1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10),
7 2.6.18-8.1.8.EL5 (RHEL 5).
8 2.6.18.8 vanilla (kernel.org)
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.9-42.0.10.EL (RHEL 4),
12 2.6.16 - 2.6.21 vanilla (kernel.org)
13 * Recommended e2fsprogs version: 1.40.2-cfs1
14 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
18 Description: don't use GFP_* in generic Lustre code.
19 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
23 Description: obdclass is limited by single OBD_ALLOC(idarray)
24 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
28 Description: warnings with build patchless client with vanila 2.6.19 and up
29 Details : change old ctl_table style and replace ctl_table/ctl_table_header
30 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
34 Frequency : Only for SLES
35 Description: sanity_quota fail test_1
36 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
37 file, which makes TSTID[2] inunique.
41 Frequency : Only for SLES9
42 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
43 Details : Update target/ChangeLog/which_patch .
47 Frequency : Only for test_10 in sanity-quota.sh
48 Description: a bug in quota test code
49 Details : checking a test flag in a wrong place
53 Frequency : Only for test_13 in sanity-quota.sh
54 Description: a bug in quota test script
55 Details : shouldn't check overfull allocated quota
59 Description: jbd statistics
60 Details : Port older jbd statistics patch for sles10
62 Severity : enhancement
64 Description: Add printing inode info into message about error in writepage.
69 Description: reply_lock_interpret crash due to race with it and lock cancel.
70 Details : Do not replay locks that are being cancelled. Do not reference
71 locks by their address during replay, just by their handle.
75 Frequency : When flocks are used.
76 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
77 Details : Flock locks might destroy just granted lock if it could be merged
78 with another existing flock, this is done in completion handler,
79 so teach ldlm_cli_enquque_fini that this is a valid case for
85 Description: Oops in read and write path when failing to allocate lock.
86 Details : Check if lock allocation failed and return error back.
90 Description: lstripe command fails for valid OST index
91 Details : The stripe offset is compared to lov->desc.ld_tgt_count
92 instead of lov->desc.ld_active_tgt_count.
96 Description: sanity.sh failed test 103
97 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
102 Description: df doesn't work properly if diskfs blocksize != 4K
103 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
107 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
108 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
112 Description: obd_health_check_timeout too short
113 Details : set obd_health_check_timeout as 1.5x of obd_timeout
117 Description: llapi_file_create() does not allow some changes
118 Details : add llapi_file_open() that allows specifying the mode and
119 open flags, and also returns an open file handle.
123 Description: merge and cleanup kernel patches.
124 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
128 Description: Add journal checksum support.(Kernel part)
129 Details : The journal checksum feature adds two new flags i.e
130 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
131 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
132 indicates that the commit block contains the checksum for
133 the blocks described by the descriptor blocks. Now commit
134 record can be sent to disk without waiting for descriptor
135 blocks to be written to disk. This behavior is controlled
136 using JBD2_FEATURE_ASYNC_COMMIT flag.
140 Description: OSS needs mutliple precreate threads
141 Details : Add ability to start more than one create thread per OSS.
145 Description: Sanity.sh test_65k failed: setstripe should have succeeded
146 Details : execute lfs setstripe on client
150 Description: mds_obd_create error creating tmp object
151 Details : When the user sets quota on root, llog will be affected and can't
152 create files and write files.
155 Frequency : Always on ia64 patchless client, and possibly others.
157 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
158 Details : This allows the patchless client to be loaded on architectures
163 Description: RedHat Update kernel for RHEL5
164 Details : Add the kernel config file for RHEL5.
168 Description: RedHat Update kernel for RHEL5
169 Details : Modify the kernel config file more closer RHEL5.
173 Description: Build failure against Centos5 (RHEL5)
174 Details : Define PAGE_SIZE when it isn't present.
178 Description: test_7 on sanity-quota.sh can't be used on separate machines
179 Details : add facet to handle it
181 --------------------------------------------------------------------------------
183 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
185 * Support for kernels:
186 2.6.5-7.283 (SLES 9),
187 2.6.9-55.EL (RHEL 4),
188 2.6.16.46-0.14 (SLES 10),
189 2.6.18.8 vanilla (kernel.org)
190 * Client support for unpatched kernels:
191 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
192 2.6.16 - 2.6.22 vanilla (kernel.org)
193 * Due to recently discovered recovery problems, we do not recommend
194 using patchless RHEL 4 clients with this or any earlier release.
195 * Recommended e2fsprogs version: 1.39.cfs8
196 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
197 * Starting with this release, the ldiskfs backing filesystem required
198 by Lustre is now in its own package, lustre-ldiskfs. This package
199 should be installed. It is versioned separately from Lustre and
200 may be released separately in future.
204 Description: block reactivating mgc import until all deactivates complete
205 Details : Fix race when failing back MDT/MGS to itself (testing)
207 Severity : enhancement
209 Description: add optional extra BUILD_VERSION info
210 Details : add a new environment variable (namely LUSTRE_VERS) which allows
211 to override the lustre version.
214 Frequency : 2.6.18 servers only
216 Description: ll_kern_mount() doesn't release the module reference
217 Details : The ldiskfs module reference count never drops down to 0
218 because ll_kern_mount() doesn't release the module reference.
223 Description: server LBUG when using old ost_num_threads parameter
224 Details : Accept the old ost_num_threads parameter but warn that it
225 is deprecated, and fix an off-by-one error that caused an LBUG.
230 Description: Transient SCSI error results in persistent IO issue
231 Details : iobuf->dr_error is not reinitialized to 0 between two
235 Frequency : sometimes when underlying device returns I/O errors
237 Description: OSTs not going read-only during write failures
238 Details : OSTs are not remounted read-only when the journal commit threads
239 get I/O errors because fsfilt_ext3 calls journal_start/stop()
240 instead of the ext3 wrappers.
244 Description: poor connect scaling with increasing client count
245 Details : Don't run filter_grant_sanity_check for more than 100 exports
246 to improve scaling for large numbers of clients.
249 Frequency : SLES10 only
251 Description: sanity-quota.sh quotacheck failed: rc = -22
252 Details : Quotas cannot be enabled on SLES10.
255 Frequency : liblustre clients only
257 Description: getdirentries does not give error when run on compute nodes
258 Details : getdirentries does not fail when the size specified as an argument
259 is too small to contain at least one entry
261 Severity : enhancement
263 Description: Add LNET router traceability for debug purposes
264 Details : If a checksum failure occurs with a router as part of the
265 IO path, the NID of the last router that forwarded the bulk data
266 is printed so it can be identified.
271 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
272 Details : Due to a race condition, liblustre clients were occasionally
275 Severity : enhancement
277 Description: lfs setstripe use optional parameters instead of postional
280 Severity : enhancement
282 Description: Nanosecond timestamp support for ldiskfs
283 Details : The on-disk ldiskfs filesystem has added support for nanosecond
284 resolution timestamps. There is not yet support for this at
285 the Lustre filesystem level.
288 Frequency : during server recovery
290 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
291 Details : request with rq_no_resend flag not awake l_wait_event if they get a
295 Frequency : nfs export on patchless client
297 Description: connectathon hang when test nfs export over patchless client
298 Details : Disconnected dentry cannot be found with lookup, so we do not need
299 to unhash it or make it invalid
302 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
303 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
304 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
309 Description: open req refcounting wrong on reconnect
310 Details : If reconnect happened between getting open reply from server and
311 call to mdc_set_replay_data in ll_file_open, we will schedule
312 replay for unreferenced request that we are about to free.
313 Subsequent close will crash in variety of ways.
314 Check that request is still eligible for replay in
315 mdc_set_replay_data().
320 Description: disable writes to filesystem when reading health_check file
321 Details : the default for reading the health_check proc file has changed
322 to NOT do a journal transaction and write to disk, because this
323 can cause reads of the /proc file to hang and block HA state
324 checking on a healthy but otherwise heavily loaded system. It
325 is possible to return to the previous behaviour during configure
326 with --enable-health-write.
328 Severity : enhancement
330 Description: 64-bit inode version
331 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
332 to the inode. This will be required for version-based recovery.
337 Description: MDS fails to start if a duplicate client export is detected
338 Details : in some rare cases it was possible for a client to connect to
339 an MDS multiple times. Upon recovery the MDS would detect this
340 and fail during startup. Handle this more gracefully.
342 Severity : enhancement
344 Description: Add -o localflock option to simulate old noflock
346 Details : This will achieve local-only flock/fcntl locks
352 Description: log_commit_thread vs filter_destroy race leads to crash
353 Details : Take import reference before releasing llog record semaphore
358 Description: Wrong request locking in request set processing
359 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
360 imp_delayed_list, in this place should be used imp_lock.
363 Frequency : when reconnection
365 Description: Grant Leak when osc reconnect to OST
366 Details : When osc reconnect ost, OST(filter) should check whether it
367 should grant more space to client by comparing fed_grant and
368 cl_avail_grant, and return the granted space to client instead
369 of "new granted" space, because client will call osc_init_grant
370 to update the client grant space info.
373 Frequency : when client reconnect to OST
375 Description: Grant Leak when osc do resend and replay bulk write
376 Details : When osc reconnect to OST, OST(filter)should clear grant info of
377 bulk write request, because the grant info while be sync between
378 OSC and OST when reconnect, and we should ignore the grant info
379 these of resend/replay write req.
384 Description: Grant space more than avaiable left space sometimes.
385 Details : When then OST is about to be full, if two bulk writing from
386 different clients came to OST. Accord the avaliable space of the
387 OST, the first req should be permitted, and the second one
388 should be denied by ENOSPC. But if the seconde arrived before
389 the first one is commited. The OST might wrongly permit second
390 writing, which will cause grant space > avaiable space.
393 Frequency : when client is evicted
395 Description: Grant might be wrongly erased when osc is evicted by OST
396 Details : when the import is evicted by server, it will fork another
397 thread ptlrpc_invalidate_import_thread to invalidate the
398 import, where the grant will be set to 0. While the original
399 thread will update the grant it got when connecting. So if
400 the former happened latter, the grant will be wrongly errased
401 because of this race.
406 Description: Checking Stale with correct fid
407 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
408 to check whether it is stale, because sometimes, we want the
409 enqueue happened anyway, and op_data.fid2 will not be initialized.
411 Severity : enhancement
413 Description: update patchless client
414 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
417 Frequency : only with 2.4 kernel
419 Description: random memory corruption
420 Details : size of struct ll_inode_info is to big for union inode.u and this
421 can be cause of random memory corruption.
426 Description: Memory leak in recovery
427 Details : Lov_mds_md was not free in an error handler in mds_create_object.
428 It should also check obd_fail before fsfilt_start, otherwise if
429 fsfilt_start return -EROFS,(failover mds during mds recovery).
430 then the req will return with repmsg->transno = 0 and rc = EROFS.
431 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
432 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
433 be freed no matter whether fsfilt_commit success or not.
436 Frequency : only with huge count clients
438 Description: Prevents from taking the superblock lock in llap_from_page for
440 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
441 for llap_from_page call in ll_removepage prevents from taking the
442 superblock lock for a soon died page.
447 Description: Not check open intent error before release open handle
448 Details : in some rare cases, the open intent error is not checked before
449 release open handle, which may cause
450 ASSERTION(open_req->rq_transno != 0), because it tries to release
451 the failed open handle.
456 Description: Set cat log bitmap only after create log success.
457 Details : in some rare cases, the cat log bitmap is set too early. and it
458 should be set only after create log success.
462 Description: Accessing a block bevice can re-enable I/O when Lustre is
463 tearing down a device.
464 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
469 Description: service threads may hog cpus when there are a lot of requests
471 Details : Insert cond_resched to give other threads a chance to use some of
477 Description: the cat log was not initialized in recovery
478 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
479 unlink log on mds will not be initialized until mds post
480 recovery. And also in mds post recovery, the unlink log will
481 initialization will be done asynchronausly, so there will be race
482 between add unlink log and unlink log initialization.
486 Description: brw_stats were being printed incorrectly
487 Details : brw_stats were being printed as log2 but all of them were not
488 recorded as log2. Also remove some code duplication arising from
489 filter_tally_{read,write}.
493 Frequency : rare, only in recovery.
494 Description: ASSERTION(req->rq_type != LI_POISON) failed
495 Details : imp_lock should be held while iterating over imp_sending_list for
496 prevent destroy request after get timeout in ptlrpc_queue_wait.
500 Description: replay-single.sh test 52 fails
501 Details : A lock's skiplist need to be cleanup when it being unlinked
502 from its resource list.
506 Description: Short directio read returns full requested size rather than
508 Details : Direct I/O operations should return actual amount of bytes
509 transferred rather than requested size.
511 Severity : enhancement
513 Description: metadata RPC reduction (e.g. for rm performance)
514 Details : decrease the amount of synchronous RPC between clients and servers
515 by canceling conflicing lock before the operation on the client side
516 and packing thier handles into the main operation RPC to server.
518 Severity : enhancement
520 Description: Async OSC create to avoid the blocking unnecessarily.
521 Details : If a OST has no remain object, system will block on the creating
522 when need to create a new object on this OST. Now, ways use
523 pre-created objects when available, instead of blocking on an
524 empty osc while others are not empty. If we must block, we block
525 for the shortest possible period of time.
527 --------------------------------------------------------------------------------
529 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
534 Frequency : on some architectures
536 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
537 Details : Uninitialized flags sometimes cause configuration commands to
541 Frequency : patchless clients only
543 Description: missing __iget() symbol export
544 Details : The __iget() symbol export is missing. To avoid the need for
545 this on patchless clients the deathrow inode reaper is turned
546 off, and we depend on the VM to clean up old inodes. This
547 dependency was during via the fix for bug 12181.
549 --------------------------------------------------------------------------------
551 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
553 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
554 INTEROPERATE with older versions automatically. In many cases a
555 special upgrade step is needed. Please read the
556 user documentation before upgrading any part of a live system.
557 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
558 version will not interoperate with 1.6 betas before beta5 (1.5.95).
559 * WARNING: Lustre configuration and startup changes are required with
560 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
566 Severity : enhancement
568 Description: MountConf
569 Details : Lustre configuration is now managed via mkfs and mount
570 commands instead of lmc and lconf. New obd types (MGS, MGC)
571 are added for dynamic configuration management. See
572 https://mail.clusterfs.com/wikis/lustre/MountConf for
575 Severity : enhancement
577 Description: dynamic OST addition
578 Details : OSTs can now be added to a live filesystem
580 Severity : enhancement
582 Description: startup order invariance
583 Details : MDTs and OSTs can be started in any order. Clients only
584 require the MDT to complete startup.
586 Severity : enhancement
588 Description: parallel, asynchronous orphan cleanup
589 Details : orphan cleanup is now performed in separate threads for each
590 OST, allowing parallel non-blocking operation.
592 Severity : enhancement
594 Description: optimized stripe assignment
595 Details : stripe assignments are now made based on ost space available,
596 ost previous usage, and OSS previous usage, in order to try
597 to optimize storage space and networking resources.
599 Severity : enhancement
601 Description: Permanently set tunables
602 Details : All writable /proc/fs/lustre tunables can now be permanently
603 set on a per-server basis, at mkfs time or on a live
606 Severity : enhancement
608 Description: Lustre message v2
609 Details : Add lustre message format v2.
611 Severity : enhancement
613 Description: client OST exclusion list
614 Details : Clients can be started with a list of OSTs that should be
615 declared "inactive" for known non-responsive OSTs.
619 Description: SPEC SFS validation failure on NFS v2 over lustre.
620 Details : Changes the blocksize for regular files to be 2x RPC size,
621 and not depend on stripe size.
623 Severity : enhancement
625 Description: Multiple MD RPCs in flight.
626 Details : Further unserialise some read-only MDS RPCs - learn about intents.
627 To avoid overly-overloading MDS, introduce a limit on number of
628 MDS RPCs in flight for a single client and add /proc controls
629 to adjust this limit.
631 Severity : enhancement
633 Description: client read/write statistics
634 Details : Add client read/write call usage stats for performance
635 analysis of user processes.
636 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
637 file access. extents_stats shows chunk size distribution.
638 extents_stats_per_process show chunk size distribution per
641 Severity : enhancement
643 Description: mds statistics
644 Details : Add detailed mds operations statistics in
645 /proc/fs/lustre/mds/*/stats.
649 Description: Failure of copying files with lustre special EAs.
650 Details : Client side always return success for setxattr call for lustre
651 special xattr (currently only "trusted.lov").
656 Description: Refcount LNET uuids
657 Details : The global LNET uuid list grew linearly with every startup;
658 refcount repeated list entries instead of always adding to
661 Severity : enhancement
663 Description: Dynamic service threads
664 Details : Within a small range, start extra service threads
665 automatically when the request queue builds up.
668 Frequency : mixed-endian client/server environments
670 Description: mixed-endian crashes
671 Details : The new msg_v2 system had some failures in mixed-endian
674 Severity : enhancement
676 Description: Easy OST removal
677 Details : OSTs can be permanently deactivated with e.g. 'lctl
678 conf_param lustre-OST0001.osc.active=0'
680 Severity : enhancement
682 Description: MGS proc entries
683 Details : Added basic proc entries for the MGS showing what filesystems
686 Severity : enhancement
688 Description: provide MGS failover
689 Details : Added config lock reacquisition after MGS server failover.
691 Severity : enhancement
693 Description: add Linux 2.4 support
694 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
698 Description: a large application tries to do I/O to the same resource and dies
700 Details : Check the req->rq_arrival time after the call to
701 ost_brw_lock_get(), but before we do anything about
702 processing it & sending the BULK transfer request. This
703 should help move old stale pending locks off the queue as
704 quickly as obd_timeout.
707 Frequency : when an incorrect nid is specified during startup
709 Description: ptlrpc connect to non-existant node causes kernel crash
710 Details : LNET can't be re-entered from an event callback, which
711 happened when we expire a message after the export has been
712 cleaned up. Instead, hand the zombie cleanup off to another
715 Severity : enhancement
717 Description: plain/inodebits lock performance improvement
718 Details : Grouping plain/inodebits in granted list by their request modes
719 and bits policy, thus improving the performance of search through
723 Frequency : only if OST filesystem is corrupted
725 Description: client incorrectly hits assertion in ptlrpc_replay_req()
726 Details : for a short time RPCs with bulk IO are in the replay list,
727 but replay of bulk IOs is unimplemented. If the OST filesystem
728 is corrupted due to disk cache incoherency and then replay is
729 started it is possible to trip an assertion. Avoid putting
730 committed RPCs into the replay list at all to avoid this issue.
733 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
735 Description: System hang on startup
736 Details : This bug allowed the liblustre (e.g. catamount) client to
737 return to the app before handling all startup RPCs. This
738 could leave the node unresponsive to lustre network traffic
739 and manifested as a server ptllnd timeout.
741 Severity : enhancement
743 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
744 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
745 boolean to control whether to print peer debug info when a
746 client's RPC times out.
749 Frequency : only for kernels with patches from Lustre below 1.4.3
751 Description: Remove old rdonly API
752 Details : Remove old rdonly API which unused from at least lustre 1.4.3
755 Frequency : only for devices with external journals
757 Description: Set external device read-only also
758 Details : During a commanded failover stop, we set the disk device
759 read-only while the server shuts down. We now also set any
760 external journal device read-only at the same time.
763 Frequency : when upgrading from 1.4 while trying to change parameters
765 Description: The wrong (new) MDC name was used when setting parameters for
766 upgraded MDT's. Also allows changing of OSC (and MDC)
767 parameters if --writeconf is specified at tunefs upgrade time.
770 Frequency : when setting specific ost indicies
772 Description: QOS code breaks on skipped indicies
773 Details : Add checks for missing OST indicies in the QOS code, so OSTs
774 created with --index need not be sequential.
776 Severity : enhancement
778 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
779 Details : The uninit_groups feature works in conjunction with the kernel
780 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
781 up the pass1 processing of e2fsck. This is a read-only feature
782 in ldiskfs2 only, so older kernels and current ldiskfs cannot
783 mount filesystems that have had this feature enabled.
785 Severity : enhancement
787 Description: Improve multi-block allocation algorithm to avoid fragmentation
788 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
789 allocation locality and avoid filesystem fragmentation.
791 ------------------------------------------------------------------------------
793 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
795 * Support for kernels:
796 2.4.21-47.0.1.EL (RHEL 3)
798 2.6.9-42.0.10.EL (RHEL 4)
799 2.6.12.6 vanilla (kernel.org)
800 2.6.16.27-0.9 (SLES 10)
801 * Recommended e2fsprogs version: 1.39.cfs5
803 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
807 Frequency : occasional, depends on client load and configuration
808 Bugzilla : 12181, 12203
809 Description: data loss for recently-modified files
811 Details : In some cases it is possible that recently written or created
812 files may not be written to disk in a timely manner (this should
813 normally be within 30s unless client IO load is very high).
814 The problem appears as zero-length files or files that are a
815 multiple of 1MB in size after a client crash or client eviction
816 that are missing data at the end of the file.
818 This problem is more likely to be hit on clients where files are
819 repeatedly created and unlinked in the same directory, clients
820 have a large amount of RAM, have many CPUs, the filesystem has
821 many OSTs, the clients are rebooted frequently, and/or the files
822 are not accessed by other nodes after being written.
824 The presence of the problem can be detected by looking at
825 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
826 smaller than the second (nr_unused) then dirty files will not
827 be flushed automatically to disk. "sync; sleep 10" should be
828 run several times on the node before unmounting it to update
829 Lustre (this is also safe to run on nodes without this problem).
831 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
832 that can cause this same problem, so customers using that kernel
833 also need to update the kernel in addition to Lustre. In order
834 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
836 It is normal that files written just before a client crash (less
837 than 30s) may not yet have been flushed to disk, even for local
841 Frequency : frequent on thin XT3 nodes
843 Description: UUID collision on thin XT3 Linux nodes
844 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
845 randomly, since we relied on an insufficiently-seeded PRNG.
850 Description: OSS hangs after "All ost request buffers busy"
851 Details : A deadlock between quota and journal operations caused OSS
852 hangs after printing "All ost request buffers busy."
855 Frequency : always on liblustre builds
857 Description: Cleanup compiler warnings on liblustre
860 Frequency : always on liblustre builds on XT3
862 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
866 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
867 > 32000 subdirectories
868 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
869 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
870 subdirectory count crosses 32000. This will aid e2fsck to
871 correctly handle more than 32000 subdirectories.
874 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
876 Description: System hang on startup
877 Details : This bug allowed the liblustre (e.g. catamount) client to
878 return to the app before handling all startup RPCs. This
879 could leave the node unresponsive to lustre network traffic
880 and manifested as a server ptllnd timeout.
882 Severity : enhancement
884 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
885 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
886 boolean to control whether to print peer debug info when a
887 client's RPC times out.
892 Description: make O_SYNC working on 2.6 kernels
893 Details : 2.6 kernels use different method for mark pages for write,
894 so need add a code to lustre for O_SYNC work.
899 Description: Failure to close file and release space on NFS
900 Details : Put inode details into lock acquired in ll_intent_file_open.
901 Use mdc_intent_lock in ll_intent_open to properly
902 detect all kind of errors unhandled by mdc_enqueue.
907 Description: proc file read during shutdown sometimes raced obd removal,
909 Details : Add lock to prevent obd access after proc file removal.
912 Frequency : Only for files larger than 4GB on 32-bit clients.
914 Description: improperly doing page alignment of locks
915 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
916 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
921 Description: RPCs being resent when they shouldn't be
922 Details : Some RPCs that should not be resent are being resent. This
923 can cause inconsistencies in the RPC state machine. Do not
924 resend such requests.
927 Frequency : rare, only with NFS export
929 Description: Crash on NFS re-export node
930 Details : under very unusual load conditions an assertion is hit in
931 ll_intent_file_open()
934 Frequency : only if OST filesystem is corrupted
936 Description: client incorrectly hits assertion in ptlrpc_replay_req()
937 Details : for a short time RPCs with bulk IO are in the replay list,
938 but replay of bulk IOs is unimplemented. If the OST filesystem
939 is corrupted due to disk cache incoherency and then replay is
940 started it is possible to trip an assertion. Avoid putting
941 committed RPCs into the replay list at all to avoid this issue.
946 Description: large O_DIRECT requests fail under memory pressure/fragmentation
947 Details : Large single O_DIRECT read and write calls can fail to allocate
948 a sufficiently large buffer to process the request. In case of
949 allocation failure the allocation is retried with a smaller
950 buffer and broken into smaller requests.
952 Severity : enhancement
954 Description: Add -o localflock option to simulate old noflock behaviour.
955 Details : This will achieve local-only flock/fcntl locks coherentness.
960 Description: versioning check is incomplete
961 Details : Checking the version difference of client vs. server, report
962 error if the gap is too big.
967 Description: add support PG_writeback bit
968 Details : add support for PG_writeback bit for Lustre, for more carefull
969 work with page cache in 2.6 kernel. This also fix some deadlocks
970 and remove hack for work O_SYNC with 2.6 kernel.
972 Severity : enhancement
974 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
975 Details : The uninit_groups feature works in conjunction with the kernel
976 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
977 up the pass1 processing of e2fsck. This is a read-only feature
978 in ldiskfs2 only, so older kernels and current ldiskfs cannot
979 mount filesystems that have had this feature enabled.
981 Severity : enhancement
983 Description: Improve multi-block allocation algorithm to avoid fragmentation
984 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
985 allocation locality and avoid filesystem fragmentation.
987 ------------------------------------------------------------------------------
989 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
991 * Support for kernels:
992 2.6.9-42.0.3.EL (RHEL 4)
994 2.4.21-47.0.1.EL (RHEL 3)
995 2.6.12.6 vanilla (kernel.org)
996 2.6.16.21-0.8 (SLES10)
997 * Recommended e2fsprogs version: 1.39.cfs2-0
999 * The backwards-compatible /proc/sys/portals symlink has been removed
1000 in this release. Before upgrading, please ensure that you change
1001 any configuration scripts or /etc/sysctl.conf files that access
1002 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1003 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1004 in advance of the upgrade on any system running Lustre 1.4.6 or
1005 newer, since /proc/sys/lnet was added in that version.
1006 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1010 Frequency : only when quota is used
1012 Description: avoid scanning export list for quota master
1013 Details : Change the algorithms to avoid scanning export list in order
1014 to improve the efficiency.
1017 Frequency : MDS failover only, very rarely
1019 Description: "went back in time" messages on mds failover
1020 Details : The greatest transno may be lost when the current operation
1021 finishes with an error (transno==0) and the client's last_rcvd
1022 record is over-written. Save the greatest transno in the
1023 mds_last_transno for this case.
1026 Frequency : always for specific kernels and striping counts
1028 Description: client may get "Matching packet too big" without ACL support
1029 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1030 when trying to access files in certain configurations. The
1031 clients should in fact be denied when mounting because they do
1032 not understand ACLs.
1035 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1037 Description: many clients connecting with IO in progress causes connect timeouts
1038 Details : Avoid synchronous journal commits to avoid delays caused by many
1039 clients connecting/disconnecting when bulk IO is in progress.
1040 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1041 OST_IO_PORTAL to avoid delays behind potentially many pending
1045 Frequency : occasionally with multiple writers to a single file
1047 Description: shared writes to file may result in wrong size reported by stat()
1048 Details : Allow growing of kms when extent lock is cancelled
1051 Frequency : always with random mmap IO to multi-striped file
1053 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1054 Details : Only if the hole is at the end of OST object so that kms is too
1055 small. Fix is to increase kms accordingly in ll_nopage.
1058 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1060 Description: writes to a missing object would leak memory on the OST
1061 Details : If there is an inconsistency between the MDS and OST filesystems,
1062 such that the MDS references an object that doesn't exist, writes
1063 to that object will leak memory due to incorrect cleanup in the
1064 error handling path, eventually running out of memory on the OST.
1069 Description: Creating too long symlink causes lustre errors
1070 Details : Check symlink and name lengths before sending requests to MDS.
1073 Frequency : only if flock is enabled (not on by default)
1075 Description: posix locks not released on fd closure on 2.6.9+
1076 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1077 kernels, this caused such locks not to be released on fd close and
1078 then assertions on fs unmount about still used locks.
1081 Frequency : MDS failover only, very rarely
1083 Description: clients may get ASSERTION(granted_lock != NULL)
1084 Details : When request was taking a long time, and a client was resending
1085 a getattr by name lock request. The were multiple lock requests
1086 with the same client lock handle and
1087 mds_getattr_name->fixup_handle_for_resent_request found one of the
1088 lock handles but later failed with ASSERTION(granted_lock != NULL).
1093 Description: handle->h_buffer_credits > 0, assertion failure
1094 Details : h_buffer_credits is zero after truncate, causing assertion
1095 failure. This patch extends the transaction or creates a new
1099 Frequency : NFS re-export or patchless client
1100 Bugzilla : 11179, 10796
1101 Description: Crash on NFS re-export node (__d_move)
1102 Details : We do not want to hash the dentry if we don't have a lock.
1103 But if this dentry is later used in d_move, we'd hit uninitialised
1104 list head d_hash, so we just do this to init d_hash field but
1105 leave dentry unhashed.
1108 Frequency : NFS re-export or patchless client
1110 Description: NFS exports has problem with symbolic link
1111 Details : lustre client didn't properly install dentry when re-exported
1112 to NFS or running patchless client.
1115 Frequency : NFS re-export or patchless client
1117 Description: Various nfs/patchless fixes.
1118 Details : fixes reuse disconected alias for lookup process - this fixes
1119 warning "find_exported_dentry: npd != pd",
1120 fix permission error with open files at nfs.
1121 fix apply umask when do revalidate.
1124 Frequency : occasional
1126 Description: Crash on NFS re-export node
1127 Details : calling clear_page() on the wrong pointer triggered oops in
1128 generic_mapping_read().
1131 Frequency : rarely, using O_DIRECT IO
1133 Description: unaligned directio crashes client with LASSERT
1134 Details : check for unaligned buffers before trying any requests.
1137 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1139 Description: stale data returned from RAID cache
1140 Details : If only a small amount of IO is done to the RAID device before
1141 reading it again it is possible to get stale data from the RAID
1142 cache instead of reading it from disk.
1145 Frequency : always for sles10 kernel
1147 Description: sles10 support
1148 Details : ll_follow_link: compile fixes and using of nd_set_link
1149 under newer kernels.
1152 Frequency : depends on arch, kernel and compiler version, always on sles10
1155 Description: recursive or deep enough symlinks cause stack overflow
1156 Details : getting rid of large stack-allocated variable in
1160 Frequency : depends on hardware
1162 Description: lustre write performance loss in the SLES10 kernel
1163 Details : the performance loss is caused by using of write barriers in the
1164 ext3 code. The SLES10 kernel turns barrier support on by
1165 default. The fix is to undo that change for ldiskfs.
1167 ------------------------------------------------------------------------------
1169 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1171 * Support for kernels:
1172 2.6.9-42.0.3EL (RHEL 4)
1173 2.6.5-7.276 (SLES 9)
1174 2.4.21-47.0.1.EL (RHEL 3)
1175 2.6.12.6 vanilla (kernel.org)
1179 Frequency : quota enabled and large files being deleted
1181 Description: releasing more than 4GB of quota at once hangs OST
1182 Details : If a user deletes more than 4GB of files on a single OST it
1183 will cause the OST to spin in an infinite loop. Release
1184 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1189 Description: statfs data retrieved from /proc may be stale or zero
1190 Details : When reading per-device statfs data from /proc, in the
1191 {kbytes,files}_{total,free,avail} files, it may appear
1192 as zero or be out of date.
1195 Frequency : systems with MD RAID1 external journal devices
1197 Description: lconf's call to blkid is confused by RAID1 journal devices
1198 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1199 of returning all block devices that match the journal UUID.
1202 Frequency : always, for aggregate stripe size over 4GB
1204 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1205 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1206 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1209 Frequency : always if "lfs find" used on a local file/directory
1211 Description: "lfs find" segfaults if used on a local file/directory
1212 Details : The case where a directory component was not specified wasn't
1213 handled correctly. Handle this properly.
1216 Frequency : always on ppc64
1218 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1219 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1220 perform endianness conversion.
1223 Frequency : rarely (truncate to non-zero file size after write under load)
1224 Bugzilla : 10730, 10687
1225 Description: Files padded with zeros to next 4K multiple
1226 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1227 it is possible that files that are truncated to a non-zero size
1228 immediately after being written are filled with zero bytes beyond
1229 the truncated size. No file data is lost.
1231 Severity : enhancement
1233 Description: Allow recovery/failover for liblustre clients.
1234 Details : liblustre clients were unaware of failover configurations until
1237 Severity : enhancement
1239 Description: user file locks should fail when not mounting with flock option
1240 Details : Set up an error-returning stub in ll_file_operations.lock field
1241 to prevent incorrect behaviour when client is mounted without
1242 flock option. Also, set up properly f_op->flock field for
1246 Frequency : always on ia64
1248 Description: "lfs df" loops on printing out MDS statfs information
1249 Details : The obd_ioctl_data was not initialized and in some systems
1250 this caused a failure during the ioctl that did not return
1251 an error. Initialize the struct and return an error on failure.
1254 Frequency : SLES 9 only
1256 Description: Error of copying files with lustre special EAs as root
1257 Details : Client side always return success for setxattr call for lustre
1258 special xattr (currently only "trusted.lov").
1261 Frequency : rarely on clusters with both ia64+i386 clients
1263 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1264 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1265 clients are doing concurrent writes to the same file it is
1266 possible that the ia64 clients may LASSERT because the OST
1267 extent locks are not PAGE_SIZE aligned. Ensure that grown
1268 locks are always aligned on the request boundary.
1271 Frequency : specific use, occasional
1273 Description: Overwriting in use executable truncates on-disk binary image
1274 Details : If one node attempts to overwrite an executable in use by
1275 another node, we now correctly return ETXTBSY instead of
1276 truncating the file.
1278 Severity : enhancement
1280 Description: Async OSC create to avoid the blocking unnecessarily.
1281 Details : If a OST has no remain object, system will block on the creating
1282 when need to create a new object on this OST. Now, ways use
1283 pre-created objects when available, instead of blocking on an
1284 empty osc while others are not empty. If we must block, we block
1285 for the shortest possible period of time.
1290 Description: chmod on Lustre root is propagated to other clients
1291 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1292 invalid by the follow_mount time.
1295 Frequency : liblustre clients only
1297 Description: Race in 'instant cancel' lock handling could lead to such locks
1298 never to be granted in case of SMP MDS
1299 Details : Do not destroy not yet granted but cbpending locks in
1303 Frequency : replay/resend of open
1305 Description: non null lock assetion failure in mds_intent_policy
1306 Details : Trying to replay/resend lockless open requests resulted in
1307 mds_open() returning 0 with no lock. Now it sets a flag if
1308 a lock is going to be returned.
1310 Severity : enhancement
1312 Description: Checksum enhancements
1313 Details : New checksum enhancements allow for resending RPCs that failed
1316 Severity : enhancement
1318 Description: Tunables on number of dirty pages in cacche
1319 Details : Allow to set limit on number of dirty pages cached.
1324 Description: client crash on unmount - lock still has references
1325 Details : In some error handling cases it was possible to leak a lock
1326 reference on a client while accessing a file. This was not
1327 harmful to the client during operation, but would cause the
1328 client to crash when the filesystem is unmounted.
1331 Frequency : specific case, rare
1333 Description: ETXTBSY on mds though file not in use by client
1334 Details : ETXTBSY is no longer incorrectly returned when attempting to
1335 chmod or chown a directory that the user previously tried to
1336 execute or a currently-executing binary.
1339 Frequency : extremely rare except on liblustre-based clients
1341 Description: Lustre space not freed when files are deleted
1342 Details : Clean up open-unlinked files after client eviction. Previously
1343 the unlink was skipped and the files remained as orphans.
1348 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1349 Details : In some cases it was possible to send a blocking callback to
1350 a client doing a glimpse, even though that client didn't get
1351 a lock granted. When the glimpse lock is cancelled on the OST
1352 the freed lock is left on the waiting list and corrupted the list.
1355 Frequency : all core dumps
1357 Description: Broke core dumps to lustre
1358 Details : Negative dentry may be unhashed if parent does not have UPDATE
1359 lock, but some callers, e.g. do_coredump, expect dentry to be
1360 hashed after successful create, hash it in ll_create_it.
1362 ------------------------------------------------------------------------------
1364 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1366 * Support for kernels:
1367 2.6.9-42.0.2.EL (RHEL 4)
1368 2.6.5-7.276 (SLES 9)
1369 2.4.21-40.EL (RHEL 3)
1370 2.6.12.6 vanilla (kernel.org)
1374 Frequency : always on RHEL 3
1376 Description: Number of open files grows over time
1377 Details : The number of open files grows over time, whether or not
1378 Lustre is started. This was due to a filp leak introduced
1379 by one of our kernel patches.
1381 ------------------------------------------------------------------------------
1383 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1385 * Support for kernels:
1386 2.6.9-42.EL (RHEL 4)
1387 2.6.5-7.276 (SLES 9)
1388 2.4.21-40.EL (RHEL 3)
1389 2.6.12.6 vanilla (kernel.org)
1394 Bugzilla : 5719, 9635, 9792, 9684
1395 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1396 Details : If a server is under heavy load and cannot reply to new
1397 connection requests before the client resends the (re)connect,
1398 the connection handling code can behave badly if two service
1399 threads are concurrently handing separate (re)connections from
1400 the same client. Add better locking to the connection handling
1401 code, and ensure that only a single connection will be processed
1402 for a given client UUID, even if the lock is dropped.
1404 Severity : enhancement
1406 Description: add TCP zero-copy support to kernel
1407 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1408 sends if the hardware supports scatter-gather and checksumming.
1409 This allows socklnd to do client-write and server-read more
1410 efficiently and reduce CPU utilization from skbuf copying.
1413 Frequency : only if NFS exporting from client
1415 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1416 Details : If setting attributes on a file created under NFS that had
1417 never been opened it would be possible to oops the client
1418 if the file had no objects.
1421 Frequency : always for liblustre
1423 Description: liblustre client does MDS+OSTs setattr RPC for each write
1424 Details : When doing a write from a liblustre client, the client
1425 incorrectly issued an RPC to the MDS and each OST the file was
1426 striped over in order to update the timestamps. When writing
1427 with small chunks and many clients this could overwhelm the MDS
1428 with RPCs. In all cases it would slow down the write because
1429 these RPCs are unnecessary.
1431 Severity : enhancement
1433 Description: allow number of MDS service threads to be changed at module load
1434 Details : It is now possible to change the number of MDS service threads
1435 running. Adding "options mds mds_num_threads={N}" to the MDS's
1436 /etc/modprobe.conf will set the number of threads for the next
1437 time Lustre is restarted (assuming the "mds" module is also
1438 reloaded at that time). The default number of threads will
1439 stay the same, 32 for most systems.
1444 Description: OST crash if filesystem is unformatted or corrupt
1445 Details : If an OST is started on a device that has never been formatted
1446 or if the filesystem is corrupt and cannot even mount then the
1447 error handling cleanup routines would dereference a NULL pointer.
1452 Description: NULL pointer deref in llap_from_page.
1453 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1454 incorrect) mapping if the page was truncated/reclaimed while it was
1455 searched for. Check for this condition and skip such pages when
1456 doing readahead. Introduce extra check to llap_from_page() to
1457 verify page->mapping->host is non-NULL (so page is not anonymous).
1460 Frequency : Sometimes when using sys_sendfile
1462 Description: "page not covered by a lock" warnings from ll_readpage
1463 Details : sendfile called ll_readpage without right page locks present.
1464 Now we introduced ll_file_sendfile that does necessary locking
1465 around call to generic_file_sendfile() much like we do in
1469 Frequency : with certain MDS communication failures at client mount time
1471 Description: NULL pointer deref after failed client mount
1472 Details : a client connection request may delayed by the network layer
1473 and not be sent until after the PTLRPC layer has timed out the
1474 request. If the client fails the mount immediately it will try
1475 to clean up before the network times out the request. Add a
1476 reference from the request import to the obd device and delay
1477 the cleanup until the network drops the request.
1480 Frequency : occasionally during client (re)connect
1482 Description: assertion failure during client (re)connect
1483 Details : processing a client connection request may be delayed by the
1484 client or server longer than the client connect timeout. This
1485 causes the client to resend the connection request. If the
1486 original connection request is replied in this interval, the
1487 client may trip an assertion failure in ptlrpc_connect_interpret()
1488 which thought it would be the only running connect process.
1491 Frequency : only with obd_echo servers and clients that are rebooted
1493 Description: kernel BUG accessing uninitialized data structure
1494 Details : When running an obd_echo server it did not start the ping_evictor
1495 thread, and when a client was evicted an uninitialized data
1496 structure was accessed. Start the ping_evictor in the RPC
1497 service startup instead of the OBD startup.
1499 Severity : enhancement
1500 Bugzilla : 10193 (patchless)
1501 Description: Remove dependency on various unexported kernel interfaces.
1502 Details : No longer need reparent_to_init, exit_mm, exit_files,
1503 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1506 Frequency : rare (only users of deprecated and unsupported LDAP config)
1508 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1509 Details : LDAP apparently contains 'lustreName' attributes instead of
1510 'name'. A simple remapping of the name is sufficient.
1513 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1515 Description: waiting_locks_callback trips kernel BUG if client is evicted
1516 Details : Running with the dump_on_timeout debug flag turned on makes
1517 it possible that the waiting_locks_callback() can try to dump
1518 the Lustre kernel debug logs from an interrupt handler. Defer
1519 this log dumping to the expired_lock_main() thread.
1521 Severity : enhancement
1523 Description: Support NFS exporting on 2.6 kernels.
1524 Details : Implement non-rawops metadata methods for NFS server to use without
1525 changing NFS server code.
1528 Frequency : very rare (synthetic metadata workload only)
1530 Description: two racing renames might cause an MDS thread to deadlock
1531 Details : Running the "racer" program may cause one MDS thread to rename
1532 a file from being the source of a rename to being the target of
1533 a rename at exactly the same time that another thread is doing
1534 so, and the second thread has already enqueued these locks after
1535 doing a lookup of the target and is trying to relock them in
1536 order. Ensure that we don't try to re-lock the same resource.
1539 Frequency : only very large systems with liblustre clients
1541 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1542 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1543 all OSTs in parallel. This allows the eviction of stale liblustre
1544 clients to proceed much faster than if they were done in series,
1545 and also offers similar improvements for other set_info RPCs.
1550 Description: excessive CPU usage during initial read phase on client
1551 Details : During the initial read phase on a client, it would agressively
1552 retry readahead on the file, consuming too much CPU and impacting
1553 performance (since 1.4.5.8). Improve the readahead algorithm
1554 to avoid this, and also improve some other common cases (read
1555 of small files in particular, where "small" is files smaller than
1556 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1561 Description: MDS crash when receiving packet with unknown intent.
1562 Details : Do not LBUG in unknown intent case, just return -EFAULT
1564 Severity : enhancement
1565 Bugzilla : 9293, 9385
1566 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1567 Details : Do not serialize getattr (non-intent version) and statfs.
1570 Frequency : occasional, when OST network is overloaded/intermittent
1572 Description: client evicted by OST after bulk IO timeout
1573 Details : If a client sends a bulk IO request (read or write) the OST
1574 may evict the client if it is unresposive to its data GET/PUT
1575 request. This is incorrect if the network is overloaded (takes
1576 too long to transfer the RPC data) or dropped the OST GET/PUT
1577 request. There is no need to evict the client at all, since
1578 the pinger and/or lock callbacks will handle this, and the
1579 client can restart the bulk request.
1582 Frequency : Always when mmapping file with no objects
1584 Description: client crashes when mmapping file with no objects
1585 Details : Check that we actually have objects in a file before doing any
1586 operations on objects in ll_vm_open, ll_vm_close and
1592 Description: Request leak when working with deleted CWD
1593 Details : Introduce advanced request refcount tracking for requests
1594 referenced from lustre intent.
1596 Severity : Enhancement
1598 Description: Cache open file handles on client.
1599 Details : MDS now will return special lock along with openhandle, if
1600 requested and client is allowed to hold openhandle, even if unused,
1601 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1602 closing files for every read/write openration.
1604 Severity : Enhancement
1606 Description: Cache open negative dentries on client when possible.
1607 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1608 negative dentries on lock revocation.
1613 Description: Remounting a client read-only wasn't possible with a zconf mount
1614 Details : It wasn't possible to remount a client read-only with llmount.
1616 Severity : enhancement
1617 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1618 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1619 MPICH is included to add Lustre-specific ADIO interfaces.
1620 This is based closely on the UFS ADIO layer and only differs
1621 in file creation, in order to allow the OST striping to be set.
1622 This is user-contributed code and not supported by CFS.
1627 Description: extended inode attributes (immutable, append-only) work improperly
1628 when 2.4 and 2.6 kernels are used on client/server or vice versa
1629 Details : Introduce kernel-independent values for these flags.
1631 Severity : enhancement
1634 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1635 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1636 granularity. Now, Fractional MB granularity is be supported,
1637 this is very useful for low memory system.
1639 Severity : enhancement
1641 Description: Getattr by fid
1642 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1643 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1644 have locks and avoids getattr rpc requests for it on every stat.
1647 Frequency : Always, for filesystems larger than 2TB
1649 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1650 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1651 filesystems larger than 2TB to be mounted with mballoc enabled.
1654 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1656 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1657 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1658 filesystems larger than 2TB is mounted then the kernel will
1659 silently corrupt the start of the filesystem. CONFIG_LBD is
1660 enabled for all CFS-supported kernels, but the possibility of
1661 this happening with a modified kernel config exists.
1663 Severity : enhancement
1665 Description: add client O_DIRECT support for 2.6 kernels
1666 Details : It is now possible to do O_DIRECT reads and writes to files
1667 in the Lustre client mountpoint on 2.6 kernel clients.
1669 Severity : enhancement
1671 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1672 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1673 parallel, not waiting for response from every OST before sending
1674 a rpc to the next OST.
1679 Description: setattr vs write race when updating file timestamps
1680 Details : Client processes that update a file timestamp into the past
1681 right after writing to the file (e.g. tar) it is possible that
1682 the updated file modification time can be reset to the current
1683 time due to a race between processing the setattr and write RPC.
1685 Severity : enhancement
1687 Description: Bring 'lfs find' closer in line with regular Linux find.
1688 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1689 -print0 options and obtains all the needed info through the lustre
1692 Severity : enhancement
1694 Description: support up to 1024 configured devices on one node
1695 Details : change obd_dev array from statically allocated to dynamically
1696 allocated structs as they are first used to reduce memory usage
1701 Description: Flush dirty partially truncated pages during truncate
1702 Details : Immediatelly flush partially truncated pages in filter_setattr,
1703 this way we completely avoid having any pages in page cache on OST
1704 and can retire ugly workarounds during writes to flush such pages.
1709 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1710 Details : protect inode from truncation within vfs_unlink() context
1711 just take a reference before calling vfs_unlink() and release it
1712 when parent's i_sem is free.
1717 Description: last_id value checked outside lock on OST caused LASSERT failure
1718 Details : If there were multiple MDS->OST object precreate requests in
1719 flight, it was possible that the OST's last object id was checked
1720 outside a lock and incorrectly tripped an assertion. Move checks
1721 inside locks, and discard old precreate requests.
1724 Frequency : always, if extents are used on OSTs
1726 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1727 Details : OSTs using the extents format would not zero the high 16 bits of
1728 the index physical block number. This is not a problem for any
1729 OST filesystems smaller than 16TB, and no kernels support ext3
1730 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1731 new/modified files) and can be fixed for existing filesystems
1732 with e2fsprogs-1.39-cfs1.
1737 Description: import connection selection may be incorrect if timer wraps
1738 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1739 import connections to be ignored if the 32-bit jiffies counter
1740 wraps. Use a 64-bit jiffies counter.
1743 Frequency : very large clusters immediately after boot
1745 Description: LNET request buffers exhausted under heavy short-term load
1746 Details : If a large number of client requests are generated on a service
1747 that has previously never seen so many requests it is possible
1748 that the request buffer growth cannot keep up with the spike in
1749 demand. Instead of dropping incoming requests, they are held in
1750 the LND until the RPC service can accept more requests.
1753 Frequency : Sometimes during replay
1755 Description: Assertion failure in ll_local_open after replay.
1756 Details : If replay happened on an open request reply before we were able
1757 to set replay handler, reply will become not swabbed tripping the
1758 assertion in ll_local_open. Now we set the handler right after
1759 recognising of open request
1762 Frequency : very rare
1764 Description: kernel reports "badness in vsnprintf"
1765 Details : Reading from the "recovery_status" /proc file in small chunks
1766 may cause a negative length in lprocfs_obd_rd_recovery_status()
1767 call to vsnprintf() (which is otherwise harmless). Exit early
1768 if there is no more space in the output buffer.
1770 Severity : enhancement
1772 Description: clear OBD RPC statistics by writing to them
1773 Details : It is now possible to clear the OBD RPC statistics by writing
1774 to the "stats" file.
1779 Description: Client mtime is not the same on different clients after utimes
1780 Details : In some cases, the client was using the utimes() syscall on
1781 a file cached on another node. The clients now validate the
1782 ctime from the MDS + OSTs to determine which one is right.
1787 Description: Inability to activate failout mode
1788 Details : lconf script incorrectly assumed that in pythong string's numeric
1789 value is used in comparisons.
1792 Frequency : always with multiple stripes per file
1794 Description: Inefficient object allocation for mutli-stripe files
1795 Details : When selecting which OSTs to stripe files over, for files with
1796 a stripe count that divides evenly into the number of OSTs,
1797 the MDS is always picking the same starting OST for each file.
1798 Return the OST selection heuristic to the original design.
1803 Description: mount failures may take full timeout to return an error
1804 Details : Under some heavy load conditions it is possible that a
1805 failed mount can wait for the full obd_timeout interval,
1806 possibly several minutes, before reporting an error.
1807 Instead return an error as soon as the status is known.
1809 Frequency : quota enabled and large files being deleted
1811 Description: releasing more than 4GB of quota at once hangs OST
1812 Details : If a user deletes more than 4GB of files on a single OST it
1813 will cause the OST to spin in an infinite loop. Release
1814 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1819 Description: statfs data retrieved from /proc may be stale or zero
1820 Details : When reading per-device statfs data from /proc, in the
1821 {kbytes,files}_{total,free,avail} files, it may appear
1822 as zero or be out of date.
1825 Frequency : systems with MD RAID1 external journal devices
1827 Description: lconf's call to blkid is confused by RAID1 journal devices
1828 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1829 of returning all block devices that match the journal UUID.
1832 Frequency : always, for aggregate stripe size over 4GB
1834 Description: assertion fails when trying to use 4GB stripe size
1835 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1836 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1839 Frequency : always on ppc64
1841 Description: the first write on an ext3 filesystem with mballoc got stuck
1842 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1843 perform endianness conversion.
1845 ------------------------------------------------------------------------------
1847 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1849 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1850 INTEROPERATE with older versions automatically. Please read the
1851 user documentation before upgrading any part of a live system.
1852 * WARNING: Lustre networking configuration changes are required with
1853 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1856 * Support for kernels:
1857 2.6.9-22.0.2.EL (RHEL 4)
1858 2.6.5-7.244 (SLES 9)
1859 2.6.12.6 vanilla (kernel.org)
1862 Severity : enhancement
1863 Bugzilla : 7981/8208
1864 Description: Introduced Lustre Networking (LNET)
1865 Details : LNET is new networking infrastructure for Lustre, it includes
1866 a reorganized network configuration mode (see the user
1867 documentation for full details) as well as support for routing
1868 between different network fabrics. Lustre Networking Devices
1869 (LNDs) for the supported network fabrics have also been
1870 created for this new infrastructure.
1872 Severity : enhancement
1873 Description: Introduced Access control lists
1874 Details : clients can set ACLs on files and directories in order to have
1875 more fine-grained permissions than the standard Unix UGO+RWX.
1876 The MDS must be started with the "-o acl" mount option.
1878 Severity : enhancement
1879 Description: Introduced filesystem quotas
1880 Details : Administrators may now establish per-user quotas on the
1883 Severity : enhancement
1885 Description: Configuration change for the XT3
1886 The PTLLND is now used to run Lustre over Portals on the XT3
1887 The configure option(s) --with-cray-portals are no longer used.
1888 Rather --with-portals=<path-to-portals-includes> is used to
1889 enable building on the XT3. In addition to enable XT3 specific
1890 features the option --enable-cray-xt3 must be used.
1895 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1896 Details : A race between allocating a new llcd and re-getting the llcd_lock
1897 allowed another thread to grab newly-allocated llcd.
1899 Severity : enhancement
1901 Description: 2.6 OST async journal commit and locking fix to improve performance
1902 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1903 2.6 kernels are now async as they already were in 2.4 kernels so
1904 that they can commit concurrently with the network bulk transfer.
1905 For block-allocated files the filter allocation semaphore is held
1906 to avoid filesystem fragmentation during allocation. BKL lock
1907 removed for 2.6 xattr operations where it is no longer needed.
1912 Description: lconf incorrectly determined whether two IP networks could talk
1913 Details : In some more complicated routing and multiple-network
1914 configurations, lconf will avoid trying to make a network
1915 connection to a disjoint part of the IP space. It was doing the
1916 math incorrectly for one set of cases.
1921 Description: Fix for potential infinite loop processing records in an llog.
1922 Details : If an llog record is corrupted/zeroed, it is possible to loop
1923 forever in llog_process(). Validate the llog record length
1924 and skip the remainder of the block on error.
1927 Frequency : occasional (liblustre only)
1929 Description: liblustre could not open files whose last component is a symlink
1930 Details : sysio_path_walk() would incorrectly pass the open intent to
1931 intermediate path components.
1934 Frequency : rare (liblustre only with non-standard tuning)
1935 Bugzilla : 7201 (7350)
1936 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1937 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1938 after it has been released, only do this for referenced locks
1940 Severity : enhancement
1942 Description: specifying an (invalid) directory default stripe_size of -1
1943 would reset the directory default striping
1944 Details : stripe_size -1 was used internally to signal directory stripe
1945 removal, now use "all default" to signal dir stripe removal
1946 as a directory striping of "all default" is not useful
1949 Frequency : common for large clusters running liblustre clients
1951 Description: doing an ls when liblustre clients are running is slow
1952 Details : sending a glimpse AST to a liblustre client waits for every AST
1953 to time out, as liblustre clients will not respond. Since they
1954 cannot cache data we refresh the OST lock LVB from disk instead.
1956 Severity : enhancement
1958 Description: doing an ls at the same time as file IO can be slow
1959 Details : enqueue and other "small" requests can be blocked behind many
1960 large IO requests. Create a new OST IO portal for non-IO
1961 requests so they can be processed faster.
1964 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1966 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1967 created file with mode 000
1968 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1969 allow an ftruncate() as a non-root user to a file with mode 000.
1970 Lustre now respects this flag to disable mode checks when
1971 truncating a file owned by the user
1974 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1977 Description: Revoking locks from clients that went dead or catatonic might take
1979 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1980 cancellation of such locks instant on servers without waiting for
1981 any reply from clients. Clients drops these locks when cancel
1982 notification from server is received without replying.
1985 Frequency : liblustre-only, when liblustre client dies or becomes busy
1987 Description: Doing ls on Linux clients can take a long time with active
1989 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1990 granting such locks to it in the first place if possible. Locks
1991 are taken by proxy on the OST during the read or write and
1992 dropped immediately afterward. Add connect flags handling, do
1993 not grant locks to liblustre clients for glimpse ASTs.
1995 Severity : enhancement
1997 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1998 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1999 sequential reads and behaves badly with applications doing random
2000 reads. Improve it by reading ahead at least read region, and
2001 avoiding excessive large RPC for small reads.
2003 Severity : enhancement
2005 Description: Creating more than 1000 files for a single job may cause a load
2006 imbalance on the OSTs if there are also a large number of OSTs.
2007 Details : qos_prep_create() uses an OST index reseed value that is an
2008 even multiple of the number of available OSTs so that if the
2009 reseed happens in the middle of the object allocation it will
2010 still utilize the OSTs as uniformly as possible.
2015 Description: OST or MDS may oops in ping_evictor_main()
2016 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2017 but doesn't restart at beginning of obd_exports_timed list
2020 Severity : enhancement
2022 Description: improve by-nid export eviction on the MDS and OST
2023 Details : allow multiple exports with the same NID to be evicted at one
2024 time without re-searching the exports list.
2027 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2029 Description: MDS may oops in groups_free()
2030 Details : in rare race conditions a newly allocated group_info struct is
2031 freed again, and this can be NULL. The 2.4 compatibility code
2032 for groups_free() checked for a NULL pointer, but 2.6 did not.
2035 Frequency : common for liblustre clients doing little filesystem IO
2036 Bugzilla : 9352, 7313
2037 Description: server may evict liblustre clients accessing contended locks
2038 Details : if a client is granted a lock or receives a completion AST
2039 with a blocking AST already set it would not reply to the AST
2040 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2043 Frequency : lfs setstripe, only systems with more than 160 OSTs
2045 Description: unable to set striping with a starting offset beyond OST 160
2046 Details : llapi_create_file() incorrectly limited the starting stripe
2047 index to the maximum single-file stripe count.
2050 Frequency : LDAP users only
2052 Description: lconf did not handle in-kernel recovery with LDAP properly
2053 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2055 Severity : enhancement
2057 Description: bind OST threads to NUMA nodes to improve performance
2058 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2059 node and do their allocations there to localize memory access
2061 Severity : enhancement
2063 Description: llmount can determine client NID directly from Myrinet (GM)
2064 Details : the client NID code from gmnalnid was moved directly into
2065 llmount, removing the need to use this or specifying the
2066 client NID explicitly when mounting GM clients with zeroconf
2069 Frequency : if client is started with down MDS
2071 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2072 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2073 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2074 ptlrpc_import_delayed_req() because we didn't check if the
2075 request was interrupted, and we also didn't break out of the
2076 event loop if there was no timeout
2081 Description: data loss during non-page-aligned writes to a single file from
2082 both multiple nodes and multiple threads on one node at same time
2083 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2084 inconsistency led to false short-reads, that were cached and later
2085 used by ->prepare_write() to fill in partially written page,
2086 leading to data loss.
2089 Frequency : always, if lconf --abort_recovery used
2091 Description: lconf --abort_recovery fails with 'Operation not supported'
2092 Details : lconf was attempting to abort recovery on the MDT device and not
2095 Severity : enhancement
2097 Description: remove cleanup logs
2098 Details : replace lconf-generated cleanup logs with lustre internal
2099 cleanup routines. Eliminates the need for client-cleanup and
2102 Severity : enhancement
2104 Description: add support for EAs (user and system) on lustre filesystems
2105 Details : it is now possible to store extended attributes in the Lustre
2106 client filesystem, and with the user_xattr mount option it
2107 is possible to allow users to store EAs on their files also
2109 Severity : enhancement
2111 Description: Add possibility (config option) to show minimal available OST free
2113 Details : When compiled with --enable-mindf configure option, statfs(2)
2114 (and so, df) will return least minimal free space available from
2115 all OSTs as amount of free space on FS, instead of summary of
2116 free spaces of all OSTs.
2118 Severity : enhancement
2120 Description: do not expand extent locks acquired on OST-side
2121 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2122 by server: they are not cached anyway.
2125 Frequency : when mmap is used/binaries executed from Lustre
2127 Description: Unmmap pages before throwing them away from read cache.
2128 Details : llap_shrink cache now attempts to unmap pages before discarding
2129 them (if unmapping failed - do not discard). SLES9 kernel has
2130 extra checks that trigger if this unmapping is not done first.
2135 Description: lconf didn't resolve symlinks before checking to see whether a
2136 given mountpoint was already in use
2139 Frequency : when migrating failover services
2140 Bugzilla : 6395, 9514
2141 Description: When migrating a subset of services from a node (e.g. failback
2142 from a failover service node) the remaining services would
2143 time out and evict clients.
2144 Details : lconf --force (implied by --failover) sets the global obd_timeout
2145 to 5 seconds in order to quickly disconnect, but this caused
2146 other RPCs to time out too quickly. Do not change the global
2147 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2149 Severity : enhancement
2150 Frequency : if MDS is started with down OST
2151 Bugzilla : 9439,5706
2152 Description: Allow startup/shutdown of an MDS without depending on the
2153 availability of the OSTs.
2154 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2155 Add appropriate locking and lov-osc refcounts for safe
2156 cleaning. Add osc abort_inflight calls in case the
2157 synchronize never started.
2160 Frequency : occasional (Cray XT3 only)
2162 Description: root not authorized to access files in CRAY_PORTALS environment
2163 Details : The client process capabilities were not honoured on the MDS in
2164 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2165 been accessed by an authorized user then root was able to access
2166 the file on the local client also. The root user capabilities
2167 are now allowed on the MDS, as this environment has secure UID.
2170 Frequency : occasional
2172 Description: ldiskfs "too long searching" message happens too often
2173 Details : A debugging message (otherwise harmless) prints too often on
2174 the OST console. This has been reduced to only happen when
2175 there are fragmentation problems on the filesystem.
2180 Description: Division by zero in statfs when all OSCs are inactive
2181 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2182 lov_statfs divides by value returned by lov_get_stripecnt().
2186 Bugzilla : 9489, 3273
2187 Description: First write from each client to each OST was only 4kB in size,
2188 to initialize client writeback cache, which caused sub-optimal
2189 RPCs and poor layout on disk for the first writen file.
2190 Details : Clients now request an initial cache grant at (re)connect time
2191 and so that they can start streaming writes to the cache right
2192 away and always do full-sized RPCs if there is enough data.
2193 If the OST is rebooted the client also re-establishes its grant
2194 so that client cached writes will be honoured under the grant.
2199 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2200 Details : Now I/O RPCs go to different portal number and (presumably) fast
2201 lock requests (and glimses) and other RPCs get their own service
2202 threads pool that should be able to service those RPCs
2205 Severity : enhancement
2207 Description: Ability to exchange lustre version between client and servers and
2208 issue warnings at client side if client is too old. Also for
2209 liblustre clients there is ability to refuse connection of too old
2211 Details : New 'version' field is added to connect data structure that is
2212 filled with version info. That info is later checked by server and
2216 Frequency : rare, liblustre only.
2217 Bugzilla : 9296, 9581
2218 Description: Two simultaneous writes from liblustre at offset within same page
2219 might proceed at the same time overwriting eachother with stale
2221 Details : I/O lock withing llu_file_prwv was released too early, before data
2222 actually was hitting the wire. Extended lock-holding time until
2223 server acknowledges receiving data.
2226 Frequency : extremely rare. Never observed in practice.
2228 Description: avoid generating lustre_handle cookie of 0.
2229 Details : class_handle_hash() generates handle cookies by incrementing
2230 global counter, and can hit 0 occasionaly (this is unlikely, but
2231 not impossible, because initial value of cookie counter is
2232 selected randonly). Value of 0 is used as a sentinel meaning
2233 "unassigned handle" --- avoid it. Also coalesce two critical
2234 sections in this function into one.
2236 Severity : enhancement
2238 Description: allow liblustre clients to delegate truncate locking to OST
2239 Details : To avoid overhead of locking, liblustre client instructs OST to
2240 take extent lock in ost_punch() on client's behalf. New connection
2241 flag is added to handle backward compatibility.
2243 Severity : enhancement
2244 Bugzilla : 4928, 7341, 9758
2245 Description: allow number of OST service threads to be specified
2246 Details : a module parameter allows the number of OST service threads
2247 to be specified via "options ost ost_num_threads={N}" in the
2248 OSS's /etc/modules.conf or /etc/modprobe.conf.
2252 Bugzilla : 6146, 9635, 9895
2253 Description: servers crash with bad pointer in target_handle_connect()
2254 Details : In rare cases when a client is reconnecting it was possible that
2255 the connection request was the last reference for that export.
2256 We would temporarily drop the export reference and get a new
2257 one, but this may have been the last reference and the export
2258 was just destroyed. Get new reference before dropping old one.
2260 Severity : enhancement
2261 Frequency : if client is started with failover MDS
2263 Description: Allow multiple MDS hostnames in the mount command
2264 Details : Try to read the configuration from all specified MDS
2265 hostnames during a client mount in case the "primary"
2268 Severity : enhancement
2270 Description: Stop sending data to evicted clients as soon as possible.
2271 Details : Check if the client we are about to send or are sending data to
2272 was evicted already. (Check is done every second of waiting,
2273 for which l_wait_event interface was extended to allow checking
2274 of exit condition at specified intervals).
2277 Frequency : rare, normally only when NFS exporting is done from client
2279 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2281 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2284 Frequency : always, if srand() is called before liblustre initialization
2286 Description: Liblustre uses system PRNG disturbing its usage by user application
2287 Details : Introduce internal to lustre fast and high-quality PRNG for
2288 lustre usage and make liblustre and some other places in generic
2289 lustre code to use it.
2291 Severity : enhancement
2292 Bugzilla : 9477, 9557, 9870
2293 Description: Verify that the MDS configuration logs are updated when xml is
2294 Details : Check if the .xml configuration logs are newer than the config
2295 logs stored on the MDS and report an error if this is the case.
2296 Request --write-conf, or allow starting with --old_conf.
2298 Severity : enhancement
2300 Description: Handle symlinks in the path when checking if Lustre is mounted.
2301 Details : Resolve intermediate symlinks when checking if a client has
2302 mounted a filesystem to avoid duplicate client mounts.
2307 Description: lconf can hit an error exception but still return success.
2308 Details : The lconf command catches the Command error exception at the top
2309 level script context and will exit with the associated exit
2310 status, but doesn't ensure that this exit status is non-zero.
2315 Description: failure of ptlrpc thread startup can cause oops
2316 Details : Starting a ptlrpc service thread can fail if there are a large
2317 number of threads or the server memory is very fragmented.
2318 Handle this without oopsing.
2321 Frequency : always, only if liblustre and non-default acceptor port was used
2323 Description: liblustre cannot connect to servers with non-default acceptor port
2324 Details : tcpnal_set_default_params() was not called and was therefore
2325 ignoring the environment varaible TCPNAL_PORT, as well as other
2326 TCPNAL_ environment variables
2331 Description: two objects could be created on the same OST for a single file
2332 Details : If an OST is down, in some cases it was possible to create two
2333 objects on a single OST for a single file. No problems other
2334 than potential performance impact and spurious error messages.
2338 Bugzilla : 5681, 9562
2339 Description: Client may oops in ll_unhash_aliases
2340 Details : Client dcache may become inconsistent in race condition.
2341 In some cases "getcwd" can fail if the current directory is
2347 Description: Inode refcounting problems in NFS export code
2348 Details : link_raw functions used to call d_instantiate without obtaining
2349 extra inode reference first.
2353 Bugzilla : 9942, 9903
2354 Description: Referencing freed requests leading to crash, memleaks with NFS.
2355 Details : We used to require that call to ll_revalidate_it was always
2356 followed by ll_lookup_it. Also with revalidate_special() it is
2357 possible to call ll_revalidate_it() twice for the same dentry
2358 even if first occurence returned success. This fix changes semantic
2359 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2360 reference on a request referred from the intent.
2361 ll_intent_release() then releases such a request.
2364 Frequency : rare, normally benchmark loads only
2366 Description: unlinked inodes were kept in memory on the client
2367 Details : If a client is repeatedly creating and unlinking files it
2368 can accumulate a lot of stale inodes in the inode slab cache.
2369 If there is no other client load running this can cause the
2370 client node to run out of memory. Instead flush old inodes
2371 from client cache that have the same inode number as a new inode.
2374 Frequency : SLES9 2.6.5 kernel and long filenames only
2375 Bugzilla : 9969, 10379
2376 Description: utime reports stale NFS file handle
2377 Details : SLES9 uses out-of-dentry names in some cases, which confused
2378 the lustre dentry revalidation. Change it to always use the
2382 Frequency : rare, unless heavy write-truncate concurrency is continuous
2383 Bugzilla : 4180, 6984, 7171, 9963, 9331
2384 Description: OST becomes very slow and/or deadlocked during object unlink
2385 Details : filter_destroy() was holding onto the parent directory lock
2386 while truncating+unlinking objects. For very large objects this
2387 may block other threads for a long time and slow overall OST
2388 responsiveness. It may also be possible to get a lock ordering
2389 deadlock in this case, or run out of journal credits because of
2390 the combined truncate+unlink. Solution is to do object truncate
2391 first in one transaction without parent lock, and then do the
2392 final unlink in a new transaction with the parent lock. This
2393 reduces the lock hold time dramatically.
2396 Frequency : rare, 2.4 kernels only
2398 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2399 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2400 before unmounting their filesystem in order to avoid deadlock.
2401 The kernel_locked() function in 2.4 kernels only checks whether
2402 the kernel lock is held, not whether it is this process that is
2403 holding it as 2.6 kernels do.
2408 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2409 Details : The client ptlrpc code may be trying to reconnect to a down
2410 server before a previous connection attempt has timed out.
2411 Increase the reconnect interval to be longer than the connection
2412 timeout interval to avoid sending duplicate connections to
2416 Frequency : echo_client brw_test command
2418 Description: fix echo_client to work with OST preallocated code
2419 Details : OST preallocation code (5137) didn't take echo_client IO path
2420 into account: echo_client calls filter methods outside of any
2421 OST thread and, hence, there is no per-thread preallocated
2422 pages and buffers to use. Solution: hijack pga pages for IO. As
2423 a byproduct, this avoids unnecessary data copying.
2427 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2428 Description: Client can oops in mdc_commit_close() after open replay
2429 Details : It was possible for the MDS to return an open request with no
2430 transaction number in mds_finish_transno() if the client was
2431 evicted, but without actually returning an error. Clients
2432 would later try to replay that open and may trip an assertion
2433 Simplify the client close codepath, and always return an error
2434 from the MDS in case the open is not successful.
2437 Frequency : rare, 2.6 OSTs only
2439 Description: OST may deadlock under high load on fragmented files
2440 Details : If there was a heavy load and highly-fragmented OST filesystems
2441 it was possible to have all the OST threads deadlock waiting on
2442 allocation of biovecs, because the biovecs were not released
2443 until the entire RPC IO was completed. Instead, release biovecs
2444 as soon as they are complete to ensure forward IO progress.
2446 Severity : enhancement
2448 Description: Support for specifying external journal device at mount
2449 Details : If an OST or MDS device is formatted with an external journal
2450 device, this device major/minor is stored in the ext3 superblock
2451 and may not be valid for failover. Allow detecting and
2452 specifying the external journal at mount time.
2457 Description: Mounting an MDS with pending unlinked files may cause oops
2458 Details : target_finish_recovery() calls mds_postrecov() which returned
2459 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2460 considers this an error and immediately begins cleaning up the
2461 lov, just after starting the mds_lov process
2463 Severity : enhancement
2465 Description: Implement 'lfs df' to report actual free space on per-OST basis
2466 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2467 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2468 usage of objects; '-h' to report in human readable format.
2470 ------------------------------------------------------------------------------
2472 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2479 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2480 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2481 at startup during orphan unlinking. Instead, skip buddy update
2482 before setup, as it will be regenerated after recovery is complete.
2487 Description: If an OST is inactive, its locks might reference stale inodes.
2488 Details : lov_change_cbdata() must iterate over all namespaces, even if
2489 they are inactive to clear inode references from the lock.
2491 Severity : enhancement
2492 Frequency : occasional, if non-standard max_dirty_mb used
2494 Description: Client will block write RPCs if not enough grant
2495 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2496 then the client will block writes while waiting for another RPC
2497 to complete instead of consuming its dirty limit. With change
2498 we get improved performance when max_dirty_mb is small.
2500 Severity : enhancement
2501 Bugzilla : 3389, 6253
2502 Description: Add support for supplementary groups on the MDS.
2503 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2504 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2505 lookups for user supplementary groups into a cache.
2509 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2510 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2511 for liblustre O_EXCL is enforced by MDS.
2515 Description: readdir never returns NULL in liblustre.
2516 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2517 dentry in struct dirent.
2521 Frequency : liblustre only, depends on application IO pattern
2522 Description: liblustre clients evicted if not contacting servers
2523 Details : Don't put liblustre clients into the ping_evictor list, so
2524 they will not be evicted by the pinger ever.
2526 Severity : enhancement
2528 Description: Add ability to evict clients by NID from MDS.
2529 Details : By echoing "nid:$NID" string into
2530 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2531 $NID would be instantly evicted from this MDS and from all active
2532 OSTs connected to it.
2536 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2537 Details : lookup_it_finish() used to query file size from OSTs that was not
2542 Description: service threads change working directory to that of init
2543 Details : Starting lustre service threads may pin the working directory
2544 of the parent thread, making that filesystem busy. Threads
2545 now change to the working directory of init to avoid this.
2549 Frequency : during shutdown only
2550 Description: shutdown with a failed MDS or OST can cause unmount to hang
2551 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2556 Frequency : 2.6 only
2557 Description: chmod/chown may include an extra supplementary group
2558 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2559 supplementary group and if none is specified this is used.
2562 Bugzilla : 5479 (6816)
2564 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2565 Details : If lookup is in progress on a file that is unlinked we might try
2566 to revalidate the inode and fail in revalidate after lookup is
2567 complete and ll_file_open() enqueues the open again but
2568 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2571 Frequency : always, if lconf --abort_recovery used
2573 Description: lconf --abort_recovery fails with 'Operation not supported'
2574 Details : lconf was attempting to abort recovery on the MDT device and not
2577 ------------------------------------------------------------------------------
2579 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2584 Frequency : rare (only unsupported configurations with a node running as an
2586 Bugzilla : 6514, 5137
2587 Description: Mounting a Lustre file system on a node running as an OST could
2589 Details : OSTs now preallocates memory needed to write out data at
2590 startup, instead of when needed, to avoid having to
2591 allocate memory in possibly low memory situations.
2592 Specifically, if the file system is mounted on on OST,
2593 memory pressure could force it to try to write out data,
2594 which it needed to allocate memory to do. Due to the low
2595 memory, it would be unable to do so and the node would
2596 become unresponsive.
2598 Severity : enhancement
2600 Description: Addition of lconf --service command line option
2601 Details : lconf now accepts a '--service <arg>' option, which is
2602 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2604 Severity : enhancement
2606 Description: Failover mode is now the default for OSTs.
2607 Details : By default, OSTs will now run in failover mode. To return to
2608 the old behaviour, add '--failout' to the lmc line for OSTs.
2610 Severity : enhancement
2612 Description: Health checks are now provided for MDS and OSTs
2613 Details : Additional detailed health check information on MSD and OSTs
2614 is now provided through the procfs health_check value.
2617 Frequency : occasional, depends on IO load
2619 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2620 after numerous create/delete cycles
2621 Details : The ext3 inode allocation policy would not allocate new inodes
2622 very well on the OSTs because there are no new directories
2623 being created. Instead we look for groups with free space if
2624 the parent directories are nearly full.
2629 Description: Network or server problems during mount may cause partially
2630 mounted clients instead of returning an error.
2631 Details : The config llog parsing code may overwrite the error return
2632 code during mount error handling, returning success instead
2638 Description: MDS can fail to allocate large reply buffers
2639 Details : After long uptimes the MDS can fail to allocate large reply
2640 buffers (e.g. zconf client mount config records) due to memory
2641 fragmentation or consumption by the buffer cache. Preallocate
2642 some large reply buffers so that these replies can be sent even
2643 under memory pressure.
2647 Frequency : rare (liblustre)
2648 Description: fsx running with liblustre complained that using truncate() to
2649 extend the file doesn't work. This patch corrects that issue.
2650 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2651 ATTR_SIZE and lsm use in llu_setattr_raw.
2655 Frequency : rare, only 2.6 kernels
2656 Description: Unusual file access patterns on the MDS may result in inode
2657 data being lost in very rare circumstances.
2658 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2659 code in the 2.6 kernel caused the inode and/or EA data not to
2660 be read from disk, causing single-file corruption.
2664 Frequency : rare, only 2.6 filesystems using extents
2665 Description: Heavy concurrent write and delete load may cause data corruption.
2666 Details : It was possible under high-load situations to have an extent
2667 metadata block in the block device cache from a just-unlinked
2668 file overwrite a newly-allocated data block. We now unmap any
2669 metadata buffers that alias just-allocated data blocks.
2673 Frequency : filesystems with default stripe_count larger than 77
2674 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2675 Details : lconf specifies an inode size of 4096 bytes when the default
2676 stripe_count is larger than 77. This conflicts with the default
2677 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2678 this case to avoid pinning too much memory for large EAs.
2680 ------------------------------------------------------------------------------
2682 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2687 Frequency : rare (extremely heavy IO load with hundreds of clients)
2689 Description: Client is evicted, gets IO error writing to file
2690 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2691 caused clients to be evicted for AST timeouts. The fixes in
2692 bug 5192 mean we no longer need to have such short AST timeouts
2693 so ldlm_timeout has been increased.
2696 Frequency : occasional during --force or --failover shutdown under load
2697 Bugzilla : 5949, 4834
2698 Description: Server oops/LBUG if stopped with --force or --failover under load
2699 Details : a collection of import/export refcount and cleanup ordering
2700 issues fixed for safer force cleanup
2703 Frequency : only filesystems larger than 120 OSTs
2704 Bugzilla : 5990, 6223
2705 Description: lfs getstripe would oops on a very large filesystem
2706 Details : lov_getconfig used kfree on vmalloc'd memory
2709 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2710 Bugzilla : 6242, 6243
2711 Description: reading from files that had been truncated to a non-zero size
2712 but never opened returned no data
2713 Details : ll_file_read() reads zeros from no-object files to EOF
2718 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2719 Details : There's little harm in aborting MDS/OSS recovery and letting it
2720 try again, so I removed the LASSERT and return an error instead.
2722 Severity : enhancement
2724 Description: New debugging infrastructure for tracking down data corruption
2725 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2726 (b) cover more of the client-side code path, and (c) try to narrow
2727 down where problems occurred
2731 Bugzilla : 3819, 4364, 4397, 6313
2732 Description: Racing close and eviction MDS could cause assertion in mds_close
2733 Details : It was possible to get multiple mfd references during close and
2734 client eviction, leading to one thread referencing a freed mfd.
2736 Severity: : enhancement
2737 Bugzilla : 3262, 6359
2738 Description: Attempts to reconnect to servers are now more aggressive.
2739 Details : This builds on the enhanced upcall-less recovery that was added
2740 in 1.4.2. When trying to reconnect to servers, clients will
2741 now try each server in the failover group every 10 seconds. By
2742 default, clients would previously try one server every 25 seconds.
2747 Description: After recovery, certain operations trigger a failed
2748 assertion on a client.
2749 Details : Failing over an mds, using lconf -d --failover, while a
2750 client was doing a readdir() call would cause the client to
2751 LBUG after recovery completed and the readdir() was resent.
2753 Severity : enhancement
2755 Description: Default groups are now added by lconf
2756 Details : You can now run lconf --group <servicename> without having to
2757 manually add groups with lmc.
2760 Frequency : occasional
2762 Description: Nodes with an elan id of 0 trigger a failed assertion
2765 Frequency : always when accessing e.g. tty/console device nodes
2767 Description: tty and some other devices nodes cannot be used on lustre
2768 Details : file's private_data field is used by device data and lustre
2769 values in there got lost. New field was added to struct file to
2770 store fs-specific private data.
2773 Frequency : when exporting Lustre via NFS
2775 Description: NFSD failed occasionally when looking up a path component
2776 Details : NFSD is looking up ".." which was broken in ext3 directories
2777 that had grown large enough to become hashed.
2780 Frequency : Clusters with multiple interfaces not on the same subnet
2782 Description: Nodes will repeatedly try to reconnect to an interface which it
2783 cannot reach and report an error to the log.
2784 Details : Extra peer list entries will be created by lconf with some peers
2785 unreachable. lconf now validates the peer before adding it.
2788 Frequency : Only if a default stripe is set on the filesystem root.
2790 Description: Setting a default stripe on the filesystem root prevented the
2791 filesystem from being remounted.
2792 Details : The client was sending extra request flags in the root getattr
2793 request and did not allocate a reply buffer for the dir EA.
2796 Frequency : occasional, higher if lots of files are accessed by one client
2797 Bugzilla : 6159, 6097
2798 Description: Client trips assertion regarding lsm mismatch/magic
2799 Details : While revalidating inodes the VFS looks up inodes with ifind()
2800 and in rare cases can find an inode that is being freed.
2801 The ll_test_inode() code will free the lsm during ifind()
2802 when it finds an existing inode and then the VFS later attaches
2803 this free lsm to a new inode.
2807 Bugzilla : 6422, 7030
2808 Description: MDS deadlock between mkdir and client eviction
2809 Details : Creating a new file via mkdir or mknod (starting a transaction
2810 and getting the ns lock) can deadlock with client eviction
2811 (gets ns lock and trying to finish a synchronous transaction).
2814 Frequency : occasional
2815 Description: While starting a server, the fsfilt_ext3 module could not be
2817 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2818 kernels. Previously, lconf would still use the ext3 name
2819 when trying to load modules. Now, it will correctly use
2820 ext3 on 2.4 and ldiskfs on 2.6.
2822 Severity : enhancement
2823 Description: The default stripe count has been changed to 1
2824 Details : The interpretation of the default stripe count (0, to lfs
2825 or lmc) has been changed to mean striping across a single
2826 OST, rather than all available. For general usage we have
2827 found a stripe count of 1 or 2 works best.
2829 Severity : enhancement
2830 Description: Add support for compiling against Cray portals.
2831 Details : Conditional compiling for some areas that are different
2835 Frequency : occasional
2836 Bugzilla : 6409, 6834
2837 Description: Creating files with an explicit stripe count may lead to
2838 a failed assertion on the MDS
2839 Details : If some OSTs are full or unavailable, creating files may
2840 trigger a failed assertion on the MDS. Now, Lustre will
2841 try to use other servers or return an error to the
2845 Frequency : occasional
2847 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2848 caused lock timeout during readahead (since 1.4.2).
2849 Details : Processes doing readahead might match a lock that hasn't been
2850 granted yet if there are overlapping and conflicting lock
2851 requests. The readahead process waits on ungranted lock
2852 (original lock is CBPENDING), while OST waits for that process
2853 to cancel CBPENDING read lock and eventually evicts client.
2855 Severity : enhancement
2857 Description: Initial enabling of flock support for clients
2858 Details : Implements fcntl advisory locking and file status functions.
2859 This feature is provided as an optional mount flag (default
2860 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2861 locking are implemented yet, and those that are are not guaranteed
2862 to be completely correct in production environments.
2863 mount -t lustre -o [flock|noflock] ...
2866 Frequency : occasional
2868 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2869 assertion in the ext3 JBD (journaling) layer.
2870 Details : The b_committed_data struct is protected by the big kernel lock
2871 in 2.4 kernels, serializing journal_commit_transaction() and
2872 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2873 access to this struct. In 2.6 kernels there is finer grained
2874 locking to improve SMP performance of the JBD layer.
2878 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2883 Description: merge and cleanup kernel patches.
2886 -----------------------------------------------------------------------------
2888 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2890 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2891 versions of Lustre. Please update all servers and clients to
2892 version 1.4.2 or later at the same time. You must also run
2893 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2894 to update the configuration logs.
2896 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2897 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2898 - swab llog records of type '0' so we get proper header size/idx (5861)
2899 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2900 - fix rename of one directory over another leaking an inode (5953)
2901 - avoid SetPageDirty on 2.6 (5981)
2902 - don't re-add just-being-destroyed locks to the waiting list (5653)
2903 - when creating new directories, inherit the parent's custom
2904 striping settings if present parent (3048)
2905 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2906 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2907 - don't hold client locks on temporary worklist from l_lru (5666)
2908 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2909 - automatically evict dead clients (5921)
2910 - Update file size properly in create+truncate+fstat case (6196)
2911 - Do not unhash mountpoint dentries, do not allow removal of
2913 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2914 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2915 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2916 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2917 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2918 - Correctly handle reads of files with no objects (6243)
2919 - lctl recover will also mark a device active if deactivate used (5933)
2921 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2922 - allow --write-conf on an MDS with different nettype than client (5619)
2923 - don't write config llogs to MDS for mounts not from that MDS (5617)
2924 - lconf should create multiple TCP connections from a client (5201)
2925 - init scripts are now turned off by default; run chkconfig --on
2926 lustre and chkconfig --on lustrefs to use them
2927 - upcalls are no longer needed for clients to recover to failover
2929 - add --abort-recovery option to lconf to abort recovery on device
2931 - add support for an arbitrary number of OSTs (3026)
2932 - Quota support protocol changes.
2933 - forward compatibility changes to wire structs (6007)
2934 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2935 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2936 - improved "lustre status" script
2937 - initialize blocksize for non-regular files (6062)
2938 - added --disable-server and --disable-client configure options (5782)
2939 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2940 - Vanilla 2.4.29 support
2941 - increase maximum number of obd devices to 520 (6242)
2942 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2943 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2944 SLES 9 2.6.5 kernels
2945 - limit stripes per file to 160 (the maximum EA size) (6093)
2947 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2950 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2951 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2952 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2953 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2954 - small ext3 extents cleanups and fixes (5733)
2955 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2956 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2957 - avoid needless client->OST connect, fix handle mismatch (5317)
2958 - fix DLM error path that led to out-of-sync client, long delays (5779)
2959 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2960 - fix several locking issues related to i_size (5492,5624,5654,5672)
2961 - don't move pending lock onto export if it is already evicted (5683)
2962 - fix kernel oops when creating .foo in unlinked directory (5548)
2963 - fix deadlock in obdfilter statistics vs. object create (5811)
2964 - use time_{before,after} to avoid timer jiffies wrap (5882)
2965 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2966 - Do not leak request if server was not able to process it (5154)
2967 - If mds_open unable to find parent dir, make that negative lookup(5154)
2968 - don't create new directories with extent-mapping (5909, 5936)
2970 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2971 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2972 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2973 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2974 - support for RHEL4 kernel on i686 (5773)
2975 - provide error messages when incompatible logs are encountered (5898)
2977 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2978 * version 1.4.0.10 (1.4.1 release candidate 1)
2980 - don't keep a lock reference when lock is not granted (4238)
2981 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2982 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2983 - drop import inflight refcount on signal_completed_replay error (5255)
2984 - unlock page after async write error during send (3677)
2985 - handle missing objects in filter_preprw_read properly (5265)
2986 - no transno return for symlink open, don't save no-trasno open (3440)
2987 - don't try to complete elan receive that already failed (4012)
2988 - free RPC server reply state on error (5406)
2989 - clean up thread from ptlrpc_start_thread() on error (5160)
2990 - readahead could read extra page into cache that wasn't ejected (5388)
2991 - prevent races in class_attach/setup/cleanup/detach (5260)
2992 - don't dereference de->d_inode after l_dput of de (5458)
2993 - use "int" for stripe value returned from lock_to_stripe (5544)
2994 - mballoc allocation and error-checking fixes in 2.6 (5504)
2995 - block device patches to fix I/O request sizes in 2.6 (5482)
2996 - look up hostnames for IB nals (5602)
2997 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2998 - don't start multiple acceptors for the same port (5277)
2999 - fix incorrect LASSERT in mds_getattr_name (5635)
3000 - export a proc file for general "ping" checking (5628)
3001 - fix "lfs check" to not block when the MDS is down (5628)
3003 - service request history (4965)
3004 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3005 - create an "evict_client" /proc entry on OSTs, like the MDS has
3006 - fix mount usage message, return errors per mount(8) (5168)
3007 - change grep [] to grep "[]" in tests so they work in more UMLs
3008 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3009 - remove ancient LOV_MAGIC_V0 EA support (5047)
3010 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3011 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3012 - allow readahead allocations to fail when low on memory (5383)
3013 - mmap locking landed again, after considerable improvement (2828)
3014 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3016 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3019 - send OST transaction number in read/write reply to free req (4966)
3020 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3021 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3022 - the watchdog thread now runs as interruptible (5246)
3023 - flock/lockf fixes (but it's still disabled, pending 5135)
3024 - don't use EXT3 constants in llite code (5094)
3025 - memory shortage at startup could cause assertion (5176)
3027 - reorganization of lov code
3028 - single portals codebase
3030 - add extents/mballoc support (5025)
3031 - direct I/O reads in the obdfilter (4048)
3032 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3034 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3037 - send OST transaction number in read/write reply to free req (4966)
3038 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3039 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3040 - the watchdog thread now runs as interruptible (5246)
3041 - handle missing objects in filter_preprw_read properly (5265)
3042 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3043 - drop import inflight refcount on signal_completed_replay error (5255)
3044 - unlock page after async write error during send (3677)
3045 - return original error code on reconstructed replies (3761)
3046 - no transno return for symlink open, don't save no-trasno open (3440)
3048 - add pid to ldlm debugging output (4922)
3049 - bump the watchdog timeouts -- we can't handle 30sec yet
3050 - extra debugging for orphan dentry/inode bug (5259)
3052 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3055 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3056 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3057 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3058 - fix NULL dereference in /proc/sys/portals/routes (4827)
3059 - allow failed mdc_close() operations to be interrupted (4561)
3060 - stop precreate on OST before MDS would time out on it (4778)
3061 - don't send partial-page writes before EOF from client (4410)
3062 - discard client grant for sub-page writes on large-page clients (4520)
3063 - don't free dentries not owned by NFS code, check generation (4806)
3064 - fix lsm leak if mds_create_objects() fails (4801)
3065 - limit debug_daemon file size, always print CERROR messages (4789)
3066 - use transno after validating reply (3892)
3067 - process timed out requests if import state changes (3754)
3068 - update mtime on OST during writes, return in glimpse (4829)
3069 - add mkfsoptions to LDAP (4679)
3070 - use ->max_readahead method instead of zapping global ra (5039)
3071 - don't interrupt __l_wait_event() during strace
3073 - add software watchdogs to catch hung threads quickly (4941)
3074 - make lustrefs init script start after nfs is mounted
3075 - fix CWARN/ERROR duplication (4930)
3076 - return async write errors to application if possible (2248)
3077 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3078 - print NAL number in %x format (4645)
3079 - update barely-supported suse-2.4.21-171 series (4842)
3080 - support for sles 9 %post scripts
3081 - support for building 2.6 kernel-source packages
3082 - support for sles km_* packages
3084 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3087 - ignore -ENOENT errors in osc_destroy (3639)
3088 - notify osc create thread that OSC is being cleaned up (4600)
3089 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3090 - reconstruct ost_handle() like mds_handle() (4657)
3091 - create a new thread to do import eviction to avoid deadlock (3969)
3092 - let lconf resolve symlinked-to devices (4629)
3093 - don't unlink "objects" from directory with default EA (4554)
3094 - hold socknal file ref over connect in case target is down (4394)
3095 - allow more than 32000 subdirectories in a single directory (3244)
3096 - fix blocks count for O_DIRECT writes (3751)
3097 - OST returns ENOSPC from object create when no space left (4539)
3098 - don't send truncate RPC if file size isn't changing (4410)
3099 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3100 - bind to privileged port in socknal and tcpnal (3689)
3102 - rate limit CERROR/CWARN console message to avoid overload (4519)
3103 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3104 - basic mmap support (3918)
3105 - kernel patch series update from b1_4 (4711)
3107 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3110 - avoid crash during MDS cleanup with OST shut down (2775)
3111 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3112 - don't use bad inodes on the MDS (3744)
3113 - dynamic object preallocation to improve recovery speed (4236)
3114 - don't hold spinlock over lock dumping or change debug flags (4401)
3115 - don't zero obd_dev when it is force cleaned (3651)
3116 - print grants to console if they go negative (4431)
3117 - "lctl deactivate" will stop automatic recovery attempts (3406)
3118 - look for existing locks in ldlm_handle_enqueue() (3764)
3119 - don't resolve lock handle twice in recovery avoiding race (4401)
3120 - revalidate should check working dir is a directory (4134)
3122 - don't always mark "slow" obdfilter messages as errors (4418)
3124 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3127 - don't close LustreDB during write_conf until it is done (3860)
3128 - fix typo in lconf for_each_profile (3821)
3129 - allow dumping logs from multiple threads at one time (3820)
3130 - don't allow multiple threads in OSC recovery (3812)
3131 - fix debug_size parameters (3864)
3132 - fix mds_postrecov to initialize import for llog ctxt (3121)
3133 - replace config semaphore with spinlock (3306)
3134 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3135 - don't allow enqueue to complete on a destroyed export (3822)
3136 - down write_lock before checking llog header bitmap (3825)
3137 - recover from lock replay timeout (3764)
3138 - up llog sem before sending rpc (3652)
3139 - reduce ns lock hold times when setting kms (3267)
3140 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3141 - fix NULL deref and obd_dev leak on setup error (3312)
3142 - replace some LBUG about llog ops with error handling (3841)
3143 - don't match INVALID dentries from d_lookup and spin (3784)
3144 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3145 - fix invalid assertion in ptlrpc_set_wait (3880)
3147 - add libwrap support for the TCP acceptor (3996)
3148 - add /proc/sys/portals/routes for non-root route listing (3994)
3149 - allow setting MDS UUID in .xml (2580)
3150 - print the stack of a process that LBUGs (4228)
3152 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3155 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3156 - make sure to unset replay flag from failed open requests (3440)
3157 - if default stripe count is 0, use OST count for inode size (3636)
3158 - update parent mtime/ctime on client for create/unlink (2611)
3159 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3160 - free recovery state on server during a forced cleanup (3571)
3161 - unregister_reply for resent reqs (3063)
3162 - loop back devices mounting and status check on 2.6 (3563)
3163 - fix resource-creation race that can provoke i_size == 0 (3513)
3164 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3165 - more debugging for page-accounting assertion (3746)
3166 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3167 - avoid deadlock after precreation failure (3758)
3168 - fix race and lock order deadlock in orphan handling (3450, 3750)
3169 - add validity checks when grabbing inodes from l_ast_data (3599)
3171 - add /proc/.../recovery_status to obdfilter (3428)
3172 - lightweight CDEBUG infrastructure, debug daemon (3668)
3173 - change default OSC RPC parameters to be better on small clusters
3174 - turn off OST read cache for files smaller than 32MB
3175 - install man pages and include them in rpms (3100)
3176 - add new init script for (un)mounting lustre filesystems (2593)
3177 - run chkconfig in %post for init scripts (3701)
3178 - drop scimac NAL (unmaintained)
3180 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3183 - clean kiobufs before and after use (3485)
3184 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3185 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3186 - change init script to fail more clearly if not run as root (1528)
3187 - allow clients to reconnect during replay (1742)
3188 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3189 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3190 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3191 - fix cond_resched (3554)
3192 - only evict unfinished clients after recovery (3515)
3193 - allow bulk resend, prevent data loss (3570)
3194 - dynamic ptlrpc request buffer allocation (2102)
3195 - don't allow unlinking open directory if it isn't empty (2904)
3196 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3197 - remove extraneous obd dereference causing LASSERT failure (3334)
3198 - don't use get_cycles() when creating temp. files on the mds (3156)
3199 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3200 - handle EEXIST for set-stripe, set proper directory name (3336)
3202 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3203 - fix ksocknal_fmb_callback() error messages (2918)
3205 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3208 - don't copy lvb into (possibly NULL) reply on error (2983)
3209 - don't deref dentry after dput, don't free lvb on error (2922)
3210 - use the kms to determine writeback rpc length (2947)
3211 - increment oti_logcookies when osc is inactive (2948)
3212 - update client's i_blocks count via lvb messages (2543)
3213 - handle intent open/close of special files properly (1557)
3214 - mount MDS with errors=remount-ro, like obdfilter (2009)
3215 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3216 - don't use cancelling-locks' kms values (2947)
3217 - use highest lock extent for kms, not last one (2925)
3218 - don't dereference ERR_PTR() dentry in error handling path (3107)
3219 - fix thread race in portals_debug_dumplog() (3122)
3220 - create lprocfs device entries at setup instead of at attach (1519)
3221 - common AST error handler, don't evict client on completion race (3145)
3222 - zero nameidata in detach_mnt in 2.6 (3118)
3223 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3224 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3225 - initialize RPC timeout timer earlier for 2.6 (3219)
3226 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3227 - print nal/nid for unknown nid (3258)
3228 - additional checks for oscc recovery before doing precreate (3284)
3229 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3230 - don't crash in mdc_close for bad permissions on open (3285)
3231 - zero i_rdev for non-device files (3147)
3232 - clear page->private before handing to FS, better assertion (3119)
3233 - tune the read pipeline (3236)
3234 - fix incorrect decref of invalidated dentry (2350)
3235 - provide read-ahead stats and refine rpc in flight stats (3328)
3236 - don't hold journal transaction open across create RPC (3313)
3237 - update atime on MDS at close time (3265)
3238 - close LDAP connection when recovering to avoid server load (3315)
3239 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3240 - don't leak open file on MDS after open resend (3325)
3241 - serialize filter_precreate and filter_destroy_precreated (3329)
3242 - loop device shouldn't call sync_dev() for nul device (3092)
3243 - clear page cache after eviction (2766)
3244 - resynchronize MDS->OST in background (2824)
3245 - refuse to mount the same filesystem twice on same mountpoint (3394)
3246 - allow llmount to create routes for mounting behind routers (3320)
3247 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3248 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3249 - fix rare problem with rename on htree directories (3417)
3251 - allow default OST striping configuration per directory (1414)
3252 - fix compilation for qswnal for 2.6 kernels (3125)
3253 - increase maximum number of MDS request buffers for large systems
3254 - change liblustreapi to be useful for external progs like lfsck (3098)
3255 - increase local configuration timeout for slow disks (3353)
3256 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3258 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3261 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3262 - don't overwrite extent policy data in reply if lock was blocked (2901)
3263 - drop filter export grants atomically with removal from device (2663)
3264 - del obd_self_export from work_list in class_disconnect_exports (2908)
3265 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3266 - swab reply message in mdc_close, other PPC fixes (2464)
3267 - fix destroying of named logs (2325)
3268 - overwrite old logs when running lconf --write_conf (2264)
3269 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3270 - fix race in target_handle_connect (2898)
3271 - mds_reint_create() should take same inode create lock (2926)
3272 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3273 - don't close files for self_export to avoid uninitialized obd (2936)
3274 - allow MDS with the same name as client node (2939)
3275 - hold dentry reference for closed log files for unlink (2325)
3276 - reserve space for all logs during transactions (2059)
3277 - don't evict page beyond end of stripe extent (2925)
3278 - don't oops on a deleted current working directory (2399)
3279 - handle hard links to targets without a parent properly (2517)
3280 - don't dereference NULL lock when racing during eviction (2867)
3281 - don't grow lock extents when lots of conflicting locks (2919)
3283 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3286 - account for cache space usage on clients to avoid data loss (974)
3287 - lfsck support in lustre kernel code (2349)
3288 - reduce journal credits needed for BRW writes (2370)
3289 - orphan handling to avoid losing space on client/server crashes
3290 - ptlrpcd can be blocked, stopping ALL progress (2477)
3291 - use lock value blocks to assist in proper KMS, faster stat (1021)
3292 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3293 - recovery for initial connections (2355)
3294 - fixes for mds_cleanup_orphans (1934)
3295 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3296 - block all file creations until orphan recovery completes (1901)
3297 - client remove rq_connection from request struct (2423)
3298 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3299 - recovery timer race (2670)
3300 - mdc_close recovey bug (2532)
3301 - ptlrpc cleanup bug (2710)
3302 - mds timeout on local locks (2588)
3303 - namespace lock held during RPCs (2431)
3304 - handle interrupted sync write properly (2503)
3305 - don't try to handle a message that hasn't been replied to (2699)
3306 - client assert failure during cleanup after abort recovery (2701)
3307 - leak mdc device after failed mount (2712)
3308 - ptlrpc_check_set allows timedout requests to complete (2714)
3309 - wait for inflight reqs when ptlrpcd finishes (2710)
3310 - make sure unregistered services are removed from the srv_list
3311 - reset bulk XID's when resending them (caught by 1138 test)
3312 - unregister_bulk after timeout
3313 - fix lconf error (2694)
3314 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3315 - readahead locks pages, leaves pending causing memory pressure (2673)
3316 - increase OST request buffers to 4096 on large machines (2729)
3317 - fix up permission of existing directories in simple_mkdir (2661)
3318 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3319 - don't assign transno to errored transactions (2742)
3320 - don't delete objects on OST if given a bogus objid from MDS (2751)
3321 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3322 - if rq_no_resend, then timeout request after recovery (2432)
3323 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3324 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3325 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3326 - MDS recovery shouldn't delete objects that it has given out (2730)
3327 - if enqueue arrives after completion, don't clobber LVB (2819)
3328 - don't unlock pages twice when trigger_group_io returns error (2814)
3329 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3330 - don't write pages to disk if there was an error (1450)
3331 - don't ping imports that have recovery disabled (2676)
3332 - take buffered bytes into account when balancing socknal conn (2817)
3333 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3334 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3335 - remove little-endian swabbing of llog records (1987)
3336 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3337 - retry reposting request buffers if they fail (1191)
3338 - grow extent at grant time to avoid granting a revoked lock (2809)
3339 - lock revoke doesn't evict page if covered by a second lock (2765)
3340 - disable VM readahead to avoid reading outside lock extents (2805)
3342 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3343 - updated kernel patches for hp-2.4.20 kernel (2681)
3345 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3348 - fix truncated write corruption (2366)
3349 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3351 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3352 - ensure len(uuid) < 37 in lmc (1171)
3353 - fix ia64 OOPS in llog_test (2255)
3354 - zero end of page at obdfilter for partial page writes (2648)
3355 - don't leave stale dentries around after renames (bug 2428)
3356 - fix timeouts when evicting a client with a single lock held (2642)
3357 - set deadline for the initial HELLO message to drain (2634)
3358 - print out dotted-quad IP addresses in the socknal (2302)
3360 - additional debugging for MDS client eviction problem (2443)
3361 - fix mkfsoptions support for osts (2603, 2604)
3363 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3366 - add series for the vanilla 2.6.0 kernel
3367 - add series for the vanilla 2.4.24 kernel
3368 - add series for a cray x86/64 UL kernel drop
3369 - fix xattr patches for the vanilla 2.4.19 series
3371 - generate true UUIDs in lmc (1171)
3372 - have portals stack dumping break in UML (2466)
3373 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3374 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3375 - wake callback waiting threads on client eviction (2460)
3376 - Add --ptldebug and --subsystem to lmc (1719)
3377 - update assertion to allow safe interrupt allocation
3378 - set rq_no_resend for cancel requests (2432)
3379 - recalculate ptlrpcd timeout after resend (2494)
3380 - call vfs_rmdir when removing pending directories (2368)
3381 - fix renaming a file to itself (2429)
3382 - lmc creates a default one-stripe lov (2454)
3383 - expand procfs space to handle large clusters (2326)
3384 - increase UML stack to avoid overflow
3385 - update lconf's list of debug and subsystem masks
3386 - fix lfs find --obd (2510)
3387 - /proc tunable for disabling filter read caching (2591)
3388 - stop rpm packages from altering slapd.conf (2301)
3389 - disable nagle in the socknal under 0conf (2578)
3390 - choose mds inode size based on stripe count (2572)
3391 - fix kernel-source rpm problems (2516)
3393 - add --disable-doc to avoid pdf generation (2421)
3394 - update documentation, tests, type-os, comments
3395 - avoid format warnings on ia64
3396 - remove the TOE NAL
3397 - tiny code cleanups by removing unused fields
3399 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3402 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3403 - remove the most blatant lies from BUILDING file (2371)
3404 - change default debug level to reasonable production setting
3405 - reduce client side cache size to reduce cache flush time
3406 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3407 - make TCP zerocopy and pinger support enabled by default (2476)
3408 - sync writes completed after process exits caused crashes (2319)
3409 - maintain correct mount count on the MDS (2356)
3410 - backout 1557, because 2316 wasn't really fixed
3411 - better file I/O statistics gathering in /proc
3412 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3413 - another kernel patch to fix zero-copy TCP function export
3414 - don't take duplicate lock when processing re-sent getattr (2420)
3415 - lctl uses obd_self_export instead of creating new conn (2353)
3416 - MDS/OST recovery case which requires object creation asserted (2425)
3417 - move lfs from /usr/sbin to /usr/bin in packages
3418 - fix race between mds_client_add and mds_client_free (2417)
3419 - use kmalloc instead of slabs in portals (2430)
3420 - don't create duplicate records when a failover MDS is present (2442)
3421 - remove unnecessary mount age check (2332)
3422 - don't remove directory inodes from locks prematurely (2451)
3423 - don't break if MDS service name is the same as hostname (2103)
3424 - fix races in client write RPC generation when cache full (2482)
3426 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3429 - remove now-unused request->rq_obd (278)
3430 - if an allocation fails, print out how much memory we've used (1933)
3431 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3432 - add the "configurable stack size" patch to most series files (1256)
3433 - ability to write large log records, for 100+ OST configs (2306)
3434 - fix NULL deref when filter_prep fails (2314)
3435 - fix operator precedence error in filter_sync
3436 - dynamic allocation of socknal TX descriptors (2315)
3437 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3438 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3439 - ensure that CWARN messages in Portals always get to the syslog
3440 - __init/__exit are not for prototype decls (ldlm_init/exit)
3441 - x86-64 compile warning fixes
3442 - fix gateway LMC keyword conflict (2318)
3443 - fix MDS lock inversions in getattr/reint paths (1844)
3444 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3445 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3446 - DLM race condition prevented some lock evictions (2328)
3447 - ENOMEM detection and retry on socknal sends (2230)
3448 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3449 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3450 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3451 - better error messages when a client is rejected during recovery (1505)
3452 - avoid cancelling locks which were never granted, after failure (2330)
3453 - fix i_sem/journal inversion in mds_client_add (2333)
3454 - fix truncate/getattr lock cycle deadlock (2334)
3455 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3456 - fix two rare exit paths which could leak an l_lock() ref (2321)
3457 - fencepost error in MDS/OST orphan recovery (2226)
3458 - make log record alignment 8 bytes (1988)
3459 - lstripe now fails when requested offset > ost_count (2237)
3460 - ensure that all kernel series have a complete list.h (1607)
3461 - fix crashes in special-file operations (2316)
3462 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3464 - allow configurable automake binary, for testing new versions
3465 - small update to the lfs documentation
3467 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3469 * fix negative export reference count in fsfilt_sync (2312)
3471 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3472 * release candidate 0.9.1
3474 - orphans are moved into the PENDING directory for possible recovery
3475 - replayed opens now open by fid for orphan/rename safety (1042)
3476 - last close of an orphan inode generates a transno (683)
3477 - chdir() and mount() now pin the directory entry (1020)
3478 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3479 - discard very old requests without processing them (1502)
3480 - remove some common, well-understood CERRORs (1505)
3481 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3482 - clear "grant" flags in OST replies until OST grant code lands (1644)
3483 - fix read performance by not clobbering i_blksize on client (1598)
3484 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3485 - make LRU size a /proc tunable, clears locks when reduced (707)
3486 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3487 - server locks take references on exports now (1558)
3488 - build fixes for 2.4.20-rh trees (1663)
3489 - return an error from lov_create if all OSCs are inactive (1751)
3490 - fix import levels when a reconnect happens without a timeout (1597)
3491 - exit early from mds_open if we get a lookup error (1749)
3492 - partial page read at EOF wouldn't wait for disk before sending (1642)
3493 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3494 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3495 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3496 - write/truncate lock inversion (1639)
3497 - Don't auto-load obdclass, portals modules during cleanup (1495)
3498 - fix timestamps from jumping to "now" (1763)
3499 - extra journal assertions (1648)
3500 - add an extra multiunlink test (1771)
3501 - fix read_record/write_record API (1776)
3502 - fix leak of offset_extent, possible incorrect i_size later (1772)
3503 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3504 - Debugging for the kqswnal_get_idle_tx problems (1820)
3505 - Allow recovery to be attempted multiple times (1536)
3506 - Write out MDS last_rcvd file after it is first created (1600)
3507 - Fix tx_descriptor leak in failed transmit situations (1827)
3508 - ext3 journaling fixes for assertion failure after IO error (1871)
3509 - class_export_put() on freed export after completion AST error (1896)
3510 - Fix revalidate looping in VFS (1322)
3511 - Don't access a freed export during MDS_REINT timeout (1521)
3512 - Add open-unlink recovery support on the MDS (1673,1764)
3513 - Return an error if no MDS data was read from last_rcvd (1946)
3514 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3515 - Don't setup a disk device that doesn't match exported UUID (317)
3516 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3517 - avoid committing NULL handle in force close
3518 - local.sh is now a one-stripe LOV configuration
3519 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3520 - fix ext3 htree duplicate directory entry corruption (1516)
3521 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3522 - update to new LOV EA format (2097)
3523 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3525 2003-06-15 Phil Schwan <phil@clusterfs.com>
3528 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3529 - per-import/export recovery handling (958, 931, 959)
3530 - multiple last-rcvd slots, for serving multiple FSes (949)
3531 - connections are again shared between multiple imp/exports (963, 964)
3532 - "umount -f" would hang if any requests needed to be sent (393, 978)
3533 - avoid pinning large req buffer by copying for queued messages (989)
3534 - add "uuid" to "lctl device" command to help upcalls (991)
3535 - "open" RPCs with transnos would confuse recovery counters (1037)
3536 - do proper endian conversion of all wire messages (288, 340, 891)
3537 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3538 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3539 - fix ptlrpc_connection leak in target_handle_connect (1174)
3540 - fix import refcounting bug in OST and MDS cleanup (1134)
3541 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3542 - fix very unlikely obd_types race condition (501)
3543 - remove osc_open hack for echo_client (1187)
3544 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3545 - a failure in read_inode2 leads to deadlock (1139)
3546 - cancel ack-locks as soon as transaction is committed (1072)
3547 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3548 - make sure to commitrw after any preprw to avoid deadlock (1162)
3549 - failing to execute a file in a lustre FS would lock inode (1203)
3550 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3551 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3552 - obd_uuid2tgt would walk off the end of the list (1255)
3553 - on IA64 the osc would give portals incorrect bulk size (1258)
3554 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3555 - fix lock inversion caused by new llite matching code (1282)
3556 - limit the number of dirty pages on a client to 10MB (1286)
3557 - timed out locks were not being corrected cancelled (1289)
3558 - fix O_DIRECT above 4GB on IA-32 (1292)
3559 * major user-visible changes
3560 - fail out/fail over policy now controlled by the upcall (993)
3562 - add OBD_PING to check server availability and failure (954)
3563 - lustre messages are now sent in sending host order (288, 340, 891)
3564 - add eadatalen to MDS getattr reply (340)
3565 - OST read replies may contain second buffer, with per-page status (593)
3567 2003-03-11 Phil Schwan <phil@clusterfs.com>
3570 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3571 - failed open()s could cause deadlock; fixed (867, 869)
3572 - stop cancelling OST locks when files are closed (481)
3573 - overlapping XID spaces caused network corruption (851, 853)
3574 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3575 - setattr_raw added, to do single-RPC, server-side setattrs
3576 - lmc/lconf syntax change for OST UUIDs
3577 - fix crashy race condition between ptlrpc_free_req and osc_close
3578 - don't use request in mdc_enqueue if we hit a timeout (889)
3579 - don't set the inode i_size for regular files from the MDS (896)
3580 - handle out of order completion AST (842)
3581 - don't LBUG if a lock request times out after receiving AST (913)
3582 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3583 - if a bad lock AST arrives, send an error instead of dropping entirely
3584 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3585 - fix leak in bulk IO when only partially completed (899, 900, 926)
3586 - fix O_DIRECT for ia64 (55)
3587 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3588 - C-z after timeout could hang a process forever; fixed (977)
3590 - client-side I/O cache (678, 924, 929, 941, 970)
3592 - READPAGE and SETATTRs which don't take server-side locks get
3595 2003-02-11 Phil Schwan <phil@clusterfs.com>
3598 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3599 - Fix fsfilt_extN_readpage() to read a full page of directory
3600 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3601 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3602 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3603 - Most intent operations were removed, in favour of a new RPC mode
3604 that does a single RPC to the server and bypasses most of the VFS
3605 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3606 - Aggressively cancel local locks on DLM servers
3607 - mds_reint_unlink sends EA to the client if it's the last nlink.
3608 client uses that EA to unlink OST objects.
3609 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3610 - recursive symlinks were fixed (439)
3611 - fixed NULL deref in DEBUG_REQ
3612 - filter_update_lastobjid no longer calls sync, which annoyed extN
3613 - fixed multi-client small-writes to a single file problem (445)
3614 - fixed mtime updates during file writes (607)
3615 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3616 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3617 - fixed Linux OST size reporting problem (444, 656)
3618 - OST now updates object mtime with writes or setattr (607, 619)
3619 - client verifies file size before zeroing page past EOF (445)
3620 - OST now writes last allocated objid to disk with allocation (108)
3621 - LOV on echo now works (409)
3623 - mds_reint_unlink sends a new buffer, with the EA included. this
3624 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3625 set if a regular file was being unlinked, and it was the last link
3626 - use PtlGet from the target for bulk writes (315)
3627 - OST now updates object mtime with writes or setattr (607, 619)
3628 - LDLM now has a grant-time callback to revalidate locked items, if
3630 - Many MDS operations were reorganized to combat race conditions
3632 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3633 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3635 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3638 - Fully reactivate OST imports after reconnection (512, others)
3639 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3640 - More graceful error handling for truncating on dead OST (515)
3641 - Don't error out unless we're actually accessing dead stripes (474)
3642 - Fix garbage sizes when stripes are missing (410)
3643 - LRU counters were broken, causing constant lock purge (433, 432)
3644 - garbage on read from stripes with failed OSTs (441)
3645 - mark OSCs as active before reconnecting during recovery (438)
3646 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3647 - lfind did not preserve OST order in output (443)
3648 - symlinks cause hung clients, incorrect data (439)
3649 - stop dereferencing request after dropping refcount (457)
3650 - don't LASSERT(spin_is_locked) on non-SMP (455)
3651 - fixes for many rename() bugs
3652 - fstat didn't correctly synchronize attributes (399)
3653 - server must handle lock cancellation during blocking AST prep (487)
3654 - bulk descriptors were free()d too soon (511)
3655 - fix paths in lconf, which would load incorrect modules (451, 507)
3656 - fix confusing lconf 'host not found' error message (386)
3657 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3658 - fix race condition in mdc_blocking_ast() for inode access (526)
3659 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3660 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3661 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3662 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3663 - fix lstripe in several ways (kernel side) (527)
3664 - fix request leak in ldlm_cli_enqueue (262)
3665 - incorrect OSC was marked inactive after OST failure
3666 - call mds_fs_cleanup before unmounting filesystem (524)
3667 - fix races between taking ns_lock and ldlm_lock_change_resource
3668 - fix races updating LOV export open file list
3669 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3670 - fix recovery NULL deref in ldlm_cli_cancel_unused
3671 - fix some DLM races by using new hash table for lock handles (419)
3672 - permit the client to specify desired inodes, at replay
3673 - duplicate requests when we queue them for replay reintegration
3674 - fix last_rcvd offset calculation
3675 - sync after each recovered transaction, so we always make progress
3676 - never, not always, ERESTART requests without transnos
3677 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3679 - skip replay if the MDS says that the client is already connected
3680 - don't check for a recovery-enabled export to match lctl's UUID
3681 - don't INC_USE_COUNT for phantom exports
3682 - don't crash when cleaning up phantom exports (567)
3683 - don't double-finish or set replay data for errored mdc_open requests
3684 - abort requests when they time out, so we don't get old replies
3685 - send/receive replies for AST messages again
3686 - if the client says that it doesn't have the lock, cancel it on the
3688 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3689 mark it as destroyed, it will all work out when decref is called
3690 - fix module use counts (22, 581)
3692 - ASTs now expect a reply (server cancels lock on error reply)
3694 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3697 - fix many simultaneous client startup (392)
3698 - fix dentry->d_it clobbering
3699 - credentials weren't being shipped for readdir/getattr operations
3700 - remove invalid assertions triggered during some concurrent MD
3702 - proper Lustre versions added (336, 389)
3703 - fix memory leak for create error case (398)
3704 - fix LOV locking bug that would get cli/srv out of sync
3705 - fix echo client over LOV (409)
3706 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3707 - fix double-O_EXCL intent crash (424)
3708 - avoid sending multiple lock CANCELs (352)
3710 - MDS can do multi-client recovery (modulo bugs in new code)
3712 - many updates, edits, cleanups
3714 2002-11-18 Phil Schwan <phil@clusterfs.com>
3717 - fix null d_it dereference (346)
3718 - fix full OST/dbench hang (333)
3719 - fix permission problem with file removal (286)
3720 - fix removal of OSCs from LOV when they fail
3721 - fix NULL deref during bulk timeout (214)
3722 - fix problems related to multiple filesystems on one MDS (241)
3723 - fixed serious subtle metadata locking bugs
3724 - free locks on clients when inodes are removed due to memory
3726 - fix inode pointer in lock data (285)
3727 - partial support for multiple MDS on a single host (241)
3728 - data locks weren't cancelled at clear_inode time (290, 311)
3729 - intent locks could lead to unbounded lock growth (205)
3730 - added a maximum lock count, an LRU list, and a flusher
3731 - fix multiple rename (365)
3732 - properly abstracted the echo client
3733 - OSC locked 1 byte too many; fixed
3734 - rewrote brw callback code:
3735 - fixed recovery bugs related to LOVs (306)
3736 - fixed too-many-pages-in-one-write crash (191)
3737 - fixed (again) crash in sync_io_timeout (214)
3738 - probably fixed callback-related race (385)
3740 - Add capability to MDS protocol
3741 - LDLM cancellations and callbacks on different portals
3743 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3746 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3747 - defer open object destruction to close time (601981, 138)
3748 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3749 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3750 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3751 - don't call obd_setattr in ll_file_release for destroyed objects
3752 * protocol change to lustre_msg: move |version| and add |flags|
3753 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3754 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3755 * added replay of create, unlink, link and rename operations during
3756 MDS failover; recovery should be much more robust now
3757 * remove failed OSCs from LOVs (only lov_create uses this so far)
3758 * the lustre-HOWTO was brought (more) up to date (582544)
3760 2002-10-23 Phil Schwan <phil@clusterfs.com>
3763 - in-use dentries weren't being reused properly (617851)
3764 - prevent multiple LDLM setup (599178)
3765 - fix LOV size calculations for truncate (617853)
3766 - fix client handling of MDS intent errors (POSIX)
3767 - fix permission bug in lovstripe.c test (624321)
3768 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3769 - truncate past end of file could corrupt data
3770 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3771 - a race in recovery could return ETIMEDOUT to apps (623947)
3772 - building outside the source directory was fixed
3773 * the lustre-HOWTO was brought (more) up to date (582544)
3774 * major progress was made on recovery functionality
3776 2002-10-10 Phil Schwan <phil@clusterfs.com>
3779 - recovery deadlock fix
3780 - rm -rf causes LBUG fix (617817)
3781 - file open by multiple tasks fix (618962)
3782 - directory permissions bugs (602707 and 620007)
3783 - journal_stop fixed with locking (611313)
3784 - O_APPEND failures resolved (618273, perhaps 614459)
3785 - lconf PATH fix (619770)
3786 - IA64 build fix (621450)
3787 - RPC buffer sizes scale with amount of memory
3789 2002-10-01 Phil Schwan <phil@clusterfs.com>
3792 - locks would be cancelled without throwing away data pages,
3793 resulting in inconsistent data (605627)
3794 - inode attributes were not always being refreshed (605627, 612449)
3795 - lconf now continues to cleanup after lctl reports an error
3796 - MDS now enforces user permissions (602707)
3797 - lprocfs cleanup fixed, but not yet enabled (614157)
3798 - fixed infinite server hang, should a client not respond to an AST
3799 - avoid going into recovery if user calls readlink() with a buffer
3800 that's too small (613941)
3801 - AST RPCs no longer require replies (614867) -- this may be changed
3802 - don't crash server if client sends an IOV that's too big (611336)
3803 - fixed lock conversion deadlock (611892)
3804 - fixed the following of symlinks (614622)
3805 * recovery: the server can remove locks from a client that dies, other
3806 clients can make progress
3807 * more extN patch fixes
3808 * compile-time configurable ptlrpc buffer allocations
3810 - collaborative read cache document
3811 - Lustre Lite Performance CDR document-in-progress
3813 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3816 - fix typo in patch-2.4.18
3818 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3821 - clear ptlrpc request each time in handle_incoming_request()
3822 - unlink of files now destroys the object on the OST
3824 2002-09-19 Peter Braam <braam@clusterfs.com>
3826 * add hard link support
3827 * change obdfile creation method
3828 * kernel patch changed
3830 2002-09-19 Peter Braam <braam@clusterfs.com>
3833 - stack overflow bug in extN fixed
3835 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3837 * documentation updates
3838 - add man pages for config tools
3839 - update tests/README to describe testing with new config tools
3840 - finish metadata API descriptions
3841 * bug fixes and cleanups
3842 - statfs workaround for 16TB limit
3843 - LOV stripe allocation improved, can stripe on subset of OSTs
3844 - LOV file size/IO offset was wrong for files > 4GB in size
3845 - object EA data was being dropped, caused files to be unreadable
3846 - memory overflow with non-LOV OST caused memory corruption
3847 - fixed regression tests to work with new config tools, obdfilter
3848 - fixed bug when directory size became larger than 1 block
3849 - fixed bug (for single client case) when PWD was deleted
3850 - invalidate local directory pages when doing intent-based ops
3851 - avoid LDLM oops when lock callback contained bad data
3853 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3855 * documentation updates
3856 * bug fixes and cleanups
3857 - configuration tools
3860 - 64-bit compile warnings
3861 - 64-bit internal statfs data
3863 * test_brw on persistent OST devices
3865 * lprocfs (disabled)
3867 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3869 * documentation updates
3870 * bug fixes and cleanups
3871 * configuration tools
3873 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3876 * many small fixes to 0_5_4
3877 * io/network handling
3878 * thinkos in MDS operations
3880 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3883 * crucial basic fixes to 0.5.3
3884 * IOR, Iozone work over Elan
3887 2002-08-07 Phil Schwan <phil@clusterfs.com>
3888 * version 0_5_3, our first alpha
3889 * we use the new Portals iovs
3890 * documentation updates
3891 * bug fixes and cleanups
3892 * small changes in the DLM wire protocol
3894 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3895 * version 0_5_1 with some initial stability,
3896 * locking on MD and file I/O.
3897 * documentation updates
3898 * several bug fixes since 0.5.0
3899 * small changes in wire protocol
3901 2002-07-18 Phil Schwan <phil@clusterfs.com>
3903 * delivered as Lustre Light Alpha
3904 * fixed a crash after handling invalid MDS requests
3905 * fixed directory pages for architectures with non-4k pages sizes
3907 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3908 * release version v0_4_4
3909 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3911 * Fixes a number of interruption problems with OST operations.
3912 * Update documentation for portals header changes
3913 * Move all wire protocol structs/defines to lustre_idl.h
3914 * Fixes symlink length bug.
3915 * Add tcpdump to repository.
3917 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3918 * release version v0_4_3
3919 * Fixes statfs for inodes on extN.
3920 * Fixes bug in runtests which would delete /etc/hosts.
3921 * Use 64-bit object IDs wherever possible (not into VFS though)
3922 Remove ost_get_info, which is unused by lustre, and out of date.
3924 2002-07-03 Peter Braam <braam@clusterfs.com>
3925 * release version v0_4_2 Fixes a lookup error (type not passed)
3926 * move forward to head of Portals
3927 * move forward to latest Lustre kernel
3929 2002-06-25 Peter Braam <braam@clusterfs.com>
3930 * release version v0_4_1. Hopefully stable on single node use.