1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10)
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4)
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.39.cfs8
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: don't use GFP_* in generic Lustre code.
18 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
22 Description: obdclass is limited by single OBD_ALLOC(idarray)
23 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
27 Description: warnings with build patchless client with vanila 2.6.19 and up
28 Details : change old ctl_table style and replace ctl_table/ctl_table_header
29 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
33 Frequency : Only for SLES
34 Description: sanity_quota fail test_1
35 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
36 file, which makes TSTID[2] inunique.
40 Frequency : Only for SLES9
41 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
42 Details : Update target/ChangeLog/which_patch .
46 Frequency : Only for test_10 in sanity-quota.sh
47 Description: a bug in quota test code
48 Details : checking a test flag in a wrong place
52 Frequency : Only for test_13 in sanity-quota.sh
53 Description: a bug in quota test script
54 Details : shouldn't check overfull allocated quota
58 Description: jbd statistics
59 Details : Port older jbd statistics patch for sles10
61 Severity : enhancement
63 Description: Add printing inode info into message about error in writepage.
68 Description: reply_lock_interpret crash due to race with it and lock cancel.
69 Details : Do not replay locks that are being cancelled. Do not reference
70 locks by their address during replay, just by their handle.
72 --------------------------------------------------------------------------------
74 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
76 * Support for kernels:
79 2.6.16.46-0.14 (SLES 10)
80 2.6.18.8 vanilla (kernel.org)
81 * Client support for unpatched kernels:
82 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
83 2.6.16 - 2.6.22 vanilla (kernel.org)
84 * Due to recently discovered recovery problems, we do not recommend
85 using patchless RHEL 4 clients with this or any earlier release.
86 * Recommended e2fsprogs version: 1.39.cfs8
87 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
88 * Starting with this release, the ldiskfs backing filesystem required
89 by Lustre is now in its own package, lustre-ldiskfs. This package
90 should be installed. It is versioned separately from Lustre and
91 may be released separately in future.
95 Description: block reactivating mgc import until all deactivates complete
96 Details : Fix race when failing back MDT/MGS to itself (testing)
98 Severity : enhancement
100 Description: add optional extra BUILD_VERSION info
101 Details : add a new environment variable (namely LUSTRE_VERS) which allows
102 to override the lustre version.
105 Frequency : 2.6.18 servers only
107 Description: ll_kern_mount() doesn't release the module reference
108 Details : The ldiskfs module reference count never drops down to 0
109 because ll_kern_mount() doesn't release the module reference.
114 Description: server LBUG when using old ost_num_threads parameter
115 Details : Accept the old ost_num_threads parameter but warn that it
116 is deprecated, and fix an off-by-one error that caused an LBUG.
121 Description: Transient SCSI error results in persistent IO issue
122 Details : iobuf->dr_error is not reinitialized to 0 between two
126 Frequency : sometimes when underlying device returns I/O errors
128 Description: OSTs not going read-only during write failures
129 Details : OSTs are not remounted read-only when the journal commit threads
130 get I/O errors because fsfilt_ext3 calls journal_start/stop()
131 instead of the ext3 wrappers.
135 Description: poor connect scaling with increasing client count
136 Details : Don't run filter_grant_sanity_check for more than 100 exports
137 to improve scaling for large numbers of clients.
140 Frequency : SLES10 only
142 Description: sanity-quota.sh quotacheck failed: rc = -22
143 Details : Quotas cannot be enabled on SLES10.
146 Frequency : liblustre clients only
148 Description: getdirentries does not give error when run on compute nodes
149 Details : getdirentries does not fail when the size specified as an argument
150 is too small to contain at least one entry
152 Severity : enhancement
154 Description: Add LNET router traceability for debug purposes
155 Details : If a checksum failure occurs with a router as part of the
156 IO path, the NID of the last router that forwarded the bulk data
157 is printed so it can be identified.
162 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
163 Details : Due to a race condition, liblustre clients were occasionally
166 Severity : enhancement
168 Description: lfs setstripe use optional parameters instead of postional
171 Severity : enhancement
173 Description: Nanosecond timestamp support for ldiskfs
174 Details : The on-disk ldiskfs filesystem has added support for nanosecond
175 resolution timestamps. There is not yet support for this at
176 the Lustre filesystem level.
179 Frequency : during server recovery
181 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
182 Details : request with rq_no_resend flag not awake l_wait_event if they get a
186 Frequency : nfs export on patchless client
188 Description: connectathon hang when test nfs export over patchless client
189 Details : Disconnected dentry cannot be found with lookup, so we do not need
190 to unhash it or make it invalid
193 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
194 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
195 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
200 Description: open req refcounting wrong on reconnect
201 Details : If reconnect happened between getting open reply from server and
202 call to mdc_set_replay_data in ll_file_open, we will schedule
203 replay for unreferenced request that we are about to free.
204 Subsequent close will crash in variety of ways.
205 Check that request is still eligible for replay in
206 mdc_set_replay_data().
211 Description: disable writes to filesystem when reading health_check file
212 Details : the default for reading the health_check proc file has changed
213 to NOT do a journal transaction and write to disk, because this
214 can cause reads of the /proc file to hang and block HA state
215 checking on a healthy but otherwise heavily loaded system. It
216 is possible to return to the previous behaviour during configure
217 with --enable-health-write.
219 Severity : enhancement
221 Description: 64-bit inode version
222 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
223 to the inode. This will be required for version-based recovery.
228 Description: MDS fails to start if a duplicate client export is detected
229 Details : in some rare cases it was possible for a client to connect to
230 an MDS multiple times. Upon recovery the MDS would detect this
231 and fail during startup. Handle this more gracefully.
233 Severity : enhancement
235 Description: Add -o localflock option to simulate old noflock
237 Details : This will achieve local-only flock/fcntl locks
243 Description: log_commit_thread vs filter_destroy race leads to crash
244 Details : Take import reference before releasing llog record semaphore
249 Description: Wrong request locking in request set processing
250 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
251 imp_delayed_list, in this place should be used imp_lock.
254 Frequency : when reconnection
256 Description: Grant Leak when osc reconnect to OST
257 Details : When osc reconnect ost, OST(filter) should check whether it
258 should grant more space to client by comparing fed_grant and
259 cl_avail_grant, and return the granted space to client instead
260 of "new granted" space, because client will call osc_init_grant
261 to update the client grant space info.
264 Frequency : when client reconnect to OST
266 Description: Grant Leak when osc do resend and replay bulk write
267 Details : When osc reconnect to OST, OST(filter)should clear grant info of
268 bulk write request, because the grant info while be sync between
269 OSC and OST when reconnect, and we should ignore the grant info
270 these of resend/replay write req.
275 Description: Grant space more than avaiable left space sometimes.
276 Details : When then OST is about to be full, if two bulk writing from
277 different clients came to OST. Accord the avaliable space of the
278 OST, the first req should be permitted, and the second one
279 should be denied by ENOSPC. But if the seconde arrived before
280 the first one is commited. The OST might wrongly permit second
281 writing, which will cause grant space > avaiable space.
284 Frequency : when client is evicted
286 Description: Grant might be wrongly erased when osc is evicted by OST
287 Details : when the import is evicted by server, it will fork another
288 thread ptlrpc_invalidate_import_thread to invalidate the
289 import, where the grant will be set to 0. While the original
290 thread will update the grant it got when connecting. So if
291 the former happened latter, the grant will be wrongly errased
292 because of this race.
297 Description: Checking Stale with correct fid
298 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
299 to check whether it is stale, because sometimes, we want the
300 enqueue happened anyway, and op_data.fid2 will not be initialized.
302 Severity : enhancement
304 Description: update patchless client
305 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
308 Frequency : only with 2.4 kernel
310 Description: random memory corruption
311 Details : size of struct ll_inode_info is to big for union inode.u and this
312 can be cause of random memory corruption.
317 Description: Memory leak in recovery
318 Details : Lov_mds_md was not free in an error handler in mds_create_object.
319 It should also check obd_fail before fsfilt_start, otherwise if
320 fsfilt_start return -EROFS,(failover mds during mds recovery).
321 then the req will return with repmsg->transno = 0 and rc = EROFS.
322 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
323 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
324 be freed no matter whether fsfilt_commit success or not.
327 Frequency : only with huge count clients
329 Description: Prevents from taking the superblock lock in llap_from_page for
331 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
332 for llap_from_page call in ll_removepage prevents from taking the
333 superblock lock for a soon died page.
338 Description: Not check open intent error before release open handle
339 Details : in some rare cases, the open intent error is not checked before
340 release open handle, which may cause
341 ASSERTION(open_req->rq_transno != 0), because it tries to release
342 the failed open handle.
347 Description: Set cat log bitmap only after create log success.
348 Details : in some rare cases, the cat log bitmap is set too early. and it
349 should be set only after create log success.
353 Description: Accessing a block bevice can re-enable I/O when Lustre is
354 tearing down a device.
355 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
360 Description: service threads may hog cpus when there are a lot of requests
362 Details : Insert cond_resched to give other threads a chance to use some of
368 Description: the cat log was not initialized in recovery
369 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
370 unlink log on mds will not be initialized until mds post
371 recovery. And also in mds post recovery, the unlink log will
372 initialization will be done asynchronausly, so there will be race
373 between add unlink log and unlink log initialization.
377 Description: brw_stats were being printed incorrectly
378 Details : brw_stats were being printed as log2 but all of them were not
379 recorded as log2. Also remove some code duplication arising from
380 filter_tally_{read,write}.
384 Frequency : rare, only in recovery.
385 Description: ASSERTION(req->rq_type != LI_POISON) failed
386 Details : imp_lock should be held while iterating over imp_sending_list for
387 prevent destroy request after get timeout in ptlrpc_queue_wait.
391 Description: replay-single.sh test 52 fails
392 Details : A lock's skiplist need to be cleanup when it being unlinked
393 from its resource list.
397 Description: Short directio read returns full requested size rather than
399 Details : Direct I/O operations should return actual amount of bytes
400 transferred rather than requested size.
402 Severity : enhancement
404 Description: metadata RPC reduction (e.g. for rm performance)
405 Details : decrease the amount of synchronous RPC between clients and servers
406 by canceling conflicing lock before the operation on the client side
407 and packing thier handles into the main operation RPC to server.
409 --------------------------------------------------------------------------------
411 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
416 Frequency : on some architectures
418 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
419 Details : Uninitialized flags sometimes cause configuration commands to
423 Frequency : patchless clients only
425 Description: missing __iget() symbol export
426 Details : The __iget() symbol export is missing. To avoid the need for
427 this on patchless clients the deathrow inode reaper is turned
428 off, and we depend on the VM to clean up old inodes. This
429 dependency was during via the fix for bug 12181.
431 --------------------------------------------------------------------------------
433 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
435 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
436 INTEROPERATE with older versions automatically. In many cases a
437 special upgrade step is needed. Please read the
438 user documentation before upgrading any part of a live system.
439 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
440 version will not interoperate with 1.6 betas before beta5 (1.5.95).
441 * WARNING: Lustre configuration and startup changes are required with
442 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
447 Severity : enhancement
449 Description: MountConf
450 Details : Lustre configuration is now managed via mkfs and mount
451 commands instead of lmc and lconf. New obd types (MGS, MGC)
452 are added for dynamic configuration management. See
453 https://mail.clusterfs.com/wikis/lustre/MountConf for
456 Severity : enhancement
458 Description: dynamic OST addition
459 Details : OSTs can now be added to a live filesystem
461 Severity : enhancement
463 Description: startup order invariance
464 Details : MDTs and OSTs can be started in any order. Clients only
465 require the MDT to complete startup.
467 Severity : enhancement
469 Description: parallel, asynchronous orphan cleanup
470 Details : orphan cleanup is now performed in separate threads for each
471 OST, allowing parallel non-blocking operation.
473 Severity : enhancement
475 Description: optimized stripe assignment
476 Details : stripe assignments are now made based on ost space available,
477 ost previous usage, and OSS previous usage, in order to try
478 to optimize storage space and networking resources.
480 Severity : enhancement
482 Description: Permanently set tunables
483 Details : All writable /proc/fs/lustre tunables can now be permanently
484 set on a per-server basis, at mkfs time or on a live
487 Severity : enhancement
489 Description: Lustre message v2
490 Details : Add lustre message format v2.
492 Severity : enhancement
494 Description: client OST exclusion list
495 Details : Clients can be started with a list of OSTs that should be
496 declared "inactive" for known non-responsive OSTs.
500 Description: SPEC SFS validation failure on NFS v2 over lustre.
501 Details : Changes the blocksize for regular files to be 2x RPC size,
502 and not depend on stripe size.
504 Severity : enhancement
506 Description: Multiple MD RPCs in flight.
507 Details : Further unserialise some read-only MDS RPCs - learn about intents.
508 To avoid overly-overloading MDS, introduce a limit on number of
509 MDS RPCs in flight for a single client and add /proc controls
510 to adjust this limit.
512 Severity : enhancement
514 Description: client read/write statistics
515 Details : Add client read/write call usage stats for performance
516 analysis of user processes.
517 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
518 file access. extents_stats shows chunk size distribution.
519 extents_stats_per_process show chunk size distribution per
522 Severity : enhancement
524 Description: mds statistics
525 Details : Add detailed mds operations statistics in
526 /proc/fs/lustre/mds/*/stats.
530 Description: Failure of copying files with lustre special EAs.
531 Details : Client side always return success for setxattr call for lustre
532 special xattr (currently only "trusted.lov").
537 Description: Refcount LNET uuids
538 Details : The global LNET uuid list grew linearly with every startup;
539 refcount repeated list entries instead of always adding to
542 Severity : enhancement
544 Description: Dynamic service threads
545 Details : Within a small range, start extra service threads
546 automatically when the request queue builds up.
549 Frequency : mixed-endian client/server environments
551 Description: mixed-endian crashes
552 Details : The new msg_v2 system had some failures in mixed-endian
555 Severity : enhancement
557 Description: Easy OST removal
558 Details : OSTs can be permanently deactivated with e.g. 'lctl
559 conf_param lustre-OST0001.osc.active=0'
561 Severity : enhancement
563 Description: MGS proc entries
564 Details : Added basic proc entries for the MGS showing what filesystems
567 Severity : enhancement
569 Description: provide MGS failover
570 Details : Added config lock reacquisition after MGS server failover.
572 Severity : enhancement
574 Description: add Linux 2.4 support
575 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
579 Description: a large application tries to do I/O to the same resource and dies
581 Details : Check the req->rq_arrival time after the call to
582 ost_brw_lock_get(), but before we do anything about
583 processing it & sending the BULK transfer request. This
584 should help move old stale pending locks off the queue as
585 quickly as obd_timeout.
588 Frequency : when an incorrect nid is specified during startup
590 Description: ptlrpc connect to non-existant node causes kernel crash
591 Details : LNET can't be re-entered from an event callback, which
592 happened when we expire a message after the export has been
593 cleaned up. Instead, hand the zombie cleanup off to another
596 Severity : enhancement
598 Description: plain/inodebits lock performance improvement
599 Details : Grouping plain/inodebits in granted list by their request modes
600 and bits policy, thus improving the performance of search through
604 Frequency : only if OST filesystem is corrupted
606 Description: client incorrectly hits assertion in ptlrpc_replay_req()
607 Details : for a short time RPCs with bulk IO are in the replay list,
608 but replay of bulk IOs is unimplemented. If the OST filesystem
609 is corrupted due to disk cache incoherency and then replay is
610 started it is possible to trip an assertion. Avoid putting
611 committed RPCs into the replay list at all to avoid this issue.
614 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
616 Description: System hang on startup
617 Details : This bug allowed the liblustre (e.g. catamount) client to
618 return to the app before handling all startup RPCs. This
619 could leave the node unresponsive to lustre network traffic
620 and manifested as a server ptllnd timeout.
622 Severity : enhancement
624 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
625 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
626 boolean to control whether to print peer debug info when a
627 client's RPC times out.
630 Frequency : only for kernels with patches from Lustre below 1.4.3
632 Description: Remove old rdonly API
633 Details : Remove old rdonly API which unused from at least lustre 1.4.3
636 Frequency : only for devices with external journals
638 Description: Set external device read-only also
639 Details : During a commanded failover stop, we set the disk device
640 read-only while the server shuts down. We now also set any
641 external journal device read-only at the same time.
644 Frequency : when upgrading from 1.4 while trying to change parameters
646 Description: The wrong (new) MDC name was used when setting parameters for
647 upgraded MDT's. Also allows changing of OSC (and MDC)
648 parameters if --writeconf is specified at tunefs upgrade time.
651 Frequency : when setting specific ost indicies
653 Description: QOS code breaks on skipped indicies
654 Details : Add checks for missing OST indicies in the QOS code, so OSTs
655 created with --index need not be sequential.
657 Severity : enhancement
659 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
660 Details : The uninit_groups feature works in conjunction with the kernel
661 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
662 up the pass1 processing of e2fsck. This is a read-only feature
663 in ldiskfs2 only, so older kernels and current ldiskfs cannot
664 mount filesystems that have had this feature enabled.
666 Severity : enhancement
668 Description: Improve multi-block allocation algorithm to avoid fragmentation
669 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
670 allocation locality and avoid filesystem fragmentation.
672 ------------------------------------------------------------------------------
674 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
676 * Support for kernels:
677 2.4.21-47.0.1.EL (RHEL 3)
679 2.6.9-42.0.10.EL (RHEL 4)
680 2.6.12.6 vanilla (kernel.org)
681 2.6.16.27-0.9 (SLES 10)
682 * Recommended e2fsprogs version: 1.39.cfs5
684 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
688 Frequency : occasional, depends on client load and configuration
689 Bugzilla : 12181, 12203
690 Description: data loss for recently-modified files
692 Details : In some cases it is possible that recently written or created
693 files may not be written to disk in a timely manner (this should
694 normally be within 30s unless client IO load is very high).
695 The problem appears as zero-length files or files that are a
696 multiple of 1MB in size after a client crash or client eviction
697 that are missing data at the end of the file.
699 This problem is more likely to be hit on clients where files are
700 repeatedly created and unlinked in the same directory, clients
701 have a large amount of RAM, have many CPUs, the filesystem has
702 many OSTs, the clients are rebooted frequently, and/or the files
703 are not accessed by other nodes after being written.
705 The presence of the problem can be detected by looking at
706 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
707 smaller than the second (nr_unused) then dirty files will not
708 be flushed automatically to disk. "sync; sleep 10" should be
709 run several times on the node before unmounting it to update
710 Lustre (this is also safe to run on nodes without this problem).
712 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
713 that can cause this same problem, so customers using that kernel
714 also need to update the kernel in addition to Lustre. In order
715 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
717 It is normal that files written just before a client crash (less
718 than 30s) may not yet have been flushed to disk, even for local
722 Frequency : frequent on thin XT3 nodes
724 Description: UUID collision on thin XT3 Linux nodes
725 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
726 randomly, since we relied on an insufficiently-seeded PRNG.
731 Description: OSS hangs after "All ost request buffers busy"
732 Details : A deadlock between quota and journal operations caused OSS
733 hangs after printing "All ost request buffers busy."
736 Frequency : always on liblustre builds
738 Description: Cleanup compiler warnings on liblustre
741 Frequency : always on liblustre builds on XT3
743 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
747 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
748 > 32000 subdirectories
749 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
750 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
751 subdirectory count crosses 32000. This will aid e2fsck to
752 correctly handle more than 32000 subdirectories.
755 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
757 Description: System hang on startup
758 Details : This bug allowed the liblustre (e.g. catamount) client to
759 return to the app before handling all startup RPCs. This
760 could leave the node unresponsive to lustre network traffic
761 and manifested as a server ptllnd timeout.
763 Severity : enhancement
765 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
766 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
767 boolean to control whether to print peer debug info when a
768 client's RPC times out.
773 Description: make O_SYNC working on 2.6 kernels
774 Details : 2.6 kernels use different method for mark pages for write,
775 so need add a code to lustre for O_SYNC work.
780 Description: Failure to close file and release space on NFS
781 Details : Put inode details into lock acquired in ll_intent_file_open.
782 Use mdc_intent_lock in ll_intent_open to properly
783 detect all kind of errors unhandled by mdc_enqueue.
788 Description: proc file read during shutdown sometimes raced obd removal,
790 Details : Add lock to prevent obd access after proc file removal.
793 Frequency : Only for files larger than 4GB on 32-bit clients.
795 Description: improperly doing page alignment of locks
796 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
797 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
802 Description: RPCs being resent when they shouldn't be
803 Details : Some RPCs that should not be resent are being resent. This
804 can cause inconsistencies in the RPC state machine. Do not
805 resend such requests.
808 Frequency : rare, only with NFS export
810 Description: Crash on NFS re-export node
811 Details : under very unusual load conditions an assertion is hit in
812 ll_intent_file_open()
815 Frequency : only if OST filesystem is corrupted
817 Description: client incorrectly hits assertion in ptlrpc_replay_req()
818 Details : for a short time RPCs with bulk IO are in the replay list,
819 but replay of bulk IOs is unimplemented. If the OST filesystem
820 is corrupted due to disk cache incoherency and then replay is
821 started it is possible to trip an assertion. Avoid putting
822 committed RPCs into the replay list at all to avoid this issue.
827 Description: large O_DIRECT requests fail under memory pressure/fragmentation
828 Details : Large single O_DIRECT read and write calls can fail to allocate
829 a sufficiently large buffer to process the request. In case of
830 allocation failure the allocation is retried with a smaller
831 buffer and broken into smaller requests.
833 Severity : enhancement
835 Description: Add -o localflock option to simulate old noflock behaviour.
836 Details : This will achieve local-only flock/fcntl locks coherentness.
841 Description: versioning check is incomplete
842 Details : Checking the version difference of client vs. server, report
843 error if the gap is too big.
848 Description: add support PG_writeback bit
849 Details : add support for PG_writeback bit for Lustre, for more carefull
850 work with page cache in 2.6 kernel. This also fix some deadlocks
851 and remove hack for work O_SYNC with 2.6 kernel.
853 Severity : enhancement
855 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
856 Details : The uninit_groups feature works in conjunction with the kernel
857 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
858 up the pass1 processing of e2fsck. This is a read-only feature
859 in ldiskfs2 only, so older kernels and current ldiskfs cannot
860 mount filesystems that have had this feature enabled.
862 Severity : enhancement
864 Description: Improve multi-block allocation algorithm to avoid fragmentation
865 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
866 allocation locality and avoid filesystem fragmentation.
868 ------------------------------------------------------------------------------
870 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
872 * Support for kernels:
873 2.6.9-42.0.3.EL (RHEL 4)
875 2.4.21-47.0.1.EL (RHEL 3)
876 2.6.12.6 vanilla (kernel.org)
877 2.6.16.21-0.8 (SLES10)
878 * Recommended e2fsprogs version: 1.39.cfs2-0
880 * The backwards-compatible /proc/sys/portals symlink has been removed
881 in this release. Before upgrading, please ensure that you change
882 any configuration scripts or /etc/sysctl.conf files that access
883 /proc/sys/portals/* or sysctl portals.* to use the corresponding
884 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
885 in advance of the upgrade on any system running Lustre 1.4.6 or
886 newer, since /proc/sys/lnet was added in that version.
887 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
891 Frequency : only when quota is used
893 Description: avoid scanning export list for quota master
894 Details : Change the algorithms to avoid scanning export list in order
895 to improve the efficiency.
898 Frequency : MDS failover only, very rarely
900 Description: "went back in time" messages on mds failover
901 Details : The greatest transno may be lost when the current operation
902 finishes with an error (transno==0) and the client's last_rcvd
903 record is over-written. Save the greatest transno in the
904 mds_last_transno for this case.
907 Frequency : always for specific kernels and striping counts
909 Description: client may get "Matching packet too big" without ACL support
910 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
911 when trying to access files in certain configurations. The
912 clients should in fact be denied when mounting because they do
916 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
918 Description: many clients connecting with IO in progress causes connect timeouts
919 Details : Avoid synchronous journal commits to avoid delays caused by many
920 clients connecting/disconnecting when bulk IO is in progress.
921 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
922 OST_IO_PORTAL to avoid delays behind potentially many pending
926 Frequency : occasionally with multiple writers to a single file
928 Description: shared writes to file may result in wrong size reported by stat()
929 Details : Allow growing of kms when extent lock is cancelled
932 Frequency : always with random mmap IO to multi-striped file
934 Description: mmap write might be lost if we are writing to a 'hole' in stripe
935 Details : Only if the hole is at the end of OST object so that kms is too
936 small. Fix is to increase kms accordingly in ll_nopage.
939 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
941 Description: writes to a missing object would leak memory on the OST
942 Details : If there is an inconsistency between the MDS and OST filesystems,
943 such that the MDS references an object that doesn't exist, writes
944 to that object will leak memory due to incorrect cleanup in the
945 error handling path, eventually running out of memory on the OST.
950 Description: Creating too long symlink causes lustre errors
951 Details : Check symlink and name lengths before sending requests to MDS.
954 Frequency : only if flock is enabled (not on by default)
956 Description: posix locks not released on fd closure on 2.6.9+
957 Details : We failed to add posix locks to list of inode locks on 2.6.9+
958 kernels, this caused such locks not to be released on fd close and
959 then assertions on fs unmount about still used locks.
962 Frequency : MDS failover only, very rarely
964 Description: clients may get ASSERTION(granted_lock != NULL)
965 Details : When request was taking a long time, and a client was resending
966 a getattr by name lock request. The were multiple lock requests
967 with the same client lock handle and
968 mds_getattr_name->fixup_handle_for_resent_request found one of the
969 lock handles but later failed with ASSERTION(granted_lock != NULL).
974 Description: handle->h_buffer_credits > 0, assertion failure
975 Details : h_buffer_credits is zero after truncate, causing assertion
976 failure. This patch extends the transaction or creates a new
980 Frequency : NFS re-export or patchless client
981 Bugzilla : 11179, 10796
982 Description: Crash on NFS re-export node (__d_move)
983 Details : We do not want to hash the dentry if we don't have a lock.
984 But if this dentry is later used in d_move, we'd hit uninitialised
985 list head d_hash, so we just do this to init d_hash field but
986 leave dentry unhashed.
989 Frequency : NFS re-export or patchless client
991 Description: NFS exports has problem with symbolic link
992 Details : lustre client didn't properly install dentry when re-exported
993 to NFS or running patchless client.
996 Frequency : NFS re-export or patchless client
998 Description: Various nfs/patchless fixes.
999 Details : fixes reuse disconected alias for lookup process - this fixes
1000 warning "find_exported_dentry: npd != pd",
1001 fix permission error with open files at nfs.
1002 fix apply umask when do revalidate.
1005 Frequency : occasional
1007 Description: Crash on NFS re-export node
1008 Details : calling clear_page() on the wrong pointer triggered oops in
1009 generic_mapping_read().
1012 Frequency : rarely, using O_DIRECT IO
1014 Description: unaligned directio crashes client with LASSERT
1015 Details : check for unaligned buffers before trying any requests.
1018 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1020 Description: stale data returned from RAID cache
1021 Details : If only a small amount of IO is done to the RAID device before
1022 reading it again it is possible to get stale data from the RAID
1023 cache instead of reading it from disk.
1026 Frequency : always for sles10 kernel
1028 Description: sles10 support
1029 Details : ll_follow_link: compile fixes and using of nd_set_link
1030 under newer kernels.
1033 Frequency : depends on arch, kernel and compiler version, always on sles10
1036 Description: recursive or deep enough symlinks cause stack overflow
1037 Details : getting rid of large stack-allocated variable in
1041 Frequency : depends on hardware
1043 Description: lustre write performance loss in the SLES10 kernel
1044 Details : the performance loss is caused by using of write barriers in the
1045 ext3 code. The SLES10 kernel turns barrier support on by
1046 default. The fix is to undo that change for ldiskfs.
1048 ------------------------------------------------------------------------------
1050 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1052 * Support for kernels:
1053 2.6.9-42.0.3EL (RHEL 4)
1054 2.6.5-7.276 (SLES 9)
1055 2.4.21-47.0.1.EL (RHEL 3)
1056 2.6.12.6 vanilla (kernel.org)
1060 Frequency : quota enabled and large files being deleted
1062 Description: releasing more than 4GB of quota at once hangs OST
1063 Details : If a user deletes more than 4GB of files on a single OST it
1064 will cause the OST to spin in an infinite loop. Release
1065 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1070 Description: statfs data retrieved from /proc may be stale or zero
1071 Details : When reading per-device statfs data from /proc, in the
1072 {kbytes,files}_{total,free,avail} files, it may appear
1073 as zero or be out of date.
1076 Frequency : systems with MD RAID1 external journal devices
1078 Description: lconf's call to blkid is confused by RAID1 journal devices
1079 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1080 of returning all block devices that match the journal UUID.
1083 Frequency : always, for aggregate stripe size over 4GB
1085 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1086 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1087 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1090 Frequency : always if "lfs find" used on a local file/directory
1092 Description: "lfs find" segfaults if used on a local file/directory
1093 Details : The case where a directory component was not specified wasn't
1094 handled correctly. Handle this properly.
1097 Frequency : always on ppc64
1099 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1100 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1101 perform endianness conversion.
1104 Frequency : rarely (truncate to non-zero file size after write under load)
1105 Bugzilla : 10730, 10687
1106 Description: Files padded with zeros to next 4K multiple
1107 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1108 it is possible that files that are truncated to a non-zero size
1109 immediately after being written are filled with zero bytes beyond
1110 the truncated size. No file data is lost.
1112 Severity : enhancement
1114 Description: Allow recovery/failover for liblustre clients.
1115 Details : liblustre clients were unaware of failover configurations until
1118 Severity : enhancement
1120 Description: user file locks should fail when not mounting with flock option
1121 Details : Set up an error-returning stub in ll_file_operations.lock field
1122 to prevent incorrect behaviour when client is mounted without
1123 flock option. Also, set up properly f_op->flock field for
1127 Frequency : always on ia64
1129 Description: "lfs df" loops on printing out MDS statfs information
1130 Details : The obd_ioctl_data was not initialized and in some systems
1131 this caused a failure during the ioctl that did not return
1132 an error. Initialize the struct and return an error on failure.
1135 Frequency : SLES 9 only
1137 Description: Error of copying files with lustre special EAs as root
1138 Details : Client side always return success for setxattr call for lustre
1139 special xattr (currently only "trusted.lov").
1142 Frequency : rarely on clusters with both ia64+i386 clients
1144 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1145 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1146 clients are doing concurrent writes to the same file it is
1147 possible that the ia64 clients may LASSERT because the OST
1148 extent locks are not PAGE_SIZE aligned. Ensure that grown
1149 locks are always aligned on the request boundary.
1152 Frequency : specific use, occasional
1154 Description: Overwriting in use executable truncates on-disk binary image
1155 Details : If one node attempts to overwrite an executable in use by
1156 another node, we now correctly return ETXTBSY instead of
1157 truncating the file.
1159 Severity : enhancement
1161 Description: Async OSC create to avoid the blocking unnecessarily.
1162 Details : If a OST has no remain object, system will block on the creating
1163 when need to create a new object on this OST. Now, ways use
1164 pre-created objects when available, instead of blocking on an
1165 empty osc while others are not empty. If we must block, we block
1166 for the shortest possible period of time.
1171 Description: chmod on Lustre root is propagated to other clients
1172 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1173 invalid by the follow_mount time.
1176 Frequency : liblustre clients only
1178 Description: Race in 'instant cancel' lock handling could lead to such locks
1179 never to be granted in case of SMP MDS
1180 Details : Do not destroy not yet granted but cbpending locks in
1184 Frequency : replay/resend of open
1186 Description: non null lock assetion failure in mds_intent_policy
1187 Details : Trying to replay/resend lockless open requests resulted in
1188 mds_open() returning 0 with no lock. Now it sets a flag if
1189 a lock is going to be returned.
1191 Severity : enhancement
1193 Description: Checksum enhancements
1194 Details : New checksum enhancements allow for resending RPCs that failed
1197 Severity : enhancement
1199 Description: Tunables on number of dirty pages in cacche
1200 Details : Allow to set limit on number of dirty pages cached.
1205 Description: client crash on unmount - lock still has references
1206 Details : In some error handling cases it was possible to leak a lock
1207 reference on a client while accessing a file. This was not
1208 harmful to the client during operation, but would cause the
1209 client to crash when the filesystem is unmounted.
1212 Frequency : specific case, rare
1214 Description: ETXTBSY on mds though file not in use by client
1215 Details : ETXTBSY is no longer incorrectly returned when attempting to
1216 chmod or chown a directory that the user previously tried to
1217 execute or a currently-executing binary.
1220 Frequency : extremely rare except on liblustre-based clients
1222 Description: Lustre space not freed when files are deleted
1223 Details : Clean up open-unlinked files after client eviction. Previously
1224 the unlink was skipped and the files remained as orphans.
1229 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1230 Details : In some cases it was possible to send a blocking callback to
1231 a client doing a glimpse, even though that client didn't get
1232 a lock granted. When the glimpse lock is cancelled on the OST
1233 the freed lock is left on the waiting list and corrupted the list.
1236 Frequency : all core dumps
1238 Description: Broke core dumps to lustre
1239 Details : Negative dentry may be unhashed if parent does not have UPDATE
1240 lock, but some callers, e.g. do_coredump, expect dentry to be
1241 hashed after successful create, hash it in ll_create_it.
1243 ------------------------------------------------------------------------------
1245 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1247 * Support for kernels:
1248 2.6.9-42.0.2.EL (RHEL 4)
1249 2.6.5-7.276 (SLES 9)
1250 2.4.21-40.EL (RHEL 3)
1251 2.6.12.6 vanilla (kernel.org)
1255 Frequency : always on RHEL 3
1257 Description: Number of open files grows over time
1258 Details : The number of open files grows over time, whether or not
1259 Lustre is started. This was due to a filp leak introduced
1260 by one of our kernel patches.
1262 ------------------------------------------------------------------------------
1264 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1266 * Support for kernels:
1267 2.6.9-42.EL (RHEL 4)
1268 2.6.5-7.276 (SLES 9)
1269 2.4.21-40.EL (RHEL 3)
1270 2.6.12.6 vanilla (kernel.org)
1275 Bugzilla : 5719, 9635, 9792, 9684
1276 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1277 Details : If a server is under heavy load and cannot reply to new
1278 connection requests before the client resends the (re)connect,
1279 the connection handling code can behave badly if two service
1280 threads are concurrently handing separate (re)connections from
1281 the same client. Add better locking to the connection handling
1282 code, and ensure that only a single connection will be processed
1283 for a given client UUID, even if the lock is dropped.
1285 Severity : enhancement
1287 Description: add TCP zero-copy support to kernel
1288 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1289 sends if the hardware supports scatter-gather and checksumming.
1290 This allows socklnd to do client-write and server-read more
1291 efficiently and reduce CPU utilization from skbuf copying.
1294 Frequency : only if NFS exporting from client
1296 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1297 Details : If setting attributes on a file created under NFS that had
1298 never been opened it would be possible to oops the client
1299 if the file had no objects.
1302 Frequency : always for liblustre
1304 Description: liblustre client does MDS+OSTs setattr RPC for each write
1305 Details : When doing a write from a liblustre client, the client
1306 incorrectly issued an RPC to the MDS and each OST the file was
1307 striped over in order to update the timestamps. When writing
1308 with small chunks and many clients this could overwhelm the MDS
1309 with RPCs. In all cases it would slow down the write because
1310 these RPCs are unnecessary.
1312 Severity : enhancement
1314 Description: allow number of MDS service threads to be changed at module load
1315 Details : It is now possible to change the number of MDS service threads
1316 running. Adding "options mds mds_num_threads={N}" to the MDS's
1317 /etc/modprobe.conf will set the number of threads for the next
1318 time Lustre is restarted (assuming the "mds" module is also
1319 reloaded at that time). The default number of threads will
1320 stay the same, 32 for most systems.
1325 Description: OST crash if filesystem is unformatted or corrupt
1326 Details : If an OST is started on a device that has never been formatted
1327 or if the filesystem is corrupt and cannot even mount then the
1328 error handling cleanup routines would dereference a NULL pointer.
1333 Description: NULL pointer deref in llap_from_page.
1334 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1335 incorrect) mapping if the page was truncated/reclaimed while it was
1336 searched for. Check for this condition and skip such pages when
1337 doing readahead. Introduce extra check to llap_from_page() to
1338 verify page->mapping->host is non-NULL (so page is not anonymous).
1341 Frequency : Sometimes when using sys_sendfile
1343 Description: "page not covered by a lock" warnings from ll_readpage
1344 Details : sendfile called ll_readpage without right page locks present.
1345 Now we introduced ll_file_sendfile that does necessary locking
1346 around call to generic_file_sendfile() much like we do in
1350 Frequency : with certain MDS communication failures at client mount time
1352 Description: NULL pointer deref after failed client mount
1353 Details : a client connection request may delayed by the network layer
1354 and not be sent until after the PTLRPC layer has timed out the
1355 request. If the client fails the mount immediately it will try
1356 to clean up before the network times out the request. Add a
1357 reference from the request import to the obd device and delay
1358 the cleanup until the network drops the request.
1361 Frequency : occasionally during client (re)connect
1363 Description: assertion failure during client (re)connect
1364 Details : processing a client connection request may be delayed by the
1365 client or server longer than the client connect timeout. This
1366 causes the client to resend the connection request. If the
1367 original connection request is replied in this interval, the
1368 client may trip an assertion failure in ptlrpc_connect_interpret()
1369 which thought it would be the only running connect process.
1372 Frequency : only with obd_echo servers and clients that are rebooted
1374 Description: kernel BUG accessing uninitialized data structure
1375 Details : When running an obd_echo server it did not start the ping_evictor
1376 thread, and when a client was evicted an uninitialized data
1377 structure was accessed. Start the ping_evictor in the RPC
1378 service startup instead of the OBD startup.
1380 Severity : enhancement
1381 Bugzilla : 10193 (patchless)
1382 Description: Remove dependency on various unexported kernel interfaces.
1383 Details : No longer need reparent_to_init, exit_mm, exit_files,
1384 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1387 Frequency : rare (only users of deprecated and unsupported LDAP config)
1389 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1390 Details : LDAP apparently contains 'lustreName' attributes instead of
1391 'name'. A simple remapping of the name is sufficient.
1394 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1396 Description: waiting_locks_callback trips kernel BUG if client is evicted
1397 Details : Running with the dump_on_timeout debug flag turned on makes
1398 it possible that the waiting_locks_callback() can try to dump
1399 the Lustre kernel debug logs from an interrupt handler. Defer
1400 this log dumping to the expired_lock_main() thread.
1402 Severity : enhancement
1404 Description: Support NFS exporting on 2.6 kernels.
1405 Details : Implement non-rawops metadata methods for NFS server to use without
1406 changing NFS server code.
1409 Frequency : very rare (synthetic metadata workload only)
1411 Description: two racing renames might cause an MDS thread to deadlock
1412 Details : Running the "racer" program may cause one MDS thread to rename
1413 a file from being the source of a rename to being the target of
1414 a rename at exactly the same time that another thread is doing
1415 so, and the second thread has already enqueued these locks after
1416 doing a lookup of the target and is trying to relock them in
1417 order. Ensure that we don't try to re-lock the same resource.
1420 Frequency : only very large systems with liblustre clients
1422 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1423 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1424 all OSTs in parallel. This allows the eviction of stale liblustre
1425 clients to proceed much faster than if they were done in series,
1426 and also offers similar improvements for other set_info RPCs.
1431 Description: excessive CPU usage during initial read phase on client
1432 Details : During the initial read phase on a client, it would agressively
1433 retry readahead on the file, consuming too much CPU and impacting
1434 performance (since 1.4.5.8). Improve the readahead algorithm
1435 to avoid this, and also improve some other common cases (read
1436 of small files in particular, where "small" is files smaller than
1437 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1442 Description: MDS crash when receiving packet with unknown intent.
1443 Details : Do not LBUG in unknown intent case, just return -EFAULT
1445 Severity : enhancement
1446 Bugzilla : 9293, 9385
1447 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1448 Details : Do not serialize getattr (non-intent version) and statfs.
1451 Frequency : occasional, when OST network is overloaded/intermittent
1453 Description: client evicted by OST after bulk IO timeout
1454 Details : If a client sends a bulk IO request (read or write) the OST
1455 may evict the client if it is unresposive to its data GET/PUT
1456 request. This is incorrect if the network is overloaded (takes
1457 too long to transfer the RPC data) or dropped the OST GET/PUT
1458 request. There is no need to evict the client at all, since
1459 the pinger and/or lock callbacks will handle this, and the
1460 client can restart the bulk request.
1463 Frequency : Always when mmapping file with no objects
1465 Description: client crashes when mmapping file with no objects
1466 Details : Check that we actually have objects in a file before doing any
1467 operations on objects in ll_vm_open, ll_vm_close and
1473 Description: Request leak when working with deleted CWD
1474 Details : Introduce advanced request refcount tracking for requests
1475 referenced from lustre intent.
1477 Severity : Enhancement
1479 Description: Cache open file handles on client.
1480 Details : MDS now will return special lock along with openhandle, if
1481 requested and client is allowed to hold openhandle, even if unused,
1482 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1483 closing files for every read/write openration.
1485 Severity : Enhancement
1487 Description: Cache open negative dentries on client when possible.
1488 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1489 negative dentries on lock revocation.
1494 Description: Remounting a client read-only wasn't possible with a zconf mount
1495 Details : It wasn't possible to remount a client read-only with llmount.
1497 Severity : enhancement
1498 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1499 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1500 MPICH is included to add Lustre-specific ADIO interfaces.
1501 This is based closely on the UFS ADIO layer and only differs
1502 in file creation, in order to allow the OST striping to be set.
1503 This is user-contributed code and not supported by CFS.
1508 Description: extended inode attributes (immutable, append-only) work improperly
1509 when 2.4 and 2.6 kernels are used on client/server or vice versa
1510 Details : Introduce kernel-independent values for these flags.
1512 Severity : enhancement
1515 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1516 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1517 granularity. Now, Fractional MB granularity is be supported,
1518 this is very useful for low memory system.
1520 Severity : enhancement
1522 Description: Getattr by fid
1523 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1524 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1525 have locks and avoids getattr rpc requests for it on every stat.
1528 Frequency : Always, for filesystems larger than 2TB
1530 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1531 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1532 filesystems larger than 2TB to be mounted with mballoc enabled.
1535 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1537 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1538 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1539 filesystems larger than 2TB is mounted then the kernel will
1540 silently corrupt the start of the filesystem. CONFIG_LBD is
1541 enabled for all CFS-supported kernels, but the possibility of
1542 this happening with a modified kernel config exists.
1544 Severity : enhancement
1546 Description: add client O_DIRECT support for 2.6 kernels
1547 Details : It is now possible to do O_DIRECT reads and writes to files
1548 in the Lustre client mountpoint on 2.6 kernel clients.
1550 Severity : enhancement
1552 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1553 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1554 parallel, not waiting for response from every OST before sending
1555 a rpc to the next OST.
1560 Description: setattr vs write race when updating file timestamps
1561 Details : Client processes that update a file timestamp into the past
1562 right after writing to the file (e.g. tar) it is possible that
1563 the updated file modification time can be reset to the current
1564 time due to a race between processing the setattr and write RPC.
1566 Severity : enhancement
1568 Description: Bring 'lfs find' closer in line with regular Linux find.
1569 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1570 -print0 options and obtains all the needed info through the lustre
1573 Severity : enhancement
1575 Description: support up to 1024 configured devices on one node
1576 Details : change obd_dev array from statically allocated to dynamically
1577 allocated structs as they are first used to reduce memory usage
1582 Description: Flush dirty partially truncated pages during truncate
1583 Details : Immediatelly flush partially truncated pages in filter_setattr,
1584 this way we completely avoid having any pages in page cache on OST
1585 and can retire ugly workarounds during writes to flush such pages.
1590 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1591 Details : protect inode from truncation within vfs_unlink() context
1592 just take a reference before calling vfs_unlink() and release it
1593 when parent's i_sem is free.
1598 Description: last_id value checked outside lock on OST caused LASSERT failure
1599 Details : If there were multiple MDS->OST object precreate requests in
1600 flight, it was possible that the OST's last object id was checked
1601 outside a lock and incorrectly tripped an assertion. Move checks
1602 inside locks, and discard old precreate requests.
1605 Frequency : always, if extents are used on OSTs
1607 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1608 Details : OSTs using the extents format would not zero the high 16 bits of
1609 the index physical block number. This is not a problem for any
1610 OST filesystems smaller than 16TB, and no kernels support ext3
1611 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1612 new/modified files) and can be fixed for existing filesystems
1613 with e2fsprogs-1.39-cfs1.
1618 Description: import connection selection may be incorrect if timer wraps
1619 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1620 import connections to be ignored if the 32-bit jiffies counter
1621 wraps. Use a 64-bit jiffies counter.
1624 Frequency : very large clusters immediately after boot
1626 Description: LNET request buffers exhausted under heavy short-term load
1627 Details : If a large number of client requests are generated on a service
1628 that has previously never seen so many requests it is possible
1629 that the request buffer growth cannot keep up with the spike in
1630 demand. Instead of dropping incoming requests, they are held in
1631 the LND until the RPC service can accept more requests.
1634 Frequency : Sometimes during replay
1636 Description: Assertion failure in ll_local_open after replay.
1637 Details : If replay happened on an open request reply before we were able
1638 to set replay handler, reply will become not swabbed tripping the
1639 assertion in ll_local_open. Now we set the handler right after
1640 recognising of open request
1643 Frequency : very rare
1645 Description: kernel reports "badness in vsnprintf"
1646 Details : Reading from the "recovery_status" /proc file in small chunks
1647 may cause a negative length in lprocfs_obd_rd_recovery_status()
1648 call to vsnprintf() (which is otherwise harmless). Exit early
1649 if there is no more space in the output buffer.
1651 Severity : enhancement
1653 Description: clear OBD RPC statistics by writing to them
1654 Details : It is now possible to clear the OBD RPC statistics by writing
1655 to the "stats" file.
1660 Description: Client mtime is not the same on different clients after utimes
1661 Details : In some cases, the client was using the utimes() syscall on
1662 a file cached on another node. The clients now validate the
1663 ctime from the MDS + OSTs to determine which one is right.
1668 Description: Inability to activate failout mode
1669 Details : lconf script incorrectly assumed that in pythong string's numeric
1670 value is used in comparisons.
1673 Frequency : always with multiple stripes per file
1675 Description: Inefficient object allocation for mutli-stripe files
1676 Details : When selecting which OSTs to stripe files over, for files with
1677 a stripe count that divides evenly into the number of OSTs,
1678 the MDS is always picking the same starting OST for each file.
1679 Return the OST selection heuristic to the original design.
1684 Description: mount failures may take full timeout to return an error
1685 Details : Under some heavy load conditions it is possible that a
1686 failed mount can wait for the full obd_timeout interval,
1687 possibly several minutes, before reporting an error.
1688 Instead return an error as soon as the status is known.
1690 Frequency : quota enabled and large files being deleted
1692 Description: releasing more than 4GB of quota at once hangs OST
1693 Details : If a user deletes more than 4GB of files on a single OST it
1694 will cause the OST to spin in an infinite loop. Release
1695 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1700 Description: statfs data retrieved from /proc may be stale or zero
1701 Details : When reading per-device statfs data from /proc, in the
1702 {kbytes,files}_{total,free,avail} files, it may appear
1703 as zero or be out of date.
1706 Frequency : systems with MD RAID1 external journal devices
1708 Description: lconf's call to blkid is confused by RAID1 journal devices
1709 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1710 of returning all block devices that match the journal UUID.
1713 Frequency : always, for aggregate stripe size over 4GB
1715 Description: assertion fails when trying to use 4GB stripe size
1716 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1717 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1720 Frequency : always on ppc64
1722 Description: the first write on an ext3 filesystem with mballoc got stuck
1723 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1724 perform endianness conversion.
1726 ------------------------------------------------------------------------------
1728 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1730 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1731 INTEROPERATE with older versions automatically. Please read the
1732 user documentation before upgrading any part of a live system.
1733 * WARNING: Lustre networking configuration changes are required with
1734 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1737 * Support for kernels:
1738 2.6.9-22.0.2.EL (RHEL 4)
1739 2.6.5-7.244 (SLES 9)
1740 2.6.12.6 vanilla (kernel.org)
1743 Severity : enhancement
1744 Bugzilla : 7981/8208
1745 Description: Introduced Lustre Networking (LNET)
1746 Details : LNET is new networking infrastructure for Lustre, it includes
1747 a reorganized network configuration mode (see the user
1748 documentation for full details) as well as support for routing
1749 between different network fabrics. Lustre Networking Devices
1750 (LNDs) for the supported network fabrics have also been
1751 created for this new infrastructure.
1753 Severity : enhancement
1754 Description: Introduced Access control lists
1755 Details : clients can set ACLs on files and directories in order to have
1756 more fine-grained permissions than the standard Unix UGO+RWX.
1757 The MDS must be started with the "-o acl" mount option.
1759 Severity : enhancement
1760 Description: Introduced filesystem quotas
1761 Details : Administrators may now establish per-user quotas on the
1764 Severity : enhancement
1766 Description: Configuration change for the XT3
1767 The PTLLND is now used to run Lustre over Portals on the XT3
1768 The configure option(s) --with-cray-portals are no longer used.
1769 Rather --with-portals=<path-to-portals-includes> is used to
1770 enable building on the XT3. In addition to enable XT3 specific
1771 features the option --enable-cray-xt3 must be used.
1776 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1777 Details : A race between allocating a new llcd and re-getting the llcd_lock
1778 allowed another thread to grab newly-allocated llcd.
1780 Severity : enhancement
1782 Description: 2.6 OST async journal commit and locking fix to improve performance
1783 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1784 2.6 kernels are now async as they already were in 2.4 kernels so
1785 that they can commit concurrently with the network bulk transfer.
1786 For block-allocated files the filter allocation semaphore is held
1787 to avoid filesystem fragmentation during allocation. BKL lock
1788 removed for 2.6 xattr operations where it is no longer needed.
1793 Description: lconf incorrectly determined whether two IP networks could talk
1794 Details : In some more complicated routing and multiple-network
1795 configurations, lconf will avoid trying to make a network
1796 connection to a disjoint part of the IP space. It was doing the
1797 math incorrectly for one set of cases.
1802 Description: Fix for potential infinite loop processing records in an llog.
1803 Details : If an llog record is corrupted/zeroed, it is possible to loop
1804 forever in llog_process(). Validate the llog record length
1805 and skip the remainder of the block on error.
1808 Frequency : occasional (liblustre only)
1810 Description: liblustre could not open files whose last component is a symlink
1811 Details : sysio_path_walk() would incorrectly pass the open intent to
1812 intermediate path components.
1815 Frequency : rare (liblustre only with non-standard tuning)
1816 Bugzilla : 7201 (7350)
1817 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1818 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1819 after it has been released, only do this for referenced locks
1821 Severity : enhancement
1823 Description: specifying an (invalid) directory default stripe_size of -1
1824 would reset the directory default striping
1825 Details : stripe_size -1 was used internally to signal directory stripe
1826 removal, now use "all default" to signal dir stripe removal
1827 as a directory striping of "all default" is not useful
1830 Frequency : common for large clusters running liblustre clients
1832 Description: doing an ls when liblustre clients are running is slow
1833 Details : sending a glimpse AST to a liblustre client waits for every AST
1834 to time out, as liblustre clients will not respond. Since they
1835 cannot cache data we refresh the OST lock LVB from disk instead.
1837 Severity : enhancement
1839 Description: doing an ls at the same time as file IO can be slow
1840 Details : enqueue and other "small" requests can be blocked behind many
1841 large IO requests. Create a new OST IO portal for non-IO
1842 requests so they can be processed faster.
1845 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1847 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1848 created file with mode 000
1849 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1850 allow an ftruncate() as a non-root user to a file with mode 000.
1851 Lustre now respects this flag to disable mode checks when
1852 truncating a file owned by the user
1855 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1858 Description: Revoking locks from clients that went dead or catatonic might take
1860 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1861 cancellation of such locks instant on servers without waiting for
1862 any reply from clients. Clients drops these locks when cancel
1863 notification from server is received without replying.
1866 Frequency : liblustre-only, when liblustre client dies or becomes busy
1868 Description: Doing ls on Linux clients can take a long time with active
1870 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1871 granting such locks to it in the first place if possible. Locks
1872 are taken by proxy on the OST during the read or write and
1873 dropped immediately afterward. Add connect flags handling, do
1874 not grant locks to liblustre clients for glimpse ASTs.
1876 Severity : enhancement
1878 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1879 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1880 sequential reads and behaves badly with applications doing random
1881 reads. Improve it by reading ahead at least read region, and
1882 avoiding excessive large RPC for small reads.
1884 Severity : enhancement
1886 Description: Creating more than 1000 files for a single job may cause a load
1887 imbalance on the OSTs if there are also a large number of OSTs.
1888 Details : qos_prep_create() uses an OST index reseed value that is an
1889 even multiple of the number of available OSTs so that if the
1890 reseed happens in the middle of the object allocation it will
1891 still utilize the OSTs as uniformly as possible.
1896 Description: OST or MDS may oops in ping_evictor_main()
1897 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1898 but doesn't restart at beginning of obd_exports_timed list
1901 Severity : enhancement
1903 Description: improve by-nid export eviction on the MDS and OST
1904 Details : allow multiple exports with the same NID to be evicted at one
1905 time without re-searching the exports list.
1908 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1910 Description: MDS may oops in groups_free()
1911 Details : in rare race conditions a newly allocated group_info struct is
1912 freed again, and this can be NULL. The 2.4 compatibility code
1913 for groups_free() checked for a NULL pointer, but 2.6 did not.
1916 Frequency : common for liblustre clients doing little filesystem IO
1917 Bugzilla : 9352, 7313
1918 Description: server may evict liblustre clients accessing contended locks
1919 Details : if a client is granted a lock or receives a completion AST
1920 with a blocking AST already set it would not reply to the AST
1921 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1924 Frequency : lfs setstripe, only systems with more than 160 OSTs
1926 Description: unable to set striping with a starting offset beyond OST 160
1927 Details : llapi_create_file() incorrectly limited the starting stripe
1928 index to the maximum single-file stripe count.
1931 Frequency : LDAP users only
1933 Description: lconf did not handle in-kernel recovery with LDAP properly
1934 Details : lconf/LustreDB get_refs() is searching the wrong namespace
1936 Severity : enhancement
1938 Description: bind OST threads to NUMA nodes to improve performance
1939 Details : all OST threads are uniformly bound to CPUs on a single NUMA
1940 node and do their allocations there to localize memory access
1942 Severity : enhancement
1944 Description: llmount can determine client NID directly from Myrinet (GM)
1945 Details : the client NID code from gmnalnid was moved directly into
1946 llmount, removing the need to use this or specifying the
1947 client NID explicitly when mounting GM clients with zeroconf
1950 Frequency : if client is started with down MDS
1952 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
1953 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
1954 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
1955 ptlrpc_import_delayed_req() because we didn't check if the
1956 request was interrupted, and we also didn't break out of the
1957 event loop if there was no timeout
1962 Description: data loss during non-page-aligned writes to a single file from
1963 both multiple nodes and multiple threads on one node at same time
1964 Details : updates to KMS and lsm weren't protected by common lock. Resulting
1965 inconsistency led to false short-reads, that were cached and later
1966 used by ->prepare_write() to fill in partially written page,
1967 leading to data loss.
1970 Frequency : always, if lconf --abort_recovery used
1972 Description: lconf --abort_recovery fails with 'Operation not supported'
1973 Details : lconf was attempting to abort recovery on the MDT device and not
1976 Severity : enhancement
1978 Description: remove cleanup logs
1979 Details : replace lconf-generated cleanup logs with lustre internal
1980 cleanup routines. Eliminates the need for client-cleanup and
1983 Severity : enhancement
1985 Description: add support for EAs (user and system) on lustre filesystems
1986 Details : it is now possible to store extended attributes in the Lustre
1987 client filesystem, and with the user_xattr mount option it
1988 is possible to allow users to store EAs on their files also
1990 Severity : enhancement
1992 Description: Add possibility (config option) to show minimal available OST free
1994 Details : When compiled with --enable-mindf configure option, statfs(2)
1995 (and so, df) will return least minimal free space available from
1996 all OSTs as amount of free space on FS, instead of summary of
1997 free spaces of all OSTs.
1999 Severity : enhancement
2001 Description: do not expand extent locks acquired on OST-side
2002 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2003 by server: they are not cached anyway.
2006 Frequency : when mmap is used/binaries executed from Lustre
2008 Description: Unmmap pages before throwing them away from read cache.
2009 Details : llap_shrink cache now attempts to unmap pages before discarding
2010 them (if unmapping failed - do not discard). SLES9 kernel has
2011 extra checks that trigger if this unmapping is not done first.
2016 Description: lconf didn't resolve symlinks before checking to see whether a
2017 given mountpoint was already in use
2020 Frequency : when migrating failover services
2021 Bugzilla : 6395, 9514
2022 Description: When migrating a subset of services from a node (e.g. failback
2023 from a failover service node) the remaining services would
2024 time out and evict clients.
2025 Details : lconf --force (implied by --failover) sets the global obd_timeout
2026 to 5 seconds in order to quickly disconnect, but this caused
2027 other RPCs to time out too quickly. Do not change the global
2028 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2030 Severity : enhancement
2031 Frequency : if MDS is started with down OST
2032 Bugzilla : 9439,5706
2033 Description: Allow startup/shutdown of an MDS without depending on the
2034 availability of the OSTs.
2035 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2036 Add appropriate locking and lov-osc refcounts for safe
2037 cleaning. Add osc abort_inflight calls in case the
2038 synchronize never started.
2041 Frequency : occasional (Cray XT3 only)
2043 Description: root not authorized to access files in CRAY_PORTALS environment
2044 Details : The client process capabilities were not honoured on the MDS in
2045 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2046 been accessed by an authorized user then root was able to access
2047 the file on the local client also. The root user capabilities
2048 are now allowed on the MDS, as this environment has secure UID.
2051 Frequency : occasional
2053 Description: ldiskfs "too long searching" message happens too often
2054 Details : A debugging message (otherwise harmless) prints too often on
2055 the OST console. This has been reduced to only happen when
2056 there are fragmentation problems on the filesystem.
2061 Description: Division by zero in statfs when all OSCs are inactive
2062 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2063 lov_statfs divides by value returned by lov_get_stripecnt().
2067 Bugzilla : 9489, 3273
2068 Description: First write from each client to each OST was only 4kB in size,
2069 to initialize client writeback cache, which caused sub-optimal
2070 RPCs and poor layout on disk for the first writen file.
2071 Details : Clients now request an initial cache grant at (re)connect time
2072 and so that they can start streaming writes to the cache right
2073 away and always do full-sized RPCs if there is enough data.
2074 If the OST is rebooted the client also re-establishes its grant
2075 so that client cached writes will be honoured under the grant.
2080 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2081 Details : Now I/O RPCs go to different portal number and (presumably) fast
2082 lock requests (and glimses) and other RPCs get their own service
2083 threads pool that should be able to service those RPCs
2086 Severity : enhancement
2088 Description: Ability to exchange lustre version between client and servers and
2089 issue warnings at client side if client is too old. Also for
2090 liblustre clients there is ability to refuse connection of too old
2092 Details : New 'version' field is added to connect data structure that is
2093 filled with version info. That info is later checked by server and
2097 Frequency : rare, liblustre only.
2098 Bugzilla : 9296, 9581
2099 Description: Two simultaneous writes from liblustre at offset within same page
2100 might proceed at the same time overwriting eachother with stale
2102 Details : I/O lock withing llu_file_prwv was released too early, before data
2103 actually was hitting the wire. Extended lock-holding time until
2104 server acknowledges receiving data.
2107 Frequency : extremely rare. Never observed in practice.
2109 Description: avoid generating lustre_handle cookie of 0.
2110 Details : class_handle_hash() generates handle cookies by incrementing
2111 global counter, and can hit 0 occasionaly (this is unlikely, but
2112 not impossible, because initial value of cookie counter is
2113 selected randonly). Value of 0 is used as a sentinel meaning
2114 "unassigned handle" --- avoid it. Also coalesce two critical
2115 sections in this function into one.
2117 Severity : enhancement
2119 Description: allow liblustre clients to delegate truncate locking to OST
2120 Details : To avoid overhead of locking, liblustre client instructs OST to
2121 take extent lock in ost_punch() on client's behalf. New connection
2122 flag is added to handle backward compatibility.
2124 Severity : enhancement
2125 Bugzilla : 4928, 7341, 9758
2126 Description: allow number of OST service threads to be specified
2127 Details : a module parameter allows the number of OST service threads
2128 to be specified via "options ost ost_num_threads={N}" in the
2129 OSS's /etc/modules.conf or /etc/modprobe.conf.
2133 Bugzilla : 6146, 9635, 9895
2134 Description: servers crash with bad pointer in target_handle_connect()
2135 Details : In rare cases when a client is reconnecting it was possible that
2136 the connection request was the last reference for that export.
2137 We would temporarily drop the export reference and get a new
2138 one, but this may have been the last reference and the export
2139 was just destroyed. Get new reference before dropping old one.
2141 Severity : enhancement
2142 Frequency : if client is started with failover MDS
2144 Description: Allow multiple MDS hostnames in the mount command
2145 Details : Try to read the configuration from all specified MDS
2146 hostnames during a client mount in case the "primary"
2149 Severity : enhancement
2151 Description: Stop sending data to evicted clients as soon as possible.
2152 Details : Check if the client we are about to send or are sending data to
2153 was evicted already. (Check is done every second of waiting,
2154 for which l_wait_event interface was extended to allow checking
2155 of exit condition at specified intervals).
2158 Frequency : rare, normally only when NFS exporting is done from client
2160 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2162 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2165 Frequency : always, if srand() is called before liblustre initialization
2167 Description: Liblustre uses system PRNG disturbing its usage by user application
2168 Details : Introduce internal to lustre fast and high-quality PRNG for
2169 lustre usage and make liblustre and some other places in generic
2170 lustre code to use it.
2172 Severity : enhancement
2173 Bugzilla : 9477, 9557, 9870
2174 Description: Verify that the MDS configuration logs are updated when xml is
2175 Details : Check if the .xml configuration logs are newer than the config
2176 logs stored on the MDS and report an error if this is the case.
2177 Request --write-conf, or allow starting with --old_conf.
2179 Severity : enhancement
2181 Description: Handle symlinks in the path when checking if Lustre is mounted.
2182 Details : Resolve intermediate symlinks when checking if a client has
2183 mounted a filesystem to avoid duplicate client mounts.
2188 Description: lconf can hit an error exception but still return success.
2189 Details : The lconf command catches the Command error exception at the top
2190 level script context and will exit with the associated exit
2191 status, but doesn't ensure that this exit status is non-zero.
2196 Description: failure of ptlrpc thread startup can cause oops
2197 Details : Starting a ptlrpc service thread can fail if there are a large
2198 number of threads or the server memory is very fragmented.
2199 Handle this without oopsing.
2202 Frequency : always, only if liblustre and non-default acceptor port was used
2204 Description: liblustre cannot connect to servers with non-default acceptor port
2205 Details : tcpnal_set_default_params() was not called and was therefore
2206 ignoring the environment varaible TCPNAL_PORT, as well as other
2207 TCPNAL_ environment variables
2212 Description: two objects could be created on the same OST for a single file
2213 Details : If an OST is down, in some cases it was possible to create two
2214 objects on a single OST for a single file. No problems other
2215 than potential performance impact and spurious error messages.
2219 Bugzilla : 5681, 9562
2220 Description: Client may oops in ll_unhash_aliases
2221 Details : Client dcache may become inconsistent in race condition.
2222 In some cases "getcwd" can fail if the current directory is
2228 Description: Inode refcounting problems in NFS export code
2229 Details : link_raw functions used to call d_instantiate without obtaining
2230 extra inode reference first.
2234 Bugzilla : 9942, 9903
2235 Description: Referencing freed requests leading to crash, memleaks with NFS.
2236 Details : We used to require that call to ll_revalidate_it was always
2237 followed by ll_lookup_it. Also with revalidate_special() it is
2238 possible to call ll_revalidate_it() twice for the same dentry
2239 even if first occurence returned success. This fix changes semantic
2240 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2241 reference on a request referred from the intent.
2242 ll_intent_release() then releases such a request.
2245 Frequency : rare, normally benchmark loads only
2247 Description: unlinked inodes were kept in memory on the client
2248 Details : If a client is repeatedly creating and unlinking files it
2249 can accumulate a lot of stale inodes in the inode slab cache.
2250 If there is no other client load running this can cause the
2251 client node to run out of memory. Instead flush old inodes
2252 from client cache that have the same inode number as a new inode.
2255 Frequency : SLES9 2.6.5 kernel and long filenames only
2256 Bugzilla : 9969, 10379
2257 Description: utime reports stale NFS file handle
2258 Details : SLES9 uses out-of-dentry names in some cases, which confused
2259 the lustre dentry revalidation. Change it to always use the
2263 Frequency : rare, unless heavy write-truncate concurrency is continuous
2264 Bugzilla : 4180, 6984, 7171, 9963, 9331
2265 Description: OST becomes very slow and/or deadlocked during object unlink
2266 Details : filter_destroy() was holding onto the parent directory lock
2267 while truncating+unlinking objects. For very large objects this
2268 may block other threads for a long time and slow overall OST
2269 responsiveness. It may also be possible to get a lock ordering
2270 deadlock in this case, or run out of journal credits because of
2271 the combined truncate+unlink. Solution is to do object truncate
2272 first in one transaction without parent lock, and then do the
2273 final unlink in a new transaction with the parent lock. This
2274 reduces the lock hold time dramatically.
2277 Frequency : rare, 2.4 kernels only
2279 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2280 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2281 before unmounting their filesystem in order to avoid deadlock.
2282 The kernel_locked() function in 2.4 kernels only checks whether
2283 the kernel lock is held, not whether it is this process that is
2284 holding it as 2.6 kernels do.
2289 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2290 Details : The client ptlrpc code may be trying to reconnect to a down
2291 server before a previous connection attempt has timed out.
2292 Increase the reconnect interval to be longer than the connection
2293 timeout interval to avoid sending duplicate connections to
2297 Frequency : echo_client brw_test command
2299 Description: fix echo_client to work with OST preallocated code
2300 Details : OST preallocation code (5137) didn't take echo_client IO path
2301 into account: echo_client calls filter methods outside of any
2302 OST thread and, hence, there is no per-thread preallocated
2303 pages and buffers to use. Solution: hijack pga pages for IO. As
2304 a byproduct, this avoids unnecessary data copying.
2308 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2309 Description: Client can oops in mdc_commit_close() after open replay
2310 Details : It was possible for the MDS to return an open request with no
2311 transaction number in mds_finish_transno() if the client was
2312 evicted, but without actually returning an error. Clients
2313 would later try to replay that open and may trip an assertion
2314 Simplify the client close codepath, and always return an error
2315 from the MDS in case the open is not successful.
2318 Frequency : rare, 2.6 OSTs only
2320 Description: OST may deadlock under high load on fragmented files
2321 Details : If there was a heavy load and highly-fragmented OST filesystems
2322 it was possible to have all the OST threads deadlock waiting on
2323 allocation of biovecs, because the biovecs were not released
2324 until the entire RPC IO was completed. Instead, release biovecs
2325 as soon as they are complete to ensure forward IO progress.
2327 Severity : enhancement
2329 Description: Support for specifying external journal device at mount
2330 Details : If an OST or MDS device is formatted with an external journal
2331 device, this device major/minor is stored in the ext3 superblock
2332 and may not be valid for failover. Allow detecting and
2333 specifying the external journal at mount time.
2338 Description: Mounting an MDS with pending unlinked files may cause oops
2339 Details : target_finish_recovery() calls mds_postrecov() which returned
2340 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2341 considers this an error and immediately begins cleaning up the
2342 lov, just after starting the mds_lov process
2344 Severity : enhancement
2346 Description: Implement 'lfs df' to report actual free space on per-OST basis
2347 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2348 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2349 usage of objects; '-h' to report in human readable format.
2351 ------------------------------------------------------------------------------
2353 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2360 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2361 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2362 at startup during orphan unlinking. Instead, skip buddy update
2363 before setup, as it will be regenerated after recovery is complete.
2368 Description: If an OST is inactive, its locks might reference stale inodes.
2369 Details : lov_change_cbdata() must iterate over all namespaces, even if
2370 they are inactive to clear inode references from the lock.
2372 Severity : enhancement
2373 Frequency : occasional, if non-standard max_dirty_mb used
2375 Description: Client will block write RPCs if not enough grant
2376 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2377 then the client will block writes while waiting for another RPC
2378 to complete instead of consuming its dirty limit. With change
2379 we get improved performance when max_dirty_mb is small.
2381 Severity : enhancement
2382 Bugzilla : 3389, 6253
2383 Description: Add support for supplementary groups on the MDS.
2384 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2385 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2386 lookups for user supplementary groups into a cache.
2390 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2391 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2392 for liblustre O_EXCL is enforced by MDS.
2396 Description: readdir never returns NULL in liblustre.
2397 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2398 dentry in struct dirent.
2402 Frequency : liblustre only, depends on application IO pattern
2403 Description: liblustre clients evicted if not contacting servers
2404 Details : Don't put liblustre clients into the ping_evictor list, so
2405 they will not be evicted by the pinger ever.
2407 Severity : enhancement
2409 Description: Add ability to evict clients by NID from MDS.
2410 Details : By echoing "nid:$NID" string into
2411 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2412 $NID would be instantly evicted from this MDS and from all active
2413 OSTs connected to it.
2417 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2418 Details : lookup_it_finish() used to query file size from OSTs that was not
2423 Description: service threads change working directory to that of init
2424 Details : Starting lustre service threads may pin the working directory
2425 of the parent thread, making that filesystem busy. Threads
2426 now change to the working directory of init to avoid this.
2430 Frequency : during shutdown only
2431 Description: shutdown with a failed MDS or OST can cause unmount to hang
2432 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2437 Frequency : 2.6 only
2438 Description: chmod/chown may include an extra supplementary group
2439 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2440 supplementary group and if none is specified this is used.
2443 Bugzilla : 5479 (6816)
2445 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2446 Details : If lookup is in progress on a file that is unlinked we might try
2447 to revalidate the inode and fail in revalidate after lookup is
2448 complete and ll_file_open() enqueues the open again but
2449 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2452 Frequency : always, if lconf --abort_recovery used
2454 Description: lconf --abort_recovery fails with 'Operation not supported'
2455 Details : lconf was attempting to abort recovery on the MDT device and not
2458 ------------------------------------------------------------------------------
2460 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2465 Frequency : rare (only unsupported configurations with a node running as an
2467 Bugzilla : 6514, 5137
2468 Description: Mounting a Lustre file system on a node running as an OST could
2470 Details : OSTs now preallocates memory needed to write out data at
2471 startup, instead of when needed, to avoid having to
2472 allocate memory in possibly low memory situations.
2473 Specifically, if the file system is mounted on on OST,
2474 memory pressure could force it to try to write out data,
2475 which it needed to allocate memory to do. Due to the low
2476 memory, it would be unable to do so and the node would
2477 become unresponsive.
2479 Severity : enhancement
2481 Description: Addition of lconf --service command line option
2482 Details : lconf now accepts a '--service <arg>' option, which is
2483 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2485 Severity : enhancement
2487 Description: Failover mode is now the default for OSTs.
2488 Details : By default, OSTs will now run in failover mode. To return to
2489 the old behaviour, add '--failout' to the lmc line for OSTs.
2491 Severity : enhancement
2493 Description: Health checks are now provided for MDS and OSTs
2494 Details : Additional detailed health check information on MSD and OSTs
2495 is now provided through the procfs health_check value.
2498 Frequency : occasional, depends on IO load
2500 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2501 after numerous create/delete cycles
2502 Details : The ext3 inode allocation policy would not allocate new inodes
2503 very well on the OSTs because there are no new directories
2504 being created. Instead we look for groups with free space if
2505 the parent directories are nearly full.
2510 Description: Network or server problems during mount may cause partially
2511 mounted clients instead of returning an error.
2512 Details : The config llog parsing code may overwrite the error return
2513 code during mount error handling, returning success instead
2519 Description: MDS can fail to allocate large reply buffers
2520 Details : After long uptimes the MDS can fail to allocate large reply
2521 buffers (e.g. zconf client mount config records) due to memory
2522 fragmentation or consumption by the buffer cache. Preallocate
2523 some large reply buffers so that these replies can be sent even
2524 under memory pressure.
2528 Frequency : rare (liblustre)
2529 Description: fsx running with liblustre complained that using truncate() to
2530 extend the file doesn't work. This patch corrects that issue.
2531 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2532 ATTR_SIZE and lsm use in llu_setattr_raw.
2536 Frequency : rare, only 2.6 kernels
2537 Description: Unusual file access patterns on the MDS may result in inode
2538 data being lost in very rare circumstances.
2539 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2540 code in the 2.6 kernel caused the inode and/or EA data not to
2541 be read from disk, causing single-file corruption.
2545 Frequency : rare, only 2.6 filesystems using extents
2546 Description: Heavy concurrent write and delete load may cause data corruption.
2547 Details : It was possible under high-load situations to have an extent
2548 metadata block in the block device cache from a just-unlinked
2549 file overwrite a newly-allocated data block. We now unmap any
2550 metadata buffers that alias just-allocated data blocks.
2554 Frequency : filesystems with default stripe_count larger than 77
2555 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2556 Details : lconf specifies an inode size of 4096 bytes when the default
2557 stripe_count is larger than 77. This conflicts with the default
2558 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2559 this case to avoid pinning too much memory for large EAs.
2561 ------------------------------------------------------------------------------
2563 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2568 Frequency : rare (extremely heavy IO load with hundreds of clients)
2570 Description: Client is evicted, gets IO error writing to file
2571 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2572 caused clients to be evicted for AST timeouts. The fixes in
2573 bug 5192 mean we no longer need to have such short AST timeouts
2574 so ldlm_timeout has been increased.
2577 Frequency : occasional during --force or --failover shutdown under load
2578 Bugzilla : 5949, 4834
2579 Description: Server oops/LBUG if stopped with --force or --failover under load
2580 Details : a collection of import/export refcount and cleanup ordering
2581 issues fixed for safer force cleanup
2584 Frequency : only filesystems larger than 120 OSTs
2585 Bugzilla : 5990, 6223
2586 Description: lfs getstripe would oops on a very large filesystem
2587 Details : lov_getconfig used kfree on vmalloc'd memory
2590 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2591 Bugzilla : 6242, 6243
2592 Description: reading from files that had been truncated to a non-zero size
2593 but never opened returned no data
2594 Details : ll_file_read() reads zeros from no-object files to EOF
2599 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2600 Details : There's little harm in aborting MDS/OSS recovery and letting it
2601 try again, so I removed the LASSERT and return an error instead.
2603 Severity : enhancement
2605 Description: New debugging infrastructure for tracking down data corruption
2606 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2607 (b) cover more of the client-side code path, and (c) try to narrow
2608 down where problems occurred
2612 Bugzilla : 3819, 4364, 4397, 6313
2613 Description: Racing close and eviction MDS could cause assertion in mds_close
2614 Details : It was possible to get multiple mfd references during close and
2615 client eviction, leading to one thread referencing a freed mfd.
2617 Severity: : enhancement
2618 Bugzilla : 3262, 6359
2619 Description: Attempts to reconnect to servers are now more aggressive.
2620 Details : This builds on the enhanced upcall-less recovery that was added
2621 in 1.4.2. When trying to reconnect to servers, clients will
2622 now try each server in the failover group every 10 seconds. By
2623 default, clients would previously try one server every 25 seconds.
2628 Description: After recovery, certain operations trigger a failed
2629 assertion on a client.
2630 Details : Failing over an mds, using lconf -d --failover, while a
2631 client was doing a readdir() call would cause the client to
2632 LBUG after recovery completed and the readdir() was resent.
2634 Severity : enhancement
2636 Description: Default groups are now added by lconf
2637 Details : You can now run lconf --group <servicename> without having to
2638 manually add groups with lmc.
2641 Frequency : occasional
2643 Description: Nodes with an elan id of 0 trigger a failed assertion
2646 Frequency : always when accessing e.g. tty/console device nodes
2648 Description: tty and some other devices nodes cannot be used on lustre
2649 Details : file's private_data field is used by device data and lustre
2650 values in there got lost. New field was added to struct file to
2651 store fs-specific private data.
2654 Frequency : when exporting Lustre via NFS
2656 Description: NFSD failed occasionally when looking up a path component
2657 Details : NFSD is looking up ".." which was broken in ext3 directories
2658 that had grown large enough to become hashed.
2661 Frequency : Clusters with multiple interfaces not on the same subnet
2663 Description: Nodes will repeatedly try to reconnect to an interface which it
2664 cannot reach and report an error to the log.
2665 Details : Extra peer list entries will be created by lconf with some peers
2666 unreachable. lconf now validates the peer before adding it.
2669 Frequency : Only if a default stripe is set on the filesystem root.
2671 Description: Setting a default stripe on the filesystem root prevented the
2672 filesystem from being remounted.
2673 Details : The client was sending extra request flags in the root getattr
2674 request and did not allocate a reply buffer for the dir EA.
2677 Frequency : occasional, higher if lots of files are accessed by one client
2678 Bugzilla : 6159, 6097
2679 Description: Client trips assertion regarding lsm mismatch/magic
2680 Details : While revalidating inodes the VFS looks up inodes with ifind()
2681 and in rare cases can find an inode that is being freed.
2682 The ll_test_inode() code will free the lsm during ifind()
2683 when it finds an existing inode and then the VFS later attaches
2684 this free lsm to a new inode.
2688 Bugzilla : 6422, 7030
2689 Description: MDS deadlock between mkdir and client eviction
2690 Details : Creating a new file via mkdir or mknod (starting a transaction
2691 and getting the ns lock) can deadlock with client eviction
2692 (gets ns lock and trying to finish a synchronous transaction).
2695 Frequency : occasional
2696 Description: While starting a server, the fsfilt_ext3 module could not be
2698 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2699 kernels. Previously, lconf would still use the ext3 name
2700 when trying to load modules. Now, it will correctly use
2701 ext3 on 2.4 and ldiskfs on 2.6.
2703 Severity : enhancement
2704 Description: The default stripe count has been changed to 1
2705 Details : The interpretation of the default stripe count (0, to lfs
2706 or lmc) has been changed to mean striping across a single
2707 OST, rather than all available. For general usage we have
2708 found a stripe count of 1 or 2 works best.
2710 Severity : enhancement
2711 Description: Add support for compiling against Cray portals.
2712 Details : Conditional compiling for some areas that are different
2716 Frequency : occasional
2717 Bugzilla : 6409, 6834
2718 Description: Creating files with an explicit stripe count may lead to
2719 a failed assertion on the MDS
2720 Details : If some OSTs are full or unavailable, creating files may
2721 trigger a failed assertion on the MDS. Now, Lustre will
2722 try to use other servers or return an error to the
2726 Frequency : occasional
2728 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2729 caused lock timeout during readahead (since 1.4.2).
2730 Details : Processes doing readahead might match a lock that hasn't been
2731 granted yet if there are overlapping and conflicting lock
2732 requests. The readahead process waits on ungranted lock
2733 (original lock is CBPENDING), while OST waits for that process
2734 to cancel CBPENDING read lock and eventually evicts client.
2736 Severity : enhancement
2738 Description: Initial enabling of flock support for clients
2739 Details : Implements fcntl advisory locking and file status functions.
2740 This feature is provided as an optional mount flag (default
2741 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2742 locking are implemented yet, and those that are are not guaranteed
2743 to be completely correct in production environments.
2744 mount -t lustre -o [flock|noflock] ...
2747 Frequency : occasional
2749 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2750 assertion in the ext3 JBD (journaling) layer.
2751 Details : The b_committed_data struct is protected by the big kernel lock
2752 in 2.4 kernels, serializing journal_commit_transaction() and
2753 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2754 access to this struct. In 2.6 kernels there is finer grained
2755 locking to improve SMP performance of the JBD layer.
2759 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2761 -----------------------------------------------------------------------------
2763 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2765 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2766 versions of Lustre. Please update all servers and clients to
2767 version 1.4.2 or later at the same time. You must also run
2768 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2769 to update the configuration logs.
2771 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2772 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2773 - swab llog records of type '0' so we get proper header size/idx (5861)
2774 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2775 - fix rename of one directory over another leaking an inode (5953)
2776 - avoid SetPageDirty on 2.6 (5981)
2777 - don't re-add just-being-destroyed locks to the waiting list (5653)
2778 - when creating new directories, inherit the parent's custom
2779 striping settings if present parent (3048)
2780 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2781 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2782 - don't hold client locks on temporary worklist from l_lru (5666)
2783 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2784 - automatically evict dead clients (5921)
2785 - Update file size properly in create+truncate+fstat case (6196)
2786 - Do not unhash mountpoint dentries, do not allow removal of
2788 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2789 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2790 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2791 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2792 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2793 - Correctly handle reads of files with no objects (6243)
2794 - lctl recover will also mark a device active if deactivate used (5933)
2796 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2797 - allow --write-conf on an MDS with different nettype than client (5619)
2798 - don't write config llogs to MDS for mounts not from that MDS (5617)
2799 - lconf should create multiple TCP connections from a client (5201)
2800 - init scripts are now turned off by default; run chkconfig --on
2801 lustre and chkconfig --on lustrefs to use them
2802 - upcalls are no longer needed for clients to recover to failover
2804 - add --abort-recovery option to lconf to abort recovery on device
2806 - add support for an arbitrary number of OSTs (3026)
2807 - Quota support protocol changes.
2808 - forward compatibility changes to wire structs (6007)
2809 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2810 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2811 - improved "lustre status" script
2812 - initialize blocksize for non-regular files (6062)
2813 - added --disable-server and --disable-client configure options (5782)
2814 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2815 - Vanilla 2.4.29 support
2816 - increase maximum number of obd devices to 520 (6242)
2817 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2818 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2819 SLES 9 2.6.5 kernels
2820 - limit stripes per file to 160 (the maximum EA size) (6093)
2822 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2825 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2826 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2827 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2828 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2829 - small ext3 extents cleanups and fixes (5733)
2830 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2831 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2832 - avoid needless client->OST connect, fix handle mismatch (5317)
2833 - fix DLM error path that led to out-of-sync client, long delays (5779)
2834 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2835 - fix several locking issues related to i_size (5492,5624,5654,5672)
2836 - don't move pending lock onto export if it is already evicted (5683)
2837 - fix kernel oops when creating .foo in unlinked directory (5548)
2838 - fix deadlock in obdfilter statistics vs. object create (5811)
2839 - use time_{before,after} to avoid timer jiffies wrap (5882)
2840 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2841 - Do not leak request if server was not able to process it (5154)
2842 - If mds_open unable to find parent dir, make that negative lookup(5154)
2843 - don't create new directories with extent-mapping (5909, 5936)
2845 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2846 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2847 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2848 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2849 - support for RHEL4 kernel on i686 (5773)
2850 - provide error messages when incompatible logs are encountered (5898)
2852 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2853 * version 1.4.0.10 (1.4.1 release candidate 1)
2855 - don't keep a lock reference when lock is not granted (4238)
2856 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2857 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2858 - drop import inflight refcount on signal_completed_replay error (5255)
2859 - unlock page after async write error during send (3677)
2860 - handle missing objects in filter_preprw_read properly (5265)
2861 - no transno return for symlink open, don't save no-trasno open (3440)
2862 - don't try to complete elan receive that already failed (4012)
2863 - free RPC server reply state on error (5406)
2864 - clean up thread from ptlrpc_start_thread() on error (5160)
2865 - readahead could read extra page into cache that wasn't ejected (5388)
2866 - prevent races in class_attach/setup/cleanup/detach (5260)
2867 - don't dereference de->d_inode after l_dput of de (5458)
2868 - use "int" for stripe value returned from lock_to_stripe (5544)
2869 - mballoc allocation and error-checking fixes in 2.6 (5504)
2870 - block device patches to fix I/O request sizes in 2.6 (5482)
2871 - look up hostnames for IB nals (5602)
2872 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2873 - don't start multiple acceptors for the same port (5277)
2874 - fix incorrect LASSERT in mds_getattr_name (5635)
2875 - export a proc file for general "ping" checking (5628)
2876 - fix "lfs check" to not block when the MDS is down (5628)
2878 - service request history (4965)
2879 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2880 - create an "evict_client" /proc entry on OSTs, like the MDS has
2881 - fix mount usage message, return errors per mount(8) (5168)
2882 - change grep [] to grep "[]" in tests so they work in more UMLs
2883 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2884 - remove ancient LOV_MAGIC_V0 EA support (5047)
2885 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2886 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2887 - allow readahead allocations to fail when low on memory (5383)
2888 - mmap locking landed again, after considerable improvement (2828)
2889 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2891 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2894 - send OST transaction number in read/write reply to free req (4966)
2895 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2896 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2897 - the watchdog thread now runs as interruptible (5246)
2898 - flock/lockf fixes (but it's still disabled, pending 5135)
2899 - don't use EXT3 constants in llite code (5094)
2900 - memory shortage at startup could cause assertion (5176)
2902 - reorganization of lov code
2903 - single portals codebase
2905 - add extents/mballoc support (5025)
2906 - direct I/O reads in the obdfilter (4048)
2907 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2909 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2912 - send OST transaction number in read/write reply to free req (4966)
2913 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2914 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2915 - the watchdog thread now runs as interruptible (5246)
2916 - handle missing objects in filter_preprw_read properly (5265)
2917 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2918 - drop import inflight refcount on signal_completed_replay error (5255)
2919 - unlock page after async write error during send (3677)
2920 - return original error code on reconstructed replies (3761)
2921 - no transno return for symlink open, don't save no-trasno open (3440)
2923 - add pid to ldlm debugging output (4922)
2924 - bump the watchdog timeouts -- we can't handle 30sec yet
2925 - extra debugging for orphan dentry/inode bug (5259)
2927 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
2930 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
2931 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
2932 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
2933 - fix NULL dereference in /proc/sys/portals/routes (4827)
2934 - allow failed mdc_close() operations to be interrupted (4561)
2935 - stop precreate on OST before MDS would time out on it (4778)
2936 - don't send partial-page writes before EOF from client (4410)
2937 - discard client grant for sub-page writes on large-page clients (4520)
2938 - don't free dentries not owned by NFS code, check generation (4806)
2939 - fix lsm leak if mds_create_objects() fails (4801)
2940 - limit debug_daemon file size, always print CERROR messages (4789)
2941 - use transno after validating reply (3892)
2942 - process timed out requests if import state changes (3754)
2943 - update mtime on OST during writes, return in glimpse (4829)
2944 - add mkfsoptions to LDAP (4679)
2945 - use ->max_readahead method instead of zapping global ra (5039)
2946 - don't interrupt __l_wait_event() during strace
2948 - add software watchdogs to catch hung threads quickly (4941)
2949 - make lustrefs init script start after nfs is mounted
2950 - fix CWARN/ERROR duplication (4930)
2951 - return async write errors to application if possible (2248)
2952 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
2953 - print NAL number in %x format (4645)
2954 - update barely-supported suse-2.4.21-171 series (4842)
2955 - support for sles 9 %post scripts
2956 - support for building 2.6 kernel-source packages
2957 - support for sles km_* packages
2959 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
2962 - ignore -ENOENT errors in osc_destroy (3639)
2963 - notify osc create thread that OSC is being cleaned up (4600)
2964 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
2965 - reconstruct ost_handle() like mds_handle() (4657)
2966 - create a new thread to do import eviction to avoid deadlock (3969)
2967 - let lconf resolve symlinked-to devices (4629)
2968 - don't unlink "objects" from directory with default EA (4554)
2969 - hold socknal file ref over connect in case target is down (4394)
2970 - allow more than 32000 subdirectories in a single directory (3244)
2971 - fix blocks count for O_DIRECT writes (3751)
2972 - OST returns ENOSPC from object create when no space left (4539)
2973 - don't send truncate RPC if file size isn't changing (4410)
2974 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
2975 - bind to privileged port in socknal and tcpnal (3689)
2977 - rate limit CERROR/CWARN console message to avoid overload (4519)
2978 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
2979 - basic mmap support (3918)
2980 - kernel patch series update from b1_4 (4711)
2982 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
2985 - avoid crash during MDS cleanup with OST shut down (2775)
2986 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
2987 - don't use bad inodes on the MDS (3744)
2988 - dynamic object preallocation to improve recovery speed (4236)
2989 - don't hold spinlock over lock dumping or change debug flags (4401)
2990 - don't zero obd_dev when it is force cleaned (3651)
2991 - print grants to console if they go negative (4431)
2992 - "lctl deactivate" will stop automatic recovery attempts (3406)
2993 - look for existing locks in ldlm_handle_enqueue() (3764)
2994 - don't resolve lock handle twice in recovery avoiding race (4401)
2995 - revalidate should check working dir is a directory (4134)
2997 - don't always mark "slow" obdfilter messages as errors (4418)
2999 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3002 - don't close LustreDB during write_conf until it is done (3860)
3003 - fix typo in lconf for_each_profile (3821)
3004 - allow dumping logs from multiple threads at one time (3820)
3005 - don't allow multiple threads in OSC recovery (3812)
3006 - fix debug_size parameters (3864)
3007 - fix mds_postrecov to initialize import for llog ctxt (3121)
3008 - replace config semaphore with spinlock (3306)
3009 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3010 - don't allow enqueue to complete on a destroyed export (3822)
3011 - down write_lock before checking llog header bitmap (3825)
3012 - recover from lock replay timeout (3764)
3013 - up llog sem before sending rpc (3652)
3014 - reduce ns lock hold times when setting kms (3267)
3015 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3016 - fix NULL deref and obd_dev leak on setup error (3312)
3017 - replace some LBUG about llog ops with error handling (3841)
3018 - don't match INVALID dentries from d_lookup and spin (3784)
3019 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3020 - fix invalid assertion in ptlrpc_set_wait (3880)
3022 - add libwrap support for the TCP acceptor (3996)
3023 - add /proc/sys/portals/routes for non-root route listing (3994)
3024 - allow setting MDS UUID in .xml (2580)
3025 - print the stack of a process that LBUGs (4228)
3027 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3030 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3031 - make sure to unset replay flag from failed open requests (3440)
3032 - if default stripe count is 0, use OST count for inode size (3636)
3033 - update parent mtime/ctime on client for create/unlink (2611)
3034 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3035 - free recovery state on server during a forced cleanup (3571)
3036 - unregister_reply for resent reqs (3063)
3037 - loop back devices mounting and status check on 2.6 (3563)
3038 - fix resource-creation race that can provoke i_size == 0 (3513)
3039 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3040 - more debugging for page-accounting assertion (3746)
3041 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3042 - avoid deadlock after precreation failure (3758)
3043 - fix race and lock order deadlock in orphan handling (3450, 3750)
3044 - add validity checks when grabbing inodes from l_ast_data (3599)
3046 - add /proc/.../recovery_status to obdfilter (3428)
3047 - lightweight CDEBUG infrastructure, debug daemon (3668)
3048 - change default OSC RPC parameters to be better on small clusters
3049 - turn off OST read cache for files smaller than 32MB
3050 - install man pages and include them in rpms (3100)
3051 - add new init script for (un)mounting lustre filesystems (2593)
3052 - run chkconfig in %post for init scripts (3701)
3053 - drop scimac NAL (unmaintained)
3055 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3058 - clean kiobufs before and after use (3485)
3059 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3060 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3061 - change init script to fail more clearly if not run as root (1528)
3062 - allow clients to reconnect during replay (1742)
3063 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3064 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3065 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3066 - fix cond_resched (3554)
3067 - only evict unfinished clients after recovery (3515)
3068 - allow bulk resend, prevent data loss (3570)
3069 - dynamic ptlrpc request buffer allocation (2102)
3070 - don't allow unlinking open directory if it isn't empty (2904)
3071 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3072 - remove extraneous obd dereference causing LASSERT failure (3334)
3073 - don't use get_cycles() when creating temp. files on the mds (3156)
3074 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3075 - handle EEXIST for set-stripe, set proper directory name (3336)
3077 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3078 - fix ksocknal_fmb_callback() error messages (2918)
3080 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3083 - don't copy lvb into (possibly NULL) reply on error (2983)
3084 - don't deref dentry after dput, don't free lvb on error (2922)
3085 - use the kms to determine writeback rpc length (2947)
3086 - increment oti_logcookies when osc is inactive (2948)
3087 - update client's i_blocks count via lvb messages (2543)
3088 - handle intent open/close of special files properly (1557)
3089 - mount MDS with errors=remount-ro, like obdfilter (2009)
3090 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3091 - don't use cancelling-locks' kms values (2947)
3092 - use highest lock extent for kms, not last one (2925)
3093 - don't dereference ERR_PTR() dentry in error handling path (3107)
3094 - fix thread race in portals_debug_dumplog() (3122)
3095 - create lprocfs device entries at setup instead of at attach (1519)
3096 - common AST error handler, don't evict client on completion race (3145)
3097 - zero nameidata in detach_mnt in 2.6 (3118)
3098 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3099 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3100 - initialize RPC timeout timer earlier for 2.6 (3219)
3101 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3102 - print nal/nid for unknown nid (3258)
3103 - additional checks for oscc recovery before doing precreate (3284)
3104 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3105 - don't crash in mdc_close for bad permissions on open (3285)
3106 - zero i_rdev for non-device files (3147)
3107 - clear page->private before handing to FS, better assertion (3119)
3108 - tune the read pipeline (3236)
3109 - fix incorrect decref of invalidated dentry (2350)
3110 - provide read-ahead stats and refine rpc in flight stats (3328)
3111 - don't hold journal transaction open across create RPC (3313)
3112 - update atime on MDS at close time (3265)
3113 - close LDAP connection when recovering to avoid server load (3315)
3114 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3115 - don't leak open file on MDS after open resend (3325)
3116 - serialize filter_precreate and filter_destroy_precreated (3329)
3117 - loop device shouldn't call sync_dev() for nul device (3092)
3118 - clear page cache after eviction (2766)
3119 - resynchronize MDS->OST in background (2824)
3120 - refuse to mount the same filesystem twice on same mountpoint (3394)
3121 - allow llmount to create routes for mounting behind routers (3320)
3122 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3123 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3124 - fix rare problem with rename on htree directories (3417)
3126 - allow default OST striping configuration per directory (1414)
3127 - fix compilation for qswnal for 2.6 kernels (3125)
3128 - increase maximum number of MDS request buffers for large systems
3129 - change liblustreapi to be useful for external progs like lfsck (3098)
3130 - increase local configuration timeout for slow disks (3353)
3131 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3133 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3136 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3137 - don't overwrite extent policy data in reply if lock was blocked (2901)
3138 - drop filter export grants atomically with removal from device (2663)
3139 - del obd_self_export from work_list in class_disconnect_exports (2908)
3140 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3141 - swab reply message in mdc_close, other PPC fixes (2464)
3142 - fix destroying of named logs (2325)
3143 - overwrite old logs when running lconf --write_conf (2264)
3144 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3145 - fix race in target_handle_connect (2898)
3146 - mds_reint_create() should take same inode create lock (2926)
3147 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3148 - don't close files for self_export to avoid uninitialized obd (2936)
3149 - allow MDS with the same name as client node (2939)
3150 - hold dentry reference for closed log files for unlink (2325)
3151 - reserve space for all logs during transactions (2059)
3152 - don't evict page beyond end of stripe extent (2925)
3153 - don't oops on a deleted current working directory (2399)
3154 - handle hard links to targets without a parent properly (2517)
3155 - don't dereference NULL lock when racing during eviction (2867)
3156 - don't grow lock extents when lots of conflicting locks (2919)
3158 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3161 - account for cache space usage on clients to avoid data loss (974)
3162 - lfsck support in lustre kernel code (2349)
3163 - reduce journal credits needed for BRW writes (2370)
3164 - orphan handling to avoid losing space on client/server crashes
3165 - ptlrpcd can be blocked, stopping ALL progress (2477)
3166 - use lock value blocks to assist in proper KMS, faster stat (1021)
3167 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3168 - recovery for initial connections (2355)
3169 - fixes for mds_cleanup_orphans (1934)
3170 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3171 - block all file creations until orphan recovery completes (1901)
3172 - client remove rq_connection from request struct (2423)
3173 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3174 - recovery timer race (2670)
3175 - mdc_close recovey bug (2532)
3176 - ptlrpc cleanup bug (2710)
3177 - mds timeout on local locks (2588)
3178 - namespace lock held during RPCs (2431)
3179 - handle interrupted sync write properly (2503)
3180 - don't try to handle a message that hasn't been replied to (2699)
3181 - client assert failure during cleanup after abort recovery (2701)
3182 - leak mdc device after failed mount (2712)
3183 - ptlrpc_check_set allows timedout requests to complete (2714)
3184 - wait for inflight reqs when ptlrpcd finishes (2710)
3185 - make sure unregistered services are removed from the srv_list
3186 - reset bulk XID's when resending them (caught by 1138 test)
3187 - unregister_bulk after timeout
3188 - fix lconf error (2694)
3189 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3190 - readahead locks pages, leaves pending causing memory pressure (2673)
3191 - increase OST request buffers to 4096 on large machines (2729)
3192 - fix up permission of existing directories in simple_mkdir (2661)
3193 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3194 - don't assign transno to errored transactions (2742)
3195 - don't delete objects on OST if given a bogus objid from MDS (2751)
3196 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3197 - if rq_no_resend, then timeout request after recovery (2432)
3198 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3199 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3200 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3201 - MDS recovery shouldn't delete objects that it has given out (2730)
3202 - if enqueue arrives after completion, don't clobber LVB (2819)
3203 - don't unlock pages twice when trigger_group_io returns error (2814)
3204 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3205 - don't write pages to disk if there was an error (1450)
3206 - don't ping imports that have recovery disabled (2676)
3207 - take buffered bytes into account when balancing socknal conn (2817)
3208 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3209 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3210 - remove little-endian swabbing of llog records (1987)
3211 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3212 - retry reposting request buffers if they fail (1191)
3213 - grow extent at grant time to avoid granting a revoked lock (2809)
3214 - lock revoke doesn't evict page if covered by a second lock (2765)
3215 - disable VM readahead to avoid reading outside lock extents (2805)
3217 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3218 - updated kernel patches for hp-2.4.20 kernel (2681)
3220 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3223 - fix truncated write corruption (2366)
3224 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3226 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3227 - ensure len(uuid) < 37 in lmc (1171)
3228 - fix ia64 OOPS in llog_test (2255)
3229 - zero end of page at obdfilter for partial page writes (2648)
3230 - don't leave stale dentries around after renames (bug 2428)
3231 - fix timeouts when evicting a client with a single lock held (2642)
3232 - set deadline for the initial HELLO message to drain (2634)
3233 - print out dotted-quad IP addresses in the socknal (2302)
3235 - additional debugging for MDS client eviction problem (2443)
3236 - fix mkfsoptions support for osts (2603, 2604)
3238 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3241 - add series for the vanilla 2.6.0 kernel
3242 - add series for the vanilla 2.4.24 kernel
3243 - add series for a cray x86/64 UL kernel drop
3244 - fix xattr patches for the vanilla 2.4.19 series
3246 - generate true UUIDs in lmc (1171)
3247 - have portals stack dumping break in UML (2466)
3248 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3249 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3250 - wake callback waiting threads on client eviction (2460)
3251 - Add --ptldebug and --subsystem to lmc (1719)
3252 - update assertion to allow safe interrupt allocation
3253 - set rq_no_resend for cancel requests (2432)
3254 - recalculate ptlrpcd timeout after resend (2494)
3255 - call vfs_rmdir when removing pending directories (2368)
3256 - fix renaming a file to itself (2429)
3257 - lmc creates a default one-stripe lov (2454)
3258 - expand procfs space to handle large clusters (2326)
3259 - increase UML stack to avoid overflow
3260 - update lconf's list of debug and subsystem masks
3261 - fix lfs find --obd (2510)
3262 - /proc tunable for disabling filter read caching (2591)
3263 - stop rpm packages from altering slapd.conf (2301)
3264 - disable nagle in the socknal under 0conf (2578)
3265 - choose mds inode size based on stripe count (2572)
3266 - fix kernel-source rpm problems (2516)
3268 - add --disable-doc to avoid pdf generation (2421)
3269 - update documentation, tests, type-os, comments
3270 - avoid format warnings on ia64
3271 - remove the TOE NAL
3272 - tiny code cleanups by removing unused fields
3274 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3277 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3278 - remove the most blatant lies from BUILDING file (2371)
3279 - change default debug level to reasonable production setting
3280 - reduce client side cache size to reduce cache flush time
3281 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3282 - make TCP zerocopy and pinger support enabled by default (2476)
3283 - sync writes completed after process exits caused crashes (2319)
3284 - maintain correct mount count on the MDS (2356)
3285 - backout 1557, because 2316 wasn't really fixed
3286 - better file I/O statistics gathering in /proc
3287 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3288 - another kernel patch to fix zero-copy TCP function export
3289 - don't take duplicate lock when processing re-sent getattr (2420)
3290 - lctl uses obd_self_export instead of creating new conn (2353)
3291 - MDS/OST recovery case which requires object creation asserted (2425)
3292 - move lfs from /usr/sbin to /usr/bin in packages
3293 - fix race between mds_client_add and mds_client_free (2417)
3294 - use kmalloc instead of slabs in portals (2430)
3295 - don't create duplicate records when a failover MDS is present (2442)
3296 - remove unnecessary mount age check (2332)
3297 - don't remove directory inodes from locks prematurely (2451)
3298 - don't break if MDS service name is the same as hostname (2103)
3299 - fix races in client write RPC generation when cache full (2482)
3301 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3304 - remove now-unused request->rq_obd (278)
3305 - if an allocation fails, print out how much memory we've used (1933)
3306 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3307 - add the "configurable stack size" patch to most series files (1256)
3308 - ability to write large log records, for 100+ OST configs (2306)
3309 - fix NULL deref when filter_prep fails (2314)
3310 - fix operator precedence error in filter_sync
3311 - dynamic allocation of socknal TX descriptors (2315)
3312 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3313 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3314 - ensure that CWARN messages in Portals always get to the syslog
3315 - __init/__exit are not for prototype decls (ldlm_init/exit)
3316 - x86-64 compile warning fixes
3317 - fix gateway LMC keyword conflict (2318)
3318 - fix MDS lock inversions in getattr/reint paths (1844)
3319 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3320 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3321 - DLM race condition prevented some lock evictions (2328)
3322 - ENOMEM detection and retry on socknal sends (2230)
3323 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3324 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3325 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3326 - better error messages when a client is rejected during recovery (1505)
3327 - avoid cancelling locks which were never granted, after failure (2330)
3328 - fix i_sem/journal inversion in mds_client_add (2333)
3329 - fix truncate/getattr lock cycle deadlock (2334)
3330 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3331 - fix two rare exit paths which could leak an l_lock() ref (2321)
3332 - fencepost error in MDS/OST orphan recovery (2226)
3333 - make log record alignment 8 bytes (1988)
3334 - lstripe now fails when requested offset > ost_count (2237)
3335 - ensure that all kernel series have a complete list.h (1607)
3336 - fix crashes in special-file operations (2316)
3337 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3339 - allow configurable automake binary, for testing new versions
3340 - small update to the lfs documentation
3342 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3344 * fix negative export reference count in fsfilt_sync (2312)
3346 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3347 * release candidate 0.9.1
3349 - orphans are moved into the PENDING directory for possible recovery
3350 - replayed opens now open by fid for orphan/rename safety (1042)
3351 - last close of an orphan inode generates a transno (683)
3352 - chdir() and mount() now pin the directory entry (1020)
3353 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3354 - discard very old requests without processing them (1502)
3355 - remove some common, well-understood CERRORs (1505)
3356 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3357 - clear "grant" flags in OST replies until OST grant code lands (1644)
3358 - fix read performance by not clobbering i_blksize on client (1598)
3359 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3360 - make LRU size a /proc tunable, clears locks when reduced (707)
3361 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3362 - server locks take references on exports now (1558)
3363 - build fixes for 2.4.20-rh trees (1663)
3364 - return an error from lov_create if all OSCs are inactive (1751)
3365 - fix import levels when a reconnect happens without a timeout (1597)
3366 - exit early from mds_open if we get a lookup error (1749)
3367 - partial page read at EOF wouldn't wait for disk before sending (1642)
3368 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3369 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3370 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3371 - write/truncate lock inversion (1639)
3372 - Don't auto-load obdclass, portals modules during cleanup (1495)
3373 - fix timestamps from jumping to "now" (1763)
3374 - extra journal assertions (1648)
3375 - add an extra multiunlink test (1771)
3376 - fix read_record/write_record API (1776)
3377 - fix leak of offset_extent, possible incorrect i_size later (1772)
3378 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3379 - Debugging for the kqswnal_get_idle_tx problems (1820)
3380 - Allow recovery to be attempted multiple times (1536)
3381 - Write out MDS last_rcvd file after it is first created (1600)
3382 - Fix tx_descriptor leak in failed transmit situations (1827)
3383 - ext3 journaling fixes for assertion failure after IO error (1871)
3384 - class_export_put() on freed export after completion AST error (1896)
3385 - Fix revalidate looping in VFS (1322)
3386 - Don't access a freed export during MDS_REINT timeout (1521)
3387 - Add open-unlink recovery support on the MDS (1673,1764)
3388 - Return an error if no MDS data was read from last_rcvd (1946)
3389 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3390 - Don't setup a disk device that doesn't match exported UUID (317)
3391 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3392 - avoid committing NULL handle in force close
3393 - local.sh is now a one-stripe LOV configuration
3394 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3395 - fix ext3 htree duplicate directory entry corruption (1516)
3396 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3397 - update to new LOV EA format (2097)
3398 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3400 2003-06-15 Phil Schwan <phil@clusterfs.com>
3403 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3404 - per-import/export recovery handling (958, 931, 959)
3405 - multiple last-rcvd slots, for serving multiple FSes (949)
3406 - connections are again shared between multiple imp/exports (963, 964)
3407 - "umount -f" would hang if any requests needed to be sent (393, 978)
3408 - avoid pinning large req buffer by copying for queued messages (989)
3409 - add "uuid" to "lctl device" command to help upcalls (991)
3410 - "open" RPCs with transnos would confuse recovery counters (1037)
3411 - do proper endian conversion of all wire messages (288, 340, 891)
3412 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3413 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3414 - fix ptlrpc_connection leak in target_handle_connect (1174)
3415 - fix import refcounting bug in OST and MDS cleanup (1134)
3416 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3417 - fix very unlikely obd_types race condition (501)
3418 - remove osc_open hack for echo_client (1187)
3419 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3420 - a failure in read_inode2 leads to deadlock (1139)
3421 - cancel ack-locks as soon as transaction is committed (1072)
3422 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3423 - make sure to commitrw after any preprw to avoid deadlock (1162)
3424 - failing to execute a file in a lustre FS would lock inode (1203)
3425 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3426 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3427 - obd_uuid2tgt would walk off the end of the list (1255)
3428 - on IA64 the osc would give portals incorrect bulk size (1258)
3429 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3430 - fix lock inversion caused by new llite matching code (1282)
3431 - limit the number of dirty pages on a client to 10MB (1286)
3432 - timed out locks were not being corrected cancelled (1289)
3433 - fix O_DIRECT above 4GB on IA-32 (1292)
3434 * major user-visible changes
3435 - fail out/fail over policy now controlled by the upcall (993)
3437 - add OBD_PING to check server availability and failure (954)
3438 - lustre messages are now sent in sending host order (288, 340, 891)
3439 - add eadatalen to MDS getattr reply (340)
3440 - OST read replies may contain second buffer, with per-page status (593)
3442 2003-03-11 Phil Schwan <phil@clusterfs.com>
3445 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3446 - failed open()s could cause deadlock; fixed (867, 869)
3447 - stop cancelling OST locks when files are closed (481)
3448 - overlapping XID spaces caused network corruption (851, 853)
3449 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3450 - setattr_raw added, to do single-RPC, server-side setattrs
3451 - lmc/lconf syntax change for OST UUIDs
3452 - fix crashy race condition between ptlrpc_free_req and osc_close
3453 - don't use request in mdc_enqueue if we hit a timeout (889)
3454 - don't set the inode i_size for regular files from the MDS (896)
3455 - handle out of order completion AST (842)
3456 - don't LBUG if a lock request times out after receiving AST (913)
3457 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3458 - if a bad lock AST arrives, send an error instead of dropping entirely
3459 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3460 - fix leak in bulk IO when only partially completed (899, 900, 926)
3461 - fix O_DIRECT for ia64 (55)
3462 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3463 - C-z after timeout could hang a process forever; fixed (977)
3465 - client-side I/O cache (678, 924, 929, 941, 970)
3467 - READPAGE and SETATTRs which don't take server-side locks get
3470 2003-02-11 Phil Schwan <phil@clusterfs.com>
3473 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3474 - Fix fsfilt_extN_readpage() to read a full page of directory
3475 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3476 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3477 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3478 - Most intent operations were removed, in favour of a new RPC mode
3479 that does a single RPC to the server and bypasses most of the VFS
3480 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3481 - Aggressively cancel local locks on DLM servers
3482 - mds_reint_unlink sends EA to the client if it's the last nlink.
3483 client uses that EA to unlink OST objects.
3484 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3485 - recursive symlinks were fixed (439)
3486 - fixed NULL deref in DEBUG_REQ
3487 - filter_update_lastobjid no longer calls sync, which annoyed extN
3488 - fixed multi-client small-writes to a single file problem (445)
3489 - fixed mtime updates during file writes (607)
3490 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3491 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3492 - fixed Linux OST size reporting problem (444, 656)
3493 - OST now updates object mtime with writes or setattr (607, 619)
3494 - client verifies file size before zeroing page past EOF (445)
3495 - OST now writes last allocated objid to disk with allocation (108)
3496 - LOV on echo now works (409)
3498 - mds_reint_unlink sends a new buffer, with the EA included. this
3499 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3500 set if a regular file was being unlinked, and it was the last link
3501 - use PtlGet from the target for bulk writes (315)
3502 - OST now updates object mtime with writes or setattr (607, 619)
3503 - LDLM now has a grant-time callback to revalidate locked items, if
3505 - Many MDS operations were reorganized to combat race conditions
3507 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3508 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3510 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3513 - Fully reactivate OST imports after reconnection (512, others)
3514 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3515 - More graceful error handling for truncating on dead OST (515)
3516 - Don't error out unless we're actually accessing dead stripes (474)
3517 - Fix garbage sizes when stripes are missing (410)
3518 - LRU counters were broken, causing constant lock purge (433, 432)
3519 - garbage on read from stripes with failed OSTs (441)
3520 - mark OSCs as active before reconnecting during recovery (438)
3521 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3522 - lfind did not preserve OST order in output (443)
3523 - symlinks cause hung clients, incorrect data (439)
3524 - stop dereferencing request after dropping refcount (457)
3525 - don't LASSERT(spin_is_locked) on non-SMP (455)
3526 - fixes for many rename() bugs
3527 - fstat didn't correctly synchronize attributes (399)
3528 - server must handle lock cancellation during blocking AST prep (487)
3529 - bulk descriptors were free()d too soon (511)
3530 - fix paths in lconf, which would load incorrect modules (451, 507)
3531 - fix confusing lconf 'host not found' error message (386)
3532 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3533 - fix race condition in mdc_blocking_ast() for inode access (526)
3534 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3535 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3536 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3537 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3538 - fix lstripe in several ways (kernel side) (527)
3539 - fix request leak in ldlm_cli_enqueue (262)
3540 - incorrect OSC was marked inactive after OST failure
3541 - call mds_fs_cleanup before unmounting filesystem (524)
3542 - fix races between taking ns_lock and ldlm_lock_change_resource
3543 - fix races updating LOV export open file list
3544 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3545 - fix recovery NULL deref in ldlm_cli_cancel_unused
3546 - fix some DLM races by using new hash table for lock handles (419)
3547 - permit the client to specify desired inodes, at replay
3548 - duplicate requests when we queue them for replay reintegration
3549 - fix last_rcvd offset calculation
3550 - sync after each recovered transaction, so we always make progress
3551 - never, not always, ERESTART requests without transnos
3552 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3554 - skip replay if the MDS says that the client is already connected
3555 - don't check for a recovery-enabled export to match lctl's UUID
3556 - don't INC_USE_COUNT for phantom exports
3557 - don't crash when cleaning up phantom exports (567)
3558 - don't double-finish or set replay data for errored mdc_open requests
3559 - abort requests when they time out, so we don't get old replies
3560 - send/receive replies for AST messages again
3561 - if the client says that it doesn't have the lock, cancel it on the
3563 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3564 mark it as destroyed, it will all work out when decref is called
3565 - fix module use counts (22, 581)
3567 - ASTs now expect a reply (server cancels lock on error reply)
3569 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3572 - fix many simultaneous client startup (392)
3573 - fix dentry->d_it clobbering
3574 - credentials weren't being shipped for readdir/getattr operations
3575 - remove invalid assertions triggered during some concurrent MD
3577 - proper Lustre versions added (336, 389)
3578 - fix memory leak for create error case (398)
3579 - fix LOV locking bug that would get cli/srv out of sync
3580 - fix echo client over LOV (409)
3581 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3582 - fix double-O_EXCL intent crash (424)
3583 - avoid sending multiple lock CANCELs (352)
3585 - MDS can do multi-client recovery (modulo bugs in new code)
3587 - many updates, edits, cleanups
3589 2002-11-18 Phil Schwan <phil@clusterfs.com>
3592 - fix null d_it dereference (346)
3593 - fix full OST/dbench hang (333)
3594 - fix permission problem with file removal (286)
3595 - fix removal of OSCs from LOV when they fail
3596 - fix NULL deref during bulk timeout (214)
3597 - fix problems related to multiple filesystems on one MDS (241)
3598 - fixed serious subtle metadata locking bugs
3599 - free locks on clients when inodes are removed due to memory
3601 - fix inode pointer in lock data (285)
3602 - partial support for multiple MDS on a single host (241)
3603 - data locks weren't cancelled at clear_inode time (290, 311)
3604 - intent locks could lead to unbounded lock growth (205)
3605 - added a maximum lock count, an LRU list, and a flusher
3606 - fix multiple rename (365)
3607 - properly abstracted the echo client
3608 - OSC locked 1 byte too many; fixed
3609 - rewrote brw callback code:
3610 - fixed recovery bugs related to LOVs (306)
3611 - fixed too-many-pages-in-one-write crash (191)
3612 - fixed (again) crash in sync_io_timeout (214)
3613 - probably fixed callback-related race (385)
3615 - Add capability to MDS protocol
3616 - LDLM cancellations and callbacks on different portals
3618 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3621 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3622 - defer open object destruction to close time (601981, 138)
3623 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3624 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3625 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3626 - don't call obd_setattr in ll_file_release for destroyed objects
3627 * protocol change to lustre_msg: move |version| and add |flags|
3628 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3629 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3630 * added replay of create, unlink, link and rename operations during
3631 MDS failover; recovery should be much more robust now
3632 * remove failed OSCs from LOVs (only lov_create uses this so far)
3633 * the lustre-HOWTO was brought (more) up to date (582544)
3635 2002-10-23 Phil Schwan <phil@clusterfs.com>
3638 - in-use dentries weren't being reused properly (617851)
3639 - prevent multiple LDLM setup (599178)
3640 - fix LOV size calculations for truncate (617853)
3641 - fix client handling of MDS intent errors (POSIX)
3642 - fix permission bug in lovstripe.c test (624321)
3643 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3644 - truncate past end of file could corrupt data
3645 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3646 - a race in recovery could return ETIMEDOUT to apps (623947)
3647 - building outside the source directory was fixed
3648 * the lustre-HOWTO was brought (more) up to date (582544)
3649 * major progress was made on recovery functionality
3651 2002-10-10 Phil Schwan <phil@clusterfs.com>
3654 - recovery deadlock fix
3655 - rm -rf causes LBUG fix (617817)
3656 - file open by multiple tasks fix (618962)
3657 - directory permissions bugs (602707 and 620007)
3658 - journal_stop fixed with locking (611313)
3659 - O_APPEND failures resolved (618273, perhaps 614459)
3660 - lconf PATH fix (619770)
3661 - IA64 build fix (621450)
3662 - RPC buffer sizes scale with amount of memory
3664 2002-10-01 Phil Schwan <phil@clusterfs.com>
3667 - locks would be cancelled without throwing away data pages,
3668 resulting in inconsistent data (605627)
3669 - inode attributes were not always being refreshed (605627, 612449)
3670 - lconf now continues to cleanup after lctl reports an error
3671 - MDS now enforces user permissions (602707)
3672 - lprocfs cleanup fixed, but not yet enabled (614157)
3673 - fixed infinite server hang, should a client not respond to an AST
3674 - avoid going into recovery if user calls readlink() with a buffer
3675 that's too small (613941)
3676 - AST RPCs no longer require replies (614867) -- this may be changed
3677 - don't crash server if client sends an IOV that's too big (611336)
3678 - fixed lock conversion deadlock (611892)
3679 - fixed the following of symlinks (614622)
3680 * recovery: the server can remove locks from a client that dies, other
3681 clients can make progress
3682 * more extN patch fixes
3683 * compile-time configurable ptlrpc buffer allocations
3685 - collaborative read cache document
3686 - Lustre Lite Performance CDR document-in-progress
3688 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3691 - fix typo in patch-2.4.18
3693 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3696 - clear ptlrpc request each time in handle_incoming_request()
3697 - unlink of files now destroys the object on the OST
3699 2002-09-19 Peter Braam <braam@clusterfs.com>
3701 * add hard link support
3702 * change obdfile creation method
3703 * kernel patch changed
3705 2002-09-19 Peter Braam <braam@clusterfs.com>
3708 - stack overflow bug in extN fixed
3710 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3712 * documentation updates
3713 - add man pages for config tools
3714 - update tests/README to describe testing with new config tools
3715 - finish metadata API descriptions
3716 * bug fixes and cleanups
3717 - statfs workaround for 16TB limit
3718 - LOV stripe allocation improved, can stripe on subset of OSTs
3719 - LOV file size/IO offset was wrong for files > 4GB in size
3720 - object EA data was being dropped, caused files to be unreadable
3721 - memory overflow with non-LOV OST caused memory corruption
3722 - fixed regression tests to work with new config tools, obdfilter
3723 - fixed bug when directory size became larger than 1 block
3724 - fixed bug (for single client case) when PWD was deleted
3725 - invalidate local directory pages when doing intent-based ops
3726 - avoid LDLM oops when lock callback contained bad data
3728 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3730 * documentation updates
3731 * bug fixes and cleanups
3732 - configuration tools
3735 - 64-bit compile warnings
3736 - 64-bit internal statfs data
3738 * test_brw on persistent OST devices
3740 * lprocfs (disabled)
3742 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3744 * documentation updates
3745 * bug fixes and cleanups
3746 * configuration tools
3748 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3751 * many small fixes to 0_5_4
3752 * io/network handling
3753 * thinkos in MDS operations
3755 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3758 * crucial basic fixes to 0.5.3
3759 * IOR, Iozone work over Elan
3762 2002-08-07 Phil Schwan <phil@clusterfs.com>
3763 * version 0_5_3, our first alpha
3764 * we use the new Portals iovs
3765 * documentation updates
3766 * bug fixes and cleanups
3767 * small changes in the DLM wire protocol
3769 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3770 * version 0_5_1 with some initial stability,
3771 * locking on MD and file I/O.
3772 * documentation updates
3773 * several bug fixes since 0.5.0
3774 * small changes in wire protocol
3776 2002-07-18 Phil Schwan <phil@clusterfs.com>
3778 * delivered as Lustre Light Alpha
3779 * fixed a crash after handling invalid MDS requests
3780 * fixed directory pages for architectures with non-4k pages sizes
3782 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3783 * release version v0_4_4
3784 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3786 * Fixes a number of interruption problems with OST operations.
3787 * Update documentation for portals header changes
3788 * Move all wire protocol structs/defines to lustre_idl.h
3789 * Fixes symlink length bug.
3790 * Add tcpdump to repository.
3792 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3793 * release version v0_4_3
3794 * Fixes statfs for inodes on extN.
3795 * Fixes bug in runtests which would delete /etc/hosts.
3796 * Use 64-bit object IDs wherever possible (not into VFS though)
3797 Remove ost_get_info, which is unused by lustre, and out of date.
3799 2002-07-03 Peter Braam <braam@clusterfs.com>
3800 * release version v0_4_2 Fixes a lookup error (type not passed)
3801 * move forward to head of Portals
3802 * move forward to latest Lustre kernel
3804 2002-06-25 Peter Braam <braam@clusterfs.com>
3805 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob