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: 1.40.2-cfs1
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
101 Description: df doesn't work properly if diskfs blocksize != 4K
102 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
106 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
107 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
111 Description: obd_health_check_timeout too short
112 Details : set obd_health_check_timeout as 1.5x of obd_timeout
116 Description: llapi_file_create() does not allow some changes
117 Details : add llapi_file_open() that allows specifying the mode and
118 open flags, and also returns an open file handle.
122 Description: merge and cleanup kernel patches.
123 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
127 Description: Add journal checksum support.(Kernel part)
128 Details : The journal checksum feature adds two new flags i.e
129 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
130 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
131 indicates that the commit block contains the checksum for
132 the blocks described by the descriptor blocks. Now commit
133 record can be sent to disk without waiting for descriptor
134 blocks to be written to disk. This behavior is controlled
135 using JBD2_FEATURE_ASYNC_COMMIT flag.
137 --------------------------------------------------------------------------------
139 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
141 * Support for kernels:
142 2.6.5-7.283 (SLES 9),
143 2.6.9-55.EL (RHEL 4),
144 2.6.16.46-0.14 (SLES 10),
145 2.6.18.8 vanilla (kernel.org)
146 * Client support for unpatched kernels:
147 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
148 2.6.16 - 2.6.22 vanilla (kernel.org)
149 * Due to recently discovered recovery problems, we do not recommend
150 using patchless RHEL 4 clients with this or any earlier release.
151 * Recommended e2fsprogs version: 1.39.cfs8
152 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
153 * Starting with this release, the ldiskfs backing filesystem required
154 by Lustre is now in its own package, lustre-ldiskfs. This package
155 should be installed. It is versioned separately from Lustre and
156 may be released separately in future.
160 Description: block reactivating mgc import until all deactivates complete
161 Details : Fix race when failing back MDT/MGS to itself (testing)
163 Severity : enhancement
165 Description: add optional extra BUILD_VERSION info
166 Details : add a new environment variable (namely LUSTRE_VERS) which allows
167 to override the lustre version.
170 Frequency : 2.6.18 servers only
172 Description: ll_kern_mount() doesn't release the module reference
173 Details : The ldiskfs module reference count never drops down to 0
174 because ll_kern_mount() doesn't release the module reference.
179 Description: server LBUG when using old ost_num_threads parameter
180 Details : Accept the old ost_num_threads parameter but warn that it
181 is deprecated, and fix an off-by-one error that caused an LBUG.
186 Description: Transient SCSI error results in persistent IO issue
187 Details : iobuf->dr_error is not reinitialized to 0 between two
191 Frequency : sometimes when underlying device returns I/O errors
193 Description: OSTs not going read-only during write failures
194 Details : OSTs are not remounted read-only when the journal commit threads
195 get I/O errors because fsfilt_ext3 calls journal_start/stop()
196 instead of the ext3 wrappers.
200 Description: poor connect scaling with increasing client count
201 Details : Don't run filter_grant_sanity_check for more than 100 exports
202 to improve scaling for large numbers of clients.
205 Frequency : SLES10 only
207 Description: sanity-quota.sh quotacheck failed: rc = -22
208 Details : Quotas cannot be enabled on SLES10.
211 Frequency : liblustre clients only
213 Description: getdirentries does not give error when run on compute nodes
214 Details : getdirentries does not fail when the size specified as an argument
215 is too small to contain at least one entry
217 Severity : enhancement
219 Description: Add LNET router traceability for debug purposes
220 Details : If a checksum failure occurs with a router as part of the
221 IO path, the NID of the last router that forwarded the bulk data
222 is printed so it can be identified.
227 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
228 Details : Due to a race condition, liblustre clients were occasionally
231 Severity : enhancement
233 Description: lfs setstripe use optional parameters instead of postional
236 Severity : enhancement
238 Description: Nanosecond timestamp support for ldiskfs
239 Details : The on-disk ldiskfs filesystem has added support for nanosecond
240 resolution timestamps. There is not yet support for this at
241 the Lustre filesystem level.
244 Frequency : during server recovery
246 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
247 Details : request with rq_no_resend flag not awake l_wait_event if they get a
251 Frequency : nfs export on patchless client
253 Description: connectathon hang when test nfs export over patchless client
254 Details : Disconnected dentry cannot be found with lookup, so we do not need
255 to unhash it or make it invalid
258 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
259 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
260 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
265 Description: open req refcounting wrong on reconnect
266 Details : If reconnect happened between getting open reply from server and
267 call to mdc_set_replay_data in ll_file_open, we will schedule
268 replay for unreferenced request that we are about to free.
269 Subsequent close will crash in variety of ways.
270 Check that request is still eligible for replay in
271 mdc_set_replay_data().
276 Description: disable writes to filesystem when reading health_check file
277 Details : the default for reading the health_check proc file has changed
278 to NOT do a journal transaction and write to disk, because this
279 can cause reads of the /proc file to hang and block HA state
280 checking on a healthy but otherwise heavily loaded system. It
281 is possible to return to the previous behaviour during configure
282 with --enable-health-write.
284 Severity : enhancement
286 Description: 64-bit inode version
287 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
288 to the inode. This will be required for version-based recovery.
293 Description: MDS fails to start if a duplicate client export is detected
294 Details : in some rare cases it was possible for a client to connect to
295 an MDS multiple times. Upon recovery the MDS would detect this
296 and fail during startup. Handle this more gracefully.
298 Severity : enhancement
300 Description: Add -o localflock option to simulate old noflock
302 Details : This will achieve local-only flock/fcntl locks
308 Description: log_commit_thread vs filter_destroy race leads to crash
309 Details : Take import reference before releasing llog record semaphore
314 Description: Wrong request locking in request set processing
315 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
316 imp_delayed_list, in this place should be used imp_lock.
319 Frequency : when reconnection
321 Description: Grant Leak when osc reconnect to OST
322 Details : When osc reconnect ost, OST(filter) should check whether it
323 should grant more space to client by comparing fed_grant and
324 cl_avail_grant, and return the granted space to client instead
325 of "new granted" space, because client will call osc_init_grant
326 to update the client grant space info.
329 Frequency : when client reconnect to OST
331 Description: Grant Leak when osc do resend and replay bulk write
332 Details : When osc reconnect to OST, OST(filter)should clear grant info of
333 bulk write request, because the grant info while be sync between
334 OSC and OST when reconnect, and we should ignore the grant info
335 these of resend/replay write req.
340 Description: Grant space more than avaiable left space sometimes.
341 Details : When then OST is about to be full, if two bulk writing from
342 different clients came to OST. Accord the avaliable space of the
343 OST, the first req should be permitted, and the second one
344 should be denied by ENOSPC. But if the seconde arrived before
345 the first one is commited. The OST might wrongly permit second
346 writing, which will cause grant space > avaiable space.
349 Frequency : when client is evicted
351 Description: Grant might be wrongly erased when osc is evicted by OST
352 Details : when the import is evicted by server, it will fork another
353 thread ptlrpc_invalidate_import_thread to invalidate the
354 import, where the grant will be set to 0. While the original
355 thread will update the grant it got when connecting. So if
356 the former happened latter, the grant will be wrongly errased
357 because of this race.
362 Description: Checking Stale with correct fid
363 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
364 to check whether it is stale, because sometimes, we want the
365 enqueue happened anyway, and op_data.fid2 will not be initialized.
367 Severity : enhancement
369 Description: update patchless client
370 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
373 Frequency : only with 2.4 kernel
375 Description: random memory corruption
376 Details : size of struct ll_inode_info is to big for union inode.u and this
377 can be cause of random memory corruption.
382 Description: Memory leak in recovery
383 Details : Lov_mds_md was not free in an error handler in mds_create_object.
384 It should also check obd_fail before fsfilt_start, otherwise if
385 fsfilt_start return -EROFS,(failover mds during mds recovery).
386 then the req will return with repmsg->transno = 0 and rc = EROFS.
387 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
388 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
389 be freed no matter whether fsfilt_commit success or not.
392 Frequency : only with huge count clients
394 Description: Prevents from taking the superblock lock in llap_from_page for
396 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
397 for llap_from_page call in ll_removepage prevents from taking the
398 superblock lock for a soon died page.
403 Description: Not check open intent error before release open handle
404 Details : in some rare cases, the open intent error is not checked before
405 release open handle, which may cause
406 ASSERTION(open_req->rq_transno != 0), because it tries to release
407 the failed open handle.
412 Description: Set cat log bitmap only after create log success.
413 Details : in some rare cases, the cat log bitmap is set too early. and it
414 should be set only after create log success.
418 Description: Accessing a block bevice can re-enable I/O when Lustre is
419 tearing down a device.
420 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
425 Description: service threads may hog cpus when there are a lot of requests
427 Details : Insert cond_resched to give other threads a chance to use some of
433 Description: the cat log was not initialized in recovery
434 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
435 unlink log on mds will not be initialized until mds post
436 recovery. And also in mds post recovery, the unlink log will
437 initialization will be done asynchronausly, so there will be race
438 between add unlink log and unlink log initialization.
442 Description: brw_stats were being printed incorrectly
443 Details : brw_stats were being printed as log2 but all of them were not
444 recorded as log2. Also remove some code duplication arising from
445 filter_tally_{read,write}.
449 Frequency : rare, only in recovery.
450 Description: ASSERTION(req->rq_type != LI_POISON) failed
451 Details : imp_lock should be held while iterating over imp_sending_list for
452 prevent destroy request after get timeout in ptlrpc_queue_wait.
456 Description: replay-single.sh test 52 fails
457 Details : A lock's skiplist need to be cleanup when it being unlinked
458 from its resource list.
462 Description: Short directio read returns full requested size rather than
464 Details : Direct I/O operations should return actual amount of bytes
465 transferred rather than requested size.
467 Severity : enhancement
469 Description: metadata RPC reduction (e.g. for rm performance)
470 Details : decrease the amount of synchronous RPC between clients and servers
471 by canceling conflicing lock before the operation on the client side
472 and packing thier handles into the main operation RPC to server.
474 Severity : enhancement
476 Description: Async OSC create to avoid the blocking unnecessarily.
477 Details : If a OST has no remain object, system will block on the creating
478 when need to create a new object on this OST. Now, ways use
479 pre-created objects when available, instead of blocking on an
480 empty osc while others are not empty. If we must block, we block
481 for the shortest possible period of time.
483 --------------------------------------------------------------------------------
485 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
490 Frequency : on some architectures
492 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
493 Details : Uninitialized flags sometimes cause configuration commands to
497 Frequency : patchless clients only
499 Description: missing __iget() symbol export
500 Details : The __iget() symbol export is missing. To avoid the need for
501 this on patchless clients the deathrow inode reaper is turned
502 off, and we depend on the VM to clean up old inodes. This
503 dependency was during via the fix for bug 12181.
505 --------------------------------------------------------------------------------
507 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
509 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
510 INTEROPERATE with older versions automatically. In many cases a
511 special upgrade step is needed. Please read the
512 user documentation before upgrading any part of a live system.
513 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
514 version will not interoperate with 1.6 betas before beta5 (1.5.95).
515 * WARNING: Lustre configuration and startup changes are required with
516 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
522 Severity : enhancement
524 Description: MountConf
525 Details : Lustre configuration is now managed via mkfs and mount
526 commands instead of lmc and lconf. New obd types (MGS, MGC)
527 are added for dynamic configuration management. See
528 https://mail.clusterfs.com/wikis/lustre/MountConf for
531 Severity : enhancement
533 Description: dynamic OST addition
534 Details : OSTs can now be added to a live filesystem
536 Severity : enhancement
538 Description: startup order invariance
539 Details : MDTs and OSTs can be started in any order. Clients only
540 require the MDT to complete startup.
542 Severity : enhancement
544 Description: parallel, asynchronous orphan cleanup
545 Details : orphan cleanup is now performed in separate threads for each
546 OST, allowing parallel non-blocking operation.
548 Severity : enhancement
550 Description: optimized stripe assignment
551 Details : stripe assignments are now made based on ost space available,
552 ost previous usage, and OSS previous usage, in order to try
553 to optimize storage space and networking resources.
555 Severity : enhancement
557 Description: Permanently set tunables
558 Details : All writable /proc/fs/lustre tunables can now be permanently
559 set on a per-server basis, at mkfs time or on a live
562 Severity : enhancement
564 Description: Lustre message v2
565 Details : Add lustre message format v2.
567 Severity : enhancement
569 Description: client OST exclusion list
570 Details : Clients can be started with a list of OSTs that should be
571 declared "inactive" for known non-responsive OSTs.
575 Description: SPEC SFS validation failure on NFS v2 over lustre.
576 Details : Changes the blocksize for regular files to be 2x RPC size,
577 and not depend on stripe size.
579 Severity : enhancement
581 Description: Multiple MD RPCs in flight.
582 Details : Further unserialise some read-only MDS RPCs - learn about intents.
583 To avoid overly-overloading MDS, introduce a limit on number of
584 MDS RPCs in flight for a single client and add /proc controls
585 to adjust this limit.
587 Severity : enhancement
589 Description: client read/write statistics
590 Details : Add client read/write call usage stats for performance
591 analysis of user processes.
592 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
593 file access. extents_stats shows chunk size distribution.
594 extents_stats_per_process show chunk size distribution per
597 Severity : enhancement
599 Description: mds statistics
600 Details : Add detailed mds operations statistics in
601 /proc/fs/lustre/mds/*/stats.
605 Description: Failure of copying files with lustre special EAs.
606 Details : Client side always return success for setxattr call for lustre
607 special xattr (currently only "trusted.lov").
612 Description: Refcount LNET uuids
613 Details : The global LNET uuid list grew linearly with every startup;
614 refcount repeated list entries instead of always adding to
617 Severity : enhancement
619 Description: Dynamic service threads
620 Details : Within a small range, start extra service threads
621 automatically when the request queue builds up.
624 Frequency : mixed-endian client/server environments
626 Description: mixed-endian crashes
627 Details : The new msg_v2 system had some failures in mixed-endian
630 Severity : enhancement
632 Description: Easy OST removal
633 Details : OSTs can be permanently deactivated with e.g. 'lctl
634 conf_param lustre-OST0001.osc.active=0'
636 Severity : enhancement
638 Description: MGS proc entries
639 Details : Added basic proc entries for the MGS showing what filesystems
642 Severity : enhancement
644 Description: provide MGS failover
645 Details : Added config lock reacquisition after MGS server failover.
647 Severity : enhancement
649 Description: add Linux 2.4 support
650 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
654 Description: a large application tries to do I/O to the same resource and dies
656 Details : Check the req->rq_arrival time after the call to
657 ost_brw_lock_get(), but before we do anything about
658 processing it & sending the BULK transfer request. This
659 should help move old stale pending locks off the queue as
660 quickly as obd_timeout.
663 Frequency : when an incorrect nid is specified during startup
665 Description: ptlrpc connect to non-existant node causes kernel crash
666 Details : LNET can't be re-entered from an event callback, which
667 happened when we expire a message after the export has been
668 cleaned up. Instead, hand the zombie cleanup off to another
671 Severity : enhancement
673 Description: plain/inodebits lock performance improvement
674 Details : Grouping plain/inodebits in granted list by their request modes
675 and bits policy, thus improving the performance of search through
679 Frequency : only if OST filesystem is corrupted
681 Description: client incorrectly hits assertion in ptlrpc_replay_req()
682 Details : for a short time RPCs with bulk IO are in the replay list,
683 but replay of bulk IOs is unimplemented. If the OST filesystem
684 is corrupted due to disk cache incoherency and then replay is
685 started it is possible to trip an assertion. Avoid putting
686 committed RPCs into the replay list at all to avoid this issue.
689 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
691 Description: System hang on startup
692 Details : This bug allowed the liblustre (e.g. catamount) client to
693 return to the app before handling all startup RPCs. This
694 could leave the node unresponsive to lustre network traffic
695 and manifested as a server ptllnd timeout.
697 Severity : enhancement
699 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
700 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
701 boolean to control whether to print peer debug info when a
702 client's RPC times out.
705 Frequency : only for kernels with patches from Lustre below 1.4.3
707 Description: Remove old rdonly API
708 Details : Remove old rdonly API which unused from at least lustre 1.4.3
711 Frequency : only for devices with external journals
713 Description: Set external device read-only also
714 Details : During a commanded failover stop, we set the disk device
715 read-only while the server shuts down. We now also set any
716 external journal device read-only at the same time.
719 Frequency : when upgrading from 1.4 while trying to change parameters
721 Description: The wrong (new) MDC name was used when setting parameters for
722 upgraded MDT's. Also allows changing of OSC (and MDC)
723 parameters if --writeconf is specified at tunefs upgrade time.
726 Frequency : when setting specific ost indicies
728 Description: QOS code breaks on skipped indicies
729 Details : Add checks for missing OST indicies in the QOS code, so OSTs
730 created with --index need not be sequential.
732 Severity : enhancement
734 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
735 Details : The uninit_groups feature works in conjunction with the kernel
736 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
737 up the pass1 processing of e2fsck. This is a read-only feature
738 in ldiskfs2 only, so older kernels and current ldiskfs cannot
739 mount filesystems that have had this feature enabled.
741 Severity : enhancement
743 Description: Improve multi-block allocation algorithm to avoid fragmentation
744 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
745 allocation locality and avoid filesystem fragmentation.
747 ------------------------------------------------------------------------------
749 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
751 * Support for kernels:
752 2.4.21-47.0.1.EL (RHEL 3)
754 2.6.9-42.0.10.EL (RHEL 4)
755 2.6.12.6 vanilla (kernel.org)
756 2.6.16.27-0.9 (SLES 10)
757 * Recommended e2fsprogs version: 1.39.cfs5
759 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
763 Frequency : occasional, depends on client load and configuration
764 Bugzilla : 12181, 12203
765 Description: data loss for recently-modified files
767 Details : In some cases it is possible that recently written or created
768 files may not be written to disk in a timely manner (this should
769 normally be within 30s unless client IO load is very high).
770 The problem appears as zero-length files or files that are a
771 multiple of 1MB in size after a client crash or client eviction
772 that are missing data at the end of the file.
774 This problem is more likely to be hit on clients where files are
775 repeatedly created and unlinked in the same directory, clients
776 have a large amount of RAM, have many CPUs, the filesystem has
777 many OSTs, the clients are rebooted frequently, and/or the files
778 are not accessed by other nodes after being written.
780 The presence of the problem can be detected by looking at
781 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
782 smaller than the second (nr_unused) then dirty files will not
783 be flushed automatically to disk. "sync; sleep 10" should be
784 run several times on the node before unmounting it to update
785 Lustre (this is also safe to run on nodes without this problem).
787 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
788 that can cause this same problem, so customers using that kernel
789 also need to update the kernel in addition to Lustre. In order
790 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
792 It is normal that files written just before a client crash (less
793 than 30s) may not yet have been flushed to disk, even for local
797 Frequency : frequent on thin XT3 nodes
799 Description: UUID collision on thin XT3 Linux nodes
800 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
801 randomly, since we relied on an insufficiently-seeded PRNG.
806 Description: OSS hangs after "All ost request buffers busy"
807 Details : A deadlock between quota and journal operations caused OSS
808 hangs after printing "All ost request buffers busy."
811 Frequency : always on liblustre builds
813 Description: Cleanup compiler warnings on liblustre
816 Frequency : always on liblustre builds on XT3
818 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
822 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
823 > 32000 subdirectories
824 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
825 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
826 subdirectory count crosses 32000. This will aid e2fsck to
827 correctly handle more than 32000 subdirectories.
830 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
832 Description: System hang on startup
833 Details : This bug allowed the liblustre (e.g. catamount) client to
834 return to the app before handling all startup RPCs. This
835 could leave the node unresponsive to lustre network traffic
836 and manifested as a server ptllnd timeout.
838 Severity : enhancement
840 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
841 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
842 boolean to control whether to print peer debug info when a
843 client's RPC times out.
848 Description: make O_SYNC working on 2.6 kernels
849 Details : 2.6 kernels use different method for mark pages for write,
850 so need add a code to lustre for O_SYNC work.
855 Description: Failure to close file and release space on NFS
856 Details : Put inode details into lock acquired in ll_intent_file_open.
857 Use mdc_intent_lock in ll_intent_open to properly
858 detect all kind of errors unhandled by mdc_enqueue.
863 Description: proc file read during shutdown sometimes raced obd removal,
865 Details : Add lock to prevent obd access after proc file removal.
868 Frequency : Only for files larger than 4GB on 32-bit clients.
870 Description: improperly doing page alignment of locks
871 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
872 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
877 Description: RPCs being resent when they shouldn't be
878 Details : Some RPCs that should not be resent are being resent. This
879 can cause inconsistencies in the RPC state machine. Do not
880 resend such requests.
883 Frequency : rare, only with NFS export
885 Description: Crash on NFS re-export node
886 Details : under very unusual load conditions an assertion is hit in
887 ll_intent_file_open()
890 Frequency : only if OST filesystem is corrupted
892 Description: client incorrectly hits assertion in ptlrpc_replay_req()
893 Details : for a short time RPCs with bulk IO are in the replay list,
894 but replay of bulk IOs is unimplemented. If the OST filesystem
895 is corrupted due to disk cache incoherency and then replay is
896 started it is possible to trip an assertion. Avoid putting
897 committed RPCs into the replay list at all to avoid this issue.
902 Description: large O_DIRECT requests fail under memory pressure/fragmentation
903 Details : Large single O_DIRECT read and write calls can fail to allocate
904 a sufficiently large buffer to process the request. In case of
905 allocation failure the allocation is retried with a smaller
906 buffer and broken into smaller requests.
908 Severity : enhancement
910 Description: Add -o localflock option to simulate old noflock behaviour.
911 Details : This will achieve local-only flock/fcntl locks coherentness.
916 Description: versioning check is incomplete
917 Details : Checking the version difference of client vs. server, report
918 error if the gap is too big.
923 Description: add support PG_writeback bit
924 Details : add support for PG_writeback bit for Lustre, for more carefull
925 work with page cache in 2.6 kernel. This also fix some deadlocks
926 and remove hack for work O_SYNC with 2.6 kernel.
928 Severity : enhancement
930 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
931 Details : The uninit_groups feature works in conjunction with the kernel
932 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
933 up the pass1 processing of e2fsck. This is a read-only feature
934 in ldiskfs2 only, so older kernels and current ldiskfs cannot
935 mount filesystems that have had this feature enabled.
937 Severity : enhancement
939 Description: Improve multi-block allocation algorithm to avoid fragmentation
940 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
941 allocation locality and avoid filesystem fragmentation.
943 ------------------------------------------------------------------------------
945 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
947 * Support for kernels:
948 2.6.9-42.0.3.EL (RHEL 4)
950 2.4.21-47.0.1.EL (RHEL 3)
951 2.6.12.6 vanilla (kernel.org)
952 2.6.16.21-0.8 (SLES10)
953 * Recommended e2fsprogs version: 1.39.cfs2-0
955 * The backwards-compatible /proc/sys/portals symlink has been removed
956 in this release. Before upgrading, please ensure that you change
957 any configuration scripts or /etc/sysctl.conf files that access
958 /proc/sys/portals/* or sysctl portals.* to use the corresponding
959 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
960 in advance of the upgrade on any system running Lustre 1.4.6 or
961 newer, since /proc/sys/lnet was added in that version.
962 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
966 Frequency : only when quota is used
968 Description: avoid scanning export list for quota master
969 Details : Change the algorithms to avoid scanning export list in order
970 to improve the efficiency.
973 Frequency : MDS failover only, very rarely
975 Description: "went back in time" messages on mds failover
976 Details : The greatest transno may be lost when the current operation
977 finishes with an error (transno==0) and the client's last_rcvd
978 record is over-written. Save the greatest transno in the
979 mds_last_transno for this case.
982 Frequency : always for specific kernels and striping counts
984 Description: client may get "Matching packet too big" without ACL support
985 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
986 when trying to access files in certain configurations. The
987 clients should in fact be denied when mounting because they do
991 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
993 Description: many clients connecting with IO in progress causes connect timeouts
994 Details : Avoid synchronous journal commits to avoid delays caused by many
995 clients connecting/disconnecting when bulk IO is in progress.
996 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
997 OST_IO_PORTAL to avoid delays behind potentially many pending
1001 Frequency : occasionally with multiple writers to a single file
1003 Description: shared writes to file may result in wrong size reported by stat()
1004 Details : Allow growing of kms when extent lock is cancelled
1007 Frequency : always with random mmap IO to multi-striped file
1009 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1010 Details : Only if the hole is at the end of OST object so that kms is too
1011 small. Fix is to increase kms accordingly in ll_nopage.
1014 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1016 Description: writes to a missing object would leak memory on the OST
1017 Details : If there is an inconsistency between the MDS and OST filesystems,
1018 such that the MDS references an object that doesn't exist, writes
1019 to that object will leak memory due to incorrect cleanup in the
1020 error handling path, eventually running out of memory on the OST.
1025 Description: Creating too long symlink causes lustre errors
1026 Details : Check symlink and name lengths before sending requests to MDS.
1029 Frequency : only if flock is enabled (not on by default)
1031 Description: posix locks not released on fd closure on 2.6.9+
1032 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1033 kernels, this caused such locks not to be released on fd close and
1034 then assertions on fs unmount about still used locks.
1037 Frequency : MDS failover only, very rarely
1039 Description: clients may get ASSERTION(granted_lock != NULL)
1040 Details : When request was taking a long time, and a client was resending
1041 a getattr by name lock request. The were multiple lock requests
1042 with the same client lock handle and
1043 mds_getattr_name->fixup_handle_for_resent_request found one of the
1044 lock handles but later failed with ASSERTION(granted_lock != NULL).
1049 Description: handle->h_buffer_credits > 0, assertion failure
1050 Details : h_buffer_credits is zero after truncate, causing assertion
1051 failure. This patch extends the transaction or creates a new
1055 Frequency : NFS re-export or patchless client
1056 Bugzilla : 11179, 10796
1057 Description: Crash on NFS re-export node (__d_move)
1058 Details : We do not want to hash the dentry if we don't have a lock.
1059 But if this dentry is later used in d_move, we'd hit uninitialised
1060 list head d_hash, so we just do this to init d_hash field but
1061 leave dentry unhashed.
1064 Frequency : NFS re-export or patchless client
1066 Description: NFS exports has problem with symbolic link
1067 Details : lustre client didn't properly install dentry when re-exported
1068 to NFS or running patchless client.
1071 Frequency : NFS re-export or patchless client
1073 Description: Various nfs/patchless fixes.
1074 Details : fixes reuse disconected alias for lookup process - this fixes
1075 warning "find_exported_dentry: npd != pd",
1076 fix permission error with open files at nfs.
1077 fix apply umask when do revalidate.
1080 Frequency : occasional
1082 Description: Crash on NFS re-export node
1083 Details : calling clear_page() on the wrong pointer triggered oops in
1084 generic_mapping_read().
1087 Frequency : rarely, using O_DIRECT IO
1089 Description: unaligned directio crashes client with LASSERT
1090 Details : check for unaligned buffers before trying any requests.
1093 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1095 Description: stale data returned from RAID cache
1096 Details : If only a small amount of IO is done to the RAID device before
1097 reading it again it is possible to get stale data from the RAID
1098 cache instead of reading it from disk.
1101 Frequency : always for sles10 kernel
1103 Description: sles10 support
1104 Details : ll_follow_link: compile fixes and using of nd_set_link
1105 under newer kernels.
1108 Frequency : depends on arch, kernel and compiler version, always on sles10
1111 Description: recursive or deep enough symlinks cause stack overflow
1112 Details : getting rid of large stack-allocated variable in
1116 Frequency : depends on hardware
1118 Description: lustre write performance loss in the SLES10 kernel
1119 Details : the performance loss is caused by using of write barriers in the
1120 ext3 code. The SLES10 kernel turns barrier support on by
1121 default. The fix is to undo that change for ldiskfs.
1123 ------------------------------------------------------------------------------
1125 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1127 * Support for kernels:
1128 2.6.9-42.0.3EL (RHEL 4)
1129 2.6.5-7.276 (SLES 9)
1130 2.4.21-47.0.1.EL (RHEL 3)
1131 2.6.12.6 vanilla (kernel.org)
1135 Frequency : quota enabled and large files being deleted
1137 Description: releasing more than 4GB of quota at once hangs OST
1138 Details : If a user deletes more than 4GB of files on a single OST it
1139 will cause the OST to spin in an infinite loop. Release
1140 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1145 Description: statfs data retrieved from /proc may be stale or zero
1146 Details : When reading per-device statfs data from /proc, in the
1147 {kbytes,files}_{total,free,avail} files, it may appear
1148 as zero or be out of date.
1151 Frequency : systems with MD RAID1 external journal devices
1153 Description: lconf's call to blkid is confused by RAID1 journal devices
1154 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1155 of returning all block devices that match the journal UUID.
1158 Frequency : always, for aggregate stripe size over 4GB
1160 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1161 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1162 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1165 Frequency : always if "lfs find" used on a local file/directory
1167 Description: "lfs find" segfaults if used on a local file/directory
1168 Details : The case where a directory component was not specified wasn't
1169 handled correctly. Handle this properly.
1172 Frequency : always on ppc64
1174 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1175 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1176 perform endianness conversion.
1179 Frequency : rarely (truncate to non-zero file size after write under load)
1180 Bugzilla : 10730, 10687
1181 Description: Files padded with zeros to next 4K multiple
1182 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1183 it is possible that files that are truncated to a non-zero size
1184 immediately after being written are filled with zero bytes beyond
1185 the truncated size. No file data is lost.
1187 Severity : enhancement
1189 Description: Allow recovery/failover for liblustre clients.
1190 Details : liblustre clients were unaware of failover configurations until
1193 Severity : enhancement
1195 Description: user file locks should fail when not mounting with flock option
1196 Details : Set up an error-returning stub in ll_file_operations.lock field
1197 to prevent incorrect behaviour when client is mounted without
1198 flock option. Also, set up properly f_op->flock field for
1202 Frequency : always on ia64
1204 Description: "lfs df" loops on printing out MDS statfs information
1205 Details : The obd_ioctl_data was not initialized and in some systems
1206 this caused a failure during the ioctl that did not return
1207 an error. Initialize the struct and return an error on failure.
1210 Frequency : SLES 9 only
1212 Description: Error of copying files with lustre special EAs as root
1213 Details : Client side always return success for setxattr call for lustre
1214 special xattr (currently only "trusted.lov").
1217 Frequency : rarely on clusters with both ia64+i386 clients
1219 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1220 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1221 clients are doing concurrent writes to the same file it is
1222 possible that the ia64 clients may LASSERT because the OST
1223 extent locks are not PAGE_SIZE aligned. Ensure that grown
1224 locks are always aligned on the request boundary.
1227 Frequency : specific use, occasional
1229 Description: Overwriting in use executable truncates on-disk binary image
1230 Details : If one node attempts to overwrite an executable in use by
1231 another node, we now correctly return ETXTBSY instead of
1232 truncating the file.
1234 Severity : enhancement
1236 Description: Async OSC create to avoid the blocking unnecessarily.
1237 Details : If a OST has no remain object, system will block on the creating
1238 when need to create a new object on this OST. Now, ways use
1239 pre-created objects when available, instead of blocking on an
1240 empty osc while others are not empty. If we must block, we block
1241 for the shortest possible period of time.
1246 Description: chmod on Lustre root is propagated to other clients
1247 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1248 invalid by the follow_mount time.
1251 Frequency : liblustre clients only
1253 Description: Race in 'instant cancel' lock handling could lead to such locks
1254 never to be granted in case of SMP MDS
1255 Details : Do not destroy not yet granted but cbpending locks in
1259 Frequency : replay/resend of open
1261 Description: non null lock assetion failure in mds_intent_policy
1262 Details : Trying to replay/resend lockless open requests resulted in
1263 mds_open() returning 0 with no lock. Now it sets a flag if
1264 a lock is going to be returned.
1266 Severity : enhancement
1268 Description: Checksum enhancements
1269 Details : New checksum enhancements allow for resending RPCs that failed
1272 Severity : enhancement
1274 Description: Tunables on number of dirty pages in cacche
1275 Details : Allow to set limit on number of dirty pages cached.
1280 Description: client crash on unmount - lock still has references
1281 Details : In some error handling cases it was possible to leak a lock
1282 reference on a client while accessing a file. This was not
1283 harmful to the client during operation, but would cause the
1284 client to crash when the filesystem is unmounted.
1287 Frequency : specific case, rare
1289 Description: ETXTBSY on mds though file not in use by client
1290 Details : ETXTBSY is no longer incorrectly returned when attempting to
1291 chmod or chown a directory that the user previously tried to
1292 execute or a currently-executing binary.
1295 Frequency : extremely rare except on liblustre-based clients
1297 Description: Lustre space not freed when files are deleted
1298 Details : Clean up open-unlinked files after client eviction. Previously
1299 the unlink was skipped and the files remained as orphans.
1304 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1305 Details : In some cases it was possible to send a blocking callback to
1306 a client doing a glimpse, even though that client didn't get
1307 a lock granted. When the glimpse lock is cancelled on the OST
1308 the freed lock is left on the waiting list and corrupted the list.
1311 Frequency : all core dumps
1313 Description: Broke core dumps to lustre
1314 Details : Negative dentry may be unhashed if parent does not have UPDATE
1315 lock, but some callers, e.g. do_coredump, expect dentry to be
1316 hashed after successful create, hash it in ll_create_it.
1318 ------------------------------------------------------------------------------
1320 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1322 * Support for kernels:
1323 2.6.9-42.0.2.EL (RHEL 4)
1324 2.6.5-7.276 (SLES 9)
1325 2.4.21-40.EL (RHEL 3)
1326 2.6.12.6 vanilla (kernel.org)
1330 Frequency : always on RHEL 3
1332 Description: Number of open files grows over time
1333 Details : The number of open files grows over time, whether or not
1334 Lustre is started. This was due to a filp leak introduced
1335 by one of our kernel patches.
1337 ------------------------------------------------------------------------------
1339 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1341 * Support for kernels:
1342 2.6.9-42.EL (RHEL 4)
1343 2.6.5-7.276 (SLES 9)
1344 2.4.21-40.EL (RHEL 3)
1345 2.6.12.6 vanilla (kernel.org)
1350 Bugzilla : 5719, 9635, 9792, 9684
1351 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1352 Details : If a server is under heavy load and cannot reply to new
1353 connection requests before the client resends the (re)connect,
1354 the connection handling code can behave badly if two service
1355 threads are concurrently handing separate (re)connections from
1356 the same client. Add better locking to the connection handling
1357 code, and ensure that only a single connection will be processed
1358 for a given client UUID, even if the lock is dropped.
1360 Severity : enhancement
1362 Description: add TCP zero-copy support to kernel
1363 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1364 sends if the hardware supports scatter-gather and checksumming.
1365 This allows socklnd to do client-write and server-read more
1366 efficiently and reduce CPU utilization from skbuf copying.
1369 Frequency : only if NFS exporting from client
1371 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1372 Details : If setting attributes on a file created under NFS that had
1373 never been opened it would be possible to oops the client
1374 if the file had no objects.
1377 Frequency : always for liblustre
1379 Description: liblustre client does MDS+OSTs setattr RPC for each write
1380 Details : When doing a write from a liblustre client, the client
1381 incorrectly issued an RPC to the MDS and each OST the file was
1382 striped over in order to update the timestamps. When writing
1383 with small chunks and many clients this could overwhelm the MDS
1384 with RPCs. In all cases it would slow down the write because
1385 these RPCs are unnecessary.
1387 Severity : enhancement
1389 Description: allow number of MDS service threads to be changed at module load
1390 Details : It is now possible to change the number of MDS service threads
1391 running. Adding "options mds mds_num_threads={N}" to the MDS's
1392 /etc/modprobe.conf will set the number of threads for the next
1393 time Lustre is restarted (assuming the "mds" module is also
1394 reloaded at that time). The default number of threads will
1395 stay the same, 32 for most systems.
1400 Description: OST crash if filesystem is unformatted or corrupt
1401 Details : If an OST is started on a device that has never been formatted
1402 or if the filesystem is corrupt and cannot even mount then the
1403 error handling cleanup routines would dereference a NULL pointer.
1408 Description: NULL pointer deref in llap_from_page.
1409 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1410 incorrect) mapping if the page was truncated/reclaimed while it was
1411 searched for. Check for this condition and skip such pages when
1412 doing readahead. Introduce extra check to llap_from_page() to
1413 verify page->mapping->host is non-NULL (so page is not anonymous).
1416 Frequency : Sometimes when using sys_sendfile
1418 Description: "page not covered by a lock" warnings from ll_readpage
1419 Details : sendfile called ll_readpage without right page locks present.
1420 Now we introduced ll_file_sendfile that does necessary locking
1421 around call to generic_file_sendfile() much like we do in
1425 Frequency : with certain MDS communication failures at client mount time
1427 Description: NULL pointer deref after failed client mount
1428 Details : a client connection request may delayed by the network layer
1429 and not be sent until after the PTLRPC layer has timed out the
1430 request. If the client fails the mount immediately it will try
1431 to clean up before the network times out the request. Add a
1432 reference from the request import to the obd device and delay
1433 the cleanup until the network drops the request.
1436 Frequency : occasionally during client (re)connect
1438 Description: assertion failure during client (re)connect
1439 Details : processing a client connection request may be delayed by the
1440 client or server longer than the client connect timeout. This
1441 causes the client to resend the connection request. If the
1442 original connection request is replied in this interval, the
1443 client may trip an assertion failure in ptlrpc_connect_interpret()
1444 which thought it would be the only running connect process.
1447 Frequency : only with obd_echo servers and clients that are rebooted
1449 Description: kernel BUG accessing uninitialized data structure
1450 Details : When running an obd_echo server it did not start the ping_evictor
1451 thread, and when a client was evicted an uninitialized data
1452 structure was accessed. Start the ping_evictor in the RPC
1453 service startup instead of the OBD startup.
1455 Severity : enhancement
1456 Bugzilla : 10193 (patchless)
1457 Description: Remove dependency on various unexported kernel interfaces.
1458 Details : No longer need reparent_to_init, exit_mm, exit_files,
1459 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1462 Frequency : rare (only users of deprecated and unsupported LDAP config)
1464 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1465 Details : LDAP apparently contains 'lustreName' attributes instead of
1466 'name'. A simple remapping of the name is sufficient.
1469 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1471 Description: waiting_locks_callback trips kernel BUG if client is evicted
1472 Details : Running with the dump_on_timeout debug flag turned on makes
1473 it possible that the waiting_locks_callback() can try to dump
1474 the Lustre kernel debug logs from an interrupt handler. Defer
1475 this log dumping to the expired_lock_main() thread.
1477 Severity : enhancement
1479 Description: Support NFS exporting on 2.6 kernels.
1480 Details : Implement non-rawops metadata methods for NFS server to use without
1481 changing NFS server code.
1484 Frequency : very rare (synthetic metadata workload only)
1486 Description: two racing renames might cause an MDS thread to deadlock
1487 Details : Running the "racer" program may cause one MDS thread to rename
1488 a file from being the source of a rename to being the target of
1489 a rename at exactly the same time that another thread is doing
1490 so, and the second thread has already enqueued these locks after
1491 doing a lookup of the target and is trying to relock them in
1492 order. Ensure that we don't try to re-lock the same resource.
1495 Frequency : only very large systems with liblustre clients
1497 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1498 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1499 all OSTs in parallel. This allows the eviction of stale liblustre
1500 clients to proceed much faster than if they were done in series,
1501 and also offers similar improvements for other set_info RPCs.
1506 Description: excessive CPU usage during initial read phase on client
1507 Details : During the initial read phase on a client, it would agressively
1508 retry readahead on the file, consuming too much CPU and impacting
1509 performance (since 1.4.5.8). Improve the readahead algorithm
1510 to avoid this, and also improve some other common cases (read
1511 of small files in particular, where "small" is files smaller than
1512 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1517 Description: MDS crash when receiving packet with unknown intent.
1518 Details : Do not LBUG in unknown intent case, just return -EFAULT
1520 Severity : enhancement
1521 Bugzilla : 9293, 9385
1522 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1523 Details : Do not serialize getattr (non-intent version) and statfs.
1526 Frequency : occasional, when OST network is overloaded/intermittent
1528 Description: client evicted by OST after bulk IO timeout
1529 Details : If a client sends a bulk IO request (read or write) the OST
1530 may evict the client if it is unresposive to its data GET/PUT
1531 request. This is incorrect if the network is overloaded (takes
1532 too long to transfer the RPC data) or dropped the OST GET/PUT
1533 request. There is no need to evict the client at all, since
1534 the pinger and/or lock callbacks will handle this, and the
1535 client can restart the bulk request.
1538 Frequency : Always when mmapping file with no objects
1540 Description: client crashes when mmapping file with no objects
1541 Details : Check that we actually have objects in a file before doing any
1542 operations on objects in ll_vm_open, ll_vm_close and
1548 Description: Request leak when working with deleted CWD
1549 Details : Introduce advanced request refcount tracking for requests
1550 referenced from lustre intent.
1552 Severity : Enhancement
1554 Description: Cache open file handles on client.
1555 Details : MDS now will return special lock along with openhandle, if
1556 requested and client is allowed to hold openhandle, even if unused,
1557 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1558 closing files for every read/write openration.
1560 Severity : Enhancement
1562 Description: Cache open negative dentries on client when possible.
1563 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1564 negative dentries on lock revocation.
1569 Description: Remounting a client read-only wasn't possible with a zconf mount
1570 Details : It wasn't possible to remount a client read-only with llmount.
1572 Severity : enhancement
1573 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1574 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1575 MPICH is included to add Lustre-specific ADIO interfaces.
1576 This is based closely on the UFS ADIO layer and only differs
1577 in file creation, in order to allow the OST striping to be set.
1578 This is user-contributed code and not supported by CFS.
1583 Description: extended inode attributes (immutable, append-only) work improperly
1584 when 2.4 and 2.6 kernels are used on client/server or vice versa
1585 Details : Introduce kernel-independent values for these flags.
1587 Severity : enhancement
1590 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1591 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1592 granularity. Now, Fractional MB granularity is be supported,
1593 this is very useful for low memory system.
1595 Severity : enhancement
1597 Description: Getattr by fid
1598 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1599 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1600 have locks and avoids getattr rpc requests for it on every stat.
1603 Frequency : Always, for filesystems larger than 2TB
1605 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1606 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1607 filesystems larger than 2TB to be mounted with mballoc enabled.
1610 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1612 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1613 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1614 filesystems larger than 2TB is mounted then the kernel will
1615 silently corrupt the start of the filesystem. CONFIG_LBD is
1616 enabled for all CFS-supported kernels, but the possibility of
1617 this happening with a modified kernel config exists.
1619 Severity : enhancement
1621 Description: add client O_DIRECT support for 2.6 kernels
1622 Details : It is now possible to do O_DIRECT reads and writes to files
1623 in the Lustre client mountpoint on 2.6 kernel clients.
1625 Severity : enhancement
1627 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1628 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1629 parallel, not waiting for response from every OST before sending
1630 a rpc to the next OST.
1635 Description: setattr vs write race when updating file timestamps
1636 Details : Client processes that update a file timestamp into the past
1637 right after writing to the file (e.g. tar) it is possible that
1638 the updated file modification time can be reset to the current
1639 time due to a race between processing the setattr and write RPC.
1641 Severity : enhancement
1643 Description: Bring 'lfs find' closer in line with regular Linux find.
1644 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1645 -print0 options and obtains all the needed info through the lustre
1648 Severity : enhancement
1650 Description: support up to 1024 configured devices on one node
1651 Details : change obd_dev array from statically allocated to dynamically
1652 allocated structs as they are first used to reduce memory usage
1657 Description: Flush dirty partially truncated pages during truncate
1658 Details : Immediatelly flush partially truncated pages in filter_setattr,
1659 this way we completely avoid having any pages in page cache on OST
1660 and can retire ugly workarounds during writes to flush such pages.
1665 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1666 Details : protect inode from truncation within vfs_unlink() context
1667 just take a reference before calling vfs_unlink() and release it
1668 when parent's i_sem is free.
1673 Description: last_id value checked outside lock on OST caused LASSERT failure
1674 Details : If there were multiple MDS->OST object precreate requests in
1675 flight, it was possible that the OST's last object id was checked
1676 outside a lock and incorrectly tripped an assertion. Move checks
1677 inside locks, and discard old precreate requests.
1680 Frequency : always, if extents are used on OSTs
1682 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1683 Details : OSTs using the extents format would not zero the high 16 bits of
1684 the index physical block number. This is not a problem for any
1685 OST filesystems smaller than 16TB, and no kernels support ext3
1686 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1687 new/modified files) and can be fixed for existing filesystems
1688 with e2fsprogs-1.39-cfs1.
1693 Description: import connection selection may be incorrect if timer wraps
1694 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1695 import connections to be ignored if the 32-bit jiffies counter
1696 wraps. Use a 64-bit jiffies counter.
1699 Frequency : very large clusters immediately after boot
1701 Description: LNET request buffers exhausted under heavy short-term load
1702 Details : If a large number of client requests are generated on a service
1703 that has previously never seen so many requests it is possible
1704 that the request buffer growth cannot keep up with the spike in
1705 demand. Instead of dropping incoming requests, they are held in
1706 the LND until the RPC service can accept more requests.
1709 Frequency : Sometimes during replay
1711 Description: Assertion failure in ll_local_open after replay.
1712 Details : If replay happened on an open request reply before we were able
1713 to set replay handler, reply will become not swabbed tripping the
1714 assertion in ll_local_open. Now we set the handler right after
1715 recognising of open request
1718 Frequency : very rare
1720 Description: kernel reports "badness in vsnprintf"
1721 Details : Reading from the "recovery_status" /proc file in small chunks
1722 may cause a negative length in lprocfs_obd_rd_recovery_status()
1723 call to vsnprintf() (which is otherwise harmless). Exit early
1724 if there is no more space in the output buffer.
1726 Severity : enhancement
1728 Description: clear OBD RPC statistics by writing to them
1729 Details : It is now possible to clear the OBD RPC statistics by writing
1730 to the "stats" file.
1735 Description: Client mtime is not the same on different clients after utimes
1736 Details : In some cases, the client was using the utimes() syscall on
1737 a file cached on another node. The clients now validate the
1738 ctime from the MDS + OSTs to determine which one is right.
1743 Description: Inability to activate failout mode
1744 Details : lconf script incorrectly assumed that in pythong string's numeric
1745 value is used in comparisons.
1748 Frequency : always with multiple stripes per file
1750 Description: Inefficient object allocation for mutli-stripe files
1751 Details : When selecting which OSTs to stripe files over, for files with
1752 a stripe count that divides evenly into the number of OSTs,
1753 the MDS is always picking the same starting OST for each file.
1754 Return the OST selection heuristic to the original design.
1759 Description: mount failures may take full timeout to return an error
1760 Details : Under some heavy load conditions it is possible that a
1761 failed mount can wait for the full obd_timeout interval,
1762 possibly several minutes, before reporting an error.
1763 Instead return an error as soon as the status is known.
1765 Frequency : quota enabled and large files being deleted
1767 Description: releasing more than 4GB of quota at once hangs OST
1768 Details : If a user deletes more than 4GB of files on a single OST it
1769 will cause the OST to spin in an infinite loop. Release
1770 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1775 Description: statfs data retrieved from /proc may be stale or zero
1776 Details : When reading per-device statfs data from /proc, in the
1777 {kbytes,files}_{total,free,avail} files, it may appear
1778 as zero or be out of date.
1781 Frequency : systems with MD RAID1 external journal devices
1783 Description: lconf's call to blkid is confused by RAID1 journal devices
1784 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1785 of returning all block devices that match the journal UUID.
1788 Frequency : always, for aggregate stripe size over 4GB
1790 Description: assertion fails when trying to use 4GB stripe size
1791 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1792 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1795 Frequency : always on ppc64
1797 Description: the first write on an ext3 filesystem with mballoc got stuck
1798 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1799 perform endianness conversion.
1801 ------------------------------------------------------------------------------
1803 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1805 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1806 INTEROPERATE with older versions automatically. Please read the
1807 user documentation before upgrading any part of a live system.
1808 * WARNING: Lustre networking configuration changes are required with
1809 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1812 * Support for kernels:
1813 2.6.9-22.0.2.EL (RHEL 4)
1814 2.6.5-7.244 (SLES 9)
1815 2.6.12.6 vanilla (kernel.org)
1818 Severity : enhancement
1819 Bugzilla : 7981/8208
1820 Description: Introduced Lustre Networking (LNET)
1821 Details : LNET is new networking infrastructure for Lustre, it includes
1822 a reorganized network configuration mode (see the user
1823 documentation for full details) as well as support for routing
1824 between different network fabrics. Lustre Networking Devices
1825 (LNDs) for the supported network fabrics have also been
1826 created for this new infrastructure.
1828 Severity : enhancement
1829 Description: Introduced Access control lists
1830 Details : clients can set ACLs on files and directories in order to have
1831 more fine-grained permissions than the standard Unix UGO+RWX.
1832 The MDS must be started with the "-o acl" mount option.
1834 Severity : enhancement
1835 Description: Introduced filesystem quotas
1836 Details : Administrators may now establish per-user quotas on the
1839 Severity : enhancement
1841 Description: Configuration change for the XT3
1842 The PTLLND is now used to run Lustre over Portals on the XT3
1843 The configure option(s) --with-cray-portals are no longer used.
1844 Rather --with-portals=<path-to-portals-includes> is used to
1845 enable building on the XT3. In addition to enable XT3 specific
1846 features the option --enable-cray-xt3 must be used.
1851 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1852 Details : A race between allocating a new llcd and re-getting the llcd_lock
1853 allowed another thread to grab newly-allocated llcd.
1855 Severity : enhancement
1857 Description: 2.6 OST async journal commit and locking fix to improve performance
1858 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1859 2.6 kernels are now async as they already were in 2.4 kernels so
1860 that they can commit concurrently with the network bulk transfer.
1861 For block-allocated files the filter allocation semaphore is held
1862 to avoid filesystem fragmentation during allocation. BKL lock
1863 removed for 2.6 xattr operations where it is no longer needed.
1868 Description: lconf incorrectly determined whether two IP networks could talk
1869 Details : In some more complicated routing and multiple-network
1870 configurations, lconf will avoid trying to make a network
1871 connection to a disjoint part of the IP space. It was doing the
1872 math incorrectly for one set of cases.
1877 Description: Fix for potential infinite loop processing records in an llog.
1878 Details : If an llog record is corrupted/zeroed, it is possible to loop
1879 forever in llog_process(). Validate the llog record length
1880 and skip the remainder of the block on error.
1883 Frequency : occasional (liblustre only)
1885 Description: liblustre could not open files whose last component is a symlink
1886 Details : sysio_path_walk() would incorrectly pass the open intent to
1887 intermediate path components.
1890 Frequency : rare (liblustre only with non-standard tuning)
1891 Bugzilla : 7201 (7350)
1892 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1893 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1894 after it has been released, only do this for referenced locks
1896 Severity : enhancement
1898 Description: specifying an (invalid) directory default stripe_size of -1
1899 would reset the directory default striping
1900 Details : stripe_size -1 was used internally to signal directory stripe
1901 removal, now use "all default" to signal dir stripe removal
1902 as a directory striping of "all default" is not useful
1905 Frequency : common for large clusters running liblustre clients
1907 Description: doing an ls when liblustre clients are running is slow
1908 Details : sending a glimpse AST to a liblustre client waits for every AST
1909 to time out, as liblustre clients will not respond. Since they
1910 cannot cache data we refresh the OST lock LVB from disk instead.
1912 Severity : enhancement
1914 Description: doing an ls at the same time as file IO can be slow
1915 Details : enqueue and other "small" requests can be blocked behind many
1916 large IO requests. Create a new OST IO portal for non-IO
1917 requests so they can be processed faster.
1920 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1922 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1923 created file with mode 000
1924 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1925 allow an ftruncate() as a non-root user to a file with mode 000.
1926 Lustre now respects this flag to disable mode checks when
1927 truncating a file owned by the user
1930 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1933 Description: Revoking locks from clients that went dead or catatonic might take
1935 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1936 cancellation of such locks instant on servers without waiting for
1937 any reply from clients. Clients drops these locks when cancel
1938 notification from server is received without replying.
1941 Frequency : liblustre-only, when liblustre client dies or becomes busy
1943 Description: Doing ls on Linux clients can take a long time with active
1945 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1946 granting such locks to it in the first place if possible. Locks
1947 are taken by proxy on the OST during the read or write and
1948 dropped immediately afterward. Add connect flags handling, do
1949 not grant locks to liblustre clients for glimpse ASTs.
1951 Severity : enhancement
1953 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1954 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1955 sequential reads and behaves badly with applications doing random
1956 reads. Improve it by reading ahead at least read region, and
1957 avoiding excessive large RPC for small reads.
1959 Severity : enhancement
1961 Description: Creating more than 1000 files for a single job may cause a load
1962 imbalance on the OSTs if there are also a large number of OSTs.
1963 Details : qos_prep_create() uses an OST index reseed value that is an
1964 even multiple of the number of available OSTs so that if the
1965 reseed happens in the middle of the object allocation it will
1966 still utilize the OSTs as uniformly as possible.
1971 Description: OST or MDS may oops in ping_evictor_main()
1972 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1973 but doesn't restart at beginning of obd_exports_timed list
1976 Severity : enhancement
1978 Description: improve by-nid export eviction on the MDS and OST
1979 Details : allow multiple exports with the same NID to be evicted at one
1980 time without re-searching the exports list.
1983 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1985 Description: MDS may oops in groups_free()
1986 Details : in rare race conditions a newly allocated group_info struct is
1987 freed again, and this can be NULL. The 2.4 compatibility code
1988 for groups_free() checked for a NULL pointer, but 2.6 did not.
1991 Frequency : common for liblustre clients doing little filesystem IO
1992 Bugzilla : 9352, 7313
1993 Description: server may evict liblustre clients accessing contended locks
1994 Details : if a client is granted a lock or receives a completion AST
1995 with a blocking AST already set it would not reply to the AST
1996 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1999 Frequency : lfs setstripe, only systems with more than 160 OSTs
2001 Description: unable to set striping with a starting offset beyond OST 160
2002 Details : llapi_create_file() incorrectly limited the starting stripe
2003 index to the maximum single-file stripe count.
2006 Frequency : LDAP users only
2008 Description: lconf did not handle in-kernel recovery with LDAP properly
2009 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2011 Severity : enhancement
2013 Description: bind OST threads to NUMA nodes to improve performance
2014 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2015 node and do their allocations there to localize memory access
2017 Severity : enhancement
2019 Description: llmount can determine client NID directly from Myrinet (GM)
2020 Details : the client NID code from gmnalnid was moved directly into
2021 llmount, removing the need to use this or specifying the
2022 client NID explicitly when mounting GM clients with zeroconf
2025 Frequency : if client is started with down MDS
2027 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2028 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2029 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2030 ptlrpc_import_delayed_req() because we didn't check if the
2031 request was interrupted, and we also didn't break out of the
2032 event loop if there was no timeout
2037 Description: data loss during non-page-aligned writes to a single file from
2038 both multiple nodes and multiple threads on one node at same time
2039 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2040 inconsistency led to false short-reads, that were cached and later
2041 used by ->prepare_write() to fill in partially written page,
2042 leading to data loss.
2045 Frequency : always, if lconf --abort_recovery used
2047 Description: lconf --abort_recovery fails with 'Operation not supported'
2048 Details : lconf was attempting to abort recovery on the MDT device and not
2051 Severity : enhancement
2053 Description: remove cleanup logs
2054 Details : replace lconf-generated cleanup logs with lustre internal
2055 cleanup routines. Eliminates the need for client-cleanup and
2058 Severity : enhancement
2060 Description: add support for EAs (user and system) on lustre filesystems
2061 Details : it is now possible to store extended attributes in the Lustre
2062 client filesystem, and with the user_xattr mount option it
2063 is possible to allow users to store EAs on their files also
2065 Severity : enhancement
2067 Description: Add possibility (config option) to show minimal available OST free
2069 Details : When compiled with --enable-mindf configure option, statfs(2)
2070 (and so, df) will return least minimal free space available from
2071 all OSTs as amount of free space on FS, instead of summary of
2072 free spaces of all OSTs.
2074 Severity : enhancement
2076 Description: do not expand extent locks acquired on OST-side
2077 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2078 by server: they are not cached anyway.
2081 Frequency : when mmap is used/binaries executed from Lustre
2083 Description: Unmmap pages before throwing them away from read cache.
2084 Details : llap_shrink cache now attempts to unmap pages before discarding
2085 them (if unmapping failed - do not discard). SLES9 kernel has
2086 extra checks that trigger if this unmapping is not done first.
2091 Description: lconf didn't resolve symlinks before checking to see whether a
2092 given mountpoint was already in use
2095 Frequency : when migrating failover services
2096 Bugzilla : 6395, 9514
2097 Description: When migrating a subset of services from a node (e.g. failback
2098 from a failover service node) the remaining services would
2099 time out and evict clients.
2100 Details : lconf --force (implied by --failover) sets the global obd_timeout
2101 to 5 seconds in order to quickly disconnect, but this caused
2102 other RPCs to time out too quickly. Do not change the global
2103 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2105 Severity : enhancement
2106 Frequency : if MDS is started with down OST
2107 Bugzilla : 9439,5706
2108 Description: Allow startup/shutdown of an MDS without depending on the
2109 availability of the OSTs.
2110 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2111 Add appropriate locking and lov-osc refcounts for safe
2112 cleaning. Add osc abort_inflight calls in case the
2113 synchronize never started.
2116 Frequency : occasional (Cray XT3 only)
2118 Description: root not authorized to access files in CRAY_PORTALS environment
2119 Details : The client process capabilities were not honoured on the MDS in
2120 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2121 been accessed by an authorized user then root was able to access
2122 the file on the local client also. The root user capabilities
2123 are now allowed on the MDS, as this environment has secure UID.
2126 Frequency : occasional
2128 Description: ldiskfs "too long searching" message happens too often
2129 Details : A debugging message (otherwise harmless) prints too often on
2130 the OST console. This has been reduced to only happen when
2131 there are fragmentation problems on the filesystem.
2136 Description: Division by zero in statfs when all OSCs are inactive
2137 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2138 lov_statfs divides by value returned by lov_get_stripecnt().
2142 Bugzilla : 9489, 3273
2143 Description: First write from each client to each OST was only 4kB in size,
2144 to initialize client writeback cache, which caused sub-optimal
2145 RPCs and poor layout on disk for the first writen file.
2146 Details : Clients now request an initial cache grant at (re)connect time
2147 and so that they can start streaming writes to the cache right
2148 away and always do full-sized RPCs if there is enough data.
2149 If the OST is rebooted the client also re-establishes its grant
2150 so that client cached writes will be honoured under the grant.
2155 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2156 Details : Now I/O RPCs go to different portal number and (presumably) fast
2157 lock requests (and glimses) and other RPCs get their own service
2158 threads pool that should be able to service those RPCs
2161 Severity : enhancement
2163 Description: Ability to exchange lustre version between client and servers and
2164 issue warnings at client side if client is too old. Also for
2165 liblustre clients there is ability to refuse connection of too old
2167 Details : New 'version' field is added to connect data structure that is
2168 filled with version info. That info is later checked by server and
2172 Frequency : rare, liblustre only.
2173 Bugzilla : 9296, 9581
2174 Description: Two simultaneous writes from liblustre at offset within same page
2175 might proceed at the same time overwriting eachother with stale
2177 Details : I/O lock withing llu_file_prwv was released too early, before data
2178 actually was hitting the wire. Extended lock-holding time until
2179 server acknowledges receiving data.
2182 Frequency : extremely rare. Never observed in practice.
2184 Description: avoid generating lustre_handle cookie of 0.
2185 Details : class_handle_hash() generates handle cookies by incrementing
2186 global counter, and can hit 0 occasionaly (this is unlikely, but
2187 not impossible, because initial value of cookie counter is
2188 selected randonly). Value of 0 is used as a sentinel meaning
2189 "unassigned handle" --- avoid it. Also coalesce two critical
2190 sections in this function into one.
2192 Severity : enhancement
2194 Description: allow liblustre clients to delegate truncate locking to OST
2195 Details : To avoid overhead of locking, liblustre client instructs OST to
2196 take extent lock in ost_punch() on client's behalf. New connection
2197 flag is added to handle backward compatibility.
2199 Severity : enhancement
2200 Bugzilla : 4928, 7341, 9758
2201 Description: allow number of OST service threads to be specified
2202 Details : a module parameter allows the number of OST service threads
2203 to be specified via "options ost ost_num_threads={N}" in the
2204 OSS's /etc/modules.conf or /etc/modprobe.conf.
2208 Bugzilla : 6146, 9635, 9895
2209 Description: servers crash with bad pointer in target_handle_connect()
2210 Details : In rare cases when a client is reconnecting it was possible that
2211 the connection request was the last reference for that export.
2212 We would temporarily drop the export reference and get a new
2213 one, but this may have been the last reference and the export
2214 was just destroyed. Get new reference before dropping old one.
2216 Severity : enhancement
2217 Frequency : if client is started with failover MDS
2219 Description: Allow multiple MDS hostnames in the mount command
2220 Details : Try to read the configuration from all specified MDS
2221 hostnames during a client mount in case the "primary"
2224 Severity : enhancement
2226 Description: Stop sending data to evicted clients as soon as possible.
2227 Details : Check if the client we are about to send or are sending data to
2228 was evicted already. (Check is done every second of waiting,
2229 for which l_wait_event interface was extended to allow checking
2230 of exit condition at specified intervals).
2233 Frequency : rare, normally only when NFS exporting is done from client
2235 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2237 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2240 Frequency : always, if srand() is called before liblustre initialization
2242 Description: Liblustre uses system PRNG disturbing its usage by user application
2243 Details : Introduce internal to lustre fast and high-quality PRNG for
2244 lustre usage and make liblustre and some other places in generic
2245 lustre code to use it.
2247 Severity : enhancement
2248 Bugzilla : 9477, 9557, 9870
2249 Description: Verify that the MDS configuration logs are updated when xml is
2250 Details : Check if the .xml configuration logs are newer than the config
2251 logs stored on the MDS and report an error if this is the case.
2252 Request --write-conf, or allow starting with --old_conf.
2254 Severity : enhancement
2256 Description: Handle symlinks in the path when checking if Lustre is mounted.
2257 Details : Resolve intermediate symlinks when checking if a client has
2258 mounted a filesystem to avoid duplicate client mounts.
2263 Description: lconf can hit an error exception but still return success.
2264 Details : The lconf command catches the Command error exception at the top
2265 level script context and will exit with the associated exit
2266 status, but doesn't ensure that this exit status is non-zero.
2271 Description: failure of ptlrpc thread startup can cause oops
2272 Details : Starting a ptlrpc service thread can fail if there are a large
2273 number of threads or the server memory is very fragmented.
2274 Handle this without oopsing.
2277 Frequency : always, only if liblustre and non-default acceptor port was used
2279 Description: liblustre cannot connect to servers with non-default acceptor port
2280 Details : tcpnal_set_default_params() was not called and was therefore
2281 ignoring the environment varaible TCPNAL_PORT, as well as other
2282 TCPNAL_ environment variables
2287 Description: two objects could be created on the same OST for a single file
2288 Details : If an OST is down, in some cases it was possible to create two
2289 objects on a single OST for a single file. No problems other
2290 than potential performance impact and spurious error messages.
2294 Bugzilla : 5681, 9562
2295 Description: Client may oops in ll_unhash_aliases
2296 Details : Client dcache may become inconsistent in race condition.
2297 In some cases "getcwd" can fail if the current directory is
2303 Description: Inode refcounting problems in NFS export code
2304 Details : link_raw functions used to call d_instantiate without obtaining
2305 extra inode reference first.
2309 Bugzilla : 9942, 9903
2310 Description: Referencing freed requests leading to crash, memleaks with NFS.
2311 Details : We used to require that call to ll_revalidate_it was always
2312 followed by ll_lookup_it. Also with revalidate_special() it is
2313 possible to call ll_revalidate_it() twice for the same dentry
2314 even if first occurence returned success. This fix changes semantic
2315 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2316 reference on a request referred from the intent.
2317 ll_intent_release() then releases such a request.
2320 Frequency : rare, normally benchmark loads only
2322 Description: unlinked inodes were kept in memory on the client
2323 Details : If a client is repeatedly creating and unlinking files it
2324 can accumulate a lot of stale inodes in the inode slab cache.
2325 If there is no other client load running this can cause the
2326 client node to run out of memory. Instead flush old inodes
2327 from client cache that have the same inode number as a new inode.
2330 Frequency : SLES9 2.6.5 kernel and long filenames only
2331 Bugzilla : 9969, 10379
2332 Description: utime reports stale NFS file handle
2333 Details : SLES9 uses out-of-dentry names in some cases, which confused
2334 the lustre dentry revalidation. Change it to always use the
2338 Frequency : rare, unless heavy write-truncate concurrency is continuous
2339 Bugzilla : 4180, 6984, 7171, 9963, 9331
2340 Description: OST becomes very slow and/or deadlocked during object unlink
2341 Details : filter_destroy() was holding onto the parent directory lock
2342 while truncating+unlinking objects. For very large objects this
2343 may block other threads for a long time and slow overall OST
2344 responsiveness. It may also be possible to get a lock ordering
2345 deadlock in this case, or run out of journal credits because of
2346 the combined truncate+unlink. Solution is to do object truncate
2347 first in one transaction without parent lock, and then do the
2348 final unlink in a new transaction with the parent lock. This
2349 reduces the lock hold time dramatically.
2352 Frequency : rare, 2.4 kernels only
2354 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2355 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2356 before unmounting their filesystem in order to avoid deadlock.
2357 The kernel_locked() function in 2.4 kernels only checks whether
2358 the kernel lock is held, not whether it is this process that is
2359 holding it as 2.6 kernels do.
2364 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2365 Details : The client ptlrpc code may be trying to reconnect to a down
2366 server before a previous connection attempt has timed out.
2367 Increase the reconnect interval to be longer than the connection
2368 timeout interval to avoid sending duplicate connections to
2372 Frequency : echo_client brw_test command
2374 Description: fix echo_client to work with OST preallocated code
2375 Details : OST preallocation code (5137) didn't take echo_client IO path
2376 into account: echo_client calls filter methods outside of any
2377 OST thread and, hence, there is no per-thread preallocated
2378 pages and buffers to use. Solution: hijack pga pages for IO. As
2379 a byproduct, this avoids unnecessary data copying.
2383 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2384 Description: Client can oops in mdc_commit_close() after open replay
2385 Details : It was possible for the MDS to return an open request with no
2386 transaction number in mds_finish_transno() if the client was
2387 evicted, but without actually returning an error. Clients
2388 would later try to replay that open and may trip an assertion
2389 Simplify the client close codepath, and always return an error
2390 from the MDS in case the open is not successful.
2393 Frequency : rare, 2.6 OSTs only
2395 Description: OST may deadlock under high load on fragmented files
2396 Details : If there was a heavy load and highly-fragmented OST filesystems
2397 it was possible to have all the OST threads deadlock waiting on
2398 allocation of biovecs, because the biovecs were not released
2399 until the entire RPC IO was completed. Instead, release biovecs
2400 as soon as they are complete to ensure forward IO progress.
2402 Severity : enhancement
2404 Description: Support for specifying external journal device at mount
2405 Details : If an OST or MDS device is formatted with an external journal
2406 device, this device major/minor is stored in the ext3 superblock
2407 and may not be valid for failover. Allow detecting and
2408 specifying the external journal at mount time.
2413 Description: Mounting an MDS with pending unlinked files may cause oops
2414 Details : target_finish_recovery() calls mds_postrecov() which returned
2415 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2416 considers this an error and immediately begins cleaning up the
2417 lov, just after starting the mds_lov process
2419 Severity : enhancement
2421 Description: Implement 'lfs df' to report actual free space on per-OST basis
2422 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2423 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2424 usage of objects; '-h' to report in human readable format.
2426 ------------------------------------------------------------------------------
2428 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2435 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2436 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2437 at startup during orphan unlinking. Instead, skip buddy update
2438 before setup, as it will be regenerated after recovery is complete.
2443 Description: If an OST is inactive, its locks might reference stale inodes.
2444 Details : lov_change_cbdata() must iterate over all namespaces, even if
2445 they are inactive to clear inode references from the lock.
2447 Severity : enhancement
2448 Frequency : occasional, if non-standard max_dirty_mb used
2450 Description: Client will block write RPCs if not enough grant
2451 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2452 then the client will block writes while waiting for another RPC
2453 to complete instead of consuming its dirty limit. With change
2454 we get improved performance when max_dirty_mb is small.
2456 Severity : enhancement
2457 Bugzilla : 3389, 6253
2458 Description: Add support for supplementary groups on the MDS.
2459 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2460 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2461 lookups for user supplementary groups into a cache.
2465 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2466 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2467 for liblustre O_EXCL is enforced by MDS.
2471 Description: readdir never returns NULL in liblustre.
2472 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2473 dentry in struct dirent.
2477 Frequency : liblustre only, depends on application IO pattern
2478 Description: liblustre clients evicted if not contacting servers
2479 Details : Don't put liblustre clients into the ping_evictor list, so
2480 they will not be evicted by the pinger ever.
2482 Severity : enhancement
2484 Description: Add ability to evict clients by NID from MDS.
2485 Details : By echoing "nid:$NID" string into
2486 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2487 $NID would be instantly evicted from this MDS and from all active
2488 OSTs connected to it.
2492 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2493 Details : lookup_it_finish() used to query file size from OSTs that was not
2498 Description: service threads change working directory to that of init
2499 Details : Starting lustre service threads may pin the working directory
2500 of the parent thread, making that filesystem busy. Threads
2501 now change to the working directory of init to avoid this.
2505 Frequency : during shutdown only
2506 Description: shutdown with a failed MDS or OST can cause unmount to hang
2507 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2512 Frequency : 2.6 only
2513 Description: chmod/chown may include an extra supplementary group
2514 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2515 supplementary group and if none is specified this is used.
2518 Bugzilla : 5479 (6816)
2520 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2521 Details : If lookup is in progress on a file that is unlinked we might try
2522 to revalidate the inode and fail in revalidate after lookup is
2523 complete and ll_file_open() enqueues the open again but
2524 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2527 Frequency : always, if lconf --abort_recovery used
2529 Description: lconf --abort_recovery fails with 'Operation not supported'
2530 Details : lconf was attempting to abort recovery on the MDT device and not
2533 ------------------------------------------------------------------------------
2535 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2540 Frequency : rare (only unsupported configurations with a node running as an
2542 Bugzilla : 6514, 5137
2543 Description: Mounting a Lustre file system on a node running as an OST could
2545 Details : OSTs now preallocates memory needed to write out data at
2546 startup, instead of when needed, to avoid having to
2547 allocate memory in possibly low memory situations.
2548 Specifically, if the file system is mounted on on OST,
2549 memory pressure could force it to try to write out data,
2550 which it needed to allocate memory to do. Due to the low
2551 memory, it would be unable to do so and the node would
2552 become unresponsive.
2554 Severity : enhancement
2556 Description: Addition of lconf --service command line option
2557 Details : lconf now accepts a '--service <arg>' option, which is
2558 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2560 Severity : enhancement
2562 Description: Failover mode is now the default for OSTs.
2563 Details : By default, OSTs will now run in failover mode. To return to
2564 the old behaviour, add '--failout' to the lmc line for OSTs.
2566 Severity : enhancement
2568 Description: Health checks are now provided for MDS and OSTs
2569 Details : Additional detailed health check information on MSD and OSTs
2570 is now provided through the procfs health_check value.
2573 Frequency : occasional, depends on IO load
2575 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2576 after numerous create/delete cycles
2577 Details : The ext3 inode allocation policy would not allocate new inodes
2578 very well on the OSTs because there are no new directories
2579 being created. Instead we look for groups with free space if
2580 the parent directories are nearly full.
2585 Description: Network or server problems during mount may cause partially
2586 mounted clients instead of returning an error.
2587 Details : The config llog parsing code may overwrite the error return
2588 code during mount error handling, returning success instead
2594 Description: MDS can fail to allocate large reply buffers
2595 Details : After long uptimes the MDS can fail to allocate large reply
2596 buffers (e.g. zconf client mount config records) due to memory
2597 fragmentation or consumption by the buffer cache. Preallocate
2598 some large reply buffers so that these replies can be sent even
2599 under memory pressure.
2603 Frequency : rare (liblustre)
2604 Description: fsx running with liblustre complained that using truncate() to
2605 extend the file doesn't work. This patch corrects that issue.
2606 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2607 ATTR_SIZE and lsm use in llu_setattr_raw.
2611 Frequency : rare, only 2.6 kernels
2612 Description: Unusual file access patterns on the MDS may result in inode
2613 data being lost in very rare circumstances.
2614 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2615 code in the 2.6 kernel caused the inode and/or EA data not to
2616 be read from disk, causing single-file corruption.
2620 Frequency : rare, only 2.6 filesystems using extents
2621 Description: Heavy concurrent write and delete load may cause data corruption.
2622 Details : It was possible under high-load situations to have an extent
2623 metadata block in the block device cache from a just-unlinked
2624 file overwrite a newly-allocated data block. We now unmap any
2625 metadata buffers that alias just-allocated data blocks.
2629 Frequency : filesystems with default stripe_count larger than 77
2630 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2631 Details : lconf specifies an inode size of 4096 bytes when the default
2632 stripe_count is larger than 77. This conflicts with the default
2633 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2634 this case to avoid pinning too much memory for large EAs.
2636 ------------------------------------------------------------------------------
2638 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2643 Frequency : rare (extremely heavy IO load with hundreds of clients)
2645 Description: Client is evicted, gets IO error writing to file
2646 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2647 caused clients to be evicted for AST timeouts. The fixes in
2648 bug 5192 mean we no longer need to have such short AST timeouts
2649 so ldlm_timeout has been increased.
2652 Frequency : occasional during --force or --failover shutdown under load
2653 Bugzilla : 5949, 4834
2654 Description: Server oops/LBUG if stopped with --force or --failover under load
2655 Details : a collection of import/export refcount and cleanup ordering
2656 issues fixed for safer force cleanup
2659 Frequency : only filesystems larger than 120 OSTs
2660 Bugzilla : 5990, 6223
2661 Description: lfs getstripe would oops on a very large filesystem
2662 Details : lov_getconfig used kfree on vmalloc'd memory
2665 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2666 Bugzilla : 6242, 6243
2667 Description: reading from files that had been truncated to a non-zero size
2668 but never opened returned no data
2669 Details : ll_file_read() reads zeros from no-object files to EOF
2674 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2675 Details : There's little harm in aborting MDS/OSS recovery and letting it
2676 try again, so I removed the LASSERT and return an error instead.
2678 Severity : enhancement
2680 Description: New debugging infrastructure for tracking down data corruption
2681 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2682 (b) cover more of the client-side code path, and (c) try to narrow
2683 down where problems occurred
2687 Bugzilla : 3819, 4364, 4397, 6313
2688 Description: Racing close and eviction MDS could cause assertion in mds_close
2689 Details : It was possible to get multiple mfd references during close and
2690 client eviction, leading to one thread referencing a freed mfd.
2692 Severity: : enhancement
2693 Bugzilla : 3262, 6359
2694 Description: Attempts to reconnect to servers are now more aggressive.
2695 Details : This builds on the enhanced upcall-less recovery that was added
2696 in 1.4.2. When trying to reconnect to servers, clients will
2697 now try each server in the failover group every 10 seconds. By
2698 default, clients would previously try one server every 25 seconds.
2703 Description: After recovery, certain operations trigger a failed
2704 assertion on a client.
2705 Details : Failing over an mds, using lconf -d --failover, while a
2706 client was doing a readdir() call would cause the client to
2707 LBUG after recovery completed and the readdir() was resent.
2709 Severity : enhancement
2711 Description: Default groups are now added by lconf
2712 Details : You can now run lconf --group <servicename> without having to
2713 manually add groups with lmc.
2716 Frequency : occasional
2718 Description: Nodes with an elan id of 0 trigger a failed assertion
2721 Frequency : always when accessing e.g. tty/console device nodes
2723 Description: tty and some other devices nodes cannot be used on lustre
2724 Details : file's private_data field is used by device data and lustre
2725 values in there got lost. New field was added to struct file to
2726 store fs-specific private data.
2729 Frequency : when exporting Lustre via NFS
2731 Description: NFSD failed occasionally when looking up a path component
2732 Details : NFSD is looking up ".." which was broken in ext3 directories
2733 that had grown large enough to become hashed.
2736 Frequency : Clusters with multiple interfaces not on the same subnet
2738 Description: Nodes will repeatedly try to reconnect to an interface which it
2739 cannot reach and report an error to the log.
2740 Details : Extra peer list entries will be created by lconf with some peers
2741 unreachable. lconf now validates the peer before adding it.
2744 Frequency : Only if a default stripe is set on the filesystem root.
2746 Description: Setting a default stripe on the filesystem root prevented the
2747 filesystem from being remounted.
2748 Details : The client was sending extra request flags in the root getattr
2749 request and did not allocate a reply buffer for the dir EA.
2752 Frequency : occasional, higher if lots of files are accessed by one client
2753 Bugzilla : 6159, 6097
2754 Description: Client trips assertion regarding lsm mismatch/magic
2755 Details : While revalidating inodes the VFS looks up inodes with ifind()
2756 and in rare cases can find an inode that is being freed.
2757 The ll_test_inode() code will free the lsm during ifind()
2758 when it finds an existing inode and then the VFS later attaches
2759 this free lsm to a new inode.
2763 Bugzilla : 6422, 7030
2764 Description: MDS deadlock between mkdir and client eviction
2765 Details : Creating a new file via mkdir or mknod (starting a transaction
2766 and getting the ns lock) can deadlock with client eviction
2767 (gets ns lock and trying to finish a synchronous transaction).
2770 Frequency : occasional
2771 Description: While starting a server, the fsfilt_ext3 module could not be
2773 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2774 kernels. Previously, lconf would still use the ext3 name
2775 when trying to load modules. Now, it will correctly use
2776 ext3 on 2.4 and ldiskfs on 2.6.
2778 Severity : enhancement
2779 Description: The default stripe count has been changed to 1
2780 Details : The interpretation of the default stripe count (0, to lfs
2781 or lmc) has been changed to mean striping across a single
2782 OST, rather than all available. For general usage we have
2783 found a stripe count of 1 or 2 works best.
2785 Severity : enhancement
2786 Description: Add support for compiling against Cray portals.
2787 Details : Conditional compiling for some areas that are different
2791 Frequency : occasional
2792 Bugzilla : 6409, 6834
2793 Description: Creating files with an explicit stripe count may lead to
2794 a failed assertion on the MDS
2795 Details : If some OSTs are full or unavailable, creating files may
2796 trigger a failed assertion on the MDS. Now, Lustre will
2797 try to use other servers or return an error to the
2801 Frequency : occasional
2803 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2804 caused lock timeout during readahead (since 1.4.2).
2805 Details : Processes doing readahead might match a lock that hasn't been
2806 granted yet if there are overlapping and conflicting lock
2807 requests. The readahead process waits on ungranted lock
2808 (original lock is CBPENDING), while OST waits for that process
2809 to cancel CBPENDING read lock and eventually evicts client.
2811 Severity : enhancement
2813 Description: Initial enabling of flock support for clients
2814 Details : Implements fcntl advisory locking and file status functions.
2815 This feature is provided as an optional mount flag (default
2816 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2817 locking are implemented yet, and those that are are not guaranteed
2818 to be completely correct in production environments.
2819 mount -t lustre -o [flock|noflock] ...
2822 Frequency : occasional
2824 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2825 assertion in the ext3 JBD (journaling) layer.
2826 Details : The b_committed_data struct is protected by the big kernel lock
2827 in 2.4 kernels, serializing journal_commit_transaction() and
2828 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2829 access to this struct. In 2.6 kernels there is finer grained
2830 locking to improve SMP performance of the JBD layer.
2834 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2839 Description: merge and cleanup kernel patches.
2842 -----------------------------------------------------------------------------
2844 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2846 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2847 versions of Lustre. Please update all servers and clients to
2848 version 1.4.2 or later at the same time. You must also run
2849 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2850 to update the configuration logs.
2852 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2853 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2854 - swab llog records of type '0' so we get proper header size/idx (5861)
2855 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2856 - fix rename of one directory over another leaking an inode (5953)
2857 - avoid SetPageDirty on 2.6 (5981)
2858 - don't re-add just-being-destroyed locks to the waiting list (5653)
2859 - when creating new directories, inherit the parent's custom
2860 striping settings if present parent (3048)
2861 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2862 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2863 - don't hold client locks on temporary worklist from l_lru (5666)
2864 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2865 - automatically evict dead clients (5921)
2866 - Update file size properly in create+truncate+fstat case (6196)
2867 - Do not unhash mountpoint dentries, do not allow removal of
2869 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2870 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2871 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2872 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2873 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2874 - Correctly handle reads of files with no objects (6243)
2875 - lctl recover will also mark a device active if deactivate used (5933)
2877 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2878 - allow --write-conf on an MDS with different nettype than client (5619)
2879 - don't write config llogs to MDS for mounts not from that MDS (5617)
2880 - lconf should create multiple TCP connections from a client (5201)
2881 - init scripts are now turned off by default; run chkconfig --on
2882 lustre and chkconfig --on lustrefs to use them
2883 - upcalls are no longer needed for clients to recover to failover
2885 - add --abort-recovery option to lconf to abort recovery on device
2887 - add support for an arbitrary number of OSTs (3026)
2888 - Quota support protocol changes.
2889 - forward compatibility changes to wire structs (6007)
2890 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2891 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2892 - improved "lustre status" script
2893 - initialize blocksize for non-regular files (6062)
2894 - added --disable-server and --disable-client configure options (5782)
2895 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2896 - Vanilla 2.4.29 support
2897 - increase maximum number of obd devices to 520 (6242)
2898 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2899 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2900 SLES 9 2.6.5 kernels
2901 - limit stripes per file to 160 (the maximum EA size) (6093)
2903 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2906 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2907 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2908 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2909 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2910 - small ext3 extents cleanups and fixes (5733)
2911 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2912 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2913 - avoid needless client->OST connect, fix handle mismatch (5317)
2914 - fix DLM error path that led to out-of-sync client, long delays (5779)
2915 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2916 - fix several locking issues related to i_size (5492,5624,5654,5672)
2917 - don't move pending lock onto export if it is already evicted (5683)
2918 - fix kernel oops when creating .foo in unlinked directory (5548)
2919 - fix deadlock in obdfilter statistics vs. object create (5811)
2920 - use time_{before,after} to avoid timer jiffies wrap (5882)
2921 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2922 - Do not leak request if server was not able to process it (5154)
2923 - If mds_open unable to find parent dir, make that negative lookup(5154)
2924 - don't create new directories with extent-mapping (5909, 5936)
2926 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2927 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2928 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2929 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2930 - support for RHEL4 kernel on i686 (5773)
2931 - provide error messages when incompatible logs are encountered (5898)
2933 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2934 * version 1.4.0.10 (1.4.1 release candidate 1)
2936 - don't keep a lock reference when lock is not granted (4238)
2937 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2938 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2939 - drop import inflight refcount on signal_completed_replay error (5255)
2940 - unlock page after async write error during send (3677)
2941 - handle missing objects in filter_preprw_read properly (5265)
2942 - no transno return for symlink open, don't save no-trasno open (3440)
2943 - don't try to complete elan receive that already failed (4012)
2944 - free RPC server reply state on error (5406)
2945 - clean up thread from ptlrpc_start_thread() on error (5160)
2946 - readahead could read extra page into cache that wasn't ejected (5388)
2947 - prevent races in class_attach/setup/cleanup/detach (5260)
2948 - don't dereference de->d_inode after l_dput of de (5458)
2949 - use "int" for stripe value returned from lock_to_stripe (5544)
2950 - mballoc allocation and error-checking fixes in 2.6 (5504)
2951 - block device patches to fix I/O request sizes in 2.6 (5482)
2952 - look up hostnames for IB nals (5602)
2953 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2954 - don't start multiple acceptors for the same port (5277)
2955 - fix incorrect LASSERT in mds_getattr_name (5635)
2956 - export a proc file for general "ping" checking (5628)
2957 - fix "lfs check" to not block when the MDS is down (5628)
2959 - service request history (4965)
2960 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2961 - create an "evict_client" /proc entry on OSTs, like the MDS has
2962 - fix mount usage message, return errors per mount(8) (5168)
2963 - change grep [] to grep "[]" in tests so they work in more UMLs
2964 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2965 - remove ancient LOV_MAGIC_V0 EA support (5047)
2966 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2967 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2968 - allow readahead allocations to fail when low on memory (5383)
2969 - mmap locking landed again, after considerable improvement (2828)
2970 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2972 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2975 - send OST transaction number in read/write reply to free req (4966)
2976 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2977 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2978 - the watchdog thread now runs as interruptible (5246)
2979 - flock/lockf fixes (but it's still disabled, pending 5135)
2980 - don't use EXT3 constants in llite code (5094)
2981 - memory shortage at startup could cause assertion (5176)
2983 - reorganization of lov code
2984 - single portals codebase
2986 - add extents/mballoc support (5025)
2987 - direct I/O reads in the obdfilter (4048)
2988 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2990 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2993 - send OST transaction number in read/write reply to free req (4966)
2994 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2995 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2996 - the watchdog thread now runs as interruptible (5246)
2997 - handle missing objects in filter_preprw_read properly (5265)
2998 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2999 - drop import inflight refcount on signal_completed_replay error (5255)
3000 - unlock page after async write error during send (3677)
3001 - return original error code on reconstructed replies (3761)
3002 - no transno return for symlink open, don't save no-trasno open (3440)
3004 - add pid to ldlm debugging output (4922)
3005 - bump the watchdog timeouts -- we can't handle 30sec yet
3006 - extra debugging for orphan dentry/inode bug (5259)
3008 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3011 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3012 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3013 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3014 - fix NULL dereference in /proc/sys/portals/routes (4827)
3015 - allow failed mdc_close() operations to be interrupted (4561)
3016 - stop precreate on OST before MDS would time out on it (4778)
3017 - don't send partial-page writes before EOF from client (4410)
3018 - discard client grant for sub-page writes on large-page clients (4520)
3019 - don't free dentries not owned by NFS code, check generation (4806)
3020 - fix lsm leak if mds_create_objects() fails (4801)
3021 - limit debug_daemon file size, always print CERROR messages (4789)
3022 - use transno after validating reply (3892)
3023 - process timed out requests if import state changes (3754)
3024 - update mtime on OST during writes, return in glimpse (4829)
3025 - add mkfsoptions to LDAP (4679)
3026 - use ->max_readahead method instead of zapping global ra (5039)
3027 - don't interrupt __l_wait_event() during strace
3029 - add software watchdogs to catch hung threads quickly (4941)
3030 - make lustrefs init script start after nfs is mounted
3031 - fix CWARN/ERROR duplication (4930)
3032 - return async write errors to application if possible (2248)
3033 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3034 - print NAL number in %x format (4645)
3035 - update barely-supported suse-2.4.21-171 series (4842)
3036 - support for sles 9 %post scripts
3037 - support for building 2.6 kernel-source packages
3038 - support for sles km_* packages
3040 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3043 - ignore -ENOENT errors in osc_destroy (3639)
3044 - notify osc create thread that OSC is being cleaned up (4600)
3045 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3046 - reconstruct ost_handle() like mds_handle() (4657)
3047 - create a new thread to do import eviction to avoid deadlock (3969)
3048 - let lconf resolve symlinked-to devices (4629)
3049 - don't unlink "objects" from directory with default EA (4554)
3050 - hold socknal file ref over connect in case target is down (4394)
3051 - allow more than 32000 subdirectories in a single directory (3244)
3052 - fix blocks count for O_DIRECT writes (3751)
3053 - OST returns ENOSPC from object create when no space left (4539)
3054 - don't send truncate RPC if file size isn't changing (4410)
3055 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3056 - bind to privileged port in socknal and tcpnal (3689)
3058 - rate limit CERROR/CWARN console message to avoid overload (4519)
3059 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3060 - basic mmap support (3918)
3061 - kernel patch series update from b1_4 (4711)
3063 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3066 - avoid crash during MDS cleanup with OST shut down (2775)
3067 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3068 - don't use bad inodes on the MDS (3744)
3069 - dynamic object preallocation to improve recovery speed (4236)
3070 - don't hold spinlock over lock dumping or change debug flags (4401)
3071 - don't zero obd_dev when it is force cleaned (3651)
3072 - print grants to console if they go negative (4431)
3073 - "lctl deactivate" will stop automatic recovery attempts (3406)
3074 - look for existing locks in ldlm_handle_enqueue() (3764)
3075 - don't resolve lock handle twice in recovery avoiding race (4401)
3076 - revalidate should check working dir is a directory (4134)
3078 - don't always mark "slow" obdfilter messages as errors (4418)
3080 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3083 - don't close LustreDB during write_conf until it is done (3860)
3084 - fix typo in lconf for_each_profile (3821)
3085 - allow dumping logs from multiple threads at one time (3820)
3086 - don't allow multiple threads in OSC recovery (3812)
3087 - fix debug_size parameters (3864)
3088 - fix mds_postrecov to initialize import for llog ctxt (3121)
3089 - replace config semaphore with spinlock (3306)
3090 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3091 - don't allow enqueue to complete on a destroyed export (3822)
3092 - down write_lock before checking llog header bitmap (3825)
3093 - recover from lock replay timeout (3764)
3094 - up llog sem before sending rpc (3652)
3095 - reduce ns lock hold times when setting kms (3267)
3096 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3097 - fix NULL deref and obd_dev leak on setup error (3312)
3098 - replace some LBUG about llog ops with error handling (3841)
3099 - don't match INVALID dentries from d_lookup and spin (3784)
3100 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3101 - fix invalid assertion in ptlrpc_set_wait (3880)
3103 - add libwrap support for the TCP acceptor (3996)
3104 - add /proc/sys/portals/routes for non-root route listing (3994)
3105 - allow setting MDS UUID in .xml (2580)
3106 - print the stack of a process that LBUGs (4228)
3108 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3111 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3112 - make sure to unset replay flag from failed open requests (3440)
3113 - if default stripe count is 0, use OST count for inode size (3636)
3114 - update parent mtime/ctime on client for create/unlink (2611)
3115 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3116 - free recovery state on server during a forced cleanup (3571)
3117 - unregister_reply for resent reqs (3063)
3118 - loop back devices mounting and status check on 2.6 (3563)
3119 - fix resource-creation race that can provoke i_size == 0 (3513)
3120 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3121 - more debugging for page-accounting assertion (3746)
3122 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3123 - avoid deadlock after precreation failure (3758)
3124 - fix race and lock order deadlock in orphan handling (3450, 3750)
3125 - add validity checks when grabbing inodes from l_ast_data (3599)
3127 - add /proc/.../recovery_status to obdfilter (3428)
3128 - lightweight CDEBUG infrastructure, debug daemon (3668)
3129 - change default OSC RPC parameters to be better on small clusters
3130 - turn off OST read cache for files smaller than 32MB
3131 - install man pages and include them in rpms (3100)
3132 - add new init script for (un)mounting lustre filesystems (2593)
3133 - run chkconfig in %post for init scripts (3701)
3134 - drop scimac NAL (unmaintained)
3136 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3139 - clean kiobufs before and after use (3485)
3140 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3141 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3142 - change init script to fail more clearly if not run as root (1528)
3143 - allow clients to reconnect during replay (1742)
3144 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3145 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3146 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3147 - fix cond_resched (3554)
3148 - only evict unfinished clients after recovery (3515)
3149 - allow bulk resend, prevent data loss (3570)
3150 - dynamic ptlrpc request buffer allocation (2102)
3151 - don't allow unlinking open directory if it isn't empty (2904)
3152 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3153 - remove extraneous obd dereference causing LASSERT failure (3334)
3154 - don't use get_cycles() when creating temp. files on the mds (3156)
3155 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3156 - handle EEXIST for set-stripe, set proper directory name (3336)
3158 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3159 - fix ksocknal_fmb_callback() error messages (2918)
3161 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3164 - don't copy lvb into (possibly NULL) reply on error (2983)
3165 - don't deref dentry after dput, don't free lvb on error (2922)
3166 - use the kms to determine writeback rpc length (2947)
3167 - increment oti_logcookies when osc is inactive (2948)
3168 - update client's i_blocks count via lvb messages (2543)
3169 - handle intent open/close of special files properly (1557)
3170 - mount MDS with errors=remount-ro, like obdfilter (2009)
3171 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3172 - don't use cancelling-locks' kms values (2947)
3173 - use highest lock extent for kms, not last one (2925)
3174 - don't dereference ERR_PTR() dentry in error handling path (3107)
3175 - fix thread race in portals_debug_dumplog() (3122)
3176 - create lprocfs device entries at setup instead of at attach (1519)
3177 - common AST error handler, don't evict client on completion race (3145)
3178 - zero nameidata in detach_mnt in 2.6 (3118)
3179 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3180 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3181 - initialize RPC timeout timer earlier for 2.6 (3219)
3182 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3183 - print nal/nid for unknown nid (3258)
3184 - additional checks for oscc recovery before doing precreate (3284)
3185 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3186 - don't crash in mdc_close for bad permissions on open (3285)
3187 - zero i_rdev for non-device files (3147)
3188 - clear page->private before handing to FS, better assertion (3119)
3189 - tune the read pipeline (3236)
3190 - fix incorrect decref of invalidated dentry (2350)
3191 - provide read-ahead stats and refine rpc in flight stats (3328)
3192 - don't hold journal transaction open across create RPC (3313)
3193 - update atime on MDS at close time (3265)
3194 - close LDAP connection when recovering to avoid server load (3315)
3195 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3196 - don't leak open file on MDS after open resend (3325)
3197 - serialize filter_precreate and filter_destroy_precreated (3329)
3198 - loop device shouldn't call sync_dev() for nul device (3092)
3199 - clear page cache after eviction (2766)
3200 - resynchronize MDS->OST in background (2824)
3201 - refuse to mount the same filesystem twice on same mountpoint (3394)
3202 - allow llmount to create routes for mounting behind routers (3320)
3203 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3204 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3205 - fix rare problem with rename on htree directories (3417)
3207 - allow default OST striping configuration per directory (1414)
3208 - fix compilation for qswnal for 2.6 kernels (3125)
3209 - increase maximum number of MDS request buffers for large systems
3210 - change liblustreapi to be useful for external progs like lfsck (3098)
3211 - increase local configuration timeout for slow disks (3353)
3212 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3214 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3217 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3218 - don't overwrite extent policy data in reply if lock was blocked (2901)
3219 - drop filter export grants atomically with removal from device (2663)
3220 - del obd_self_export from work_list in class_disconnect_exports (2908)
3221 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3222 - swab reply message in mdc_close, other PPC fixes (2464)
3223 - fix destroying of named logs (2325)
3224 - overwrite old logs when running lconf --write_conf (2264)
3225 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3226 - fix race in target_handle_connect (2898)
3227 - mds_reint_create() should take same inode create lock (2926)
3228 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3229 - don't close files for self_export to avoid uninitialized obd (2936)
3230 - allow MDS with the same name as client node (2939)
3231 - hold dentry reference for closed log files for unlink (2325)
3232 - reserve space for all logs during transactions (2059)
3233 - don't evict page beyond end of stripe extent (2925)
3234 - don't oops on a deleted current working directory (2399)
3235 - handle hard links to targets without a parent properly (2517)
3236 - don't dereference NULL lock when racing during eviction (2867)
3237 - don't grow lock extents when lots of conflicting locks (2919)
3239 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3242 - account for cache space usage on clients to avoid data loss (974)
3243 - lfsck support in lustre kernel code (2349)
3244 - reduce journal credits needed for BRW writes (2370)
3245 - orphan handling to avoid losing space on client/server crashes
3246 - ptlrpcd can be blocked, stopping ALL progress (2477)
3247 - use lock value blocks to assist in proper KMS, faster stat (1021)
3248 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3249 - recovery for initial connections (2355)
3250 - fixes for mds_cleanup_orphans (1934)
3251 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3252 - block all file creations until orphan recovery completes (1901)
3253 - client remove rq_connection from request struct (2423)
3254 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3255 - recovery timer race (2670)
3256 - mdc_close recovey bug (2532)
3257 - ptlrpc cleanup bug (2710)
3258 - mds timeout on local locks (2588)
3259 - namespace lock held during RPCs (2431)
3260 - handle interrupted sync write properly (2503)
3261 - don't try to handle a message that hasn't been replied to (2699)
3262 - client assert failure during cleanup after abort recovery (2701)
3263 - leak mdc device after failed mount (2712)
3264 - ptlrpc_check_set allows timedout requests to complete (2714)
3265 - wait for inflight reqs when ptlrpcd finishes (2710)
3266 - make sure unregistered services are removed from the srv_list
3267 - reset bulk XID's when resending them (caught by 1138 test)
3268 - unregister_bulk after timeout
3269 - fix lconf error (2694)
3270 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3271 - readahead locks pages, leaves pending causing memory pressure (2673)
3272 - increase OST request buffers to 4096 on large machines (2729)
3273 - fix up permission of existing directories in simple_mkdir (2661)
3274 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3275 - don't assign transno to errored transactions (2742)
3276 - don't delete objects on OST if given a bogus objid from MDS (2751)
3277 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3278 - if rq_no_resend, then timeout request after recovery (2432)
3279 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3280 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3281 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3282 - MDS recovery shouldn't delete objects that it has given out (2730)
3283 - if enqueue arrives after completion, don't clobber LVB (2819)
3284 - don't unlock pages twice when trigger_group_io returns error (2814)
3285 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3286 - don't write pages to disk if there was an error (1450)
3287 - don't ping imports that have recovery disabled (2676)
3288 - take buffered bytes into account when balancing socknal conn (2817)
3289 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3290 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3291 - remove little-endian swabbing of llog records (1987)
3292 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3293 - retry reposting request buffers if they fail (1191)
3294 - grow extent at grant time to avoid granting a revoked lock (2809)
3295 - lock revoke doesn't evict page if covered by a second lock (2765)
3296 - disable VM readahead to avoid reading outside lock extents (2805)
3298 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3299 - updated kernel patches for hp-2.4.20 kernel (2681)
3301 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3304 - fix truncated write corruption (2366)
3305 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3307 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3308 - ensure len(uuid) < 37 in lmc (1171)
3309 - fix ia64 OOPS in llog_test (2255)
3310 - zero end of page at obdfilter for partial page writes (2648)
3311 - don't leave stale dentries around after renames (bug 2428)
3312 - fix timeouts when evicting a client with a single lock held (2642)
3313 - set deadline for the initial HELLO message to drain (2634)
3314 - print out dotted-quad IP addresses in the socknal (2302)
3316 - additional debugging for MDS client eviction problem (2443)
3317 - fix mkfsoptions support for osts (2603, 2604)
3319 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3322 - add series for the vanilla 2.6.0 kernel
3323 - add series for the vanilla 2.4.24 kernel
3324 - add series for a cray x86/64 UL kernel drop
3325 - fix xattr patches for the vanilla 2.4.19 series
3327 - generate true UUIDs in lmc (1171)
3328 - have portals stack dumping break in UML (2466)
3329 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3330 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3331 - wake callback waiting threads on client eviction (2460)
3332 - Add --ptldebug and --subsystem to lmc (1719)
3333 - update assertion to allow safe interrupt allocation
3334 - set rq_no_resend for cancel requests (2432)
3335 - recalculate ptlrpcd timeout after resend (2494)
3336 - call vfs_rmdir when removing pending directories (2368)
3337 - fix renaming a file to itself (2429)
3338 - lmc creates a default one-stripe lov (2454)
3339 - expand procfs space to handle large clusters (2326)
3340 - increase UML stack to avoid overflow
3341 - update lconf's list of debug and subsystem masks
3342 - fix lfs find --obd (2510)
3343 - /proc tunable for disabling filter read caching (2591)
3344 - stop rpm packages from altering slapd.conf (2301)
3345 - disable nagle in the socknal under 0conf (2578)
3346 - choose mds inode size based on stripe count (2572)
3347 - fix kernel-source rpm problems (2516)
3349 - add --disable-doc to avoid pdf generation (2421)
3350 - update documentation, tests, type-os, comments
3351 - avoid format warnings on ia64
3352 - remove the TOE NAL
3353 - tiny code cleanups by removing unused fields
3355 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3358 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3359 - remove the most blatant lies from BUILDING file (2371)
3360 - change default debug level to reasonable production setting
3361 - reduce client side cache size to reduce cache flush time
3362 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3363 - make TCP zerocopy and pinger support enabled by default (2476)
3364 - sync writes completed after process exits caused crashes (2319)
3365 - maintain correct mount count on the MDS (2356)
3366 - backout 1557, because 2316 wasn't really fixed
3367 - better file I/O statistics gathering in /proc
3368 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3369 - another kernel patch to fix zero-copy TCP function export
3370 - don't take duplicate lock when processing re-sent getattr (2420)
3371 - lctl uses obd_self_export instead of creating new conn (2353)
3372 - MDS/OST recovery case which requires object creation asserted (2425)
3373 - move lfs from /usr/sbin to /usr/bin in packages
3374 - fix race between mds_client_add and mds_client_free (2417)
3375 - use kmalloc instead of slabs in portals (2430)
3376 - don't create duplicate records when a failover MDS is present (2442)
3377 - remove unnecessary mount age check (2332)
3378 - don't remove directory inodes from locks prematurely (2451)
3379 - don't break if MDS service name is the same as hostname (2103)
3380 - fix races in client write RPC generation when cache full (2482)
3382 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3385 - remove now-unused request->rq_obd (278)
3386 - if an allocation fails, print out how much memory we've used (1933)
3387 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3388 - add the "configurable stack size" patch to most series files (1256)
3389 - ability to write large log records, for 100+ OST configs (2306)
3390 - fix NULL deref when filter_prep fails (2314)
3391 - fix operator precedence error in filter_sync
3392 - dynamic allocation of socknal TX descriptors (2315)
3393 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3394 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3395 - ensure that CWARN messages in Portals always get to the syslog
3396 - __init/__exit are not for prototype decls (ldlm_init/exit)
3397 - x86-64 compile warning fixes
3398 - fix gateway LMC keyword conflict (2318)
3399 - fix MDS lock inversions in getattr/reint paths (1844)
3400 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3401 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3402 - DLM race condition prevented some lock evictions (2328)
3403 - ENOMEM detection and retry on socknal sends (2230)
3404 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3405 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3406 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3407 - better error messages when a client is rejected during recovery (1505)
3408 - avoid cancelling locks which were never granted, after failure (2330)
3409 - fix i_sem/journal inversion in mds_client_add (2333)
3410 - fix truncate/getattr lock cycle deadlock (2334)
3411 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3412 - fix two rare exit paths which could leak an l_lock() ref (2321)
3413 - fencepost error in MDS/OST orphan recovery (2226)
3414 - make log record alignment 8 bytes (1988)
3415 - lstripe now fails when requested offset > ost_count (2237)
3416 - ensure that all kernel series have a complete list.h (1607)
3417 - fix crashes in special-file operations (2316)
3418 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3420 - allow configurable automake binary, for testing new versions
3421 - small update to the lfs documentation
3423 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3425 * fix negative export reference count in fsfilt_sync (2312)
3427 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3428 * release candidate 0.9.1
3430 - orphans are moved into the PENDING directory for possible recovery
3431 - replayed opens now open by fid for orphan/rename safety (1042)
3432 - last close of an orphan inode generates a transno (683)
3433 - chdir() and mount() now pin the directory entry (1020)
3434 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3435 - discard very old requests without processing them (1502)
3436 - remove some common, well-understood CERRORs (1505)
3437 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3438 - clear "grant" flags in OST replies until OST grant code lands (1644)
3439 - fix read performance by not clobbering i_blksize on client (1598)
3440 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3441 - make LRU size a /proc tunable, clears locks when reduced (707)
3442 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3443 - server locks take references on exports now (1558)
3444 - build fixes for 2.4.20-rh trees (1663)
3445 - return an error from lov_create if all OSCs are inactive (1751)
3446 - fix import levels when a reconnect happens without a timeout (1597)
3447 - exit early from mds_open if we get a lookup error (1749)
3448 - partial page read at EOF wouldn't wait for disk before sending (1642)
3449 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3450 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3451 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3452 - write/truncate lock inversion (1639)
3453 - Don't auto-load obdclass, portals modules during cleanup (1495)
3454 - fix timestamps from jumping to "now" (1763)
3455 - extra journal assertions (1648)
3456 - add an extra multiunlink test (1771)
3457 - fix read_record/write_record API (1776)
3458 - fix leak of offset_extent, possible incorrect i_size later (1772)
3459 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3460 - Debugging for the kqswnal_get_idle_tx problems (1820)
3461 - Allow recovery to be attempted multiple times (1536)
3462 - Write out MDS last_rcvd file after it is first created (1600)
3463 - Fix tx_descriptor leak in failed transmit situations (1827)
3464 - ext3 journaling fixes for assertion failure after IO error (1871)
3465 - class_export_put() on freed export after completion AST error (1896)
3466 - Fix revalidate looping in VFS (1322)
3467 - Don't access a freed export during MDS_REINT timeout (1521)
3468 - Add open-unlink recovery support on the MDS (1673,1764)
3469 - Return an error if no MDS data was read from last_rcvd (1946)
3470 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3471 - Don't setup a disk device that doesn't match exported UUID (317)
3472 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3473 - avoid committing NULL handle in force close
3474 - local.sh is now a one-stripe LOV configuration
3475 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3476 - fix ext3 htree duplicate directory entry corruption (1516)
3477 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3478 - update to new LOV EA format (2097)
3479 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3481 2003-06-15 Phil Schwan <phil@clusterfs.com>
3484 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3485 - per-import/export recovery handling (958, 931, 959)
3486 - multiple last-rcvd slots, for serving multiple FSes (949)
3487 - connections are again shared between multiple imp/exports (963, 964)
3488 - "umount -f" would hang if any requests needed to be sent (393, 978)
3489 - avoid pinning large req buffer by copying for queued messages (989)
3490 - add "uuid" to "lctl device" command to help upcalls (991)
3491 - "open" RPCs with transnos would confuse recovery counters (1037)
3492 - do proper endian conversion of all wire messages (288, 340, 891)
3493 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3494 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3495 - fix ptlrpc_connection leak in target_handle_connect (1174)
3496 - fix import refcounting bug in OST and MDS cleanup (1134)
3497 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3498 - fix very unlikely obd_types race condition (501)
3499 - remove osc_open hack for echo_client (1187)
3500 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3501 - a failure in read_inode2 leads to deadlock (1139)
3502 - cancel ack-locks as soon as transaction is committed (1072)
3503 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3504 - make sure to commitrw after any preprw to avoid deadlock (1162)
3505 - failing to execute a file in a lustre FS would lock inode (1203)
3506 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3507 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3508 - obd_uuid2tgt would walk off the end of the list (1255)
3509 - on IA64 the osc would give portals incorrect bulk size (1258)
3510 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3511 - fix lock inversion caused by new llite matching code (1282)
3512 - limit the number of dirty pages on a client to 10MB (1286)
3513 - timed out locks were not being corrected cancelled (1289)
3514 - fix O_DIRECT above 4GB on IA-32 (1292)
3515 * major user-visible changes
3516 - fail out/fail over policy now controlled by the upcall (993)
3518 - add OBD_PING to check server availability and failure (954)
3519 - lustre messages are now sent in sending host order (288, 340, 891)
3520 - add eadatalen to MDS getattr reply (340)
3521 - OST read replies may contain second buffer, with per-page status (593)
3523 2003-03-11 Phil Schwan <phil@clusterfs.com>
3526 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3527 - failed open()s could cause deadlock; fixed (867, 869)
3528 - stop cancelling OST locks when files are closed (481)
3529 - overlapping XID spaces caused network corruption (851, 853)
3530 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3531 - setattr_raw added, to do single-RPC, server-side setattrs
3532 - lmc/lconf syntax change for OST UUIDs
3533 - fix crashy race condition between ptlrpc_free_req and osc_close
3534 - don't use request in mdc_enqueue if we hit a timeout (889)
3535 - don't set the inode i_size for regular files from the MDS (896)
3536 - handle out of order completion AST (842)
3537 - don't LBUG if a lock request times out after receiving AST (913)
3538 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3539 - if a bad lock AST arrives, send an error instead of dropping entirely
3540 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3541 - fix leak in bulk IO when only partially completed (899, 900, 926)
3542 - fix O_DIRECT for ia64 (55)
3543 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3544 - C-z after timeout could hang a process forever; fixed (977)
3546 - client-side I/O cache (678, 924, 929, 941, 970)
3548 - READPAGE and SETATTRs which don't take server-side locks get
3551 2003-02-11 Phil Schwan <phil@clusterfs.com>
3554 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3555 - Fix fsfilt_extN_readpage() to read a full page of directory
3556 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3557 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3558 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3559 - Most intent operations were removed, in favour of a new RPC mode
3560 that does a single RPC to the server and bypasses most of the VFS
3561 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3562 - Aggressively cancel local locks on DLM servers
3563 - mds_reint_unlink sends EA to the client if it's the last nlink.
3564 client uses that EA to unlink OST objects.
3565 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3566 - recursive symlinks were fixed (439)
3567 - fixed NULL deref in DEBUG_REQ
3568 - filter_update_lastobjid no longer calls sync, which annoyed extN
3569 - fixed multi-client small-writes to a single file problem (445)
3570 - fixed mtime updates during file writes (607)
3571 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3572 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3573 - fixed Linux OST size reporting problem (444, 656)
3574 - OST now updates object mtime with writes or setattr (607, 619)
3575 - client verifies file size before zeroing page past EOF (445)
3576 - OST now writes last allocated objid to disk with allocation (108)
3577 - LOV on echo now works (409)
3579 - mds_reint_unlink sends a new buffer, with the EA included. this
3580 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3581 set if a regular file was being unlinked, and it was the last link
3582 - use PtlGet from the target for bulk writes (315)
3583 - OST now updates object mtime with writes or setattr (607, 619)
3584 - LDLM now has a grant-time callback to revalidate locked items, if
3586 - Many MDS operations were reorganized to combat race conditions
3588 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3589 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3591 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3594 - Fully reactivate OST imports after reconnection (512, others)
3595 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3596 - More graceful error handling for truncating on dead OST (515)
3597 - Don't error out unless we're actually accessing dead stripes (474)
3598 - Fix garbage sizes when stripes are missing (410)
3599 - LRU counters were broken, causing constant lock purge (433, 432)
3600 - garbage on read from stripes with failed OSTs (441)
3601 - mark OSCs as active before reconnecting during recovery (438)
3602 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3603 - lfind did not preserve OST order in output (443)
3604 - symlinks cause hung clients, incorrect data (439)
3605 - stop dereferencing request after dropping refcount (457)
3606 - don't LASSERT(spin_is_locked) on non-SMP (455)
3607 - fixes for many rename() bugs
3608 - fstat didn't correctly synchronize attributes (399)
3609 - server must handle lock cancellation during blocking AST prep (487)
3610 - bulk descriptors were free()d too soon (511)
3611 - fix paths in lconf, which would load incorrect modules (451, 507)
3612 - fix confusing lconf 'host not found' error message (386)
3613 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3614 - fix race condition in mdc_blocking_ast() for inode access (526)
3615 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3616 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3617 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3618 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3619 - fix lstripe in several ways (kernel side) (527)
3620 - fix request leak in ldlm_cli_enqueue (262)
3621 - incorrect OSC was marked inactive after OST failure
3622 - call mds_fs_cleanup before unmounting filesystem (524)
3623 - fix races between taking ns_lock and ldlm_lock_change_resource
3624 - fix races updating LOV export open file list
3625 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3626 - fix recovery NULL deref in ldlm_cli_cancel_unused
3627 - fix some DLM races by using new hash table for lock handles (419)
3628 - permit the client to specify desired inodes, at replay
3629 - duplicate requests when we queue them for replay reintegration
3630 - fix last_rcvd offset calculation
3631 - sync after each recovered transaction, so we always make progress
3632 - never, not always, ERESTART requests without transnos
3633 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3635 - skip replay if the MDS says that the client is already connected
3636 - don't check for a recovery-enabled export to match lctl's UUID
3637 - don't INC_USE_COUNT for phantom exports
3638 - don't crash when cleaning up phantom exports (567)
3639 - don't double-finish or set replay data for errored mdc_open requests
3640 - abort requests when they time out, so we don't get old replies
3641 - send/receive replies for AST messages again
3642 - if the client says that it doesn't have the lock, cancel it on the
3644 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3645 mark it as destroyed, it will all work out when decref is called
3646 - fix module use counts (22, 581)
3648 - ASTs now expect a reply (server cancels lock on error reply)
3650 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3653 - fix many simultaneous client startup (392)
3654 - fix dentry->d_it clobbering
3655 - credentials weren't being shipped for readdir/getattr operations
3656 - remove invalid assertions triggered during some concurrent MD
3658 - proper Lustre versions added (336, 389)
3659 - fix memory leak for create error case (398)
3660 - fix LOV locking bug that would get cli/srv out of sync
3661 - fix echo client over LOV (409)
3662 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3663 - fix double-O_EXCL intent crash (424)
3664 - avoid sending multiple lock CANCELs (352)
3666 - MDS can do multi-client recovery (modulo bugs in new code)
3668 - many updates, edits, cleanups
3670 2002-11-18 Phil Schwan <phil@clusterfs.com>
3673 - fix null d_it dereference (346)
3674 - fix full OST/dbench hang (333)
3675 - fix permission problem with file removal (286)
3676 - fix removal of OSCs from LOV when they fail
3677 - fix NULL deref during bulk timeout (214)
3678 - fix problems related to multiple filesystems on one MDS (241)
3679 - fixed serious subtle metadata locking bugs
3680 - free locks on clients when inodes are removed due to memory
3682 - fix inode pointer in lock data (285)
3683 - partial support for multiple MDS on a single host (241)
3684 - data locks weren't cancelled at clear_inode time (290, 311)
3685 - intent locks could lead to unbounded lock growth (205)
3686 - added a maximum lock count, an LRU list, and a flusher
3687 - fix multiple rename (365)
3688 - properly abstracted the echo client
3689 - OSC locked 1 byte too many; fixed
3690 - rewrote brw callback code:
3691 - fixed recovery bugs related to LOVs (306)
3692 - fixed too-many-pages-in-one-write crash (191)
3693 - fixed (again) crash in sync_io_timeout (214)
3694 - probably fixed callback-related race (385)
3696 - Add capability to MDS protocol
3697 - LDLM cancellations and callbacks on different portals
3699 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3702 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3703 - defer open object destruction to close time (601981, 138)
3704 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3705 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3706 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3707 - don't call obd_setattr in ll_file_release for destroyed objects
3708 * protocol change to lustre_msg: move |version| and add |flags|
3709 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3710 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3711 * added replay of create, unlink, link and rename operations during
3712 MDS failover; recovery should be much more robust now
3713 * remove failed OSCs from LOVs (only lov_create uses this so far)
3714 * the lustre-HOWTO was brought (more) up to date (582544)
3716 2002-10-23 Phil Schwan <phil@clusterfs.com>
3719 - in-use dentries weren't being reused properly (617851)
3720 - prevent multiple LDLM setup (599178)
3721 - fix LOV size calculations for truncate (617853)
3722 - fix client handling of MDS intent errors (POSIX)
3723 - fix permission bug in lovstripe.c test (624321)
3724 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3725 - truncate past end of file could corrupt data
3726 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3727 - a race in recovery could return ETIMEDOUT to apps (623947)
3728 - building outside the source directory was fixed
3729 * the lustre-HOWTO was brought (more) up to date (582544)
3730 * major progress was made on recovery functionality
3732 2002-10-10 Phil Schwan <phil@clusterfs.com>
3735 - recovery deadlock fix
3736 - rm -rf causes LBUG fix (617817)
3737 - file open by multiple tasks fix (618962)
3738 - directory permissions bugs (602707 and 620007)
3739 - journal_stop fixed with locking (611313)
3740 - O_APPEND failures resolved (618273, perhaps 614459)
3741 - lconf PATH fix (619770)
3742 - IA64 build fix (621450)
3743 - RPC buffer sizes scale with amount of memory
3745 2002-10-01 Phil Schwan <phil@clusterfs.com>
3748 - locks would be cancelled without throwing away data pages,
3749 resulting in inconsistent data (605627)
3750 - inode attributes were not always being refreshed (605627, 612449)
3751 - lconf now continues to cleanup after lctl reports an error
3752 - MDS now enforces user permissions (602707)
3753 - lprocfs cleanup fixed, but not yet enabled (614157)
3754 - fixed infinite server hang, should a client not respond to an AST
3755 - avoid going into recovery if user calls readlink() with a buffer
3756 that's too small (613941)
3757 - AST RPCs no longer require replies (614867) -- this may be changed
3758 - don't crash server if client sends an IOV that's too big (611336)
3759 - fixed lock conversion deadlock (611892)
3760 - fixed the following of symlinks (614622)
3761 * recovery: the server can remove locks from a client that dies, other
3762 clients can make progress
3763 * more extN patch fixes
3764 * compile-time configurable ptlrpc buffer allocations
3766 - collaborative read cache document
3767 - Lustre Lite Performance CDR document-in-progress
3769 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3772 - fix typo in patch-2.4.18
3774 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3777 - clear ptlrpc request each time in handle_incoming_request()
3778 - unlink of files now destroys the object on the OST
3780 2002-09-19 Peter Braam <braam@clusterfs.com>
3782 * add hard link support
3783 * change obdfile creation method
3784 * kernel patch changed
3786 2002-09-19 Peter Braam <braam@clusterfs.com>
3789 - stack overflow bug in extN fixed
3791 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3793 * documentation updates
3794 - add man pages for config tools
3795 - update tests/README to describe testing with new config tools
3796 - finish metadata API descriptions
3797 * bug fixes and cleanups
3798 - statfs workaround for 16TB limit
3799 - LOV stripe allocation improved, can stripe on subset of OSTs
3800 - LOV file size/IO offset was wrong for files > 4GB in size
3801 - object EA data was being dropped, caused files to be unreadable
3802 - memory overflow with non-LOV OST caused memory corruption
3803 - fixed regression tests to work with new config tools, obdfilter
3804 - fixed bug when directory size became larger than 1 block
3805 - fixed bug (for single client case) when PWD was deleted
3806 - invalidate local directory pages when doing intent-based ops
3807 - avoid LDLM oops when lock callback contained bad data
3809 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3811 * documentation updates
3812 * bug fixes and cleanups
3813 - configuration tools
3816 - 64-bit compile warnings
3817 - 64-bit internal statfs data
3819 * test_brw on persistent OST devices
3821 * lprocfs (disabled)
3823 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3825 * documentation updates
3826 * bug fixes and cleanups
3827 * configuration tools
3829 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3832 * many small fixes to 0_5_4
3833 * io/network handling
3834 * thinkos in MDS operations
3836 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3839 * crucial basic fixes to 0.5.3
3840 * IOR, Iozone work over Elan
3843 2002-08-07 Phil Schwan <phil@clusterfs.com>
3844 * version 0_5_3, our first alpha
3845 * we use the new Portals iovs
3846 * documentation updates
3847 * bug fixes and cleanups
3848 * small changes in the DLM wire protocol
3850 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3851 * version 0_5_1 with some initial stability,
3852 * locking on MD and file I/O.
3853 * documentation updates
3854 * several bug fixes since 0.5.0
3855 * small changes in wire protocol
3857 2002-07-18 Phil Schwan <phil@clusterfs.com>
3859 * delivered as Lustre Light Alpha
3860 * fixed a crash after handling invalid MDS requests
3861 * fixed directory pages for architectures with non-4k pages sizes
3863 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3864 * release version v0_4_4
3865 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3867 * Fixes a number of interruption problems with OST operations.
3868 * Update documentation for portals header changes
3869 * Move all wire protocol structs/defines to lustre_idl.h
3870 * Fixes symlink length bug.
3871 * Add tcpdump to repository.
3873 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3874 * release version v0_4_3
3875 * Fixes statfs for inodes on extN.
3876 * Fixes bug in runtests which would delete /etc/hosts.
3877 * Use 64-bit object IDs wherever possible (not into VFS though)
3878 Remove ost_get_info, which is unused by lustre, and out of date.
3880 2002-07-03 Peter Braam <braam@clusterfs.com>
3881 * release version v0_4_2 Fixes a lookup error (type not passed)
3882 * move forward to head of Portals
3883 * move forward to latest Lustre kernel
3885 2002-06-25 Peter Braam <braam@clusterfs.com>
3886 * release version v0_4_1. Hopefully stable on single node use.