1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10),
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.39.cfs8
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: don't use GFP_* in generic Lustre code.
18 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
22 Description: obdclass is limited by single OBD_ALLOC(idarray)
23 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
27 Description: warnings with build patchless client with vanila 2.6.19 and up
28 Details : change old ctl_table style and replace ctl_table/ctl_table_header
29 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
33 Frequency : Only for SLES
34 Description: sanity_quota fail test_1
35 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
36 file, which makes TSTID[2] inunique.
40 Frequency : Only for SLES9
41 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
42 Details : Update target/ChangeLog/which_patch .
46 Frequency : Only for test_10 in sanity-quota.sh
47 Description: a bug in quota test code
48 Details : checking a test flag in a wrong place
52 Frequency : Only for test_13 in sanity-quota.sh
53 Description: a bug in quota test script
54 Details : shouldn't check overfull allocated quota
58 Description: jbd statistics
59 Details : Port older jbd statistics patch for sles10
61 Severity : enhancement
63 Description: Add printing inode info into message about error in writepage.
68 Description: reply_lock_interpret crash due to race with it and lock cancel.
69 Details : Do not replay locks that are being cancelled. Do not reference
70 locks by their address during replay, just by their handle.
74 Frequency : When flocks are used.
75 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
76 Details : Flock locks might destroy just granted lock if it could be merged
77 with another existing flock, this is done in completion handler,
78 so teach ldlm_cli_enquque_fini that this is a valid case for
84 Description: Oops in read and write path when failing to allocate lock.
85 Details : Check if lock allocation failed and return error back.
89 Description: lstripe command fails for valid OST index
90 Details : The stripe offset is compared to lov->desc.ld_tgt_count
91 instead of lov->desc.ld_active_tgt_count.
95 Description: sanity.sh failed test 103
96 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
99 --------------------------------------------------------------------------------
101 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
103 * Support for kernels:
104 2.6.5-7.283 (SLES 9),
105 2.6.9-55.EL (RHEL 4),
106 2.6.16.46-0.14 (SLES 10),
107 2.6.18.8 vanilla (kernel.org)
108 * Client support for unpatched kernels:
109 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
110 2.6.16 - 2.6.22 vanilla (kernel.org)
111 * Due to recently discovered recovery problems, we do not recommend
112 using patchless RHEL 4 clients with this or any earlier release.
113 * Recommended e2fsprogs version: 1.39.cfs8
114 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
115 * Starting with this release, the ldiskfs backing filesystem required
116 by Lustre is now in its own package, lustre-ldiskfs. This package
117 should be installed. It is versioned separately from Lustre and
118 may be released separately in future.
122 Description: block reactivating mgc import until all deactivates complete
123 Details : Fix race when failing back MDT/MGS to itself (testing)
125 Severity : enhancement
127 Description: add optional extra BUILD_VERSION info
128 Details : add a new environment variable (namely LUSTRE_VERS) which allows
129 to override the lustre version.
132 Frequency : 2.6.18 servers only
134 Description: ll_kern_mount() doesn't release the module reference
135 Details : The ldiskfs module reference count never drops down to 0
136 because ll_kern_mount() doesn't release the module reference.
141 Description: server LBUG when using old ost_num_threads parameter
142 Details : Accept the old ost_num_threads parameter but warn that it
143 is deprecated, and fix an off-by-one error that caused an LBUG.
148 Description: Transient SCSI error results in persistent IO issue
149 Details : iobuf->dr_error is not reinitialized to 0 between two
153 Frequency : sometimes when underlying device returns I/O errors
155 Description: OSTs not going read-only during write failures
156 Details : OSTs are not remounted read-only when the journal commit threads
157 get I/O errors because fsfilt_ext3 calls journal_start/stop()
158 instead of the ext3 wrappers.
162 Description: poor connect scaling with increasing client count
163 Details : Don't run filter_grant_sanity_check for more than 100 exports
164 to improve scaling for large numbers of clients.
167 Frequency : SLES10 only
169 Description: sanity-quota.sh quotacheck failed: rc = -22
170 Details : Quotas cannot be enabled on SLES10.
173 Frequency : liblustre clients only
175 Description: getdirentries does not give error when run on compute nodes
176 Details : getdirentries does not fail when the size specified as an argument
177 is too small to contain at least one entry
179 Severity : enhancement
181 Description: Add LNET router traceability for debug purposes
182 Details : If a checksum failure occurs with a router as part of the
183 IO path, the NID of the last router that forwarded the bulk data
184 is printed so it can be identified.
189 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
190 Details : Due to a race condition, liblustre clients were occasionally
193 Severity : enhancement
195 Description: lfs setstripe use optional parameters instead of postional
198 Severity : enhancement
200 Description: Nanosecond timestamp support for ldiskfs
201 Details : The on-disk ldiskfs filesystem has added support for nanosecond
202 resolution timestamps. There is not yet support for this at
203 the Lustre filesystem level.
206 Frequency : during server recovery
208 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
209 Details : request with rq_no_resend flag not awake l_wait_event if they get a
213 Frequency : nfs export on patchless client
215 Description: connectathon hang when test nfs export over patchless client
216 Details : Disconnected dentry cannot be found with lookup, so we do not need
217 to unhash it or make it invalid
220 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
221 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
222 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
227 Description: open req refcounting wrong on reconnect
228 Details : If reconnect happened between getting open reply from server and
229 call to mdc_set_replay_data in ll_file_open, we will schedule
230 replay for unreferenced request that we are about to free.
231 Subsequent close will crash in variety of ways.
232 Check that request is still eligible for replay in
233 mdc_set_replay_data().
238 Description: disable writes to filesystem when reading health_check file
239 Details : the default for reading the health_check proc file has changed
240 to NOT do a journal transaction and write to disk, because this
241 can cause reads of the /proc file to hang and block HA state
242 checking on a healthy but otherwise heavily loaded system. It
243 is possible to return to the previous behaviour during configure
244 with --enable-health-write.
246 Severity : enhancement
248 Description: 64-bit inode version
249 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
250 to the inode. This will be required for version-based recovery.
255 Description: MDS fails to start if a duplicate client export is detected
256 Details : in some rare cases it was possible for a client to connect to
257 an MDS multiple times. Upon recovery the MDS would detect this
258 and fail during startup. Handle this more gracefully.
260 Severity : enhancement
262 Description: Add -o localflock option to simulate old noflock
264 Details : This will achieve local-only flock/fcntl locks
270 Description: log_commit_thread vs filter_destroy race leads to crash
271 Details : Take import reference before releasing llog record semaphore
276 Description: Wrong request locking in request set processing
277 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
278 imp_delayed_list, in this place should be used imp_lock.
281 Frequency : when reconnection
283 Description: Grant Leak when osc reconnect to OST
284 Details : When osc reconnect ost, OST(filter) should check whether it
285 should grant more space to client by comparing fed_grant and
286 cl_avail_grant, and return the granted space to client instead
287 of "new granted" space, because client will call osc_init_grant
288 to update the client grant space info.
291 Frequency : when client reconnect to OST
293 Description: Grant Leak when osc do resend and replay bulk write
294 Details : When osc reconnect to OST, OST(filter)should clear grant info of
295 bulk write request, because the grant info while be sync between
296 OSC and OST when reconnect, and we should ignore the grant info
297 these of resend/replay write req.
302 Description: Grant space more than avaiable left space sometimes.
303 Details : When then OST is about to be full, if two bulk writing from
304 different clients came to OST. Accord the avaliable space of the
305 OST, the first req should be permitted, and the second one
306 should be denied by ENOSPC. But if the seconde arrived before
307 the first one is commited. The OST might wrongly permit second
308 writing, which will cause grant space > avaiable space.
311 Frequency : when client is evicted
313 Description: Grant might be wrongly erased when osc is evicted by OST
314 Details : when the import is evicted by server, it will fork another
315 thread ptlrpc_invalidate_import_thread to invalidate the
316 import, where the grant will be set to 0. While the original
317 thread will update the grant it got when connecting. So if
318 the former happened latter, the grant will be wrongly errased
319 because of this race.
324 Description: Checking Stale with correct fid
325 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
326 to check whether it is stale, because sometimes, we want the
327 enqueue happened anyway, and op_data.fid2 will not be initialized.
329 Severity : enhancement
331 Description: update patchless client
332 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
335 Frequency : only with 2.4 kernel
337 Description: random memory corruption
338 Details : size of struct ll_inode_info is to big for union inode.u and this
339 can be cause of random memory corruption.
344 Description: Memory leak in recovery
345 Details : Lov_mds_md was not free in an error handler in mds_create_object.
346 It should also check obd_fail before fsfilt_start, otherwise if
347 fsfilt_start return -EROFS,(failover mds during mds recovery).
348 then the req will return with repmsg->transno = 0 and rc = EROFS.
349 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
350 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
351 be freed no matter whether fsfilt_commit success or not.
354 Frequency : only with huge count clients
356 Description: Prevents from taking the superblock lock in llap_from_page for
358 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
359 for llap_from_page call in ll_removepage prevents from taking the
360 superblock lock for a soon died page.
365 Description: Not check open intent error before release open handle
366 Details : in some rare cases, the open intent error is not checked before
367 release open handle, which may cause
368 ASSERTION(open_req->rq_transno != 0), because it tries to release
369 the failed open handle.
374 Description: Set cat log bitmap only after create log success.
375 Details : in some rare cases, the cat log bitmap is set too early. and it
376 should be set only after create log success.
380 Description: Accessing a block bevice can re-enable I/O when Lustre is
381 tearing down a device.
382 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
387 Description: service threads may hog cpus when there are a lot of requests
389 Details : Insert cond_resched to give other threads a chance to use some of
395 Description: the cat log was not initialized in recovery
396 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
397 unlink log on mds will not be initialized until mds post
398 recovery. And also in mds post recovery, the unlink log will
399 initialization will be done asynchronausly, so there will be race
400 between add unlink log and unlink log initialization.
404 Description: brw_stats were being printed incorrectly
405 Details : brw_stats were being printed as log2 but all of them were not
406 recorded as log2. Also remove some code duplication arising from
407 filter_tally_{read,write}.
411 Frequency : rare, only in recovery.
412 Description: ASSERTION(req->rq_type != LI_POISON) failed
413 Details : imp_lock should be held while iterating over imp_sending_list for
414 prevent destroy request after get timeout in ptlrpc_queue_wait.
418 Description: replay-single.sh test 52 fails
419 Details : A lock's skiplist need to be cleanup when it being unlinked
420 from its resource list.
424 Description: Short directio read returns full requested size rather than
426 Details : Direct I/O operations should return actual amount of bytes
427 transferred rather than requested size.
429 Severity : enhancement
431 Description: metadata RPC reduction (e.g. for rm performance)
432 Details : decrease the amount of synchronous RPC between clients and servers
433 by canceling conflicing lock before the operation on the client side
434 and packing thier handles into the main operation RPC to server.
436 --------------------------------------------------------------------------------
438 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
443 Frequency : on some architectures
445 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
446 Details : Uninitialized flags sometimes cause configuration commands to
450 Frequency : patchless clients only
452 Description: missing __iget() symbol export
453 Details : The __iget() symbol export is missing. To avoid the need for
454 this on patchless clients the deathrow inode reaper is turned
455 off, and we depend on the VM to clean up old inodes. This
456 dependency was during via the fix for bug 12181.
458 --------------------------------------------------------------------------------
460 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
462 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
463 INTEROPERATE with older versions automatically. In many cases a
464 special upgrade step is needed. Please read the
465 user documentation before upgrading any part of a live system.
466 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
467 version will not interoperate with 1.6 betas before beta5 (1.5.95).
468 * WARNING: Lustre configuration and startup changes are required with
469 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
474 Severity : enhancement
476 Description: MountConf
477 Details : Lustre configuration is now managed via mkfs and mount
478 commands instead of lmc and lconf. New obd types (MGS, MGC)
479 are added for dynamic configuration management. See
480 https://mail.clusterfs.com/wikis/lustre/MountConf for
483 Severity : enhancement
485 Description: dynamic OST addition
486 Details : OSTs can now be added to a live filesystem
488 Severity : enhancement
490 Description: startup order invariance
491 Details : MDTs and OSTs can be started in any order. Clients only
492 require the MDT to complete startup.
494 Severity : enhancement
496 Description: parallel, asynchronous orphan cleanup
497 Details : orphan cleanup is now performed in separate threads for each
498 OST, allowing parallel non-blocking operation.
500 Severity : enhancement
502 Description: optimized stripe assignment
503 Details : stripe assignments are now made based on ost space available,
504 ost previous usage, and OSS previous usage, in order to try
505 to optimize storage space and networking resources.
507 Severity : enhancement
509 Description: Permanently set tunables
510 Details : All writable /proc/fs/lustre tunables can now be permanently
511 set on a per-server basis, at mkfs time or on a live
514 Severity : enhancement
516 Description: Lustre message v2
517 Details : Add lustre message format v2.
519 Severity : enhancement
521 Description: client OST exclusion list
522 Details : Clients can be started with a list of OSTs that should be
523 declared "inactive" for known non-responsive OSTs.
527 Description: SPEC SFS validation failure on NFS v2 over lustre.
528 Details : Changes the blocksize for regular files to be 2x RPC size,
529 and not depend on stripe size.
531 Severity : enhancement
533 Description: Multiple MD RPCs in flight.
534 Details : Further unserialise some read-only MDS RPCs - learn about intents.
535 To avoid overly-overloading MDS, introduce a limit on number of
536 MDS RPCs in flight for a single client and add /proc controls
537 to adjust this limit.
539 Severity : enhancement
541 Description: client read/write statistics
542 Details : Add client read/write call usage stats for performance
543 analysis of user processes.
544 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
545 file access. extents_stats shows chunk size distribution.
546 extents_stats_per_process show chunk size distribution per
549 Severity : enhancement
551 Description: mds statistics
552 Details : Add detailed mds operations statistics in
553 /proc/fs/lustre/mds/*/stats.
557 Description: Failure of copying files with lustre special EAs.
558 Details : Client side always return success for setxattr call for lustre
559 special xattr (currently only "trusted.lov").
564 Description: Refcount LNET uuids
565 Details : The global LNET uuid list grew linearly with every startup;
566 refcount repeated list entries instead of always adding to
569 Severity : enhancement
571 Description: Dynamic service threads
572 Details : Within a small range, start extra service threads
573 automatically when the request queue builds up.
576 Frequency : mixed-endian client/server environments
578 Description: mixed-endian crashes
579 Details : The new msg_v2 system had some failures in mixed-endian
582 Severity : enhancement
584 Description: Easy OST removal
585 Details : OSTs can be permanently deactivated with e.g. 'lctl
586 conf_param lustre-OST0001.osc.active=0'
588 Severity : enhancement
590 Description: MGS proc entries
591 Details : Added basic proc entries for the MGS showing what filesystems
594 Severity : enhancement
596 Description: provide MGS failover
597 Details : Added config lock reacquisition after MGS server failover.
599 Severity : enhancement
601 Description: add Linux 2.4 support
602 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
606 Description: a large application tries to do I/O to the same resource and dies
608 Details : Check the req->rq_arrival time after the call to
609 ost_brw_lock_get(), but before we do anything about
610 processing it & sending the BULK transfer request. This
611 should help move old stale pending locks off the queue as
612 quickly as obd_timeout.
615 Frequency : when an incorrect nid is specified during startup
617 Description: ptlrpc connect to non-existant node causes kernel crash
618 Details : LNET can't be re-entered from an event callback, which
619 happened when we expire a message after the export has been
620 cleaned up. Instead, hand the zombie cleanup off to another
623 Severity : enhancement
625 Description: plain/inodebits lock performance improvement
626 Details : Grouping plain/inodebits in granted list by their request modes
627 and bits policy, thus improving the performance of search through
631 Frequency : only if OST filesystem is corrupted
633 Description: client incorrectly hits assertion in ptlrpc_replay_req()
634 Details : for a short time RPCs with bulk IO are in the replay list,
635 but replay of bulk IOs is unimplemented. If the OST filesystem
636 is corrupted due to disk cache incoherency and then replay is
637 started it is possible to trip an assertion. Avoid putting
638 committed RPCs into the replay list at all to avoid this issue.
641 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
643 Description: System hang on startup
644 Details : This bug allowed the liblustre (e.g. catamount) client to
645 return to the app before handling all startup RPCs. This
646 could leave the node unresponsive to lustre network traffic
647 and manifested as a server ptllnd timeout.
649 Severity : enhancement
651 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
652 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
653 boolean to control whether to print peer debug info when a
654 client's RPC times out.
657 Frequency : only for kernels with patches from Lustre below 1.4.3
659 Description: Remove old rdonly API
660 Details : Remove old rdonly API which unused from at least lustre 1.4.3
663 Frequency : only for devices with external journals
665 Description: Set external device read-only also
666 Details : During a commanded failover stop, we set the disk device
667 read-only while the server shuts down. We now also set any
668 external journal device read-only at the same time.
671 Frequency : when upgrading from 1.4 while trying to change parameters
673 Description: The wrong (new) MDC name was used when setting parameters for
674 upgraded MDT's. Also allows changing of OSC (and MDC)
675 parameters if --writeconf is specified at tunefs upgrade time.
678 Frequency : when setting specific ost indicies
680 Description: QOS code breaks on skipped indicies
681 Details : Add checks for missing OST indicies in the QOS code, so OSTs
682 created with --index need not be sequential.
684 Severity : enhancement
686 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
687 Details : The uninit_groups feature works in conjunction with the kernel
688 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
689 up the pass1 processing of e2fsck. This is a read-only feature
690 in ldiskfs2 only, so older kernels and current ldiskfs cannot
691 mount filesystems that have had this feature enabled.
693 Severity : enhancement
695 Description: Improve multi-block allocation algorithm to avoid fragmentation
696 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
697 allocation locality and avoid filesystem fragmentation.
699 ------------------------------------------------------------------------------
701 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
703 * Support for kernels:
704 2.4.21-47.0.1.EL (RHEL 3)
706 2.6.9-42.0.10.EL (RHEL 4)
707 2.6.12.6 vanilla (kernel.org)
708 2.6.16.27-0.9 (SLES 10)
709 * Recommended e2fsprogs version: 1.39.cfs5
711 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
715 Frequency : occasional, depends on client load and configuration
716 Bugzilla : 12181, 12203
717 Description: data loss for recently-modified files
719 Details : In some cases it is possible that recently written or created
720 files may not be written to disk in a timely manner (this should
721 normally be within 30s unless client IO load is very high).
722 The problem appears as zero-length files or files that are a
723 multiple of 1MB in size after a client crash or client eviction
724 that are missing data at the end of the file.
726 This problem is more likely to be hit on clients where files are
727 repeatedly created and unlinked in the same directory, clients
728 have a large amount of RAM, have many CPUs, the filesystem has
729 many OSTs, the clients are rebooted frequently, and/or the files
730 are not accessed by other nodes after being written.
732 The presence of the problem can be detected by looking at
733 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
734 smaller than the second (nr_unused) then dirty files will not
735 be flushed automatically to disk. "sync; sleep 10" should be
736 run several times on the node before unmounting it to update
737 Lustre (this is also safe to run on nodes without this problem).
739 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
740 that can cause this same problem, so customers using that kernel
741 also need to update the kernel in addition to Lustre. In order
742 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
744 It is normal that files written just before a client crash (less
745 than 30s) may not yet have been flushed to disk, even for local
749 Frequency : frequent on thin XT3 nodes
751 Description: UUID collision on thin XT3 Linux nodes
752 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
753 randomly, since we relied on an insufficiently-seeded PRNG.
758 Description: OSS hangs after "All ost request buffers busy"
759 Details : A deadlock between quota and journal operations caused OSS
760 hangs after printing "All ost request buffers busy."
763 Frequency : always on liblustre builds
765 Description: Cleanup compiler warnings on liblustre
768 Frequency : always on liblustre builds on XT3
770 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
774 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
775 > 32000 subdirectories
776 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
777 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
778 subdirectory count crosses 32000. This will aid e2fsck to
779 correctly handle more than 32000 subdirectories.
782 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
784 Description: System hang on startup
785 Details : This bug allowed the liblustre (e.g. catamount) client to
786 return to the app before handling all startup RPCs. This
787 could leave the node unresponsive to lustre network traffic
788 and manifested as a server ptllnd timeout.
790 Severity : enhancement
792 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
793 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
794 boolean to control whether to print peer debug info when a
795 client's RPC times out.
800 Description: make O_SYNC working on 2.6 kernels
801 Details : 2.6 kernels use different method for mark pages for write,
802 so need add a code to lustre for O_SYNC work.
807 Description: Failure to close file and release space on NFS
808 Details : Put inode details into lock acquired in ll_intent_file_open.
809 Use mdc_intent_lock in ll_intent_open to properly
810 detect all kind of errors unhandled by mdc_enqueue.
815 Description: proc file read during shutdown sometimes raced obd removal,
817 Details : Add lock to prevent obd access after proc file removal.
820 Frequency : Only for files larger than 4GB on 32-bit clients.
822 Description: improperly doing page alignment of locks
823 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
824 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
829 Description: RPCs being resent when they shouldn't be
830 Details : Some RPCs that should not be resent are being resent. This
831 can cause inconsistencies in the RPC state machine. Do not
832 resend such requests.
835 Frequency : rare, only with NFS export
837 Description: Crash on NFS re-export node
838 Details : under very unusual load conditions an assertion is hit in
839 ll_intent_file_open()
842 Frequency : only if OST filesystem is corrupted
844 Description: client incorrectly hits assertion in ptlrpc_replay_req()
845 Details : for a short time RPCs with bulk IO are in the replay list,
846 but replay of bulk IOs is unimplemented. If the OST filesystem
847 is corrupted due to disk cache incoherency and then replay is
848 started it is possible to trip an assertion. Avoid putting
849 committed RPCs into the replay list at all to avoid this issue.
854 Description: large O_DIRECT requests fail under memory pressure/fragmentation
855 Details : Large single O_DIRECT read and write calls can fail to allocate
856 a sufficiently large buffer to process the request. In case of
857 allocation failure the allocation is retried with a smaller
858 buffer and broken into smaller requests.
860 Severity : enhancement
862 Description: Add -o localflock option to simulate old noflock behaviour.
863 Details : This will achieve local-only flock/fcntl locks coherentness.
868 Description: versioning check is incomplete
869 Details : Checking the version difference of client vs. server, report
870 error if the gap is too big.
875 Description: add support PG_writeback bit
876 Details : add support for PG_writeback bit for Lustre, for more carefull
877 work with page cache in 2.6 kernel. This also fix some deadlocks
878 and remove hack for work O_SYNC with 2.6 kernel.
880 Severity : enhancement
882 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
883 Details : The uninit_groups feature works in conjunction with the kernel
884 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
885 up the pass1 processing of e2fsck. This is a read-only feature
886 in ldiskfs2 only, so older kernels and current ldiskfs cannot
887 mount filesystems that have had this feature enabled.
889 Severity : enhancement
891 Description: Improve multi-block allocation algorithm to avoid fragmentation
892 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
893 allocation locality and avoid filesystem fragmentation.
895 ------------------------------------------------------------------------------
897 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
899 * Support for kernels:
900 2.6.9-42.0.3.EL (RHEL 4)
902 2.4.21-47.0.1.EL (RHEL 3)
903 2.6.12.6 vanilla (kernel.org)
904 2.6.16.21-0.8 (SLES10)
905 * Recommended e2fsprogs version: 1.39.cfs2-0
907 * The backwards-compatible /proc/sys/portals symlink has been removed
908 in this release. Before upgrading, please ensure that you change
909 any configuration scripts or /etc/sysctl.conf files that access
910 /proc/sys/portals/* or sysctl portals.* to use the corresponding
911 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
912 in advance of the upgrade on any system running Lustre 1.4.6 or
913 newer, since /proc/sys/lnet was added in that version.
914 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
918 Frequency : only when quota is used
920 Description: avoid scanning export list for quota master
921 Details : Change the algorithms to avoid scanning export list in order
922 to improve the efficiency.
925 Frequency : MDS failover only, very rarely
927 Description: "went back in time" messages on mds failover
928 Details : The greatest transno may be lost when the current operation
929 finishes with an error (transno==0) and the client's last_rcvd
930 record is over-written. Save the greatest transno in the
931 mds_last_transno for this case.
934 Frequency : always for specific kernels and striping counts
936 Description: client may get "Matching packet too big" without ACL support
937 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
938 when trying to access files in certain configurations. The
939 clients should in fact be denied when mounting because they do
943 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
945 Description: many clients connecting with IO in progress causes connect timeouts
946 Details : Avoid synchronous journal commits to avoid delays caused by many
947 clients connecting/disconnecting when bulk IO is in progress.
948 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
949 OST_IO_PORTAL to avoid delays behind potentially many pending
953 Frequency : occasionally with multiple writers to a single file
955 Description: shared writes to file may result in wrong size reported by stat()
956 Details : Allow growing of kms when extent lock is cancelled
959 Frequency : always with random mmap IO to multi-striped file
961 Description: mmap write might be lost if we are writing to a 'hole' in stripe
962 Details : Only if the hole is at the end of OST object so that kms is too
963 small. Fix is to increase kms accordingly in ll_nopage.
966 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
968 Description: writes to a missing object would leak memory on the OST
969 Details : If there is an inconsistency between the MDS and OST filesystems,
970 such that the MDS references an object that doesn't exist, writes
971 to that object will leak memory due to incorrect cleanup in the
972 error handling path, eventually running out of memory on the OST.
977 Description: Creating too long symlink causes lustre errors
978 Details : Check symlink and name lengths before sending requests to MDS.
981 Frequency : only if flock is enabled (not on by default)
983 Description: posix locks not released on fd closure on 2.6.9+
984 Details : We failed to add posix locks to list of inode locks on 2.6.9+
985 kernels, this caused such locks not to be released on fd close and
986 then assertions on fs unmount about still used locks.
989 Frequency : MDS failover only, very rarely
991 Description: clients may get ASSERTION(granted_lock != NULL)
992 Details : When request was taking a long time, and a client was resending
993 a getattr by name lock request. The were multiple lock requests
994 with the same client lock handle and
995 mds_getattr_name->fixup_handle_for_resent_request found one of the
996 lock handles but later failed with ASSERTION(granted_lock != NULL).
1001 Description: handle->h_buffer_credits > 0, assertion failure
1002 Details : h_buffer_credits is zero after truncate, causing assertion
1003 failure. This patch extends the transaction or creates a new
1007 Frequency : NFS re-export or patchless client
1008 Bugzilla : 11179, 10796
1009 Description: Crash on NFS re-export node (__d_move)
1010 Details : We do not want to hash the dentry if we don't have a lock.
1011 But if this dentry is later used in d_move, we'd hit uninitialised
1012 list head d_hash, so we just do this to init d_hash field but
1013 leave dentry unhashed.
1016 Frequency : NFS re-export or patchless client
1018 Description: NFS exports has problem with symbolic link
1019 Details : lustre client didn't properly install dentry when re-exported
1020 to NFS or running patchless client.
1023 Frequency : NFS re-export or patchless client
1025 Description: Various nfs/patchless fixes.
1026 Details : fixes reuse disconected alias for lookup process - this fixes
1027 warning "find_exported_dentry: npd != pd",
1028 fix permission error with open files at nfs.
1029 fix apply umask when do revalidate.
1032 Frequency : occasional
1034 Description: Crash on NFS re-export node
1035 Details : calling clear_page() on the wrong pointer triggered oops in
1036 generic_mapping_read().
1039 Frequency : rarely, using O_DIRECT IO
1041 Description: unaligned directio crashes client with LASSERT
1042 Details : check for unaligned buffers before trying any requests.
1045 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1047 Description: stale data returned from RAID cache
1048 Details : If only a small amount of IO is done to the RAID device before
1049 reading it again it is possible to get stale data from the RAID
1050 cache instead of reading it from disk.
1053 Frequency : always for sles10 kernel
1055 Description: sles10 support
1056 Details : ll_follow_link: compile fixes and using of nd_set_link
1057 under newer kernels.
1060 Frequency : depends on arch, kernel and compiler version, always on sles10
1063 Description: recursive or deep enough symlinks cause stack overflow
1064 Details : getting rid of large stack-allocated variable in
1068 Frequency : depends on hardware
1070 Description: lustre write performance loss in the SLES10 kernel
1071 Details : the performance loss is caused by using of write barriers in the
1072 ext3 code. The SLES10 kernel turns barrier support on by
1073 default. The fix is to undo that change for ldiskfs.
1075 ------------------------------------------------------------------------------
1077 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1079 * Support for kernels:
1080 2.6.9-42.0.3EL (RHEL 4)
1081 2.6.5-7.276 (SLES 9)
1082 2.4.21-47.0.1.EL (RHEL 3)
1083 2.6.12.6 vanilla (kernel.org)
1087 Frequency : quota enabled and large files being deleted
1089 Description: releasing more than 4GB of quota at once hangs OST
1090 Details : If a user deletes more than 4GB of files on a single OST it
1091 will cause the OST to spin in an infinite loop. Release
1092 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1097 Description: statfs data retrieved from /proc may be stale or zero
1098 Details : When reading per-device statfs data from /proc, in the
1099 {kbytes,files}_{total,free,avail} files, it may appear
1100 as zero or be out of date.
1103 Frequency : systems with MD RAID1 external journal devices
1105 Description: lconf's call to blkid is confused by RAID1 journal devices
1106 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1107 of returning all block devices that match the journal UUID.
1110 Frequency : always, for aggregate stripe size over 4GB
1112 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1113 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1114 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1117 Frequency : always if "lfs find" used on a local file/directory
1119 Description: "lfs find" segfaults if used on a local file/directory
1120 Details : The case where a directory component was not specified wasn't
1121 handled correctly. Handle this properly.
1124 Frequency : always on ppc64
1126 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1127 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1128 perform endianness conversion.
1131 Frequency : rarely (truncate to non-zero file size after write under load)
1132 Bugzilla : 10730, 10687
1133 Description: Files padded with zeros to next 4K multiple
1134 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1135 it is possible that files that are truncated to a non-zero size
1136 immediately after being written are filled with zero bytes beyond
1137 the truncated size. No file data is lost.
1139 Severity : enhancement
1141 Description: Allow recovery/failover for liblustre clients.
1142 Details : liblustre clients were unaware of failover configurations until
1145 Severity : enhancement
1147 Description: user file locks should fail when not mounting with flock option
1148 Details : Set up an error-returning stub in ll_file_operations.lock field
1149 to prevent incorrect behaviour when client is mounted without
1150 flock option. Also, set up properly f_op->flock field for
1154 Frequency : always on ia64
1156 Description: "lfs df" loops on printing out MDS statfs information
1157 Details : The obd_ioctl_data was not initialized and in some systems
1158 this caused a failure during the ioctl that did not return
1159 an error. Initialize the struct and return an error on failure.
1162 Frequency : SLES 9 only
1164 Description: Error of copying files with lustre special EAs as root
1165 Details : Client side always return success for setxattr call for lustre
1166 special xattr (currently only "trusted.lov").
1169 Frequency : rarely on clusters with both ia64+i386 clients
1171 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1172 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1173 clients are doing concurrent writes to the same file it is
1174 possible that the ia64 clients may LASSERT because the OST
1175 extent locks are not PAGE_SIZE aligned. Ensure that grown
1176 locks are always aligned on the request boundary.
1179 Frequency : specific use, occasional
1181 Description: Overwriting in use executable truncates on-disk binary image
1182 Details : If one node attempts to overwrite an executable in use by
1183 another node, we now correctly return ETXTBSY instead of
1184 truncating the file.
1186 Severity : enhancement
1188 Description: Async OSC create to avoid the blocking unnecessarily.
1189 Details : If a OST has no remain object, system will block on the creating
1190 when need to create a new object on this OST. Now, ways use
1191 pre-created objects when available, instead of blocking on an
1192 empty osc while others are not empty. If we must block, we block
1193 for the shortest possible period of time.
1198 Description: chmod on Lustre root is propagated to other clients
1199 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1200 invalid by the follow_mount time.
1203 Frequency : liblustre clients only
1205 Description: Race in 'instant cancel' lock handling could lead to such locks
1206 never to be granted in case of SMP MDS
1207 Details : Do not destroy not yet granted but cbpending locks in
1211 Frequency : replay/resend of open
1213 Description: non null lock assetion failure in mds_intent_policy
1214 Details : Trying to replay/resend lockless open requests resulted in
1215 mds_open() returning 0 with no lock. Now it sets a flag if
1216 a lock is going to be returned.
1218 Severity : enhancement
1220 Description: Checksum enhancements
1221 Details : New checksum enhancements allow for resending RPCs that failed
1224 Severity : enhancement
1226 Description: Tunables on number of dirty pages in cacche
1227 Details : Allow to set limit on number of dirty pages cached.
1232 Description: client crash on unmount - lock still has references
1233 Details : In some error handling cases it was possible to leak a lock
1234 reference on a client while accessing a file. This was not
1235 harmful to the client during operation, but would cause the
1236 client to crash when the filesystem is unmounted.
1239 Frequency : specific case, rare
1241 Description: ETXTBSY on mds though file not in use by client
1242 Details : ETXTBSY is no longer incorrectly returned when attempting to
1243 chmod or chown a directory that the user previously tried to
1244 execute or a currently-executing binary.
1247 Frequency : extremely rare except on liblustre-based clients
1249 Description: Lustre space not freed when files are deleted
1250 Details : Clean up open-unlinked files after client eviction. Previously
1251 the unlink was skipped and the files remained as orphans.
1256 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1257 Details : In some cases it was possible to send a blocking callback to
1258 a client doing a glimpse, even though that client didn't get
1259 a lock granted. When the glimpse lock is cancelled on the OST
1260 the freed lock is left on the waiting list and corrupted the list.
1263 Frequency : all core dumps
1265 Description: Broke core dumps to lustre
1266 Details : Negative dentry may be unhashed if parent does not have UPDATE
1267 lock, but some callers, e.g. do_coredump, expect dentry to be
1268 hashed after successful create, hash it in ll_create_it.
1270 ------------------------------------------------------------------------------
1272 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1274 * Support for kernels:
1275 2.6.9-42.0.2.EL (RHEL 4)
1276 2.6.5-7.276 (SLES 9)
1277 2.4.21-40.EL (RHEL 3)
1278 2.6.12.6 vanilla (kernel.org)
1282 Frequency : always on RHEL 3
1284 Description: Number of open files grows over time
1285 Details : The number of open files grows over time, whether or not
1286 Lustre is started. This was due to a filp leak introduced
1287 by one of our kernel patches.
1289 ------------------------------------------------------------------------------
1291 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1293 * Support for kernels:
1294 2.6.9-42.EL (RHEL 4)
1295 2.6.5-7.276 (SLES 9)
1296 2.4.21-40.EL (RHEL 3)
1297 2.6.12.6 vanilla (kernel.org)
1302 Bugzilla : 5719, 9635, 9792, 9684
1303 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1304 Details : If a server is under heavy load and cannot reply to new
1305 connection requests before the client resends the (re)connect,
1306 the connection handling code can behave badly if two service
1307 threads are concurrently handing separate (re)connections from
1308 the same client. Add better locking to the connection handling
1309 code, and ensure that only a single connection will be processed
1310 for a given client UUID, even if the lock is dropped.
1312 Severity : enhancement
1314 Description: add TCP zero-copy support to kernel
1315 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1316 sends if the hardware supports scatter-gather and checksumming.
1317 This allows socklnd to do client-write and server-read more
1318 efficiently and reduce CPU utilization from skbuf copying.
1321 Frequency : only if NFS exporting from client
1323 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1324 Details : If setting attributes on a file created under NFS that had
1325 never been opened it would be possible to oops the client
1326 if the file had no objects.
1329 Frequency : always for liblustre
1331 Description: liblustre client does MDS+OSTs setattr RPC for each write
1332 Details : When doing a write from a liblustre client, the client
1333 incorrectly issued an RPC to the MDS and each OST the file was
1334 striped over in order to update the timestamps. When writing
1335 with small chunks and many clients this could overwhelm the MDS
1336 with RPCs. In all cases it would slow down the write because
1337 these RPCs are unnecessary.
1339 Severity : enhancement
1341 Description: allow number of MDS service threads to be changed at module load
1342 Details : It is now possible to change the number of MDS service threads
1343 running. Adding "options mds mds_num_threads={N}" to the MDS's
1344 /etc/modprobe.conf will set the number of threads for the next
1345 time Lustre is restarted (assuming the "mds" module is also
1346 reloaded at that time). The default number of threads will
1347 stay the same, 32 for most systems.
1352 Description: OST crash if filesystem is unformatted or corrupt
1353 Details : If an OST is started on a device that has never been formatted
1354 or if the filesystem is corrupt and cannot even mount then the
1355 error handling cleanup routines would dereference a NULL pointer.
1360 Description: NULL pointer deref in llap_from_page.
1361 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1362 incorrect) mapping if the page was truncated/reclaimed while it was
1363 searched for. Check for this condition and skip such pages when
1364 doing readahead. Introduce extra check to llap_from_page() to
1365 verify page->mapping->host is non-NULL (so page is not anonymous).
1368 Frequency : Sometimes when using sys_sendfile
1370 Description: "page not covered by a lock" warnings from ll_readpage
1371 Details : sendfile called ll_readpage without right page locks present.
1372 Now we introduced ll_file_sendfile that does necessary locking
1373 around call to generic_file_sendfile() much like we do in
1377 Frequency : with certain MDS communication failures at client mount time
1379 Description: NULL pointer deref after failed client mount
1380 Details : a client connection request may delayed by the network layer
1381 and not be sent until after the PTLRPC layer has timed out the
1382 request. If the client fails the mount immediately it will try
1383 to clean up before the network times out the request. Add a
1384 reference from the request import to the obd device and delay
1385 the cleanup until the network drops the request.
1388 Frequency : occasionally during client (re)connect
1390 Description: assertion failure during client (re)connect
1391 Details : processing a client connection request may be delayed by the
1392 client or server longer than the client connect timeout. This
1393 causes the client to resend the connection request. If the
1394 original connection request is replied in this interval, the
1395 client may trip an assertion failure in ptlrpc_connect_interpret()
1396 which thought it would be the only running connect process.
1399 Frequency : only with obd_echo servers and clients that are rebooted
1401 Description: kernel BUG accessing uninitialized data structure
1402 Details : When running an obd_echo server it did not start the ping_evictor
1403 thread, and when a client was evicted an uninitialized data
1404 structure was accessed. Start the ping_evictor in the RPC
1405 service startup instead of the OBD startup.
1407 Severity : enhancement
1408 Bugzilla : 10193 (patchless)
1409 Description: Remove dependency on various unexported kernel interfaces.
1410 Details : No longer need reparent_to_init, exit_mm, exit_files,
1411 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1414 Frequency : rare (only users of deprecated and unsupported LDAP config)
1416 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1417 Details : LDAP apparently contains 'lustreName' attributes instead of
1418 'name'. A simple remapping of the name is sufficient.
1421 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1423 Description: waiting_locks_callback trips kernel BUG if client is evicted
1424 Details : Running with the dump_on_timeout debug flag turned on makes
1425 it possible that the waiting_locks_callback() can try to dump
1426 the Lustre kernel debug logs from an interrupt handler. Defer
1427 this log dumping to the expired_lock_main() thread.
1429 Severity : enhancement
1431 Description: Support NFS exporting on 2.6 kernels.
1432 Details : Implement non-rawops metadata methods for NFS server to use without
1433 changing NFS server code.
1436 Frequency : very rare (synthetic metadata workload only)
1438 Description: two racing renames might cause an MDS thread to deadlock
1439 Details : Running the "racer" program may cause one MDS thread to rename
1440 a file from being the source of a rename to being the target of
1441 a rename at exactly the same time that another thread is doing
1442 so, and the second thread has already enqueued these locks after
1443 doing a lookup of the target and is trying to relock them in
1444 order. Ensure that we don't try to re-lock the same resource.
1447 Frequency : only very large systems with liblustre clients
1449 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1450 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1451 all OSTs in parallel. This allows the eviction of stale liblustre
1452 clients to proceed much faster than if they were done in series,
1453 and also offers similar improvements for other set_info RPCs.
1458 Description: excessive CPU usage during initial read phase on client
1459 Details : During the initial read phase on a client, it would agressively
1460 retry readahead on the file, consuming too much CPU and impacting
1461 performance (since 1.4.5.8). Improve the readahead algorithm
1462 to avoid this, and also improve some other common cases (read
1463 of small files in particular, where "small" is files smaller than
1464 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1469 Description: MDS crash when receiving packet with unknown intent.
1470 Details : Do not LBUG in unknown intent case, just return -EFAULT
1472 Severity : enhancement
1473 Bugzilla : 9293, 9385
1474 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1475 Details : Do not serialize getattr (non-intent version) and statfs.
1478 Frequency : occasional, when OST network is overloaded/intermittent
1480 Description: client evicted by OST after bulk IO timeout
1481 Details : If a client sends a bulk IO request (read or write) the OST
1482 may evict the client if it is unresposive to its data GET/PUT
1483 request. This is incorrect if the network is overloaded (takes
1484 too long to transfer the RPC data) or dropped the OST GET/PUT
1485 request. There is no need to evict the client at all, since
1486 the pinger and/or lock callbacks will handle this, and the
1487 client can restart the bulk request.
1490 Frequency : Always when mmapping file with no objects
1492 Description: client crashes when mmapping file with no objects
1493 Details : Check that we actually have objects in a file before doing any
1494 operations on objects in ll_vm_open, ll_vm_close and
1500 Description: Request leak when working with deleted CWD
1501 Details : Introduce advanced request refcount tracking for requests
1502 referenced from lustre intent.
1504 Severity : Enhancement
1506 Description: Cache open file handles on client.
1507 Details : MDS now will return special lock along with openhandle, if
1508 requested and client is allowed to hold openhandle, even if unused,
1509 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1510 closing files for every read/write openration.
1512 Severity : Enhancement
1514 Description: Cache open negative dentries on client when possible.
1515 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1516 negative dentries on lock revocation.
1521 Description: Remounting a client read-only wasn't possible with a zconf mount
1522 Details : It wasn't possible to remount a client read-only with llmount.
1524 Severity : enhancement
1525 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1526 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1527 MPICH is included to add Lustre-specific ADIO interfaces.
1528 This is based closely on the UFS ADIO layer and only differs
1529 in file creation, in order to allow the OST striping to be set.
1530 This is user-contributed code and not supported by CFS.
1535 Description: extended inode attributes (immutable, append-only) work improperly
1536 when 2.4 and 2.6 kernels are used on client/server or vice versa
1537 Details : Introduce kernel-independent values for these flags.
1539 Severity : enhancement
1542 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1543 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1544 granularity. Now, Fractional MB granularity is be supported,
1545 this is very useful for low memory system.
1547 Severity : enhancement
1549 Description: Getattr by fid
1550 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1551 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1552 have locks and avoids getattr rpc requests for it on every stat.
1555 Frequency : Always, for filesystems larger than 2TB
1557 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1558 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1559 filesystems larger than 2TB to be mounted with mballoc enabled.
1562 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1564 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1565 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1566 filesystems larger than 2TB is mounted then the kernel will
1567 silently corrupt the start of the filesystem. CONFIG_LBD is
1568 enabled for all CFS-supported kernels, but the possibility of
1569 this happening with a modified kernel config exists.
1571 Severity : enhancement
1573 Description: add client O_DIRECT support for 2.6 kernels
1574 Details : It is now possible to do O_DIRECT reads and writes to files
1575 in the Lustre client mountpoint on 2.6 kernel clients.
1577 Severity : enhancement
1579 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1580 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1581 parallel, not waiting for response from every OST before sending
1582 a rpc to the next OST.
1587 Description: setattr vs write race when updating file timestamps
1588 Details : Client processes that update a file timestamp into the past
1589 right after writing to the file (e.g. tar) it is possible that
1590 the updated file modification time can be reset to the current
1591 time due to a race between processing the setattr and write RPC.
1593 Severity : enhancement
1595 Description: Bring 'lfs find' closer in line with regular Linux find.
1596 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1597 -print0 options and obtains all the needed info through the lustre
1600 Severity : enhancement
1602 Description: support up to 1024 configured devices on one node
1603 Details : change obd_dev array from statically allocated to dynamically
1604 allocated structs as they are first used to reduce memory usage
1609 Description: Flush dirty partially truncated pages during truncate
1610 Details : Immediatelly flush partially truncated pages in filter_setattr,
1611 this way we completely avoid having any pages in page cache on OST
1612 and can retire ugly workarounds during writes to flush such pages.
1617 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1618 Details : protect inode from truncation within vfs_unlink() context
1619 just take a reference before calling vfs_unlink() and release it
1620 when parent's i_sem is free.
1625 Description: last_id value checked outside lock on OST caused LASSERT failure
1626 Details : If there were multiple MDS->OST object precreate requests in
1627 flight, it was possible that the OST's last object id was checked
1628 outside a lock and incorrectly tripped an assertion. Move checks
1629 inside locks, and discard old precreate requests.
1632 Frequency : always, if extents are used on OSTs
1634 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1635 Details : OSTs using the extents format would not zero the high 16 bits of
1636 the index physical block number. This is not a problem for any
1637 OST filesystems smaller than 16TB, and no kernels support ext3
1638 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1639 new/modified files) and can be fixed for existing filesystems
1640 with e2fsprogs-1.39-cfs1.
1645 Description: import connection selection may be incorrect if timer wraps
1646 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1647 import connections to be ignored if the 32-bit jiffies counter
1648 wraps. Use a 64-bit jiffies counter.
1651 Frequency : very large clusters immediately after boot
1653 Description: LNET request buffers exhausted under heavy short-term load
1654 Details : If a large number of client requests are generated on a service
1655 that has previously never seen so many requests it is possible
1656 that the request buffer growth cannot keep up with the spike in
1657 demand. Instead of dropping incoming requests, they are held in
1658 the LND until the RPC service can accept more requests.
1661 Frequency : Sometimes during replay
1663 Description: Assertion failure in ll_local_open after replay.
1664 Details : If replay happened on an open request reply before we were able
1665 to set replay handler, reply will become not swabbed tripping the
1666 assertion in ll_local_open. Now we set the handler right after
1667 recognising of open request
1670 Frequency : very rare
1672 Description: kernel reports "badness in vsnprintf"
1673 Details : Reading from the "recovery_status" /proc file in small chunks
1674 may cause a negative length in lprocfs_obd_rd_recovery_status()
1675 call to vsnprintf() (which is otherwise harmless). Exit early
1676 if there is no more space in the output buffer.
1678 Severity : enhancement
1680 Description: clear OBD RPC statistics by writing to them
1681 Details : It is now possible to clear the OBD RPC statistics by writing
1682 to the "stats" file.
1687 Description: Client mtime is not the same on different clients after utimes
1688 Details : In some cases, the client was using the utimes() syscall on
1689 a file cached on another node. The clients now validate the
1690 ctime from the MDS + OSTs to determine which one is right.
1695 Description: Inability to activate failout mode
1696 Details : lconf script incorrectly assumed that in pythong string's numeric
1697 value is used in comparisons.
1700 Frequency : always with multiple stripes per file
1702 Description: Inefficient object allocation for mutli-stripe files
1703 Details : When selecting which OSTs to stripe files over, for files with
1704 a stripe count that divides evenly into the number of OSTs,
1705 the MDS is always picking the same starting OST for each file.
1706 Return the OST selection heuristic to the original design.
1711 Description: mount failures may take full timeout to return an error
1712 Details : Under some heavy load conditions it is possible that a
1713 failed mount can wait for the full obd_timeout interval,
1714 possibly several minutes, before reporting an error.
1715 Instead return an error as soon as the status is known.
1717 Frequency : quota enabled and large files being deleted
1719 Description: releasing more than 4GB of quota at once hangs OST
1720 Details : If a user deletes more than 4GB of files on a single OST it
1721 will cause the OST to spin in an infinite loop. Release
1722 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1727 Description: statfs data retrieved from /proc may be stale or zero
1728 Details : When reading per-device statfs data from /proc, in the
1729 {kbytes,files}_{total,free,avail} files, it may appear
1730 as zero or be out of date.
1733 Frequency : systems with MD RAID1 external journal devices
1735 Description: lconf's call to blkid is confused by RAID1 journal devices
1736 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1737 of returning all block devices that match the journal UUID.
1740 Frequency : always, for aggregate stripe size over 4GB
1742 Description: assertion fails when trying to use 4GB stripe size
1743 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1744 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1747 Frequency : always on ppc64
1749 Description: the first write on an ext3 filesystem with mballoc got stuck
1750 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1751 perform endianness conversion.
1753 ------------------------------------------------------------------------------
1755 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1757 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1758 INTEROPERATE with older versions automatically. Please read the
1759 user documentation before upgrading any part of a live system.
1760 * WARNING: Lustre networking configuration changes are required with
1761 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1764 * Support for kernels:
1765 2.6.9-22.0.2.EL (RHEL 4)
1766 2.6.5-7.244 (SLES 9)
1767 2.6.12.6 vanilla (kernel.org)
1770 Severity : enhancement
1771 Bugzilla : 7981/8208
1772 Description: Introduced Lustre Networking (LNET)
1773 Details : LNET is new networking infrastructure for Lustre, it includes
1774 a reorganized network configuration mode (see the user
1775 documentation for full details) as well as support for routing
1776 between different network fabrics. Lustre Networking Devices
1777 (LNDs) for the supported network fabrics have also been
1778 created for this new infrastructure.
1780 Severity : enhancement
1781 Description: Introduced Access control lists
1782 Details : clients can set ACLs on files and directories in order to have
1783 more fine-grained permissions than the standard Unix UGO+RWX.
1784 The MDS must be started with the "-o acl" mount option.
1786 Severity : enhancement
1787 Description: Introduced filesystem quotas
1788 Details : Administrators may now establish per-user quotas on the
1791 Severity : enhancement
1793 Description: Configuration change for the XT3
1794 The PTLLND is now used to run Lustre over Portals on the XT3
1795 The configure option(s) --with-cray-portals are no longer used.
1796 Rather --with-portals=<path-to-portals-includes> is used to
1797 enable building on the XT3. In addition to enable XT3 specific
1798 features the option --enable-cray-xt3 must be used.
1803 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1804 Details : A race between allocating a new llcd and re-getting the llcd_lock
1805 allowed another thread to grab newly-allocated llcd.
1807 Severity : enhancement
1809 Description: 2.6 OST async journal commit and locking fix to improve performance
1810 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1811 2.6 kernels are now async as they already were in 2.4 kernels so
1812 that they can commit concurrently with the network bulk transfer.
1813 For block-allocated files the filter allocation semaphore is held
1814 to avoid filesystem fragmentation during allocation. BKL lock
1815 removed for 2.6 xattr operations where it is no longer needed.
1820 Description: lconf incorrectly determined whether two IP networks could talk
1821 Details : In some more complicated routing and multiple-network
1822 configurations, lconf will avoid trying to make a network
1823 connection to a disjoint part of the IP space. It was doing the
1824 math incorrectly for one set of cases.
1829 Description: Fix for potential infinite loop processing records in an llog.
1830 Details : If an llog record is corrupted/zeroed, it is possible to loop
1831 forever in llog_process(). Validate the llog record length
1832 and skip the remainder of the block on error.
1835 Frequency : occasional (liblustre only)
1837 Description: liblustre could not open files whose last component is a symlink
1838 Details : sysio_path_walk() would incorrectly pass the open intent to
1839 intermediate path components.
1842 Frequency : rare (liblustre only with non-standard tuning)
1843 Bugzilla : 7201 (7350)
1844 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1845 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1846 after it has been released, only do this for referenced locks
1848 Severity : enhancement
1850 Description: specifying an (invalid) directory default stripe_size of -1
1851 would reset the directory default striping
1852 Details : stripe_size -1 was used internally to signal directory stripe
1853 removal, now use "all default" to signal dir stripe removal
1854 as a directory striping of "all default" is not useful
1857 Frequency : common for large clusters running liblustre clients
1859 Description: doing an ls when liblustre clients are running is slow
1860 Details : sending a glimpse AST to a liblustre client waits for every AST
1861 to time out, as liblustre clients will not respond. Since they
1862 cannot cache data we refresh the OST lock LVB from disk instead.
1864 Severity : enhancement
1866 Description: doing an ls at the same time as file IO can be slow
1867 Details : enqueue and other "small" requests can be blocked behind many
1868 large IO requests. Create a new OST IO portal for non-IO
1869 requests so they can be processed faster.
1872 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1874 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1875 created file with mode 000
1876 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1877 allow an ftruncate() as a non-root user to a file with mode 000.
1878 Lustre now respects this flag to disable mode checks when
1879 truncating a file owned by the user
1882 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1885 Description: Revoking locks from clients that went dead or catatonic might take
1887 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1888 cancellation of such locks instant on servers without waiting for
1889 any reply from clients. Clients drops these locks when cancel
1890 notification from server is received without replying.
1893 Frequency : liblustre-only, when liblustre client dies or becomes busy
1895 Description: Doing ls on Linux clients can take a long time with active
1897 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1898 granting such locks to it in the first place if possible. Locks
1899 are taken by proxy on the OST during the read or write and
1900 dropped immediately afterward. Add connect flags handling, do
1901 not grant locks to liblustre clients for glimpse ASTs.
1903 Severity : enhancement
1905 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1906 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1907 sequential reads and behaves badly with applications doing random
1908 reads. Improve it by reading ahead at least read region, and
1909 avoiding excessive large RPC for small reads.
1911 Severity : enhancement
1913 Description: Creating more than 1000 files for a single job may cause a load
1914 imbalance on the OSTs if there are also a large number of OSTs.
1915 Details : qos_prep_create() uses an OST index reseed value that is an
1916 even multiple of the number of available OSTs so that if the
1917 reseed happens in the middle of the object allocation it will
1918 still utilize the OSTs as uniformly as possible.
1923 Description: OST or MDS may oops in ping_evictor_main()
1924 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1925 but doesn't restart at beginning of obd_exports_timed list
1928 Severity : enhancement
1930 Description: improve by-nid export eviction on the MDS and OST
1931 Details : allow multiple exports with the same NID to be evicted at one
1932 time without re-searching the exports list.
1935 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1937 Description: MDS may oops in groups_free()
1938 Details : in rare race conditions a newly allocated group_info struct is
1939 freed again, and this can be NULL. The 2.4 compatibility code
1940 for groups_free() checked for a NULL pointer, but 2.6 did not.
1943 Frequency : common for liblustre clients doing little filesystem IO
1944 Bugzilla : 9352, 7313
1945 Description: server may evict liblustre clients accessing contended locks
1946 Details : if a client is granted a lock or receives a completion AST
1947 with a blocking AST already set it would not reply to the AST
1948 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1951 Frequency : lfs setstripe, only systems with more than 160 OSTs
1953 Description: unable to set striping with a starting offset beyond OST 160
1954 Details : llapi_create_file() incorrectly limited the starting stripe
1955 index to the maximum single-file stripe count.
1958 Frequency : LDAP users only
1960 Description: lconf did not handle in-kernel recovery with LDAP properly
1961 Details : lconf/LustreDB get_refs() is searching the wrong namespace
1963 Severity : enhancement
1965 Description: bind OST threads to NUMA nodes to improve performance
1966 Details : all OST threads are uniformly bound to CPUs on a single NUMA
1967 node and do their allocations there to localize memory access
1969 Severity : enhancement
1971 Description: llmount can determine client NID directly from Myrinet (GM)
1972 Details : the client NID code from gmnalnid was moved directly into
1973 llmount, removing the need to use this or specifying the
1974 client NID explicitly when mounting GM clients with zeroconf
1977 Frequency : if client is started with down MDS
1979 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
1980 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
1981 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
1982 ptlrpc_import_delayed_req() because we didn't check if the
1983 request was interrupted, and we also didn't break out of the
1984 event loop if there was no timeout
1989 Description: data loss during non-page-aligned writes to a single file from
1990 both multiple nodes and multiple threads on one node at same time
1991 Details : updates to KMS and lsm weren't protected by common lock. Resulting
1992 inconsistency led to false short-reads, that were cached and later
1993 used by ->prepare_write() to fill in partially written page,
1994 leading to data loss.
1997 Frequency : always, if lconf --abort_recovery used
1999 Description: lconf --abort_recovery fails with 'Operation not supported'
2000 Details : lconf was attempting to abort recovery on the MDT device and not
2003 Severity : enhancement
2005 Description: remove cleanup logs
2006 Details : replace lconf-generated cleanup logs with lustre internal
2007 cleanup routines. Eliminates the need for client-cleanup and
2010 Severity : enhancement
2012 Description: add support for EAs (user and system) on lustre filesystems
2013 Details : it is now possible to store extended attributes in the Lustre
2014 client filesystem, and with the user_xattr mount option it
2015 is possible to allow users to store EAs on their files also
2017 Severity : enhancement
2019 Description: Add possibility (config option) to show minimal available OST free
2021 Details : When compiled with --enable-mindf configure option, statfs(2)
2022 (and so, df) will return least minimal free space available from
2023 all OSTs as amount of free space on FS, instead of summary of
2024 free spaces of all OSTs.
2026 Severity : enhancement
2028 Description: do not expand extent locks acquired on OST-side
2029 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2030 by server: they are not cached anyway.
2033 Frequency : when mmap is used/binaries executed from Lustre
2035 Description: Unmmap pages before throwing them away from read cache.
2036 Details : llap_shrink cache now attempts to unmap pages before discarding
2037 them (if unmapping failed - do not discard). SLES9 kernel has
2038 extra checks that trigger if this unmapping is not done first.
2043 Description: lconf didn't resolve symlinks before checking to see whether a
2044 given mountpoint was already in use
2047 Frequency : when migrating failover services
2048 Bugzilla : 6395, 9514
2049 Description: When migrating a subset of services from a node (e.g. failback
2050 from a failover service node) the remaining services would
2051 time out and evict clients.
2052 Details : lconf --force (implied by --failover) sets the global obd_timeout
2053 to 5 seconds in order to quickly disconnect, but this caused
2054 other RPCs to time out too quickly. Do not change the global
2055 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2057 Severity : enhancement
2058 Frequency : if MDS is started with down OST
2059 Bugzilla : 9439,5706
2060 Description: Allow startup/shutdown of an MDS without depending on the
2061 availability of the OSTs.
2062 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2063 Add appropriate locking and lov-osc refcounts for safe
2064 cleaning. Add osc abort_inflight calls in case the
2065 synchronize never started.
2068 Frequency : occasional (Cray XT3 only)
2070 Description: root not authorized to access files in CRAY_PORTALS environment
2071 Details : The client process capabilities were not honoured on the MDS in
2072 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2073 been accessed by an authorized user then root was able to access
2074 the file on the local client also. The root user capabilities
2075 are now allowed on the MDS, as this environment has secure UID.
2078 Frequency : occasional
2080 Description: ldiskfs "too long searching" message happens too often
2081 Details : A debugging message (otherwise harmless) prints too often on
2082 the OST console. This has been reduced to only happen when
2083 there are fragmentation problems on the filesystem.
2088 Description: Division by zero in statfs when all OSCs are inactive
2089 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2090 lov_statfs divides by value returned by lov_get_stripecnt().
2094 Bugzilla : 9489, 3273
2095 Description: First write from each client to each OST was only 4kB in size,
2096 to initialize client writeback cache, which caused sub-optimal
2097 RPCs and poor layout on disk for the first writen file.
2098 Details : Clients now request an initial cache grant at (re)connect time
2099 and so that they can start streaming writes to the cache right
2100 away and always do full-sized RPCs if there is enough data.
2101 If the OST is rebooted the client also re-establishes its grant
2102 so that client cached writes will be honoured under the grant.
2107 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2108 Details : Now I/O RPCs go to different portal number and (presumably) fast
2109 lock requests (and glimses) and other RPCs get their own service
2110 threads pool that should be able to service those RPCs
2113 Severity : enhancement
2115 Description: Ability to exchange lustre version between client and servers and
2116 issue warnings at client side if client is too old. Also for
2117 liblustre clients there is ability to refuse connection of too old
2119 Details : New 'version' field is added to connect data structure that is
2120 filled with version info. That info is later checked by server and
2124 Frequency : rare, liblustre only.
2125 Bugzilla : 9296, 9581
2126 Description: Two simultaneous writes from liblustre at offset within same page
2127 might proceed at the same time overwriting eachother with stale
2129 Details : I/O lock withing llu_file_prwv was released too early, before data
2130 actually was hitting the wire. Extended lock-holding time until
2131 server acknowledges receiving data.
2134 Frequency : extremely rare. Never observed in practice.
2136 Description: avoid generating lustre_handle cookie of 0.
2137 Details : class_handle_hash() generates handle cookies by incrementing
2138 global counter, and can hit 0 occasionaly (this is unlikely, but
2139 not impossible, because initial value of cookie counter is
2140 selected randonly). Value of 0 is used as a sentinel meaning
2141 "unassigned handle" --- avoid it. Also coalesce two critical
2142 sections in this function into one.
2144 Severity : enhancement
2146 Description: allow liblustre clients to delegate truncate locking to OST
2147 Details : To avoid overhead of locking, liblustre client instructs OST to
2148 take extent lock in ost_punch() on client's behalf. New connection
2149 flag is added to handle backward compatibility.
2151 Severity : enhancement
2152 Bugzilla : 4928, 7341, 9758
2153 Description: allow number of OST service threads to be specified
2154 Details : a module parameter allows the number of OST service threads
2155 to be specified via "options ost ost_num_threads={N}" in the
2156 OSS's /etc/modules.conf or /etc/modprobe.conf.
2160 Bugzilla : 6146, 9635, 9895
2161 Description: servers crash with bad pointer in target_handle_connect()
2162 Details : In rare cases when a client is reconnecting it was possible that
2163 the connection request was the last reference for that export.
2164 We would temporarily drop the export reference and get a new
2165 one, but this may have been the last reference and the export
2166 was just destroyed. Get new reference before dropping old one.
2168 Severity : enhancement
2169 Frequency : if client is started with failover MDS
2171 Description: Allow multiple MDS hostnames in the mount command
2172 Details : Try to read the configuration from all specified MDS
2173 hostnames during a client mount in case the "primary"
2176 Severity : enhancement
2178 Description: Stop sending data to evicted clients as soon as possible.
2179 Details : Check if the client we are about to send or are sending data to
2180 was evicted already. (Check is done every second of waiting,
2181 for which l_wait_event interface was extended to allow checking
2182 of exit condition at specified intervals).
2185 Frequency : rare, normally only when NFS exporting is done from client
2187 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2189 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2192 Frequency : always, if srand() is called before liblustre initialization
2194 Description: Liblustre uses system PRNG disturbing its usage by user application
2195 Details : Introduce internal to lustre fast and high-quality PRNG for
2196 lustre usage and make liblustre and some other places in generic
2197 lustre code to use it.
2199 Severity : enhancement
2200 Bugzilla : 9477, 9557, 9870
2201 Description: Verify that the MDS configuration logs are updated when xml is
2202 Details : Check if the .xml configuration logs are newer than the config
2203 logs stored on the MDS and report an error if this is the case.
2204 Request --write-conf, or allow starting with --old_conf.
2206 Severity : enhancement
2208 Description: Handle symlinks in the path when checking if Lustre is mounted.
2209 Details : Resolve intermediate symlinks when checking if a client has
2210 mounted a filesystem to avoid duplicate client mounts.
2215 Description: lconf can hit an error exception but still return success.
2216 Details : The lconf command catches the Command error exception at the top
2217 level script context and will exit with the associated exit
2218 status, but doesn't ensure that this exit status is non-zero.
2223 Description: failure of ptlrpc thread startup can cause oops
2224 Details : Starting a ptlrpc service thread can fail if there are a large
2225 number of threads or the server memory is very fragmented.
2226 Handle this without oopsing.
2229 Frequency : always, only if liblustre and non-default acceptor port was used
2231 Description: liblustre cannot connect to servers with non-default acceptor port
2232 Details : tcpnal_set_default_params() was not called and was therefore
2233 ignoring the environment varaible TCPNAL_PORT, as well as other
2234 TCPNAL_ environment variables
2239 Description: two objects could be created on the same OST for a single file
2240 Details : If an OST is down, in some cases it was possible to create two
2241 objects on a single OST for a single file. No problems other
2242 than potential performance impact and spurious error messages.
2246 Bugzilla : 5681, 9562
2247 Description: Client may oops in ll_unhash_aliases
2248 Details : Client dcache may become inconsistent in race condition.
2249 In some cases "getcwd" can fail if the current directory is
2255 Description: Inode refcounting problems in NFS export code
2256 Details : link_raw functions used to call d_instantiate without obtaining
2257 extra inode reference first.
2261 Bugzilla : 9942, 9903
2262 Description: Referencing freed requests leading to crash, memleaks with NFS.
2263 Details : We used to require that call to ll_revalidate_it was always
2264 followed by ll_lookup_it. Also with revalidate_special() it is
2265 possible to call ll_revalidate_it() twice for the same dentry
2266 even if first occurence returned success. This fix changes semantic
2267 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2268 reference on a request referred from the intent.
2269 ll_intent_release() then releases such a request.
2272 Frequency : rare, normally benchmark loads only
2274 Description: unlinked inodes were kept in memory on the client
2275 Details : If a client is repeatedly creating and unlinking files it
2276 can accumulate a lot of stale inodes in the inode slab cache.
2277 If there is no other client load running this can cause the
2278 client node to run out of memory. Instead flush old inodes
2279 from client cache that have the same inode number as a new inode.
2282 Frequency : SLES9 2.6.5 kernel and long filenames only
2283 Bugzilla : 9969, 10379
2284 Description: utime reports stale NFS file handle
2285 Details : SLES9 uses out-of-dentry names in some cases, which confused
2286 the lustre dentry revalidation. Change it to always use the
2290 Frequency : rare, unless heavy write-truncate concurrency is continuous
2291 Bugzilla : 4180, 6984, 7171, 9963, 9331
2292 Description: OST becomes very slow and/or deadlocked during object unlink
2293 Details : filter_destroy() was holding onto the parent directory lock
2294 while truncating+unlinking objects. For very large objects this
2295 may block other threads for a long time and slow overall OST
2296 responsiveness. It may also be possible to get a lock ordering
2297 deadlock in this case, or run out of journal credits because of
2298 the combined truncate+unlink. Solution is to do object truncate
2299 first in one transaction without parent lock, and then do the
2300 final unlink in a new transaction with the parent lock. This
2301 reduces the lock hold time dramatically.
2304 Frequency : rare, 2.4 kernels only
2306 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2307 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2308 before unmounting their filesystem in order to avoid deadlock.
2309 The kernel_locked() function in 2.4 kernels only checks whether
2310 the kernel lock is held, not whether it is this process that is
2311 holding it as 2.6 kernels do.
2316 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2317 Details : The client ptlrpc code may be trying to reconnect to a down
2318 server before a previous connection attempt has timed out.
2319 Increase the reconnect interval to be longer than the connection
2320 timeout interval to avoid sending duplicate connections to
2324 Frequency : echo_client brw_test command
2326 Description: fix echo_client to work with OST preallocated code
2327 Details : OST preallocation code (5137) didn't take echo_client IO path
2328 into account: echo_client calls filter methods outside of any
2329 OST thread and, hence, there is no per-thread preallocated
2330 pages and buffers to use. Solution: hijack pga pages for IO. As
2331 a byproduct, this avoids unnecessary data copying.
2335 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2336 Description: Client can oops in mdc_commit_close() after open replay
2337 Details : It was possible for the MDS to return an open request with no
2338 transaction number in mds_finish_transno() if the client was
2339 evicted, but without actually returning an error. Clients
2340 would later try to replay that open and may trip an assertion
2341 Simplify the client close codepath, and always return an error
2342 from the MDS in case the open is not successful.
2345 Frequency : rare, 2.6 OSTs only
2347 Description: OST may deadlock under high load on fragmented files
2348 Details : If there was a heavy load and highly-fragmented OST filesystems
2349 it was possible to have all the OST threads deadlock waiting on
2350 allocation of biovecs, because the biovecs were not released
2351 until the entire RPC IO was completed. Instead, release biovecs
2352 as soon as they are complete to ensure forward IO progress.
2354 Severity : enhancement
2356 Description: Support for specifying external journal device at mount
2357 Details : If an OST or MDS device is formatted with an external journal
2358 device, this device major/minor is stored in the ext3 superblock
2359 and may not be valid for failover. Allow detecting and
2360 specifying the external journal at mount time.
2365 Description: Mounting an MDS with pending unlinked files may cause oops
2366 Details : target_finish_recovery() calls mds_postrecov() which returned
2367 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2368 considers this an error and immediately begins cleaning up the
2369 lov, just after starting the mds_lov process
2371 Severity : enhancement
2373 Description: Implement 'lfs df' to report actual free space on per-OST basis
2374 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2375 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2376 usage of objects; '-h' to report in human readable format.
2378 ------------------------------------------------------------------------------
2380 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2387 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2388 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2389 at startup during orphan unlinking. Instead, skip buddy update
2390 before setup, as it will be regenerated after recovery is complete.
2395 Description: If an OST is inactive, its locks might reference stale inodes.
2396 Details : lov_change_cbdata() must iterate over all namespaces, even if
2397 they are inactive to clear inode references from the lock.
2399 Severity : enhancement
2400 Frequency : occasional, if non-standard max_dirty_mb used
2402 Description: Client will block write RPCs if not enough grant
2403 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2404 then the client will block writes while waiting for another RPC
2405 to complete instead of consuming its dirty limit. With change
2406 we get improved performance when max_dirty_mb is small.
2408 Severity : enhancement
2409 Bugzilla : 3389, 6253
2410 Description: Add support for supplementary groups on the MDS.
2411 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2412 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2413 lookups for user supplementary groups into a cache.
2417 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2418 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2419 for liblustre O_EXCL is enforced by MDS.
2423 Description: readdir never returns NULL in liblustre.
2424 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2425 dentry in struct dirent.
2429 Frequency : liblustre only, depends on application IO pattern
2430 Description: liblustre clients evicted if not contacting servers
2431 Details : Don't put liblustre clients into the ping_evictor list, so
2432 they will not be evicted by the pinger ever.
2434 Severity : enhancement
2436 Description: Add ability to evict clients by NID from MDS.
2437 Details : By echoing "nid:$NID" string into
2438 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2439 $NID would be instantly evicted from this MDS and from all active
2440 OSTs connected to it.
2444 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2445 Details : lookup_it_finish() used to query file size from OSTs that was not
2450 Description: service threads change working directory to that of init
2451 Details : Starting lustre service threads may pin the working directory
2452 of the parent thread, making that filesystem busy. Threads
2453 now change to the working directory of init to avoid this.
2457 Frequency : during shutdown only
2458 Description: shutdown with a failed MDS or OST can cause unmount to hang
2459 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2464 Frequency : 2.6 only
2465 Description: chmod/chown may include an extra supplementary group
2466 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2467 supplementary group and if none is specified this is used.
2470 Bugzilla : 5479 (6816)
2472 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2473 Details : If lookup is in progress on a file that is unlinked we might try
2474 to revalidate the inode and fail in revalidate after lookup is
2475 complete and ll_file_open() enqueues the open again but
2476 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2479 Frequency : always, if lconf --abort_recovery used
2481 Description: lconf --abort_recovery fails with 'Operation not supported'
2482 Details : lconf was attempting to abort recovery on the MDT device and not
2485 ------------------------------------------------------------------------------
2487 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2492 Frequency : rare (only unsupported configurations with a node running as an
2494 Bugzilla : 6514, 5137
2495 Description: Mounting a Lustre file system on a node running as an OST could
2497 Details : OSTs now preallocates memory needed to write out data at
2498 startup, instead of when needed, to avoid having to
2499 allocate memory in possibly low memory situations.
2500 Specifically, if the file system is mounted on on OST,
2501 memory pressure could force it to try to write out data,
2502 which it needed to allocate memory to do. Due to the low
2503 memory, it would be unable to do so and the node would
2504 become unresponsive.
2506 Severity : enhancement
2508 Description: Addition of lconf --service command line option
2509 Details : lconf now accepts a '--service <arg>' option, which is
2510 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2512 Severity : enhancement
2514 Description: Failover mode is now the default for OSTs.
2515 Details : By default, OSTs will now run in failover mode. To return to
2516 the old behaviour, add '--failout' to the lmc line for OSTs.
2518 Severity : enhancement
2520 Description: Health checks are now provided for MDS and OSTs
2521 Details : Additional detailed health check information on MSD and OSTs
2522 is now provided through the procfs health_check value.
2525 Frequency : occasional, depends on IO load
2527 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2528 after numerous create/delete cycles
2529 Details : The ext3 inode allocation policy would not allocate new inodes
2530 very well on the OSTs because there are no new directories
2531 being created. Instead we look for groups with free space if
2532 the parent directories are nearly full.
2537 Description: Network or server problems during mount may cause partially
2538 mounted clients instead of returning an error.
2539 Details : The config llog parsing code may overwrite the error return
2540 code during mount error handling, returning success instead
2546 Description: MDS can fail to allocate large reply buffers
2547 Details : After long uptimes the MDS can fail to allocate large reply
2548 buffers (e.g. zconf client mount config records) due to memory
2549 fragmentation or consumption by the buffer cache. Preallocate
2550 some large reply buffers so that these replies can be sent even
2551 under memory pressure.
2555 Frequency : rare (liblustre)
2556 Description: fsx running with liblustre complained that using truncate() to
2557 extend the file doesn't work. This patch corrects that issue.
2558 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2559 ATTR_SIZE and lsm use in llu_setattr_raw.
2563 Frequency : rare, only 2.6 kernels
2564 Description: Unusual file access patterns on the MDS may result in inode
2565 data being lost in very rare circumstances.
2566 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2567 code in the 2.6 kernel caused the inode and/or EA data not to
2568 be read from disk, causing single-file corruption.
2572 Frequency : rare, only 2.6 filesystems using extents
2573 Description: Heavy concurrent write and delete load may cause data corruption.
2574 Details : It was possible under high-load situations to have an extent
2575 metadata block in the block device cache from a just-unlinked
2576 file overwrite a newly-allocated data block. We now unmap any
2577 metadata buffers that alias just-allocated data blocks.
2581 Frequency : filesystems with default stripe_count larger than 77
2582 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2583 Details : lconf specifies an inode size of 4096 bytes when the default
2584 stripe_count is larger than 77. This conflicts with the default
2585 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2586 this case to avoid pinning too much memory for large EAs.
2588 ------------------------------------------------------------------------------
2590 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2595 Frequency : rare (extremely heavy IO load with hundreds of clients)
2597 Description: Client is evicted, gets IO error writing to file
2598 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2599 caused clients to be evicted for AST timeouts. The fixes in
2600 bug 5192 mean we no longer need to have such short AST timeouts
2601 so ldlm_timeout has been increased.
2604 Frequency : occasional during --force or --failover shutdown under load
2605 Bugzilla : 5949, 4834
2606 Description: Server oops/LBUG if stopped with --force or --failover under load
2607 Details : a collection of import/export refcount and cleanup ordering
2608 issues fixed for safer force cleanup
2611 Frequency : only filesystems larger than 120 OSTs
2612 Bugzilla : 5990, 6223
2613 Description: lfs getstripe would oops on a very large filesystem
2614 Details : lov_getconfig used kfree on vmalloc'd memory
2617 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2618 Bugzilla : 6242, 6243
2619 Description: reading from files that had been truncated to a non-zero size
2620 but never opened returned no data
2621 Details : ll_file_read() reads zeros from no-object files to EOF
2626 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2627 Details : There's little harm in aborting MDS/OSS recovery and letting it
2628 try again, so I removed the LASSERT and return an error instead.
2630 Severity : enhancement
2632 Description: New debugging infrastructure for tracking down data corruption
2633 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2634 (b) cover more of the client-side code path, and (c) try to narrow
2635 down where problems occurred
2639 Bugzilla : 3819, 4364, 4397, 6313
2640 Description: Racing close and eviction MDS could cause assertion in mds_close
2641 Details : It was possible to get multiple mfd references during close and
2642 client eviction, leading to one thread referencing a freed mfd.
2644 Severity: : enhancement
2645 Bugzilla : 3262, 6359
2646 Description: Attempts to reconnect to servers are now more aggressive.
2647 Details : This builds on the enhanced upcall-less recovery that was added
2648 in 1.4.2. When trying to reconnect to servers, clients will
2649 now try each server in the failover group every 10 seconds. By
2650 default, clients would previously try one server every 25 seconds.
2655 Description: After recovery, certain operations trigger a failed
2656 assertion on a client.
2657 Details : Failing over an mds, using lconf -d --failover, while a
2658 client was doing a readdir() call would cause the client to
2659 LBUG after recovery completed and the readdir() was resent.
2661 Severity : enhancement
2663 Description: Default groups are now added by lconf
2664 Details : You can now run lconf --group <servicename> without having to
2665 manually add groups with lmc.
2668 Frequency : occasional
2670 Description: Nodes with an elan id of 0 trigger a failed assertion
2673 Frequency : always when accessing e.g. tty/console device nodes
2675 Description: tty and some other devices nodes cannot be used on lustre
2676 Details : file's private_data field is used by device data and lustre
2677 values in there got lost. New field was added to struct file to
2678 store fs-specific private data.
2681 Frequency : when exporting Lustre via NFS
2683 Description: NFSD failed occasionally when looking up a path component
2684 Details : NFSD is looking up ".." which was broken in ext3 directories
2685 that had grown large enough to become hashed.
2688 Frequency : Clusters with multiple interfaces not on the same subnet
2690 Description: Nodes will repeatedly try to reconnect to an interface which it
2691 cannot reach and report an error to the log.
2692 Details : Extra peer list entries will be created by lconf with some peers
2693 unreachable. lconf now validates the peer before adding it.
2696 Frequency : Only if a default stripe is set on the filesystem root.
2698 Description: Setting a default stripe on the filesystem root prevented the
2699 filesystem from being remounted.
2700 Details : The client was sending extra request flags in the root getattr
2701 request and did not allocate a reply buffer for the dir EA.
2704 Frequency : occasional, higher if lots of files are accessed by one client
2705 Bugzilla : 6159, 6097
2706 Description: Client trips assertion regarding lsm mismatch/magic
2707 Details : While revalidating inodes the VFS looks up inodes with ifind()
2708 and in rare cases can find an inode that is being freed.
2709 The ll_test_inode() code will free the lsm during ifind()
2710 when it finds an existing inode and then the VFS later attaches
2711 this free lsm to a new inode.
2715 Bugzilla : 6422, 7030
2716 Description: MDS deadlock between mkdir and client eviction
2717 Details : Creating a new file via mkdir or mknod (starting a transaction
2718 and getting the ns lock) can deadlock with client eviction
2719 (gets ns lock and trying to finish a synchronous transaction).
2722 Frequency : occasional
2723 Description: While starting a server, the fsfilt_ext3 module could not be
2725 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2726 kernels. Previously, lconf would still use the ext3 name
2727 when trying to load modules. Now, it will correctly use
2728 ext3 on 2.4 and ldiskfs on 2.6.
2730 Severity : enhancement
2731 Description: The default stripe count has been changed to 1
2732 Details : The interpretation of the default stripe count (0, to lfs
2733 or lmc) has been changed to mean striping across a single
2734 OST, rather than all available. For general usage we have
2735 found a stripe count of 1 or 2 works best.
2737 Severity : enhancement
2738 Description: Add support for compiling against Cray portals.
2739 Details : Conditional compiling for some areas that are different
2743 Frequency : occasional
2744 Bugzilla : 6409, 6834
2745 Description: Creating files with an explicit stripe count may lead to
2746 a failed assertion on the MDS
2747 Details : If some OSTs are full or unavailable, creating files may
2748 trigger a failed assertion on the MDS. Now, Lustre will
2749 try to use other servers or return an error to the
2753 Frequency : occasional
2755 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2756 caused lock timeout during readahead (since 1.4.2).
2757 Details : Processes doing readahead might match a lock that hasn't been
2758 granted yet if there are overlapping and conflicting lock
2759 requests. The readahead process waits on ungranted lock
2760 (original lock is CBPENDING), while OST waits for that process
2761 to cancel CBPENDING read lock and eventually evicts client.
2763 Severity : enhancement
2765 Description: Initial enabling of flock support for clients
2766 Details : Implements fcntl advisory locking and file status functions.
2767 This feature is provided as an optional mount flag (default
2768 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2769 locking are implemented yet, and those that are are not guaranteed
2770 to be completely correct in production environments.
2771 mount -t lustre -o [flock|noflock] ...
2774 Frequency : occasional
2776 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2777 assertion in the ext3 JBD (journaling) layer.
2778 Details : The b_committed_data struct is protected by the big kernel lock
2779 in 2.4 kernels, serializing journal_commit_transaction() and
2780 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2781 access to this struct. In 2.6 kernels there is finer grained
2782 locking to improve SMP performance of the JBD layer.
2786 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2788 -----------------------------------------------------------------------------
2790 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2792 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2793 versions of Lustre. Please update all servers and clients to
2794 version 1.4.2 or later at the same time. You must also run
2795 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2796 to update the configuration logs.
2798 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2799 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2800 - swab llog records of type '0' so we get proper header size/idx (5861)
2801 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2802 - fix rename of one directory over another leaking an inode (5953)
2803 - avoid SetPageDirty on 2.6 (5981)
2804 - don't re-add just-being-destroyed locks to the waiting list (5653)
2805 - when creating new directories, inherit the parent's custom
2806 striping settings if present parent (3048)
2807 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2808 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2809 - don't hold client locks on temporary worklist from l_lru (5666)
2810 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2811 - automatically evict dead clients (5921)
2812 - Update file size properly in create+truncate+fstat case (6196)
2813 - Do not unhash mountpoint dentries, do not allow removal of
2815 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2816 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2817 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2818 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2819 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2820 - Correctly handle reads of files with no objects (6243)
2821 - lctl recover will also mark a device active if deactivate used (5933)
2823 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2824 - allow --write-conf on an MDS with different nettype than client (5619)
2825 - don't write config llogs to MDS for mounts not from that MDS (5617)
2826 - lconf should create multiple TCP connections from a client (5201)
2827 - init scripts are now turned off by default; run chkconfig --on
2828 lustre and chkconfig --on lustrefs to use them
2829 - upcalls are no longer needed for clients to recover to failover
2831 - add --abort-recovery option to lconf to abort recovery on device
2833 - add support for an arbitrary number of OSTs (3026)
2834 - Quota support protocol changes.
2835 - forward compatibility changes to wire structs (6007)
2836 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2837 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2838 - improved "lustre status" script
2839 - initialize blocksize for non-regular files (6062)
2840 - added --disable-server and --disable-client configure options (5782)
2841 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2842 - Vanilla 2.4.29 support
2843 - increase maximum number of obd devices to 520 (6242)
2844 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2845 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2846 SLES 9 2.6.5 kernels
2847 - limit stripes per file to 160 (the maximum EA size) (6093)
2849 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2852 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2853 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2854 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2855 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2856 - small ext3 extents cleanups and fixes (5733)
2857 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2858 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2859 - avoid needless client->OST connect, fix handle mismatch (5317)
2860 - fix DLM error path that led to out-of-sync client, long delays (5779)
2861 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2862 - fix several locking issues related to i_size (5492,5624,5654,5672)
2863 - don't move pending lock onto export if it is already evicted (5683)
2864 - fix kernel oops when creating .foo in unlinked directory (5548)
2865 - fix deadlock in obdfilter statistics vs. object create (5811)
2866 - use time_{before,after} to avoid timer jiffies wrap (5882)
2867 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2868 - Do not leak request if server was not able to process it (5154)
2869 - If mds_open unable to find parent dir, make that negative lookup(5154)
2870 - don't create new directories with extent-mapping (5909, 5936)
2872 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2873 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2874 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2875 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2876 - support for RHEL4 kernel on i686 (5773)
2877 - provide error messages when incompatible logs are encountered (5898)
2879 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2880 * version 1.4.0.10 (1.4.1 release candidate 1)
2882 - don't keep a lock reference when lock is not granted (4238)
2883 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2884 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2885 - drop import inflight refcount on signal_completed_replay error (5255)
2886 - unlock page after async write error during send (3677)
2887 - handle missing objects in filter_preprw_read properly (5265)
2888 - no transno return for symlink open, don't save no-trasno open (3440)
2889 - don't try to complete elan receive that already failed (4012)
2890 - free RPC server reply state on error (5406)
2891 - clean up thread from ptlrpc_start_thread() on error (5160)
2892 - readahead could read extra page into cache that wasn't ejected (5388)
2893 - prevent races in class_attach/setup/cleanup/detach (5260)
2894 - don't dereference de->d_inode after l_dput of de (5458)
2895 - use "int" for stripe value returned from lock_to_stripe (5544)
2896 - mballoc allocation and error-checking fixes in 2.6 (5504)
2897 - block device patches to fix I/O request sizes in 2.6 (5482)
2898 - look up hostnames for IB nals (5602)
2899 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2900 - don't start multiple acceptors for the same port (5277)
2901 - fix incorrect LASSERT in mds_getattr_name (5635)
2902 - export a proc file for general "ping" checking (5628)
2903 - fix "lfs check" to not block when the MDS is down (5628)
2905 - service request history (4965)
2906 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2907 - create an "evict_client" /proc entry on OSTs, like the MDS has
2908 - fix mount usage message, return errors per mount(8) (5168)
2909 - change grep [] to grep "[]" in tests so they work in more UMLs
2910 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2911 - remove ancient LOV_MAGIC_V0 EA support (5047)
2912 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2913 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2914 - allow readahead allocations to fail when low on memory (5383)
2915 - mmap locking landed again, after considerable improvement (2828)
2916 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2918 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2921 - send OST transaction number in read/write reply to free req (4966)
2922 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2923 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2924 - the watchdog thread now runs as interruptible (5246)
2925 - flock/lockf fixes (but it's still disabled, pending 5135)
2926 - don't use EXT3 constants in llite code (5094)
2927 - memory shortage at startup could cause assertion (5176)
2929 - reorganization of lov code
2930 - single portals codebase
2932 - add extents/mballoc support (5025)
2933 - direct I/O reads in the obdfilter (4048)
2934 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2936 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2939 - send OST transaction number in read/write reply to free req (4966)
2940 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2941 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2942 - the watchdog thread now runs as interruptible (5246)
2943 - handle missing objects in filter_preprw_read properly (5265)
2944 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2945 - drop import inflight refcount on signal_completed_replay error (5255)
2946 - unlock page after async write error during send (3677)
2947 - return original error code on reconstructed replies (3761)
2948 - no transno return for symlink open, don't save no-trasno open (3440)
2950 - add pid to ldlm debugging output (4922)
2951 - bump the watchdog timeouts -- we can't handle 30sec yet
2952 - extra debugging for orphan dentry/inode bug (5259)
2954 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
2957 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
2958 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
2959 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
2960 - fix NULL dereference in /proc/sys/portals/routes (4827)
2961 - allow failed mdc_close() operations to be interrupted (4561)
2962 - stop precreate on OST before MDS would time out on it (4778)
2963 - don't send partial-page writes before EOF from client (4410)
2964 - discard client grant for sub-page writes on large-page clients (4520)
2965 - don't free dentries not owned by NFS code, check generation (4806)
2966 - fix lsm leak if mds_create_objects() fails (4801)
2967 - limit debug_daemon file size, always print CERROR messages (4789)
2968 - use transno after validating reply (3892)
2969 - process timed out requests if import state changes (3754)
2970 - update mtime on OST during writes, return in glimpse (4829)
2971 - add mkfsoptions to LDAP (4679)
2972 - use ->max_readahead method instead of zapping global ra (5039)
2973 - don't interrupt __l_wait_event() during strace
2975 - add software watchdogs to catch hung threads quickly (4941)
2976 - make lustrefs init script start after nfs is mounted
2977 - fix CWARN/ERROR duplication (4930)
2978 - return async write errors to application if possible (2248)
2979 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
2980 - print NAL number in %x format (4645)
2981 - update barely-supported suse-2.4.21-171 series (4842)
2982 - support for sles 9 %post scripts
2983 - support for building 2.6 kernel-source packages
2984 - support for sles km_* packages
2986 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
2989 - ignore -ENOENT errors in osc_destroy (3639)
2990 - notify osc create thread that OSC is being cleaned up (4600)
2991 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
2992 - reconstruct ost_handle() like mds_handle() (4657)
2993 - create a new thread to do import eviction to avoid deadlock (3969)
2994 - let lconf resolve symlinked-to devices (4629)
2995 - don't unlink "objects" from directory with default EA (4554)
2996 - hold socknal file ref over connect in case target is down (4394)
2997 - allow more than 32000 subdirectories in a single directory (3244)
2998 - fix blocks count for O_DIRECT writes (3751)
2999 - OST returns ENOSPC from object create when no space left (4539)
3000 - don't send truncate RPC if file size isn't changing (4410)
3001 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3002 - bind to privileged port in socknal and tcpnal (3689)
3004 - rate limit CERROR/CWARN console message to avoid overload (4519)
3005 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3006 - basic mmap support (3918)
3007 - kernel patch series update from b1_4 (4711)
3009 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3012 - avoid crash during MDS cleanup with OST shut down (2775)
3013 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3014 - don't use bad inodes on the MDS (3744)
3015 - dynamic object preallocation to improve recovery speed (4236)
3016 - don't hold spinlock over lock dumping or change debug flags (4401)
3017 - don't zero obd_dev when it is force cleaned (3651)
3018 - print grants to console if they go negative (4431)
3019 - "lctl deactivate" will stop automatic recovery attempts (3406)
3020 - look for existing locks in ldlm_handle_enqueue() (3764)
3021 - don't resolve lock handle twice in recovery avoiding race (4401)
3022 - revalidate should check working dir is a directory (4134)
3024 - don't always mark "slow" obdfilter messages as errors (4418)
3026 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3029 - don't close LustreDB during write_conf until it is done (3860)
3030 - fix typo in lconf for_each_profile (3821)
3031 - allow dumping logs from multiple threads at one time (3820)
3032 - don't allow multiple threads in OSC recovery (3812)
3033 - fix debug_size parameters (3864)
3034 - fix mds_postrecov to initialize import for llog ctxt (3121)
3035 - replace config semaphore with spinlock (3306)
3036 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3037 - don't allow enqueue to complete on a destroyed export (3822)
3038 - down write_lock before checking llog header bitmap (3825)
3039 - recover from lock replay timeout (3764)
3040 - up llog sem before sending rpc (3652)
3041 - reduce ns lock hold times when setting kms (3267)
3042 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3043 - fix NULL deref and obd_dev leak on setup error (3312)
3044 - replace some LBUG about llog ops with error handling (3841)
3045 - don't match INVALID dentries from d_lookup and spin (3784)
3046 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3047 - fix invalid assertion in ptlrpc_set_wait (3880)
3049 - add libwrap support for the TCP acceptor (3996)
3050 - add /proc/sys/portals/routes for non-root route listing (3994)
3051 - allow setting MDS UUID in .xml (2580)
3052 - print the stack of a process that LBUGs (4228)
3054 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3057 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3058 - make sure to unset replay flag from failed open requests (3440)
3059 - if default stripe count is 0, use OST count for inode size (3636)
3060 - update parent mtime/ctime on client for create/unlink (2611)
3061 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3062 - free recovery state on server during a forced cleanup (3571)
3063 - unregister_reply for resent reqs (3063)
3064 - loop back devices mounting and status check on 2.6 (3563)
3065 - fix resource-creation race that can provoke i_size == 0 (3513)
3066 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3067 - more debugging for page-accounting assertion (3746)
3068 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3069 - avoid deadlock after precreation failure (3758)
3070 - fix race and lock order deadlock in orphan handling (3450, 3750)
3071 - add validity checks when grabbing inodes from l_ast_data (3599)
3073 - add /proc/.../recovery_status to obdfilter (3428)
3074 - lightweight CDEBUG infrastructure, debug daemon (3668)
3075 - change default OSC RPC parameters to be better on small clusters
3076 - turn off OST read cache for files smaller than 32MB
3077 - install man pages and include them in rpms (3100)
3078 - add new init script for (un)mounting lustre filesystems (2593)
3079 - run chkconfig in %post for init scripts (3701)
3080 - drop scimac NAL (unmaintained)
3082 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3085 - clean kiobufs before and after use (3485)
3086 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3087 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3088 - change init script to fail more clearly if not run as root (1528)
3089 - allow clients to reconnect during replay (1742)
3090 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3091 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3092 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3093 - fix cond_resched (3554)
3094 - only evict unfinished clients after recovery (3515)
3095 - allow bulk resend, prevent data loss (3570)
3096 - dynamic ptlrpc request buffer allocation (2102)
3097 - don't allow unlinking open directory if it isn't empty (2904)
3098 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3099 - remove extraneous obd dereference causing LASSERT failure (3334)
3100 - don't use get_cycles() when creating temp. files on the mds (3156)
3101 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3102 - handle EEXIST for set-stripe, set proper directory name (3336)
3104 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3105 - fix ksocknal_fmb_callback() error messages (2918)
3107 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3110 - don't copy lvb into (possibly NULL) reply on error (2983)
3111 - don't deref dentry after dput, don't free lvb on error (2922)
3112 - use the kms to determine writeback rpc length (2947)
3113 - increment oti_logcookies when osc is inactive (2948)
3114 - update client's i_blocks count via lvb messages (2543)
3115 - handle intent open/close of special files properly (1557)
3116 - mount MDS with errors=remount-ro, like obdfilter (2009)
3117 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3118 - don't use cancelling-locks' kms values (2947)
3119 - use highest lock extent for kms, not last one (2925)
3120 - don't dereference ERR_PTR() dentry in error handling path (3107)
3121 - fix thread race in portals_debug_dumplog() (3122)
3122 - create lprocfs device entries at setup instead of at attach (1519)
3123 - common AST error handler, don't evict client on completion race (3145)
3124 - zero nameidata in detach_mnt in 2.6 (3118)
3125 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3126 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3127 - initialize RPC timeout timer earlier for 2.6 (3219)
3128 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3129 - print nal/nid for unknown nid (3258)
3130 - additional checks for oscc recovery before doing precreate (3284)
3131 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3132 - don't crash in mdc_close for bad permissions on open (3285)
3133 - zero i_rdev for non-device files (3147)
3134 - clear page->private before handing to FS, better assertion (3119)
3135 - tune the read pipeline (3236)
3136 - fix incorrect decref of invalidated dentry (2350)
3137 - provide read-ahead stats and refine rpc in flight stats (3328)
3138 - don't hold journal transaction open across create RPC (3313)
3139 - update atime on MDS at close time (3265)
3140 - close LDAP connection when recovering to avoid server load (3315)
3141 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3142 - don't leak open file on MDS after open resend (3325)
3143 - serialize filter_precreate and filter_destroy_precreated (3329)
3144 - loop device shouldn't call sync_dev() for nul device (3092)
3145 - clear page cache after eviction (2766)
3146 - resynchronize MDS->OST in background (2824)
3147 - refuse to mount the same filesystem twice on same mountpoint (3394)
3148 - allow llmount to create routes for mounting behind routers (3320)
3149 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3150 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3151 - fix rare problem with rename on htree directories (3417)
3153 - allow default OST striping configuration per directory (1414)
3154 - fix compilation for qswnal for 2.6 kernels (3125)
3155 - increase maximum number of MDS request buffers for large systems
3156 - change liblustreapi to be useful for external progs like lfsck (3098)
3157 - increase local configuration timeout for slow disks (3353)
3158 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3160 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3163 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3164 - don't overwrite extent policy data in reply if lock was blocked (2901)
3165 - drop filter export grants atomically with removal from device (2663)
3166 - del obd_self_export from work_list in class_disconnect_exports (2908)
3167 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3168 - swab reply message in mdc_close, other PPC fixes (2464)
3169 - fix destroying of named logs (2325)
3170 - overwrite old logs when running lconf --write_conf (2264)
3171 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3172 - fix race in target_handle_connect (2898)
3173 - mds_reint_create() should take same inode create lock (2926)
3174 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3175 - don't close files for self_export to avoid uninitialized obd (2936)
3176 - allow MDS with the same name as client node (2939)
3177 - hold dentry reference for closed log files for unlink (2325)
3178 - reserve space for all logs during transactions (2059)
3179 - don't evict page beyond end of stripe extent (2925)
3180 - don't oops on a deleted current working directory (2399)
3181 - handle hard links to targets without a parent properly (2517)
3182 - don't dereference NULL lock when racing during eviction (2867)
3183 - don't grow lock extents when lots of conflicting locks (2919)
3185 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3188 - account for cache space usage on clients to avoid data loss (974)
3189 - lfsck support in lustre kernel code (2349)
3190 - reduce journal credits needed for BRW writes (2370)
3191 - orphan handling to avoid losing space on client/server crashes
3192 - ptlrpcd can be blocked, stopping ALL progress (2477)
3193 - use lock value blocks to assist in proper KMS, faster stat (1021)
3194 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3195 - recovery for initial connections (2355)
3196 - fixes for mds_cleanup_orphans (1934)
3197 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3198 - block all file creations until orphan recovery completes (1901)
3199 - client remove rq_connection from request struct (2423)
3200 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3201 - recovery timer race (2670)
3202 - mdc_close recovey bug (2532)
3203 - ptlrpc cleanup bug (2710)
3204 - mds timeout on local locks (2588)
3205 - namespace lock held during RPCs (2431)
3206 - handle interrupted sync write properly (2503)
3207 - don't try to handle a message that hasn't been replied to (2699)
3208 - client assert failure during cleanup after abort recovery (2701)
3209 - leak mdc device after failed mount (2712)
3210 - ptlrpc_check_set allows timedout requests to complete (2714)
3211 - wait for inflight reqs when ptlrpcd finishes (2710)
3212 - make sure unregistered services are removed from the srv_list
3213 - reset bulk XID's when resending them (caught by 1138 test)
3214 - unregister_bulk after timeout
3215 - fix lconf error (2694)
3216 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3217 - readahead locks pages, leaves pending causing memory pressure (2673)
3218 - increase OST request buffers to 4096 on large machines (2729)
3219 - fix up permission of existing directories in simple_mkdir (2661)
3220 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3221 - don't assign transno to errored transactions (2742)
3222 - don't delete objects on OST if given a bogus objid from MDS (2751)
3223 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3224 - if rq_no_resend, then timeout request after recovery (2432)
3225 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3226 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3227 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3228 - MDS recovery shouldn't delete objects that it has given out (2730)
3229 - if enqueue arrives after completion, don't clobber LVB (2819)
3230 - don't unlock pages twice when trigger_group_io returns error (2814)
3231 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3232 - don't write pages to disk if there was an error (1450)
3233 - don't ping imports that have recovery disabled (2676)
3234 - take buffered bytes into account when balancing socknal conn (2817)
3235 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3236 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3237 - remove little-endian swabbing of llog records (1987)
3238 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3239 - retry reposting request buffers if they fail (1191)
3240 - grow extent at grant time to avoid granting a revoked lock (2809)
3241 - lock revoke doesn't evict page if covered by a second lock (2765)
3242 - disable VM readahead to avoid reading outside lock extents (2805)
3244 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3245 - updated kernel patches for hp-2.4.20 kernel (2681)
3247 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3250 - fix truncated write corruption (2366)
3251 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3253 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3254 - ensure len(uuid) < 37 in lmc (1171)
3255 - fix ia64 OOPS in llog_test (2255)
3256 - zero end of page at obdfilter for partial page writes (2648)
3257 - don't leave stale dentries around after renames (bug 2428)
3258 - fix timeouts when evicting a client with a single lock held (2642)
3259 - set deadline for the initial HELLO message to drain (2634)
3260 - print out dotted-quad IP addresses in the socknal (2302)
3262 - additional debugging for MDS client eviction problem (2443)
3263 - fix mkfsoptions support for osts (2603, 2604)
3265 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3268 - add series for the vanilla 2.6.0 kernel
3269 - add series for the vanilla 2.4.24 kernel
3270 - add series for a cray x86/64 UL kernel drop
3271 - fix xattr patches for the vanilla 2.4.19 series
3273 - generate true UUIDs in lmc (1171)
3274 - have portals stack dumping break in UML (2466)
3275 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3276 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3277 - wake callback waiting threads on client eviction (2460)
3278 - Add --ptldebug and --subsystem to lmc (1719)
3279 - update assertion to allow safe interrupt allocation
3280 - set rq_no_resend for cancel requests (2432)
3281 - recalculate ptlrpcd timeout after resend (2494)
3282 - call vfs_rmdir when removing pending directories (2368)
3283 - fix renaming a file to itself (2429)
3284 - lmc creates a default one-stripe lov (2454)
3285 - expand procfs space to handle large clusters (2326)
3286 - increase UML stack to avoid overflow
3287 - update lconf's list of debug and subsystem masks
3288 - fix lfs find --obd (2510)
3289 - /proc tunable for disabling filter read caching (2591)
3290 - stop rpm packages from altering slapd.conf (2301)
3291 - disable nagle in the socknal under 0conf (2578)
3292 - choose mds inode size based on stripe count (2572)
3293 - fix kernel-source rpm problems (2516)
3295 - add --disable-doc to avoid pdf generation (2421)
3296 - update documentation, tests, type-os, comments
3297 - avoid format warnings on ia64
3298 - remove the TOE NAL
3299 - tiny code cleanups by removing unused fields
3301 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3304 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3305 - remove the most blatant lies from BUILDING file (2371)
3306 - change default debug level to reasonable production setting
3307 - reduce client side cache size to reduce cache flush time
3308 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3309 - make TCP zerocopy and pinger support enabled by default (2476)
3310 - sync writes completed after process exits caused crashes (2319)
3311 - maintain correct mount count on the MDS (2356)
3312 - backout 1557, because 2316 wasn't really fixed
3313 - better file I/O statistics gathering in /proc
3314 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3315 - another kernel patch to fix zero-copy TCP function export
3316 - don't take duplicate lock when processing re-sent getattr (2420)
3317 - lctl uses obd_self_export instead of creating new conn (2353)
3318 - MDS/OST recovery case which requires object creation asserted (2425)
3319 - move lfs from /usr/sbin to /usr/bin in packages
3320 - fix race between mds_client_add and mds_client_free (2417)
3321 - use kmalloc instead of slabs in portals (2430)
3322 - don't create duplicate records when a failover MDS is present (2442)
3323 - remove unnecessary mount age check (2332)
3324 - don't remove directory inodes from locks prematurely (2451)
3325 - don't break if MDS service name is the same as hostname (2103)
3326 - fix races in client write RPC generation when cache full (2482)
3328 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3331 - remove now-unused request->rq_obd (278)
3332 - if an allocation fails, print out how much memory we've used (1933)
3333 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3334 - add the "configurable stack size" patch to most series files (1256)
3335 - ability to write large log records, for 100+ OST configs (2306)
3336 - fix NULL deref when filter_prep fails (2314)
3337 - fix operator precedence error in filter_sync
3338 - dynamic allocation of socknal TX descriptors (2315)
3339 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3340 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3341 - ensure that CWARN messages in Portals always get to the syslog
3342 - __init/__exit are not for prototype decls (ldlm_init/exit)
3343 - x86-64 compile warning fixes
3344 - fix gateway LMC keyword conflict (2318)
3345 - fix MDS lock inversions in getattr/reint paths (1844)
3346 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3347 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3348 - DLM race condition prevented some lock evictions (2328)
3349 - ENOMEM detection and retry on socknal sends (2230)
3350 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3351 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3352 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3353 - better error messages when a client is rejected during recovery (1505)
3354 - avoid cancelling locks which were never granted, after failure (2330)
3355 - fix i_sem/journal inversion in mds_client_add (2333)
3356 - fix truncate/getattr lock cycle deadlock (2334)
3357 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3358 - fix two rare exit paths which could leak an l_lock() ref (2321)
3359 - fencepost error in MDS/OST orphan recovery (2226)
3360 - make log record alignment 8 bytes (1988)
3361 - lstripe now fails when requested offset > ost_count (2237)
3362 - ensure that all kernel series have a complete list.h (1607)
3363 - fix crashes in special-file operations (2316)
3364 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3366 - allow configurable automake binary, for testing new versions
3367 - small update to the lfs documentation
3369 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3371 * fix negative export reference count in fsfilt_sync (2312)
3373 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3374 * release candidate 0.9.1
3376 - orphans are moved into the PENDING directory for possible recovery
3377 - replayed opens now open by fid for orphan/rename safety (1042)
3378 - last close of an orphan inode generates a transno (683)
3379 - chdir() and mount() now pin the directory entry (1020)
3380 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3381 - discard very old requests without processing them (1502)
3382 - remove some common, well-understood CERRORs (1505)
3383 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3384 - clear "grant" flags in OST replies until OST grant code lands (1644)
3385 - fix read performance by not clobbering i_blksize on client (1598)
3386 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3387 - make LRU size a /proc tunable, clears locks when reduced (707)
3388 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3389 - server locks take references on exports now (1558)
3390 - build fixes for 2.4.20-rh trees (1663)
3391 - return an error from lov_create if all OSCs are inactive (1751)
3392 - fix import levels when a reconnect happens without a timeout (1597)
3393 - exit early from mds_open if we get a lookup error (1749)
3394 - partial page read at EOF wouldn't wait for disk before sending (1642)
3395 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3396 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3397 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3398 - write/truncate lock inversion (1639)
3399 - Don't auto-load obdclass, portals modules during cleanup (1495)
3400 - fix timestamps from jumping to "now" (1763)
3401 - extra journal assertions (1648)
3402 - add an extra multiunlink test (1771)
3403 - fix read_record/write_record API (1776)
3404 - fix leak of offset_extent, possible incorrect i_size later (1772)
3405 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3406 - Debugging for the kqswnal_get_idle_tx problems (1820)
3407 - Allow recovery to be attempted multiple times (1536)
3408 - Write out MDS last_rcvd file after it is first created (1600)
3409 - Fix tx_descriptor leak in failed transmit situations (1827)
3410 - ext3 journaling fixes for assertion failure after IO error (1871)
3411 - class_export_put() on freed export after completion AST error (1896)
3412 - Fix revalidate looping in VFS (1322)
3413 - Don't access a freed export during MDS_REINT timeout (1521)
3414 - Add open-unlink recovery support on the MDS (1673,1764)
3415 - Return an error if no MDS data was read from last_rcvd (1946)
3416 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3417 - Don't setup a disk device that doesn't match exported UUID (317)
3418 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3419 - avoid committing NULL handle in force close
3420 - local.sh is now a one-stripe LOV configuration
3421 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3422 - fix ext3 htree duplicate directory entry corruption (1516)
3423 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3424 - update to new LOV EA format (2097)
3425 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3427 2003-06-15 Phil Schwan <phil@clusterfs.com>
3430 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3431 - per-import/export recovery handling (958, 931, 959)
3432 - multiple last-rcvd slots, for serving multiple FSes (949)
3433 - connections are again shared between multiple imp/exports (963, 964)
3434 - "umount -f" would hang if any requests needed to be sent (393, 978)
3435 - avoid pinning large req buffer by copying for queued messages (989)
3436 - add "uuid" to "lctl device" command to help upcalls (991)
3437 - "open" RPCs with transnos would confuse recovery counters (1037)
3438 - do proper endian conversion of all wire messages (288, 340, 891)
3439 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3440 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3441 - fix ptlrpc_connection leak in target_handle_connect (1174)
3442 - fix import refcounting bug in OST and MDS cleanup (1134)
3443 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3444 - fix very unlikely obd_types race condition (501)
3445 - remove osc_open hack for echo_client (1187)
3446 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3447 - a failure in read_inode2 leads to deadlock (1139)
3448 - cancel ack-locks as soon as transaction is committed (1072)
3449 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3450 - make sure to commitrw after any preprw to avoid deadlock (1162)
3451 - failing to execute a file in a lustre FS would lock inode (1203)
3452 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3453 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3454 - obd_uuid2tgt would walk off the end of the list (1255)
3455 - on IA64 the osc would give portals incorrect bulk size (1258)
3456 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3457 - fix lock inversion caused by new llite matching code (1282)
3458 - limit the number of dirty pages on a client to 10MB (1286)
3459 - timed out locks were not being corrected cancelled (1289)
3460 - fix O_DIRECT above 4GB on IA-32 (1292)
3461 * major user-visible changes
3462 - fail out/fail over policy now controlled by the upcall (993)
3464 - add OBD_PING to check server availability and failure (954)
3465 - lustre messages are now sent in sending host order (288, 340, 891)
3466 - add eadatalen to MDS getattr reply (340)
3467 - OST read replies may contain second buffer, with per-page status (593)
3469 2003-03-11 Phil Schwan <phil@clusterfs.com>
3472 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3473 - failed open()s could cause deadlock; fixed (867, 869)
3474 - stop cancelling OST locks when files are closed (481)
3475 - overlapping XID spaces caused network corruption (851, 853)
3476 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3477 - setattr_raw added, to do single-RPC, server-side setattrs
3478 - lmc/lconf syntax change for OST UUIDs
3479 - fix crashy race condition between ptlrpc_free_req and osc_close
3480 - don't use request in mdc_enqueue if we hit a timeout (889)
3481 - don't set the inode i_size for regular files from the MDS (896)
3482 - handle out of order completion AST (842)
3483 - don't LBUG if a lock request times out after receiving AST (913)
3484 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3485 - if a bad lock AST arrives, send an error instead of dropping entirely
3486 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3487 - fix leak in bulk IO when only partially completed (899, 900, 926)
3488 - fix O_DIRECT for ia64 (55)
3489 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3490 - C-z after timeout could hang a process forever; fixed (977)
3492 - client-side I/O cache (678, 924, 929, 941, 970)
3494 - READPAGE and SETATTRs which don't take server-side locks get
3497 2003-02-11 Phil Schwan <phil@clusterfs.com>
3500 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3501 - Fix fsfilt_extN_readpage() to read a full page of directory
3502 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3503 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3504 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3505 - Most intent operations were removed, in favour of a new RPC mode
3506 that does a single RPC to the server and bypasses most of the VFS
3507 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3508 - Aggressively cancel local locks on DLM servers
3509 - mds_reint_unlink sends EA to the client if it's the last nlink.
3510 client uses that EA to unlink OST objects.
3511 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3512 - recursive symlinks were fixed (439)
3513 - fixed NULL deref in DEBUG_REQ
3514 - filter_update_lastobjid no longer calls sync, which annoyed extN
3515 - fixed multi-client small-writes to a single file problem (445)
3516 - fixed mtime updates during file writes (607)
3517 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3518 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3519 - fixed Linux OST size reporting problem (444, 656)
3520 - OST now updates object mtime with writes or setattr (607, 619)
3521 - client verifies file size before zeroing page past EOF (445)
3522 - OST now writes last allocated objid to disk with allocation (108)
3523 - LOV on echo now works (409)
3525 - mds_reint_unlink sends a new buffer, with the EA included. this
3526 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3527 set if a regular file was being unlinked, and it was the last link
3528 - use PtlGet from the target for bulk writes (315)
3529 - OST now updates object mtime with writes or setattr (607, 619)
3530 - LDLM now has a grant-time callback to revalidate locked items, if
3532 - Many MDS operations were reorganized to combat race conditions
3534 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3535 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3537 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3540 - Fully reactivate OST imports after reconnection (512, others)
3541 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3542 - More graceful error handling for truncating on dead OST (515)
3543 - Don't error out unless we're actually accessing dead stripes (474)
3544 - Fix garbage sizes when stripes are missing (410)
3545 - LRU counters were broken, causing constant lock purge (433, 432)
3546 - garbage on read from stripes with failed OSTs (441)
3547 - mark OSCs as active before reconnecting during recovery (438)
3548 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3549 - lfind did not preserve OST order in output (443)
3550 - symlinks cause hung clients, incorrect data (439)
3551 - stop dereferencing request after dropping refcount (457)
3552 - don't LASSERT(spin_is_locked) on non-SMP (455)
3553 - fixes for many rename() bugs
3554 - fstat didn't correctly synchronize attributes (399)
3555 - server must handle lock cancellation during blocking AST prep (487)
3556 - bulk descriptors were free()d too soon (511)
3557 - fix paths in lconf, which would load incorrect modules (451, 507)
3558 - fix confusing lconf 'host not found' error message (386)
3559 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3560 - fix race condition in mdc_blocking_ast() for inode access (526)
3561 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3562 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3563 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3564 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3565 - fix lstripe in several ways (kernel side) (527)
3566 - fix request leak in ldlm_cli_enqueue (262)
3567 - incorrect OSC was marked inactive after OST failure
3568 - call mds_fs_cleanup before unmounting filesystem (524)
3569 - fix races between taking ns_lock and ldlm_lock_change_resource
3570 - fix races updating LOV export open file list
3571 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3572 - fix recovery NULL deref in ldlm_cli_cancel_unused
3573 - fix some DLM races by using new hash table for lock handles (419)
3574 - permit the client to specify desired inodes, at replay
3575 - duplicate requests when we queue them for replay reintegration
3576 - fix last_rcvd offset calculation
3577 - sync after each recovered transaction, so we always make progress
3578 - never, not always, ERESTART requests without transnos
3579 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3581 - skip replay if the MDS says that the client is already connected
3582 - don't check for a recovery-enabled export to match lctl's UUID
3583 - don't INC_USE_COUNT for phantom exports
3584 - don't crash when cleaning up phantom exports (567)
3585 - don't double-finish or set replay data for errored mdc_open requests
3586 - abort requests when they time out, so we don't get old replies
3587 - send/receive replies for AST messages again
3588 - if the client says that it doesn't have the lock, cancel it on the
3590 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3591 mark it as destroyed, it will all work out when decref is called
3592 - fix module use counts (22, 581)
3594 - ASTs now expect a reply (server cancels lock on error reply)
3596 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3599 - fix many simultaneous client startup (392)
3600 - fix dentry->d_it clobbering
3601 - credentials weren't being shipped for readdir/getattr operations
3602 - remove invalid assertions triggered during some concurrent MD
3604 - proper Lustre versions added (336, 389)
3605 - fix memory leak for create error case (398)
3606 - fix LOV locking bug that would get cli/srv out of sync
3607 - fix echo client over LOV (409)
3608 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3609 - fix double-O_EXCL intent crash (424)
3610 - avoid sending multiple lock CANCELs (352)
3612 - MDS can do multi-client recovery (modulo bugs in new code)
3614 - many updates, edits, cleanups
3616 2002-11-18 Phil Schwan <phil@clusterfs.com>
3619 - fix null d_it dereference (346)
3620 - fix full OST/dbench hang (333)
3621 - fix permission problem with file removal (286)
3622 - fix removal of OSCs from LOV when they fail
3623 - fix NULL deref during bulk timeout (214)
3624 - fix problems related to multiple filesystems on one MDS (241)
3625 - fixed serious subtle metadata locking bugs
3626 - free locks on clients when inodes are removed due to memory
3628 - fix inode pointer in lock data (285)
3629 - partial support for multiple MDS on a single host (241)
3630 - data locks weren't cancelled at clear_inode time (290, 311)
3631 - intent locks could lead to unbounded lock growth (205)
3632 - added a maximum lock count, an LRU list, and a flusher
3633 - fix multiple rename (365)
3634 - properly abstracted the echo client
3635 - OSC locked 1 byte too many; fixed
3636 - rewrote brw callback code:
3637 - fixed recovery bugs related to LOVs (306)
3638 - fixed too-many-pages-in-one-write crash (191)
3639 - fixed (again) crash in sync_io_timeout (214)
3640 - probably fixed callback-related race (385)
3642 - Add capability to MDS protocol
3643 - LDLM cancellations and callbacks on different portals
3645 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3648 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3649 - defer open object destruction to close time (601981, 138)
3650 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3651 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3652 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3653 - don't call obd_setattr in ll_file_release for destroyed objects
3654 * protocol change to lustre_msg: move |version| and add |flags|
3655 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3656 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3657 * added replay of create, unlink, link and rename operations during
3658 MDS failover; recovery should be much more robust now
3659 * remove failed OSCs from LOVs (only lov_create uses this so far)
3660 * the lustre-HOWTO was brought (more) up to date (582544)
3662 2002-10-23 Phil Schwan <phil@clusterfs.com>
3665 - in-use dentries weren't being reused properly (617851)
3666 - prevent multiple LDLM setup (599178)
3667 - fix LOV size calculations for truncate (617853)
3668 - fix client handling of MDS intent errors (POSIX)
3669 - fix permission bug in lovstripe.c test (624321)
3670 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3671 - truncate past end of file could corrupt data
3672 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3673 - a race in recovery could return ETIMEDOUT to apps (623947)
3674 - building outside the source directory was fixed
3675 * the lustre-HOWTO was brought (more) up to date (582544)
3676 * major progress was made on recovery functionality
3678 2002-10-10 Phil Schwan <phil@clusterfs.com>
3681 - recovery deadlock fix
3682 - rm -rf causes LBUG fix (617817)
3683 - file open by multiple tasks fix (618962)
3684 - directory permissions bugs (602707 and 620007)
3685 - journal_stop fixed with locking (611313)
3686 - O_APPEND failures resolved (618273, perhaps 614459)
3687 - lconf PATH fix (619770)
3688 - IA64 build fix (621450)
3689 - RPC buffer sizes scale with amount of memory
3691 2002-10-01 Phil Schwan <phil@clusterfs.com>
3694 - locks would be cancelled without throwing away data pages,
3695 resulting in inconsistent data (605627)
3696 - inode attributes were not always being refreshed (605627, 612449)
3697 - lconf now continues to cleanup after lctl reports an error
3698 - MDS now enforces user permissions (602707)
3699 - lprocfs cleanup fixed, but not yet enabled (614157)
3700 - fixed infinite server hang, should a client not respond to an AST
3701 - avoid going into recovery if user calls readlink() with a buffer
3702 that's too small (613941)
3703 - AST RPCs no longer require replies (614867) -- this may be changed
3704 - don't crash server if client sends an IOV that's too big (611336)
3705 - fixed lock conversion deadlock (611892)
3706 - fixed the following of symlinks (614622)
3707 * recovery: the server can remove locks from a client that dies, other
3708 clients can make progress
3709 * more extN patch fixes
3710 * compile-time configurable ptlrpc buffer allocations
3712 - collaborative read cache document
3713 - Lustre Lite Performance CDR document-in-progress
3715 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3718 - fix typo in patch-2.4.18
3720 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3723 - clear ptlrpc request each time in handle_incoming_request()
3724 - unlink of files now destroys the object on the OST
3726 2002-09-19 Peter Braam <braam@clusterfs.com>
3728 * add hard link support
3729 * change obdfile creation method
3730 * kernel patch changed
3732 2002-09-19 Peter Braam <braam@clusterfs.com>
3735 - stack overflow bug in extN fixed
3737 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3739 * documentation updates
3740 - add man pages for config tools
3741 - update tests/README to describe testing with new config tools
3742 - finish metadata API descriptions
3743 * bug fixes and cleanups
3744 - statfs workaround for 16TB limit
3745 - LOV stripe allocation improved, can stripe on subset of OSTs
3746 - LOV file size/IO offset was wrong for files > 4GB in size
3747 - object EA data was being dropped, caused files to be unreadable
3748 - memory overflow with non-LOV OST caused memory corruption
3749 - fixed regression tests to work with new config tools, obdfilter
3750 - fixed bug when directory size became larger than 1 block
3751 - fixed bug (for single client case) when PWD was deleted
3752 - invalidate local directory pages when doing intent-based ops
3753 - avoid LDLM oops when lock callback contained bad data
3755 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3757 * documentation updates
3758 * bug fixes and cleanups
3759 - configuration tools
3762 - 64-bit compile warnings
3763 - 64-bit internal statfs data
3765 * test_brw on persistent OST devices
3767 * lprocfs (disabled)
3769 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3771 * documentation updates
3772 * bug fixes and cleanups
3773 * configuration tools
3775 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3778 * many small fixes to 0_5_4
3779 * io/network handling
3780 * thinkos in MDS operations
3782 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3785 * crucial basic fixes to 0.5.3
3786 * IOR, Iozone work over Elan
3789 2002-08-07 Phil Schwan <phil@clusterfs.com>
3790 * version 0_5_3, our first alpha
3791 * we use the new Portals iovs
3792 * documentation updates
3793 * bug fixes and cleanups
3794 * small changes in the DLM wire protocol
3796 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3797 * version 0_5_1 with some initial stability,
3798 * locking on MD and file I/O.
3799 * documentation updates
3800 * several bug fixes since 0.5.0
3801 * small changes in wire protocol
3803 2002-07-18 Phil Schwan <phil@clusterfs.com>
3805 * delivered as Lustre Light Alpha
3806 * fixed a crash after handling invalid MDS requests
3807 * fixed directory pages for architectures with non-4k pages sizes
3809 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3810 * release version v0_4_4
3811 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3813 * Fixes a number of interruption problems with OST operations.
3814 * Update documentation for portals header changes
3815 * Move all wire protocol structs/defines to lustre_idl.h
3816 * Fixes symlink length bug.
3817 * Add tcpdump to repository.
3819 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3820 * release version v0_4_3
3821 * Fixes statfs for inodes on extN.
3822 * Fixes bug in runtests which would delete /etc/hosts.
3823 * Use 64-bit object IDs wherever possible (not into VFS though)
3824 Remove ost_get_info, which is unused by lustre, and out of date.
3826 2002-07-03 Peter Braam <braam@clusterfs.com>
3827 * release version v0_4_2 Fixes a lookup error (type not passed)
3828 * move forward to head of Portals
3829 * move forward to latest Lustre kernel
3831 2002-06-25 Peter Braam <braam@clusterfs.com>
3832 * release version v0_4_1. Hopefully stable on single node use.