Request leak when working with deleted CWD
1458 Details : Introduce advanced request refcount tracking for requests
1459 referenced from lustre intent.
1461 Severity : Enhancement
1463 Description: Cache open file handles on client.
1464 Details : MDS now will return special lock along with openhandle, if
1465 requested and client is allowed to hold openhandle, even if unused,
1466 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1467 closing files for every read/write openration.
1469 Severity : Enhancement
1471 Description: Cache open negative dentries on client when possible.
1472 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1473 negative dentries on lock revocation.
1478 Description: Remounting a client read-only wasn't possible with a zconf mount
1479 Details : It wasn't possible to remount a client read-only with llmount.
1481 Severity : enhancement
1482 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1483 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1484 MPICH is included to add Lustre-specific ADIO interfaces.
1485 This is based closely on the UFS ADIO layer and only differs
1486 in file creation, in order to allow the OST striping to be set.
1487 This is user-contributed code and not supported by CFS.
1492 Description: extended inode attributes (immutable, append-only) work improperly
1493 when 2.4 and 2.6 kernels are used on client/server or vice versa
1494 Details : Introduce kernel-independent values for these flags.
1496 Severity : enhancement
1499 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1500 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1501 granularity. Now, Fractional MB granularity is be supported,
1502 this is very useful for low memory system.
1504 Severity : enhancement
1506 Description: Getattr by fid
1507 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1508 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1509 have locks and avoids getattr rpc requests for it on every stat.
1512 Frequency : Always, for filesystems larger than 2TB
1514 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1515 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1516 filesystems larger than 2TB to be mounted with mballoc enabled.
1519 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1521 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1522 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1523 filesystems larger than 2TB is mounted then the kernel will
1524 silently corrupt the start of the filesystem. CONFIG_LBD is
1525 enabled for all CFS-supported kernels, but the possibility of
1526 this happening with a modified kernel config exists.
1528 Severity : enhancement
1530 Description: add client O_DIRECT support for 2.6 kernels
1531 Details : It is now possible to do O_DIRECT reads and writes to files
1532 in the Lustre client mountpoint on 2.6 kernel clients.
1534 Severity : enhancement
1536 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1537 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1538 parallel, not waiting for response from every OST before sending
1539 a rpc to the next OST.
1544 Description: setattr vs write race when updating file timestamps
1545 Details : Client processes that update a file timestamp into the past
1546 right after writing to the file (e.g. tar) it is possible that
1547 the updated file modification time can be reset to the current
1548 time due to a race between processing the setattr and write RPC.
1550 Severity : enhancement
1552 Description: Bring 'lfs find' closer in line with regular Linux find.
1553 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1554 -print0 options and obtains all the needed info through the lustre
1557 Severity : enhancement
1559 Description: support up to 1024 configured devices on one node
1560 Details : change obd_dev array from statically allocated to dynamically
1561 allocated structs as they are first used to reduce memory usage
1566 Description: Flush dirty partially truncated pages during truncate
1567 Details : Immediatelly flush partially truncated pages in filter_setattr,
1568 this way we completely avoid having any pages in page cache on OST
1569 and can retire ugly workarounds during writes to flush such pages.
1574 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1575 Details : protect inode from truncation within vfs_unlink() context
1576 just take a reference before calling vfs_unlink() and release it
1577 when parent's i_sem is free.
1582 Description: last_id value checked outside lock on OST caused LASSERT failure
1583 Details : If there were multiple MDS->OST object precreate requests in
1584 flight, it was possible that the OST's last object id was checked
1585 outside a lock and incorrectly tripped an assertion. Move checks
1586 inside locks, and discard old precreate requests.
1589 Frequency : always, if extents are used on OSTs
1591 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1592 Details : OSTs using the extents format would not zero the high 16 bits of
1593 the index physical block number. This is not a problem for any
1594 OST filesystems smaller than 16TB, and no kernels support ext3
1595 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1596 new/modified files) and can be fixed for existing filesystems
1597 with e2fsprogs-1.39-cfs1.
1602 Description: import connection selection may be incorrect if timer wraps
1603 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1604 import connections to be ignored if the 32-bit jiffies counter
1605 wraps. Use a 64-bit jiffies counter.
1608 Frequency : very large clusters immediately after boot
1610 Description: LNET request buffers exhausted under heavy short-term load
1611 Details : If a large number of client requests are generated on a service
1612 that has previously never seen so many requests it is possible
1613 that the request buffer growth cannot keep up with the spike in
1614 demand. Instead of dropping incoming requests, they are held in
1615 the LND until the RPC service can accept more requests.
1618 Frequency : Sometimes during replay
1620 Description: Assertion failure in ll_local_open after replay.
1621 Details : If replay happened on an open request reply before we were able
1622 to set replay handler, reply will become not swabbed tripping the
1623 assertion in ll_local_open. Now we set the handler right after
1624 recognising of open request
1627 Frequency : very rare
1629 Description: kernel reports "badness in vsnprintf"
1630 Details : Reading from the "recovery_status" /proc file in small chunks
1631 may cause a negative length in lprocfs_obd_rd_recovery_status()
1632 call to vsnprintf() (which is otherwise harmless). Exit early
1633 if there is no more space in the output buffer.
1635 Severity : enhancement
1637 Description: clear OBD RPC statistics by writing to them
1638 Details : It is now possible to clear the OBD RPC statistics by writing
1639 to the "stats" file.
1644 Description: Client mtime is not the same on different clients after utimes
1645 Details : In some cases, the client was using the utimes() syscall on
1646 a file cached on another node. The clients now validate the
1647 ctime from the MDS + OSTs to determine which one is right.
1652 Description: Inability to activate failout mode
1653 Details : lconf script incorrectly assumed that in pythong string's numeric
1654 value is used in comparisons.
1657 Frequency : always with multiple stripes per file
1659 Description: Inefficient object allocation for mutli-stripe files
1660 Details : When selecting which OSTs to stripe files over, for files with
1661 a stripe count that divides evenly into the number of OSTs,
1662 the MDS is always picking the same starting OST for each file.
1663 Return the OST selection heuristic to the original design.
1668 Description: mount failures may take full timeout to return an error
1669 Details : Under some heavy load conditions it is possible that a
1670 failed mount can wait for the full obd_timeout interval,
1671 possibly several minutes, before reporting an error.
1672 Instead return an error as soon as the status is known.
1674 Frequency : quota enabled and large files being deleted
1676 Description: releasing more than 4GB of quota at once hangs OST
1677 Details : If a user deletes more than 4GB of files on a single OST it
1678 will cause the OST to spin in an infinite loop. Release
1679 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1684 Description: statfs data retrieved from /proc may be stale or zero
1685 Details : When reading per-device statfs data from /proc, in the
1686 {kbytes,files}_{total,free,avail} files, it may appear
1687 as zero or be out of date.
1690 Frequency : systems with MD RAID1 external journal devices
1692 Description: lconf's call to blkid is confused by RAID1 journal devices
1693 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1694 of returning all block devices that match the journal UUID.
1697 Frequency : always, for aggregate stripe size over 4GB
1699 Description: assertion fails when trying to use 4GB stripe size
1700 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1701 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1704 Frequency : always on ppc64
1706 Description: the first write on an ext3 filesystem with mballoc got stuck
1707 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1708 perform endianness conversion.
1710 ------------------------------------------------------------------------------
1712 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1714 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1715 INTEROPERATE with older versions automatically. Please read the
1716 user documentation before upgrading any part of a live system.
1717 * WARNING: Lustre networking configuration changes are required with
1718 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1721 * Support for kernels:
1722 2.6.9-22.0.2.EL (RHEL 4)
1723 2.6.5-7.244 (SLES 9)
1724 2.6.12.6 vanilla (kernel.org)
1727 Severity : enhancement
1728 Bugzilla : 7981/8208
1729 Description: Introduced Lustre Networking (LNET)
1730 Details : LNET is new networking infrastructure for Lustre, it includes
1731 a reorganized network configuration mode (see the user
1732 documentation for full details) as well as support for routing
1733 between different network fabrics. Lustre Networking Devices
1734 (LNDs) for the supported network fabrics have also been
1735 created for this new infrastructure.
1737 Severity : enhancement
1738 Description: Introduced Access control lists
1739 Details : clients can set ACLs on files and directories in order to have
1740 more fine-grained permissions than the standard Unix UGO+RWX.
1741 The MDS must be started with the "-o acl" mount option.
1743 Severity : enhancement
1744 Description: Introduced filesystem quotas
1745 Details : Administrators may now establish per-user quotas on the
1748 Severity : enhancement
1750 Description: Configuration change for the XT3
1751 The PTLLND is now used to run Lustre over Portals on the XT3
1752 The configure option(s) --with-cray-portals are no longer used.
1753 Rather --with-portals=<path-to-portals-includes> is used to
1754 enable building on the XT3. In addition to enable XT3 specific
1755 features the option --enable-cray-xt3 must be used.
1760 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1761 Details : A race between allocating a new llcd and re-getting the llcd_lock
1762 allowed another thread to grab newly-allocated llcd.
1764 Severity : enhancement
1766 Description: 2.6 OST async journal commit and locking fix to improve performance
1767 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1768 2.6 kernels are now async as they already were in 2.4 kernels so
1769 that they can commit concurrently with the network bulk transfer.
1770 For block-allocated files the filter allocation semaphore is held
1771 to avoid filesystem fragmentation during allocation. BKL lock
1772 removed for 2.6 xattr operations where it is no longer needed.
1777 Description: lconf incorrectly determined whether two IP networks could talk
1778 Details : In some more complicated routing and multiple-network
1779 configurations, lconf will avoid trying to make a network
1780 connection to a disjoint part of the IP space. It was doing the
1781 math incorrectly for one set of cases.
1786 Description: Fix for potential infinite loop processing records in an llog.
1787 Details : If an llog record is corrupted/zeroed, it is possible to loop
1788 forever in llog_process(). Validate the llog record length
1789 and skip the remainder of the block on error.
1792 Frequency : occasional (liblustre only)
1794 Description: liblustre could not open files whose last component is a symlink
1795 Details : sysio_path_walk() would incorrectly pass the open intent to
1796 intermediate path components.
1799 Frequency : rare (liblustre only with non-standard tuning)
1800 Bugzilla : 7201 (7350)
1801 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1802 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1803 after it has been released, only do this for referenced locks
1805 Severity : enhancement
1807 Description: specifying an (invalid) directory default stripe_size of -1
1808 would reset the directory default striping
1809 Details : stripe_size -1 was used internally to signal directory stripe
1810 removal, now use "all default" to signal dir stripe removal
1811 as a directory striping of "all default" is not useful
1814 Frequency : common for large clusters running liblustre clients
1816 Description: doing an ls when liblustre clients are running is slow
1817 Details : sending a glimpse AST to a liblustre client waits for every AST
1818 to time out, as liblustre clients will not respond. Since they
1819 cannot cache data we refresh the OST lock LVB from disk instead.
1821 Severity : enhancement
1823 Description: doing an ls at the same time as file IO can be slow
1824 Details : enqueue and other "small" requests can be blocked behind many
1825 large IO requests. Create a new OST IO portal for non-IO
1826 requests so they can be processed faster.
1829 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1831 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1832 created file with mode 000
1833 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1834 allow an ftruncate() as a non-root user to a file with mode 000.
1835 Lustre now respects this flag to disable mode checks when
1836 truncating a file owned by the user
1839 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1842 Description: Revoking locks from clients that went dead or catatonic might take
1844 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1845 cancellation of such locks instant on servers without waiting for
1846 any reply from clients. Clients drops these locks when cancel
1847 notification from server is received without replying.
1850 Frequency : liblustre-only, when liblustre client dies or becomes busy
1852 Description: Doing ls on Linux clients can take a long time with active
1854 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1855 granting such locks to it in the first place if possible. Locks
1856 are taken by proxy on the OST during the read or write and
1857 dropped immediately afterward. Add connect flags handling, do
1858 not grant locks to liblustre clients for glimpse ASTs.
1860 Severity : enhancement
1862 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1863 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1864 sequential reads and behaves badly with applications doing random
1865 reads. Improve it by reading ahead at least read region, and
1866 avoiding excessive large RPC for small reads.
1868 Severity : enhancement
1870 Description: Creating more than 1000 files for a single job may cause a load
1871 imbalance on the OSTs if there are also a large number of OSTs.
1872 Details : qos_prep_create() uses an OST index reseed value that is an
1873 even multiple of the number of available OSTs so that if the
1874 reseed happens in the middle of the object allocation it will
1875 still utilize the OSTs as uniformly as possible.
1880 Description: OST or MDS may oops in ping_evictor_main()
1881 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1882 but doesn't restart at beginning of obd_exports_timed list
1885 Severity : enhancement
1887 Description: improve by-nid export eviction on the MDS and OST
1888 Details : allow multiple exports with the same NID to be evicted at one
1889 time without re-searching the exports list.
1892 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1894 Description: MDS may oops in groups_free()
1895 Details : in rare race conditions a newly allocated group_info struct is
1896 freed again, and this can be NULL. The 2.4 compatibility code
1897 for groups_free() checked for a NULL pointer, but 2.6 did not.
1900 Frequency : common for liblustre clients doing little filesystem IO
1901 Bugzilla : 9352, 7313
1902 Description: server may evict liblustre clients accessing contended locks
1903 Details : if a client is granted a lock or receives a completion AST
1904 with a blocking AST already set it would not reply to the AST
1905 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1908 Frequency : lfs setstripe, only systems with more than 160 OSTs
1910 Description: unable to set striping with a starting offset beyond OST 160
1911 Details : llapi_create_file() incorrectly limited the starting stripe
1912 index to the maximum single-file stripe count.
1915 Frequency : LDAP users only
1917 Description: lconf did not handle in-kernel recovery with LDAP properly
1918 Details : lconf/LustreDB get_refs() is searching the wrong namespace
1920 Severity : enhancement
1922 Description: bind OST threads to NUMA nodes to improve performance
1923 Details : all OST threads are uniformly bound to CPUs on a single NUMA
1924 node and do their allocations there to localize memory access
1926 Severity : enhancement
1928 Description: llmount can determine client NID directly from Myrinet (GM)
1929 Details : the client NID code from gmnalnid was moved directly into
1930 llmount, removing the need to use this or specifying the
1931 client NID explicitly when mounting GM clients with zeroconf
1934 Frequency : if client is started with down MDS
1936 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
1937 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
1938 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
1939 ptlrpc_import_delayed_req() because we didn't check if the
1940 request was interrupted, and we also didn't break out of the
1941 event loop if there was no timeout
1946 Description: data loss during non-page-aligned writes to a single file from
1947 both multiple nodes and multiple threads on one node at same time
1948 Details : updates to KMS and lsm weren't protected by common lock. Resulting
1949 inconsistency led to false short-reads, that were cached and later
1950 used by ->prepare_write() to fill in partially written page,
1951 leading to data loss.
1954 Frequency : always, if lconf --abort_recovery used
1956 Description: lconf --abort_recovery fails with 'Operation not supported'
1957 Details : lconf was attempting to abort recovery on the MDT device and not
1960 Severity : enhancement
1962 Description: remove cleanup logs
1963 Details : replace lconf-generated cleanup logs with lustre internal
1964 cleanup routines. Eliminates the need for client-cleanup and
1967 Severity : enhancement
1969 Description: add support for EAs (user and system) on lustre filesystems
1970 Details : it is now possible to store extended attributes in the Lustre
1971 client filesystem, and with the user_xattr mount option it
1972 is possible to allow users to store EAs on their files also
1974 Severity : enhancement
1976 Description: Add possibility (config option) to show minimal available OST free
1978 Details : When compiled with --enable-mindf configure option, statfs(2)
1979 (and so, df) will return least minimal free space available from
1980 all OSTs as amount of free space on FS, instead of summary of
1981 free spaces of all OSTs.
1983 Severity : enhancement
1985 Description: do not expand extent locks acquired on OST-side
1986 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
1987 by server: they are not cached anyway.
1990 Frequency : when mmap is used/binaries executed from Lustre
1992 Description: Unmmap pages before throwing them away from read cache.
1993 Details : llap_shrink cache now attempts to unmap pages before discarding
1994 them (if unmapping failed - do not discard). SLES9 kernel has
1995 extra checks that trigger if this unmapping is not done first.
2000 Description: lconf didn't resolve symlinks before checking to see whether a
2001 given mountpoint was already in use
2004 Frequency : when migrating failover services
2005 Bugzilla : 6395, 9514
2006 Description: When migrating a subset of services from a node (e.g. failback
2007 from a failover service node) the remaining services would
2008 time out and evict clients.
2009 Details : lconf --force (implied by --failover) sets the global obd_timeout
2010 to 5 seconds in order to quickly disconnect, but this caused
2011 other RPCs to time out too quickly. Do not change the global
2012 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2014 Severity : enhancement
2015 Frequency : if MDS is started with down OST
2016 Bugzilla : 9439,5706
2017 Description: Allow startup/shutdown of an MDS without depending on the
2018 availability of the OSTs.
2019 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2020 Add appropriate locking and lov-osc refcounts for safe
2021 cleaning. Add osc abort_inflight calls in case the
2022 synchronize never started.
2025 Frequency : occasional (Cray XT3 only)
2027 Description: root not authorized to access files in CRAY_PORTALS environment
2028 Details : The client process capabilities were not honoured on the MDS in
2029 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2030 been accessed by an authorized user then root was able to access
2031 the file on the local client also. The root user capabilities
2032 are now allowed on the MDS, as this environment has secure UID.
2035 Frequency : occasional
2037 Description: ldiskfs "too long searching" message happens too often
2038 Details : A debugging message (otherwise harmless) prints too often on
2039 the OST console. This has been reduced to only happen when
2040 there are fragmentation problems on the filesystem.
2045 Description: Division by zero in statfs when all OSCs are inactive
2046 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2047 lov_statfs divides by value returned by lov_get_stripecnt().
2051 Bugzilla : 9489, 3273
2052 Description: First write from each client to each OST was only 4kB in size,
2053 to initialize client writeback cache, which caused sub-optimal
2054 RPCs and poor layout on disk for the first writen file.
2055 Details : Clients now request an initial cache grant at (re)connect time
2056 and so that they can start streaming writes to the cache right
2057 away and always do full-sized RPCs if there is enough data.
2058 If the OST is rebooted the client also re-establishes its grant
2059 so that client cached writes will be honoured under the grant.
2064 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2065 Details : Now I/O RPCs go to different portal number and (presumably) fast
2066 lock requests (and glimses) and other RPCs get their own service
2067 threads pool that should be able to service those RPCs
2070 Severity : enhancement
2072 Description: Ability to exchange lustre version between client and servers and
2073 issue warnings at client side if client is too old. Also for
2074 liblustre clients there is ability to refuse connection of too old
2076 Details : New 'version' field is added to connect data structure that is
2077 filled with version info. That info is later checked by server and
2081 Frequency : rare, liblustre only.
2082 Bugzilla : 9296, 9581
2083 Description: Two simultaneous writes from liblustre at offset within same page
2084 might proceed at the same time overwriting eachother with stale
2086 Details : I/O lock withing llu_file_prwv was released too early, before data
2087 actually was hitting the wire. Extended lock-holding time until
2088 server acknowledges receiving data.
2091 Frequency : extremely rare. Never observed in practice.
2093 Description: avoid generating lustre_handle cookie of 0.
2094 Details : class_handle_hash() generates handle cookies by incrementing
2095 global counter, and can hit 0 occasionaly (this is unlikely, but
2096 not impossible, because initial value of cookie counter is
2097 selected randonly). Value of 0 is used as a sentinel meaning
2098 "unassigned handle" --- avoid it. Also coalesce two critical
2099 sections in this function into one.
2101 Severity : enhancement
2103 Description: allow liblustre clients to delegate truncate locking to OST
2104 Details : To avoid overhead of locking, liblustre client instructs OST to
2105 take extent lock in ost_punch() on client's behalf. New connection
2106 flag is added to handle backward compatibility.
2108 Severity : enhancement
2109 Bugzilla : 4928, 7341, 9758
2110 Description: allow number of OST service threads to be specified
2111 Details : a module parameter allows the number of OST service threads
2112 to be specified via "options ost ost_num_threads={N}" in the
2113 OSS's /etc/modules.conf or /etc/modprobe.conf.
2117 Bugzilla : 6146, 9635, 9895
2118 Description: servers crash with bad pointer in target_handle_connect()
2119 Details : In rare cases when a client is reconnecting it was possible that
2120 the connection request was the last reference for that export.
2121 We would temporarily drop the export reference and get a new
2122 one, but this may have been the last reference and the export
2123 was just destroyed. Get new reference before dropping old one.
2125 Severity : enhancement
2126 Frequency : if client is started with failover MDS
2128 Description: Allow multiple MDS hostnames in the mount command
2129 Details : Try to read the configuration from all specified MDS
2130 hostnames during a client mount in case the "primary"
2133 Severity : enhancement
2135 Description: Stop sending data to evicted clients as soon as possible.
2136 Details : Check if the client we are about to send or are sending data to
2137 was evicted already. (Check is done every second of waiting,
2138 for which l_wait_event interface was extended to allow checking
2139 of exit condition at specified intervals).
2142 Frequency : rare, normally only when NFS exporting is done from client
2144 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2146 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2149 Frequency : always, if srand() is called before liblustre initialization
2151 Description: Liblustre uses system PRNG disturbing its usage by user application
2152 Details : Introduce internal to lustre fast and high-quality PRNG for
2153 lustre usage and make liblustre and some other places in generic
2154 lustre code to use it.
2156 Severity : enhancement
2157 Bugzilla : 9477, 9557, 9870
2158 Description: Verify that the MDS configuration logs are updated when xml is
2159 Details : Check if the .xml configuration logs are newer than the config
2160 logs stored on the MDS and report an error if this is the case.
2161 Request --write-conf, or allow starting with --old_conf.
2163 Severity : enhancement
2165 Description: Handle symlinks in the path when checking if Lustre is mounted.
2166 Details : Resolve intermediate symlinks when checking if a client has
2167 mounted a filesystem to avoid duplicate client mounts.
2172 Description: lconf can hit an error exception but still return success.
2173 Details : The lconf command catches the Command error exception at the top
2174 level script context and will exit with the associated exit
2175 status, but doesn't ensure that this exit status is non-zero.
2180 Description: failure of ptlrpc thread startup can cause oops
2181 Details : Starting a ptlrpc service thread can fail if there are a large
2182 number of threads or the server memory is very fragmented.
2183 Handle this without oopsing.
2186 Frequency : always, only if liblustre and non-default acceptor port was used
2188 Description: liblustre cannot connect to servers with non-default acceptor port
2189 Details : tcpnal_set_default_params() was not called and was therefore
2190 ignoring the environment varaible TCPNAL_PORT, as well as other
2191 TCPNAL_ environment variables
2196 Description: two objects could be created on the same OST for a single file
2197 Details : If an OST is down, in some cases it was possible to create two
2198 objects on a single OST for a single file. No problems other
2199 than potential performance impact and spurious error messages.
2203 Bugzilla : 5681, 9562
2204 Description: Client may oops in ll_unhash_aliases
2205 Details : Client dcache may become inconsistent in race condition.
2206 In some cases "getcwd" can fail if the current directory is
2212 Description: Inode refcounting problems in NFS export code
2213 Details : link_raw functions used to call d_instantiate without obtaining
2214 extra inode reference first.
2218 Bugzilla : 9942, 9903
2219 Description: Referencing freed requests leading to crash, memleaks with NFS.
2220 Details : We used to require that call to ll_revalidate_it was always
2221 followed by ll_lookup_it. Also with revalidate_special() it is
2222 possible to call ll_revalidate_it() twice for the same dentry
2223 even if first occurence returned success. This fix changes semantic
2224 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2225 reference on a request referred from the intent.
2226 ll_intent_release() then releases such a request.
2229 Frequency : rare, normally benchmark loads only
2231 Description: unlinked inodes were kept in memory on the client
2232 Details : If a client is repeatedly creating and unlinking files it
2233 can accumulate a lot of stale inodes in the inode slab cache.
2234 If there is no other client load running this can cause the
2235 client node to run out of memory. Instead flush old inodes
2236 from client cache that have the same inode number as a new inode.
2239 Frequency : SLES9 2.6.5 kernel and long filenames only
2240 Bugzilla : 9969, 10379
2241 Description: utime reports stale NFS file handle
2242 Details : SLES9 uses out-of-dentry names in some cases, which confused
2243 the lustre dentry revalidation. Change it to always use the
2247 Frequency : rare, unless heavy write-truncate concurrency is continuous
2248 Bugzilla : 4180, 6984, 7171, 9963, 9331
2249 Description: OST becomes very slow and/or deadlocked during object unlink
2250 Details : filter_destroy() was holding onto the parent directory lock
2251 while truncating+unlinking objects. For very large objects this
2252 may block other threads for a long time and slow overall OST
2253 responsiveness. It may also be possible to get a lock ordering
2254 deadlock in this case, or run out of journal credits because of
2255 the combined truncate+unlink. Solution is to do object truncate
2256 first in one transaction without parent lock, and then do the
2257 final unlink in a new transaction with the parent lock. This
2258 reduces the lock hold time dramatically.
2261 Frequency : rare, 2.4 kernels only
2263 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2264 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2265 before unmounting their filesystem in order to avoid deadlock.
2266 The kernel_locked() function in 2.4 kernels only checks whether
2267 the kernel lock is held, not whether it is this process that is
2268 holding it as 2.6 kernels do.
2273 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2274 Details : The client ptlrpc code may be trying to reconnect to a down
2275 server before a previous connection attempt has timed out.
2276 Increase the reconnect interval to be longer than the connection
2277 timeout interval to avoid sending duplicate connections to
2281 Frequency : echo_client brw_test command
2283 Description: fix echo_client to work with OST preallocated code
2284 Details : OST preallocation code (5137) didn't take echo_client IO path
2285 into account: echo_client calls filter methods outside of any
2286 OST thread and, hence, there is no per-thread preallocated
2287 pages and buffers to use. Solution: hijack pga pages for IO. As
2288 a byproduct, this avoids unnecessary data copying.
2292 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2293 Description: Client can oops in mdc_commit_close() after open replay
2294 Details : It was possible for the MDS to return an open request with no
2295 transaction number in mds_finish_transno() if the client was
2296 evicted, but without actually returning an error. Clients
2297 would later try to replay that open and may trip an assertion
2298 Simplify the client close codepath, and always return an error
2299 from the MDS in case the open is not successful.
2302 Frequency : rare, 2.6 OSTs only
2304 Description: OST may deadlock under high load on fragmented files
2305 Details : If there was a heavy load and highly-fragmented OST filesystems
2306 it was possible to have all the OST threads deadlock waiting on
2307 allocation of biovecs, because the biovecs were not released
2308 until the entire RPC IO was completed. Instead, release biovecs
2309 as soon as they are complete to ensure forward IO progress.
2311 Severity : enhancement
2313 Description: Support for specifying external journal device at mount
2314 Details : If an OST or MDS device is formatted with an external journal
2315 device, this device major/minor is stored in the ext3 superblock
2316 and may not be valid for failover. Allow detecting and
2317 specifying the external journal at mount time.
2322 Description: Mounting an MDS with pending unlinked files may cause oops
2323 Details : target_finish_recovery() calls mds_postrecov() which returned
2324 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2325 considers this an error and immediately begins cleaning up the
2326 lov, just after starting the mds_lov process
2328 Severity : enhancement
2330 Description: Implement 'lfs df' to report actual free space on per-OST basis
2331 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2332 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2333 usage of objects; '-h' to report in human readable format.
2335 ------------------------------------------------------------------------------
2337 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2344 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2345 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2346 at startup during orphan unlinking. Instead, skip buddy update
2347 before setup, as it will be regenerated after recovery is complete.
2352 Description: If an OST is inactive, its locks might reference stale inodes.
2353 Details : lov_change_cbdata() must iterate over all namespaces, even if
2354 they are inactive to clear inode references from the lock.
2356 Severity : enhancement
2357 Frequency : occasional, if non-standard max_dirty_mb used
2359 Description: Client will block write RPCs if not enough grant
2360 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2361 then the client will block writes while waiting for another RPC
2362 to complete instead of consuming its dirty limit. With change
2363 we get improved performance when max_dirty_mb is small.
2365 Severity : enhancement
2366 Bugzilla : 3389, 6253
2367 Description: Add support for supplementary groups on the MDS.
2368 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2369 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2370 lookups for user supplementary groups into a cache.
2374 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2375 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2376 for liblustre O_EXCL is enforced by MDS.
2380 Description: readdir never returns NULL in liblustre.
2381 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2382 dentry in struct dirent.
2386 Frequency : liblustre only, depends on application IO pattern
2387 Description: liblustre clients evicted if not contacting servers
2388 Details : Don't put liblustre clients into the ping_evictor list, so
2389 they will not be evicted by the pinger ever.
2391 Severity : enhancement
2393 Description: Add ability to evict clients by NID from MDS.
2394 Details : By echoing "nid:$NID" string into
2395 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2396 $NID would be instantly evicted from this MDS and from all active
2397 OSTs connected to it.
2401 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2402 Details : lookup_it_finish() used to query file size from OSTs that was not
2407 Description: service threads change working directory to that of init
2408 Details : Starting lustre service threads may pin the working directory
2409 of the parent thread, making that filesystem busy. Threads
2410 now change to the working directory of init to avoid this.
2414 Frequency : during shutdown only
2415 Description: shutdown with a failed MDS or OST can cause unmount to hang
2416 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2421 Frequency : 2.6 only
2422 Description: chmod/chown may include an extra supplementary group
2423 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2424 supplementary group and if none is specified this is used.
2427 Bugzilla : 5479 (6816)
2429 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2430 Details : If lookup is in progress on a file that is unlinked we might try
2431 to revalidate the inode and fail in revalidate after lookup is
2432 complete and ll_file_open() enqueues the open again but
2433 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2436 Frequency : always, if lconf --abort_recovery used
2438 Description: lconf --abort_recovery fails with 'Operation not supported'
2439 Details : lconf was attempting to abort recovery on the MDT device and not
2442 ------------------------------------------------------------------------------
2444 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2449 Frequency : rare (only unsupported configurations with a node running as an
2451 Bugzilla : 6514, 5137
2452 Description: Mounting a Lustre file system on a node running as an OST could
2454 Details : OSTs now preallocates memory needed to write out data at
2455 startup, instead of when needed, to avoid having to
2456 allocate memory in possibly low memory situations.
2457 Specifically, if the file system is mounted on on OST,
2458 memory pressure could force it to try to write out data,
2459 which it needed to allocate memory to do. Due to the low
2460 memory, it would be unable to do so and the node would
2461 become unresponsive.
2463 Severity : enhancement
2465 Description: Addition of lconf --service command line option
2466 Details : lconf now accepts a '--service <arg>' option, which is
2467 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2469 Severity : enhancement
2471 Description: Failover mode is now the default for OSTs.
2472 Details : By default, OSTs will now run in failover mode. To return to
2473 the old behaviour, add '--failout' to the lmc line for OSTs.
2475 Severity : enhancement
2477 Description: Health checks are now provided for MDS and OSTs
2478 Details : Additional detailed health check information on MSD and OSTs
2479 is now provided through the procfs health_check value.
2482 Frequency : occasional, depends on IO load
2484 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2485 after numerous create/delete cycles
2486 Details : The ext3 inode allocation policy would not allocate new inodes
2487 very well on the OSTs because there are no new directories
2488 being created. Instead we look for groups with free space if
2489 the parent directories are nearly full.
2494 Description: Network or server problems during mount may cause partially
2495 mounted clients instead of returning an error.
2496 Details : The config llog parsing code may overwrite the error return
2497 code during mount error handling, returning success instead
2503 Description: MDS can fail to allocate large reply buffers
2504 Details : After long uptimes the MDS can fail to allocate large reply
2505 buffers (e.g. zconf client mount config records) due to memory
2506 fragmentation or consumption by the buffer cache. Preallocate
2507 some large reply buffers so that these replies can be sent even
2508 under memory pressure.
2512 Frequency : rare (liblustre)
2513 Description: fsx running with liblustre complained that using truncate() to
2514 extend the file doesn't work. This patch corrects that issue.
2515 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2516 ATTR_SIZE and lsm use in llu_setattr_raw.
2520 Frequency : rare, only 2.6 kernels
2521 Description: Unusual file access patterns on the MDS may result in inode
2522 data being lost in very rare circumstances.
2523 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2524 code in the 2.6 kernel caused the inode and/or EA data not to
2525 be read from disk, causing single-file corruption.
2529 Frequency : rare, only 2.6 filesystems using extents
2530 Description: Heavy concurrent write and delete load may cause data corruption.
2531 Details : It was possible under high-load situations to have an extent
2532 metadata block in the block device cache from a just-unlinked
2533 file overwrite a newly-allocated data block. We now unmap any
2534 metadata buffers that alias just-allocated data blocks.
2538 Frequency : filesystems with default stripe_count larger than 77
2539 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2540 Details : lconf specifies an inode size of 4096 bytes when the default
2541 stripe_count is larger than 77. This conflicts with the default
2542 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2543 this case to avoid pinning too much memory for large EAs.
2545 ------------------------------------------------------------------------------
2547 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2552 Frequency : rare (extremely heavy IO load with hundreds of clients)
2554 Description: Client is evicted, gets IO error writing to file
2555 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2556 caused clients to be evicted for AST timeouts. The fixes in
2557 bug 5192 mean we no longer need to have such short AST timeouts
2558 so ldlm_timeout has been increased.
2561 Frequency : occasional during --force or --failover shutdown under load
2562 Bugzilla : 5949, 4834
2563 Description: Server oops/LBUG if stopped with --force or --failover under load
2564 Details : a collection of import/export refcount and cleanup ordering
2565 issues fixed for safer force cleanup
2568 Frequency : only filesystems larger than 120 OSTs
2569 Bugzilla : 5990, 6223
2570 Description: lfs getstripe would oops on a very large filesystem
2571 Details : lov_getconfig used kfree on vmalloc'd memory
2574 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2575 Bugzilla : 6242, 6243
2576 Description: reading from files that had been truncated to a non-zero size
2577 but never opened returned no data
2578 Details : ll_file_read() reads zeros from no-object files to EOF
2583 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2584 Details : There's little harm in aborting MDS/OSS recovery and letting it
2585 try again, so I removed the LASSERT and return an error instead.
2587 Severity : enhancement
2589 Description: New debugging infrastructure for tracking down data corruption
2590 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2591 (b) cover more of the client-side code path, and (c) try to narrow
2592 down where problems occurred
2596 Bugzilla : 3819, 4364, 4397, 6313
2597 Description: Racing close and eviction MDS could cause assertion in mds_close
2598 Details : It was possible to get multiple mfd references during close and
2599 client eviction, leading to one thread referencing a freed mfd.
2601 Severity: : enhancement
2602 Bugzilla : 3262, 6359
2603 Description: Attempts to reconnect to servers are now more aggressive.
2604 Details : This builds on the enhanced upcall-less recovery that was added
2605 in 1.4.2. When trying to reconnect to servers, clients will
2606 now try each server in the failover group every 10 seconds. By
2607 default, clients would previously try one server every 25 seconds.
2612 Description: After recovery, certain operations trigger a failed
2613 assertion on a client.
2614 Details : Failing over an mds, using lconf -d --failover, while a
2615 client was doing a readdir() call would cause the client to
2616 LBUG after recovery completed and the readdir() was resent.
2618 Severity : enhancement
2620 Description: Default groups are now added by lconf
2621 Details : You can now run lconf --group <servicename> without having to
2622 manually add groups with lmc.
2625 Frequency : occasional
2627 Description: Nodes with an elan id of 0 trigger a failed assertion
2630 Frequency : always when accessing e.g. tty/console device nodes
2632 Description: tty and some other devices nodes cannot be used on lustre
2633 Details : file's private_data field is used by device data and lustre
2634 values in there got lost. New field was added to struct file to
2635 store fs-specific private data.
2638 Frequency : when exporting Lustre via NFS
2640 Description: NFSD failed occasionally when looking up a path component
2641 Details : NFSD is looking up ".." which was broken in ext3 directories
2642 that had grown large enough to become hashed.
2645 Frequency : Clusters with multiple interfaces not on the same subnet
2647 Description: Nodes will repeatedly try to reconnect to an interface which it
2648 cannot reach and report an error to the log.
2649 Details : Extra peer list entries will be created by lconf with some peers
2650 unreachable. lconf now validates the peer before adding it.
2653 Frequency : Only if a default stripe is set on the filesystem root.
2655 Description: Setting a default stripe on the filesystem root prevented the
2656 filesystem from being remounted.
2657 Details : The client was sending extra request flags in the root getattr
2658 request and did not allocate a reply buffer for the dir EA.
2661 Frequency : occasional, higher if lots of files are accessed by one client
2662 Bugzilla : 6159, 6097
2663 Description: Client trips assertion regarding lsm mismatch/magic
2664 Details : While revalidating inodes the VFS looks up inodes with ifind()
2665 and in rare cases can find an inode that is being freed.
2666 The ll_test_inode() code will free the lsm during ifind()
2667 when it finds an existing inode and then the VFS later attaches
2668 this free lsm to a new inode.
2672 Bugzilla : 6422, 7030
2673 Description: MDS deadlock between mkdir and client eviction
2674 Details : Creating a new file via mkdir or mknod (starting a transaction
2675 and getting the ns lock) can deadlock with client eviction
2676 (gets ns lock and trying to finish a synchronous transaction).
2679 Frequency : occasional
2680 Description: While starting a server, the fsfilt_ext3 module could not be
2682 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2683 kernels. Previously, lconf would still use the ext3 name
2684 when trying to load modules. Now, it will correctly use
2685 ext3 on 2.4 and ldiskfs on 2.6.
2687 Severity : enhancement
2688 Description: The default stripe count has been changed to 1
2689 Details : The interpretation of the default stripe count (0, to lfs
2690 or lmc) has been changed to mean striping across a single
2691 OST, rather than all available. For general usage we have
2692 found a stripe count of 1 or 2 works best.
2694 Severity : enhancement
2695 Description: Add support for compiling against Cray portals.
2696 Details : Conditional compiling for some areas that are different
2700 Frequency : occasional
2701 Bugzilla : 6409, 6834
2702 Description: Creating files with an explicit stripe count may lead to
2703 a failed assertion on the MDS
2704 Details : If some OSTs are full or unavailable, creating files may
2705 trigger a failed assertion on the MDS. Now, Lustre will
2706 try to use other servers or return an error to the
2710 Frequency : occasional
2712 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2713 caused lock timeout during readahead (since 1.4.2).
2714 Details : Processes doing readahead might match a lock that hasn't been
2715 granted yet if there are overlapping and conflicting lock
2716 requests. The readahead process waits on ungranted lock
2717 (original lock is CBPENDING), while OST waits for that process
2718 to cancel CBPENDING read lock and eventually evicts client.
2720 Severity : enhancement
2722 Description: Initial enabling of flock support for clients
2723 Details : Implements fcntl advisory locking and file status functions.
2724 This feature is provided as an optional mount flag (default
2725 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2726 locking are implemented yet, and those that are are not guaranteed
2727 to be completely correct in production environments.
2728 mount -t lustre -o [flock|noflock] ...
2731 Frequency : occasional
2733 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2734 assertion in the ext3 JBD (journaling) layer.
2735 Details : The b_committed_data struct is protected by the big kernel lock
2736 in 2.4 kernels, serializing journal_commit_transaction() and
2737 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2738 access to this struct. In 2.6 kernels there is finer grained
2739 locking to improve SMP performance of the JBD layer.
2743 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2745 -----------------------------------------------------------------------------
2747 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2749 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2750 versions of Lustre. Please update all servers and clients to
2751 version 1.4.2 or later at the same time. You must also run
2752 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2753 to update the configuration logs.
2755 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2756 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2757 - swab llog records of type '0' so we get proper header size/idx (5861)
2758 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2759 - fix rename of one directory over another leaking an inode (5953)
2760 - avoid SetPageDirty on 2.6 (5981)
2761 - don't re-add just-being-destroyed locks to the waiting list (5653)
2762 - when creating new directories, inherit the parent's custom
2763 striping settings if present parent (3048)
2764 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2765 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2766 - don't hold client locks on temporary worklist from l_lru (5666)
2767 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2768 - automatically evict dead clients (5921)
2769 - Update file size properly in create+truncate+fstat case (6196)
2770 - Do not unhash mountpoint dentries, do not allow removal of
2772 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2773 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2774 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2775 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2776 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2777 - Correctly handle reads of files with no objects (6243)
2778 - lctl recover will also mark a device active if deactivate used (5933)
2780 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2781 - allow --write-conf on an MDS with different nettype than client (5619)
2782 - don't write config llogs to MDS for mounts not from that MDS (5617)
2783 - lconf should create multiple TCP connections from a client (5201)
2784 - init scripts are now turned off by default; run chkconfig --on
2785 lustre and chkconfig --on lustrefs to use them
2786 - upcalls are no longer needed for clients to recover to failover
2788 - add --abort-recovery option to lconf to abort recovery on device
2790 - add support for an arbitrary number of OSTs (3026)
2791 - Quota support protocol changes.
2792 - forward compatibility changes to wire structs (6007)
2793 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2794 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2795 - improved "lustre status" script
2796 - initialize blocksize for non-regular files (6062)
2797 - added --disable-server and --disable-client configure options (5782)
2798 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2799 - Vanilla 2.4.29 support
2800 - increase maximum number of obd devices to 520 (6242)
2801 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2802 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2803 SLES 9 2.6.5 kernels
2804 - limit stripes per file to 160 (the maximum EA size) (6093)
2806 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2809 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2810 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2811 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2812 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2813 - small ext3 extents cleanups and fixes (5733)
2814 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2815 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2816 - avoid needless client->OST connect, fix handle mismatch (5317)
2817 - fix DLM error path that led to out-of-sync client, long delays (5779)
2818 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2819 - fix several locking issues related to i_size (5492,5624,5654,5672)
2820 - don't move pending lock onto export if it is already evicted (5683)
2821 - fix kernel oops when creating .foo in unlinked directory (5548)
2822 - fix deadlock in obdfilter statistics vs. object create (5811)
2823 - use time_{before,after} to avoid timer jiffies wrap (5882)
2824 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2825 - Do not leak request if server was not able to process it (5154)
2826 - If mds_open unable to find parent dir, make that negative lookup(5154)
2827 - don't create new directories with extent-mapping (5909, 5936)
2829 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2830 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2831 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2832 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2833 - support for RHEL4 kernel on i686 (5773)
2834 - provide error messages when incompatible logs are encountered (5898)
2836 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2837 * version 1.4.0.10 (1.4.1 release candidate 1)
2839 - don't keep a lock reference when lock is not granted (4238)
2840 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2841 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2842 - drop import inflight refcount on signal_completed_replay error (5255)
2843 - unlock page after async write error during send (3677)
2844 - handle missing objects in filter_preprw_read properly (5265)
2845 - no transno return for symlink open, don't save no-trasno open (3440)
2846 - don't try to complete elan receive that already failed (4012)
2847 - free RPC server reply state on error (5406)
2848 - clean up thread from ptlrpc_start_thread() on error (5160)
2849 - readahead could read extra page into cache that wasn't ejected (5388)
2850 - prevent races in class_attach/setup/cleanup/detach (5260)
2851 - don't dereference de->d_inode after l_dput of de (5458)
2852 - use "int" for stripe value returned from lock_to_stripe (5544)
2853 - mballoc allocation and error-checking fixes in 2.6 (5504)
2854 - block device patches to fix I/O request sizes in 2.6 (5482)
2855 - look up hostnames for IB nals (5602)
2856 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2857 - don't start multiple acceptors for the same port (5277)
2858 - fix incorrect LASSERT in mds_getattr_name (5635)
2859 - export a proc file for general "ping" checking (5628)
2860 - fix "lfs check" to not block when the MDS is down (5628)
2862 - service request history (4965)
2863 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2864 - create an "evict_client" /proc entry on OSTs, like the MDS has
2865 - fix mount usage message, return errors per mount(8) (5168)
2866 - change grep [] to grep "[]" in tests so they work in more UMLs
2867 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2868 - remove ancient LOV_MAGIC_V0 EA support (5047)
2869 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2870 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2871 - allow readahead allocations to fail when low on memory (5383)
2872 - mmap locking landed again, after considerable improvement (2828)
2873 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2875 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2878 - send OST transaction number in read/write reply to free req (4966)
2879 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2880 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2881 - the watchdog thread now runs as interruptible (5246)
2882 - flock/lockf fixes (but it's still disabled, pending 5135)
2883 - don't use EXT3 constants in llite code (5094)
2884 - memory shortage at startup could cause assertion (5176)
2886 - reorganization of lov code
2887 - single portals codebase
2889 - add extents/mballoc support (5025)
2890 - direct I/O reads in the obdfilter (4048)
2891 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2893 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2896 - send OST transaction number in read/write reply to free req (4966)
2897 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2898 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2899 - the watchdog thread now runs as interruptible (5246)
2900 - handle missing objects in filter_preprw_read properly (5265)
2901 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2902 - drop import inflight refcount on signal_completed_replay error (5255)
2903 - unlock page after async write error during send (3677)
2904 - return original error code on reconstructed replies (3761)
2905 - no transno return for symlink open, don't save no-trasno open (3440)
2907 - add pid to ldlm debugging output (4922)
2908 - bump the watchdog timeouts -- we can't handle 30sec yet
2909 - extra debugging for orphan dentry/inode bug (5259)
2911 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
2914 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
2915 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
2916 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
2917 - fix NULL dereference in /proc/sys/portals/routes (4827)
2918 - allow failed mdc_close() operations to be interrupted (4561)
2919 - stop precreate on OST before MDS would time out on it (4778)
2920 - don't send partial-page writes before EOF from client (4410)
2921 - discard client grant for sub-page writes on large-page clients (4520)
2922 - don't free dentries not owned by NFS code, check generation (4806)
2923 - fix lsm leak if mds_create_objects() fails (4801)
2924 - limit debug_daemon file size, always print CERROR messages (4789)
2925 - use transno after validating reply (3892)
2926 - process timed out requests if import state changes (3754)
2927 - update mtime on OST during writes, return in glimpse (4829)
2928 - add mkfsoptions to LDAP (4679)
2929 - use ->max_readahead method instead of zapping global ra (5039)
2930 - don't interrupt __l_wait_event() during strace
2932 - add software watchdogs to catch hung threads quickly (4941)
2933 - make lustrefs init script start after nfs is mounted
2934 - fix CWARN/ERROR duplication (4930)
2935 - return async write errors to application if possible (2248)
2936 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
2937 - print NAL number in %x format (4645)
2938 - update barely-supported suse-2.4.21-171 series (4842)
2939 - support for sles 9 %post scripts
2940 - support for building 2.6 kernel-source packages
2941 - support for sles km_* packages
2943 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
2946 - ignore -ENOENT errors in osc_destroy (3639)
2947 - notify osc create thread that OSC is being cleaned up (4600)
2948 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
2949 - reconstruct ost_handle() like mds_handle() (4657)
2950 - create a new thread to do import eviction to avoid deadlock (3969)
2951 - let lconf resolve symlinked-to devices (4629)
2952 - don't unlink "objects" from directory with default EA (4554)
2953 - hold socknal file ref over connect in case target is down (4394)
2954 - allow more than 32000 subdirectories in a single directory (3244)
2955 - fix blocks count for O_DIRECT writes (3751)
2956 - OST returns ENOSPC from object create when no space left (4539)
2957 - don't send truncate RPC if file size isn't changing (4410)
2958 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
2959 - bind to privileged port in socknal and tcpnal (3689)
2961 - rate limit CERROR/CWARN console message to avoid overload (4519)
2962 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
2963 - basic mmap support (3918)
2964 - kernel patch series update from b1_4 (4711)
2966 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
2969 - avoid crash during MDS cleanup with OST shut down (2775)
2970 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
2971 - don't use bad inodes on the MDS (3744)
2972 - dynamic object preallocation to improve recovery speed (4236)
2973 - don't hold spinlock over lock dumping or change debug flags (4401)
2974 - don't zero obd_dev when it is force cleaned (3651)
2975 - print grants to console if they go negative (4431)
2976 - "lctl deactivate" will stop automatic recovery attempts (3406)
2977 - look for existing locks in ldlm_handle_enqueue() (3764)
2978 - don't resolve lock handle twice in recovery avoiding race (4401)
2979 - revalidate should check working dir is a directory (4134)
2981 - don't always mark "slow" obdfilter messages as errors (4418)
2983 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
2986 - don't close LustreDB during write_conf until it is done (3860)
2987 - fix typo in lconf for_each_profile (3821)
2988 - allow dumping logs from multiple threads at one time (3820)
2989 - don't allow multiple threads in OSC recovery (3812)
2990 - fix debug_size parameters (3864)
2991 - fix mds_postrecov to initialize import for llog ctxt (3121)
2992 - replace config semaphore with spinlock (3306)
2993 - be sure to send a reply for a CANCEL rpc with bad export (3863)
2994 - don't allow enqueue to complete on a destroyed export (3822)
2995 - down write_lock before checking llog header bitmap (3825)
2996 - recover from lock replay timeout (3764)
2997 - up llog sem before sending rpc (3652)
2998 - reduce ns lock hold times when setting kms (3267)
2999 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3000 - fix NULL deref and obd_dev leak on setup error (3312)
3001 - replace some LBUG about llog ops with error handling (3841)
3002 - don't match INVALID dentries from d_lookup and spin (3784)
3003 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3004 - fix invalid assertion in ptlrpc_set_wait (3880)
3006 - add libwrap support for the TCP acceptor (3996)
3007 - add /proc/sys/portals/routes for non-root route listing (3994)
3008 - allow setting MDS UUID in .xml (2580)
3009 - print the stack of a process that LBUGs (4228)
3011 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3014 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3015 - make sure to unset replay flag from failed open requests (3440)
3016 - if default stripe count is 0, use OST count for inode size (3636)
3017 - update parent mtime/ctime on client for create/unlink (2611)
3018 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3019 - free recovery state on server during a forced cleanup (3571)
3020 - unregister_reply for resent reqs (3063)
3021 - loop back devices mounting and status check on 2.6 (3563)
3022 - fix resource-creation race that can provoke i_size == 0 (3513)
3023 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3024 - more debugging for page-accounting assertion (3746)
3025 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3026 - avoid deadlock after precreation failure (3758)
3027 - fix race and lock order deadlock in orphan handling (3450, 3750)
3028 - add validity checks when grabbing inodes from l_ast_data (3599)
3030 - add /proc/.../recovery_status to obdfilter (3428)
3031 - lightweight CDEBUG infrastructure, debug daemon (3668)
3032 - change default OSC RPC parameters to be better on small clusters
3033 - turn off OST read cache for files smaller than 32MB
3034 - install man pages and include them in rpms (3100)
3035 - add new init script for (un)mounting lustre filesystems (2593)
3036 - run chkconfig in %post for init scripts (3701)
3037 - drop scimac NAL (unmaintained)
3039 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3042 - clean kiobufs before and after use (3485)
3043 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3044 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3045 - change init script to fail more clearly if not run as root (1528)
3046 - allow clients to reconnect during replay (1742)
3047 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3048 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3049 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3050 - fix cond_resched (3554)
3051 - only evict unfinished clients after recovery (3515)
3052 - allow bulk resend, prevent data loss (3570)
3053 - dynamic ptlrpc request buffer allocation (2102)
3054 - don't allow unlinking open directory if it isn't empty (2904)
3055 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3056 - remove extraneous obd dereference causing LASSERT failure (3334)
3057 - don't use get_cycles() when creating temp. files on the mds (3156)
3058 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3059 - handle EEXIST for set-stripe, set proper directory name (3336)
3061 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3062 - fix ksocknal_fmb_callback() error messages (2918)
3064 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3067 - don't copy lvb into (possibly NULL) reply on error (2983)
3068 - don't deref dentry after dput, don't free lvb on error (2922)
3069 - use the kms to determine writeback rpc length (2947)
3070 - increment oti_logcookies when osc is inactive (2948)
3071 - update client's i_blocks count via lvb messages (2543)
3072 - handle intent open/close of special files properly (1557)
3073 - mount MDS with errors=remount-ro, like obdfilter (2009)
3074 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3075 - don't use cancelling-locks' kms values (2947)
3076 - use highest lock extent for kms, not last one (2925)
3077 - don't dereference ERR_PTR() dentry in error handling path (3107)
3078 - fix thread race in portals_debug_dumplog() (3122)
3079 - create lprocfs device entries at setup instead of at attach (1519)
3080 - common AST error handler, don't evict client on completion race (3145)
3081 - zero nameidata in detach_mnt in 2.6 (3118)
3082 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3083 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3084 - initialize RPC timeout timer earlier for 2.6 (3219)
3085 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3086 - print nal/nid for unknown nid (3258)
3087 - additional checks for oscc recovery before doing precreate (3284)
3088 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3089 - don't crash in mdc_close for bad permissions on open (3285)
3090 - zero i_rdev for non-device files (3147)
3091 - clear page->private before handing to FS, better assertion (3119)
3092 - tune the read pipeline (3236)
3093 - fix incorrect decref of invalidated dentry (2350)
3094 - provide read-ahead stats and refine rpc in flight stats (3328)
3095 - don't hold journal transaction open across create RPC (3313)
3096 - update atime on MDS at close time (3265)
3097 - close LDAP connection when recovering to avoid server load (3315)
3098 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3099 - don't leak open file on MDS after open resend (3325)
3100 - serialize filter_precreate and filter_destroy_precreated (3329)
3101 - loop device shouldn't call sync_dev() for nul device (3092)
3102 - clear page cache after eviction (2766)
3103 - resynchronize MDS->OST in background (2824)
3104 - refuse to mount the same filesystem twice on same mountpoint (3394)
3105 - allow llmount to create routes for mounting behind routers (3320)
3106 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3107 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3108 - fix rare problem with rename on htree directories (3417)
3110 - allow default OST striping configuration per directory (1414)
3111 - fix compilation for qswnal for 2.6 kernels (3125)
3112 - increase maximum number of MDS request buffers for large systems
3113 - change liblustreapi to be useful for external progs like lfsck (3098)
3114 - increase local configuration timeout for slow disks (3353)
3115 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3117 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3120 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3121 - don't overwrite extent policy data in reply if lock was blocked (2901)
3122 - drop filter export grants atomically with removal from device (2663)
3123 - del obd_self_export from work_list in class_disconnect_exports (2908)
3124 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3125 - swab reply message in mdc_close, other PPC fixes (2464)
3126 - fix destroying of named logs (2325)
3127 - overwrite old logs when running lconf --write_conf (2264)
3128 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3129 - fix race in target_handle_connect (2898)
3130 - mds_reint_create() should take same inode create lock (2926)
3131 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3132 - don't close files for self_export to avoid uninitialized obd (2936)
3133 - allow MDS with the same name as client node (2939)
3134 - hold dentry reference for closed log files for unlink (2325)
3135 - reserve space for all logs during transactions (2059)
3136 - don't evict page beyond end of stripe extent (2925)
3137 - don't oops on a deleted current working directory (2399)
3138 - handle hard links to targets without a parent properly (2517)
3139 - don't dereference NULL lock when racing during eviction (2867)
3140 - don't grow lock extents when lots of conflicting locks (2919)
3142 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3145 - account for cache space usage on clients to avoid data loss (974)
3146 - lfsck support in lustre kernel code (2349)
3147 - reduce journal credits needed for BRW writes (2370)
3148 - orphan handling to avoid losing space on client/server crashes
3149 - ptlrpcd can be blocked, stopping ALL progress (2477)
3150 - use lock value blocks to assist in proper KMS, faster stat (1021)
3151 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3152 - recovery for initial connections (2355)
3153 - fixes for mds_cleanup_orphans (1934)
3154 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3155 - block all file creations until orphan recovery completes (1901)
3156 - client remove rq_connection from request struct (2423)
3157 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3158 - recovery timer race (2670)
3159 - mdc_close recovey bug (2532)
3160 - ptlrpc cleanup bug (2710)
3161 - mds timeout on local locks (2588)
3162 - namespace lock held during RPCs (2431)
3163 - handle interrupted sync write properly (2503)
3164 - don't try to handle a message that hasn't been replied to (2699)
3165 - client assert failure during cleanup after abort recovery (2701)
3166 - leak mdc device after failed mount (2712)
3167 - ptlrpc_check_set allows timedout requests to complete (2714)
3168 - wait for inflight reqs when ptlrpcd finishes (2710)
3169 - make sure unregistered services are removed from the srv_list
3170 - reset bulk XID's when resending them (caught by 1138 test)
3171 - unregister_bulk after timeout
3172 - fix lconf error (2694)
3173 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3174 - readahead locks pages, leaves pending causing memory pressure (2673)
3175 - increase OST request buffers to 4096 on large machines (2729)
3176 - fix up permission of existing directories in simple_mkdir (2661)
3177 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3178 - don't assign transno to errored transactions (2742)
3179 - don't delete objects on OST if given a bogus objid from MDS (2751)
3180 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3181 - if rq_no_resend, then timeout request after recovery (2432)
3182 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3183 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3184 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3185 - MDS recovery shouldn't delete objects that it has given out (2730)
3186 - if enqueue arrives after completion, don't clobber LVB (2819)
3187 - don't unlock pages twice when trigger_group_io returns error (2814)
3188 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3189 - don't write pages to disk if there was an error (1450)
3190 - don't ping imports that have recovery disabled (2676)
3191 - take buffered bytes into account when balancing socknal conn (2817)
3192 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3193 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3194 - remove little-endian swabbing of llog records (1987)
3195 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3196 - retry reposting request buffers if they fail (1191)
3197 - grow extent at grant time to avoid granting a revoked lock (2809)
3198 - lock revoke doesn't evict page if covered by a second lock (2765)
3199 - disable VM readahead to avoid reading outside lock extents (2805)
3201 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3202 - updated kernel patches for hp-2.4.20 kernel (2681)
3204 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3207 - fix truncated write corruption (2366)
3208 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3210 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3211 - ensure len(uuid) < 37 in lmc (1171)
3212 - fix ia64 OOPS in llog_test (2255)
3213 - zero end of page at obdfilter for partial page writes (2648)
3214 - don't leave stale dentries around after renames (bug 2428)
3215 - fix timeouts when evicting a client with a single lock held (2642)
3216 - set deadline for the initial HELLO message to drain (2634)
3217 - print out dotted-quad IP addresses in the socknal (2302)
3219 - additional debugging for MDS client eviction problem (2443)
3220 - fix mkfsoptions support for osts (2603, 2604)
3222 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3225 - add series for the vanilla 2.6.0 kernel
3226 - add series for the vanilla 2.4.24 kernel
3227 - add series for a cray x86/64 UL kernel drop
3228 - fix xattr patches for the vanilla 2.4.19 series
3230 - generate true UUIDs in lmc (1171)
3231 - have portals stack dumping break in UML (2466)
3232 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3233 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3234 - wake callback waiting threads on client eviction (2460)
3235 - Add --ptldebug and --subsystem to lmc (1719)
3236 - update assertion to allow safe interrupt allocation
3237 - set rq_no_resend for cancel requests (2432)
3238 - recalculate ptlrpcd timeout after resend (2494)
3239 - call vfs_rmdir when removing pending directories (2368)
3240 - fix renaming a file to itself (2429)
3241 - lmc creates a default one-stripe lov (2454)
3242 - expand procfs space to handle large clusters (2326)
3243 - increase UML stack to avoid overflow
3244 - update lconf's list of debug and subsystem masks
3245 - fix lfs find --obd (2510)
3246 - /proc tunable for disabling filter read caching (2591)
3247 - stop rpm packages from altering slapd.conf (2301)
3248 - disable nagle in the socknal under 0conf (2578)
3249 - choose mds inode size based on stripe count (2572)
3250 - fix kernel-source rpm problems (2516)
3252 - add --disable-doc to avoid pdf generation (2421)
3253 - update documentation, tests, type-os, comments
3254 - avoid format warnings on ia64
3255 - remove the TOE NAL
3256 - tiny code cleanups by removing unused fields
3258 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3261 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3262 - remove the most blatant lies from BUILDING file (2371)
3263 - change default debug level to reasonable production setting
3264 - reduce client side cache size to reduce cache flush time
3265 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3266 - make TCP zerocopy and pinger support enabled by default (2476)
3267 - sync writes completed after process exits caused crashes (2319)
3268 - maintain correct mount count on the MDS (2356)
3269 - backout 1557, because 2316 wasn't really fixed
3270 - better file I/O statistics gathering in /proc
3271 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3272 - another kernel patch to fix zero-copy TCP function export
3273 - don't take duplicate lock when processing re-sent getattr (2420)
3274 - lctl uses obd_self_export instead of creating new conn (2353)
3275 - MDS/OST recovery case which requires object creation asserted (2425)
3276 - move lfs from /usr/sbin to /usr/bin in packages
3277 - fix race between mds_client_add and mds_client_free (2417)
3278 - use kmalloc instead of slabs in portals (2430)
3279 - don't create duplicate records when a failover MDS is present (2442)
3280 - remove unnecessary mount age check (2332)
3281 - don't remove directory inodes from locks prematurely (2451)
3282 - don't break if MDS service name is the same as hostname (2103)
3283 - fix races in client write RPC generation when cache full (2482)
3285 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3288 - remove now-unused request->rq_obd (278)
3289 - if an allocation fails, print out how much memory we've used (1933)
3290 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3291 - add the "configurable stack size" patch to most series files (1256)
3292 - ability to write large log records, for 100+ OST configs (2306)
3293 - fix NULL deref when filter_prep fails (2314)
3294 - fix operator precedence error in filter_sync
3295 - dynamic allocation of socknal TX descriptors (2315)
3296 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3297 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3298 - ensure that CWARN messages in Portals always get to the syslog
3299 - __init/__exit are not for prototype decls (ldlm_init/exit)
3300 - x86-64 compile warning fixes
3301 - fix gateway LMC keyword conflict (2318)
3302 - fix MDS lock inversions in getattr/reint paths (1844)
3303 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3304 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3305 - DLM race condition prevented some lock evictions (2328)
3306 - ENOMEM detection and retry on socknal sends (2230)
3307 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3308 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3309 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3310 - better error messages when a client is rejected during recovery (1505)
3311 - avoid cancelling locks which were never granted, after failure (2330)
3312 - fix i_sem/journal inversion in mds_client_add (2333)
3313 - fix truncate/getattr lock cycle deadlock (2334)
3314 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3315 - fix two rare exit paths which could leak an l_lock() ref (2321)
3316 - fencepost error in MDS/OST orphan recovery (2226)
3317 - make log record alignment 8 bytes (1988)
3318 - lstripe now fails when requested offset > ost_count (2237)
3319 - ensure that all kernel series have a complete list.h (1607)
3320 - fix crashes in special-file operations (2316)
3321 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3323 - allow configurable automake binary, for testing new versions
3324 - small update to the lfs documentation
3326 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3328 * fix negative export reference count in fsfilt_sync (2312)
3330 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3331 * release candidate 0.9.1
3333 - orphans are moved into the PENDING directory for possible recovery
3334 - replayed opens now open by fid for orphan/rename safety (1042)
3335 - last close of an orphan inode generates a transno (683)
3336 - chdir() and mount() now pin the directory entry (1020)
3337 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3338 - discard very old requests without processing them (1502)
3339 - remove some common, well-understood CERRORs (1505)
3340 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3341 - clear "grant" flags in OST replies until OST grant code lands (1644)
3342 - fix read performance by not clobbering i_blksize on client (1598)
3343 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3344 - make LRU size a /proc tunable, clears locks when reduced (707)
3345 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3346 - server locks take references on exports now (1558)
3347 - build fixes for 2.4.20-rh trees (1663)
3348 - return an error from lov_create if all OSCs are inactive (1751)
3349 - fix import levels when a reconnect happens without a timeout (1597)
3350 - exit early from mds_open if we get a lookup error (1749)
3351 - partial page read at EOF wouldn't wait for disk before sending (1642)
3352 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3353 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3354 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3355 - write/truncate lock inversion (1639)
3356 - Don't auto-load obdclass, portals modules during cleanup (1495)
3357 - fix timestamps from jumping to "now" (1763)
3358 - extra journal assertions (1648)
3359 - add an extra multiunlink test (1771)
3360 - fix read_record/write_record API (1776)
3361 - fix leak of offset_extent, possible incorrect i_size later (1772)
3362 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3363 - Debugging for the kqswnal_get_idle_tx problems (1820)
3364 - Allow recovery to be attempted multiple times (1536)
3365 - Write out MDS last_rcvd file after it is first created (1600)
3366 - Fix tx_descriptor leak in failed transmit situations (1827)
3367 - ext3 journaling fixes for assertion failure after IO error (1871)
3368 - class_export_put() on freed export after completion AST error (1896)
3369 - Fix revalidate looping in VFS (1322)
3370 - Don't access a freed export during MDS_REINT timeout (1521)
3371 - Add open-unlink recovery support on the MDS (1673,1764)
3372 - Return an error if no MDS data was read from last_rcvd (1946)
3373 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3374 - Don't setup a disk device that doesn't match exported UUID (317)
3375 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3376 - avoid committing NULL handle in force close
3377 - local.sh is now a one-stripe LOV configuration
3378 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3379 - fix ext3 htree duplicate directory entry corruption (1516)
3380 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3381 - update to new LOV EA format (2097)
3382 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3384 2003-06-15 Phil Schwan <phil@clusterfs.com>
3387 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3388 - per-import/export recovery handling (958, 931, 959)
3389 - multiple last-rcvd slots, for serving multiple FSes (949)
3390 - connections are again shared between multiple imp/exports (963, 964)
3391 - "umount -f" would hang if any requests needed to be sent (393, 978)
3392 - avoid pinning large req buffer by copying for queued messages (989)
3393 - add "uuid" to "lctl device" command to help upcalls (991)
3394 - "open" RPCs with transnos would confuse recovery counters (1037)
3395 - do proper endian conversion of all wire messages (288, 340, 891)
3396 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3397 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3398 - fix ptlrpc_connection leak in target_handle_connect (1174)
3399 - fix import refcounting bug in OST and MDS cleanup (1134)
3400 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3401 - fix very unlikely obd_types race condition (501)
3402 - remove osc_open hack for echo_client (1187)
3403 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3404 - a failure in read_inode2 leads to deadlock (1139)
3405 - cancel ack-locks as soon as transaction is committed (1072)
3406 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3407 - make sure to commitrw after any preprw to avoid deadlock (1162)
3408 - failing to execute a file in a lustre FS would lock inode (1203)
3409 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3410 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3411 - obd_uuid2tgt would walk off the end of the list (1255)
3412 - on IA64 the osc would give portals incorrect bulk size (1258)
3413 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3414 - fix lock inversion caused by new llite matching code (1282)
3415 - limit the number of dirty pages on a client to 10MB (1286)
3416 - timed out locks were not being corrected cancelled (1289)
3417 - fix O_DIRECT above 4GB on IA-32 (1292)
3418 * major user-visible changes
3419 - fail out/fail over policy now controlled by the upcall (993)
3421 - add OBD_PING to check server availability and failure (954)
3422 - lustre messages are now sent in sending host order (288, 340, 891)
3423 - add eadatalen to MDS getattr reply (340)
3424 - OST read replies may contain second buffer, with per-page status (593)
3426 2003-03-11 Phil Schwan <phil@clusterfs.com>
3429 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3430 - failed open()s could cause deadlock; fixed (867, 869)
3431 - stop cancelling OST locks when files are closed (481)
3432 - overlapping XID spaces caused network corruption (851, 853)
3433 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3434 - setattr_raw added, to do single-RPC, server-side setattrs
3435 - lmc/lconf syntax change for OST UUIDs
3436 - fix crashy race condition between ptlrpc_free_req and osc_close
3437 - don't use request in mdc_enqueue if we hit a timeout (889)
3438 - don't set the inode i_size for regular files from the MDS (896)
3439 - handle out of order completion AST (842)
3440 - don't LBUG if a lock request times out after receiving AST (913)
3441 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3442 - if a bad lock AST arrives, send an error instead of dropping entirely
3443 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3444 - fix leak in bulk IO when only partially completed (899, 900, 926)
3445 - fix O_DIRECT for ia64 (55)
3446 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3447 - C-z after timeout could hang a process forever; fixed (977)
3449 - client-side I/O cache (678, 924, 929, 941, 970)
3451 - READPAGE and SETATTRs which don't take server-side locks get
3454 2003-02-11 Phil Schwan <phil@clusterfs.com>
3457 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3458 - Fix fsfilt_extN_readpage() to read a full page of directory
3459 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3460 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3461 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3462 - Most intent operations were removed, in favour of a new RPC mode
3463 that does a single RPC to the server and bypasses most of the VFS
3464 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3465 - Aggressively cancel local locks on DLM servers
3466 - mds_reint_unlink sends EA to the client if it's the last nlink.
3467 client uses that EA to unlink OST objects.
3468 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3469 - recursive symlinks were fixed (439)
3470 - fixed NULL deref in DEBUG_REQ
3471 - filter_update_lastobjid no longer calls sync, which annoyed extN
3472 - fixed multi-client small-writes to a single file problem (445)
3473 - fixed mtime updates during file writes (607)
3474 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3475 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3476 - fixed Linux OST size reporting problem (444, 656)
3477 - OST now updates object mtime with writes or setattr (607, 619)
3478 - client verifies file size before zeroing page past EOF (445)
3479 - OST now writes last allocated objid to disk with allocation (108)
3480 - LOV on echo now works (409)
3482 - mds_reint_unlink sends a new buffer, with the EA included. this
3483 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3484 set if a regular file was being unlinked, and it was the last link
3485 - use PtlGet from the target for bulk writes (315)
3486 - OST now updates object mtime with writes or setattr (607, 619)
3487 - LDLM now has a grant-time callback to revalidate locked items, if
3489 - Many MDS operations were reorganized to combat race conditions
3491 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3492 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3494 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3497 - Fully reactivate OST imports after reconnection (512, others)
3498 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3499 - More graceful error handling for truncating on dead OST (515)
3500 - Don't error out unless we're actually accessing dead stripes (474)
3501 - Fix garbage sizes when stripes are missing (410)
3502 - LRU counters were broken, causing constant lock purge (433, 432)
3503 - garbage on read from stripes with failed OSTs (441)
3504 - mark OSCs as active before reconnecting during recovery (438)
3505 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3506 - lfind did not preserve OST order in output (443)
3507 - symlinks cause hung clients, incorrect data (439)
3508 - stop dereferencing request after dropping refcount (457)
3509 - don't LASSERT(spin_is_locked) on non-SMP (455)
3510 - fixes for many rename() bugs
3511 - fstat didn't correctly synchronize attributes (399)
3512 - server must handle lock cancellation during blocking AST prep (487)
3513 - bulk descriptors were free()d too soon (511)
3514 - fix paths in lconf, which would load incorrect modules (451, 507)
3515 - fix confusing lconf 'host not found' error message (386)
3516 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3517 - fix race condition in mdc_blocking_ast() for inode access (526)
3518 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3519 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3520 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3521 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3522 - fix lstripe in several ways (kernel side) (527)
3523 - fix request leak in ldlm_cli_enqueue (262)
3524 - incorrect OSC was marked inactive after OST failure
3525 - call mds_fs_cleanup before unmounting filesystem (524)
3526 - fix races between taking ns_lock and ldlm_lock_change_resource
3527 - fix races updating LOV export open file list
3528 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3529 - fix recovery NULL deref in ldlm_cli_cancel_unused
3530 - fix some DLM races by using new hash table for lock handles (419)
3531 - permit the client to specify desired inodes, at replay
3532 - duplicate requests when we queue them for replay reintegration
3533 - fix last_rcvd offset calculation
3534 - sync after each recovered transaction, so we always make progress
3535 - never, not always, ERESTART requests without transnos
3536 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3538 - skip replay if the MDS says that the client is already connected
3539 - don't check for a recovery-enabled export to match lctl's UUID
3540 - don't INC_USE_COUNT for phantom exports
3541 - don't crash when cleaning up phantom exports (567)
3542 - don't double-finish or set replay data for errored mdc_open requests
3543 - abort requests when they time out, so we don't get old replies
3544 - send/receive replies for AST messages again
3545 - if the client says that it doesn't have the lock, cancel it on the
3547 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3548 mark it as destroyed, it will all work out when decref is called
3549 - fix module use counts (22, 581)
3551 - ASTs now expect a reply (server cancels lock on error reply)
3553 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3556 - fix many simultaneous client startup (392)
3557 - fix dentry->d_it clobbering
3558 - credentials weren't being shipped for readdir/getattr operations
3559 - remove invalid assertions triggered during some concurrent MD
3561 - proper Lustre versions added (336, 389)
3562 - fix memory leak for create error case (398)
3563 - fix LOV locking bug that would get cli/srv out of sync
3564 - fix echo client over LOV (409)
3565 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3566 - fix double-O_EXCL intent crash (424)
3567 - avoid sending multiple lock CANCELs (352)
3569 - MDS can do multi-client recovery (modulo bugs in new code)
3571 - many updates, edits, cleanups
3573 2002-11-18 Phil Schwan <phil@clusterfs.com>
3576 - fix null d_it dereference (346)
3577 - fix full OST/dbench hang (333)
3578 - fix permission problem with file removal (286)
3579 - fix removal of OSCs from LOV when they fail
3580 - fix NULL deref during bulk timeout (214)
3581 - fix problems related to multiple filesystems on one MDS (241)
3582 - fixed serious subtle metadata locking bugs
3583 - free locks on clients when inodes are removed due to memory
3585 - fix inode pointer in lock data (285)
3586 - partial support for multiple MDS on a single host (241)
3587 - data locks weren't cancelled at clear_inode time (290, 311)
3588 - intent locks could lead to unbounded lock growth (205)
3589 - added a maximum lock count, an LRU list, and a flusher
3590 - fix multiple rename (365)
3591 - properly abstracted the echo client
3592 - OSC locked 1 byte too many; fixed
3593 - rewrote brw callback code:
3594 - fixed recovery bugs related to LOVs (306)
3595 - fixed too-many-pages-in-one-write crash (191)
3596 - fixed (again) crash in sync_io_timeout (214)
3597 - probably fixed callback-related race (385)
3599 - Add capability to MDS protocol
3600 - LDLM cancellations and callbacks on different portals
3602 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3605 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3606 - defer open object destruction to close time (601981, 138)
3607 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3608 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3609 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3610 - don't call obd_setattr in ll_file_release for destroyed objects
3611 * protocol change to lustre_msg: move |version| and add |flags|
3612 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3613 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3614 * added replay of create, unlink, link and rename operations during
3615 MDS failover; recovery should be much more robust now
3616 * remove failed OSCs from LOVs (only lov_create uses this so far)
3617 * the lustre-HOWTO was brought (more) up to date (582544)
3619 2002-10-23 Phil Schwan <phil@clusterfs.com>
3622 - in-use dentries weren't being reused properly (617851)
3623 - prevent multiple LDLM setup (599178)
3624 - fix LOV size calculations for truncate (617853)
3625 - fix client handling of MDS intent errors (POSIX)
3626 - fix permission bug in lovstripe.c test (624321)
3627 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3628 - truncate past end of file could corrupt data
3629 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3630 - a race in recovery could return ETIMEDOUT to apps (623947)
3631 - building outside the source directory was fixed
3632 * the lustre-HOWTO was brought (more) up to date (582544)
3633 * major progress was made on recovery functionality
3635 2002-10-10 Phil Schwan <phil@clusterfs.com>
3638 - recovery deadlock fix
3639 - rm -rf causes LBUG fix (617817)
3640 - file open by multiple tasks fix (618962)
3641 - directory permissions bugs (602707 and 620007)
3642 - journal_stop fixed with locking (611313)
3643 - O_APPEND failures resolved (618273, perhaps 614459)
3644 - lconf PATH fix (619770)
3645 - IA64 build fix (621450)
3646 - RPC buffer sizes scale with amount of memory
3648 2002-10-01 Phil Schwan <phil@clusterfs.com>
3651 - locks would be cancelled without throwing away data pages,
3652 resulting in inconsistent data (605627)
3653 - inode attributes were not always being refreshed (605627, 612449)
3654 - lconf now continues to cleanup after lctl reports an error
3655 - MDS now enforces user permissions (602707)
3656 - lprocfs cleanup fixed, but not yet enabled (614157)
3657 - fixed infinite server hang, should a client not respond to an AST
3658 - avoid going into recovery if user calls readlink() with a buffer
3659 that's too small (613941)
3660 - AST RPCs no longer require replies (614867) -- this may be changed
3661 - don't crash server if client sends an IOV that's too big (611336)
3662 - fixed lock conversion deadlock (611892)
3663 - fixed the following of symlinks (614622)
3664 * recovery: the server can remove locks from a client that dies, other
3665 clients can make progress
3666 * more extN patch fixes
3667 * compile-time configurable ptlrpc buffer allocations
3669 - collaborative read cache document
3670 - Lustre Lite Performance CDR document-in-progress
3672 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3675 - fix typo in patch-2.4.18
3677 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3680 - clear ptlrpc request each time in handle_incoming_request()
3681 - unlink of files now destroys the object on the OST
3683 2002-09-19 Peter Braam <braam@clusterfs.com>
3685 * add hard link support
3686 * change obdfile creation method
3687 * kernel patch changed
3689 2002-09-19 Peter Braam <braam@clusterfs.com>
3692 - stack overflow bug in extN fixed
3694 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3696 * documentation updates
3697 - add man pages for config tools
3698 - update tests/README to describe testing with new config tools
3699 - finish metadata API descriptions
3700 * bug fixes and cleanups
3701 - statfs workaround for 16TB limit
3702 - LOV stripe allocation improved, can stripe on subset of OSTs
3703 - LOV file size/IO offset was wrong for files > 4GB in size
3704 - object EA data was being dropped, caused files to be unreadable
3705 - memory overflow with non-LOV OST caused memory corruption
3706 - fixed regression tests to work with new config tools, obdfilter
3707 - fixed bug when directory size became larger than 1 block
3708 - fixed bug (for single client case) when PWD was deleted
3709 - invalidate local directory pages when doing intent-based ops
3710 - avoid LDLM oops when lock callback contained bad data
3712 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3714 * documentation updates
3715 * bug fixes and cleanups
3716 - configuration tools
3719 - 64-bit compile warnings
3720 - 64-bit internal statfs data
3722 * test_brw on persistent OST devices
3724 * lprocfs (disabled)
3726 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3728 * documentation updates
3729 * bug fixes and cleanups
3730 * configuration tools
3732 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3735 * many small fixes to 0_5_4
3736 * io/network handling
3737 * thinkos in MDS operations
3739 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3742 * crucial basic fixes to 0.5.3
3743 * IOR, Iozone work over Elan
3746 2002-08-07 Phil Schwan <phil@clusterfs.com>
3747 * version 0_5_3, our first alpha
3748 * we use the new Portals iovs
3749 * documentation updates
3750 * bug fixes and cleanups
3751 * small changes in the DLM wire protocol
3753 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3754 * version 0_5_1 with some initial stability,
3755 * locking on MD and file I/O.
3756 * documentation updates
3757 * several bug fixes since 0.5.0
3758 * small changes in wire protocol
3760 2002-07-18 Phil Schwan <phil@clusterfs.com>
3762 * delivered as Lustre Light Alpha
3763 * fixed a crash after handling invalid MDS requests
3764 * fixed directory pages for architectures with non-4k pages sizes
3766 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3767 * release version v0_4_4
3768 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3770 * Fixes a number of interruption problems with OST operations.
3771 * Update documentation for portals header changes
3772 * Move all wire protocol structs/defines to lustre_idl.h
3773 * Fixes symlink length bug.
3774 * Add tcpdump to repository.
3776 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3777 * release version v0_4_3
3778 * Fixes statfs for inodes on extN.
3779 * Fixes bug in runtests which would delete /etc/hosts.
3780 * Use 64-bit object IDs wherever possible (not into VFS though)
3781 Remove ost_get_info, which is unused by lustre, and out of date.
3783 2002-07-03 Peter Braam <braam@clusterfs.com>
3784 * release version v0_4_2 Fixes a lookup error (type not passed)
3785 * move forward to head of Portals
3786 * move forward to latest Lustre kernel
3788 2002-06-25 Peter Braam <braam@clusterfs.com>
3789 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob