1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10),
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: don't use GFP_* in generic Lustre code.
18 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
22 Description: obdclass is limited by single OBD_ALLOC(idarray)
23 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
27 Description: warnings with build patchless client with vanila 2.6.19 and up
28 Details : change old ctl_table style and replace ctl_table/ctl_table_header
29 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
33 Frequency : Only for SLES
34 Description: sanity_quota fail test_1
35 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
36 file, which makes TSTID[2] inunique.
40 Frequency : Only for SLES9
41 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
42 Details : Update target/ChangeLog/which_patch .
46 Frequency : Only for test_10 in sanity-quota.sh
47 Description: a bug in quota test code
48 Details : checking a test flag in a wrong place
52 Frequency : Only for test_13 in sanity-quota.sh
53 Description: a bug in quota test script
54 Details : shouldn't check overfull allocated quota
58 Description: jbd statistics
59 Details : Port older jbd statistics patch for sles10
61 Severity : enhancement
63 Description: Add printing inode info into message about error in writepage.
68 Description: reply_lock_interpret crash due to race with it and lock cancel.
69 Details : Do not replay locks that are being cancelled. Do not reference
70 locks by their address during replay, just by their handle.
74 Frequency : When flocks are used.
75 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
76 Details : Flock locks might destroy just granted lock if it could be merged
77 with another existing flock, this is done in completion handler,
78 so teach ldlm_cli_enquque_fini that this is a valid case for
84 Description: Oops in read and write path when failing to allocate lock.
85 Details : Check if lock allocation failed and return error back.
89 Description: lstripe command fails for valid OST index
90 Details : The stripe offset is compared to lov->desc.ld_tgt_count
91 instead of lov->desc.ld_active_tgt_count.
95 Description: sanity.sh failed test 103
96 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
101 Description: df doesn't work properly if diskfs blocksize != 4K
102 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
106 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
107 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
111 Description: obd_health_check_timeout too short
112 Details : set obd_health_check_timeout as 1.5x of obd_timeout
116 Description: llapi_file_create() does not allow some changes
117 Details : add llapi_file_open() that allows specifying the mode and
118 open flags, and also returns an open file handle.
122 Description: merge and cleanup kernel patches.
123 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
127 Description: Add journal checksum support.(Kernel part)
128 Details : The journal checksum feature adds two new flags i.e
129 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
130 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
131 indicates that the commit block contains the checksum for
132 the blocks described by the descriptor blocks. Now commit
133 record can be sent to disk without waiting for descriptor
134 blocks to be written to disk. This behavior is controlled
135 using JBD2_FEATURE_ASYNC_COMMIT flag.
139 Description: OSS needs mutliple precreate threads
140 Details : Add ability to start more than one create thread per OSS.
142 --------------------------------------------------------------------------------
144 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
146 * Support for kernels:
147 2.6.5-7.283 (SLES 9),
148 2.6.9-55.EL (RHEL 4),
149 2.6.16.46-0.14 (SLES 10),
150 2.6.18.8 vanilla (kernel.org)
151 * Client support for unpatched kernels:
152 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
153 2.6.16 - 2.6.22 vanilla (kernel.org)
154 * Due to recently discovered recovery problems, we do not recommend
155 using patchless RHEL 4 clients with this or any earlier release.
156 * Recommended e2fsprogs version: 1.39.cfs8
157 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
158 * Starting with this release, the ldiskfs backing filesystem required
159 by Lustre is now in its own package, lustre-ldiskfs. This package
160 should be installed. It is versioned separately from Lustre and
161 may be released separately in future.
165 Description: block reactivating mgc import until all deactivates complete
166 Details : Fix race when failing back MDT/MGS to itself (testing)
168 Severity : enhancement
170 Description: add optional extra BUILD_VERSION info
171 Details : add a new environment variable (namely LUSTRE_VERS) which allows
172 to override the lustre version.
175 Frequency : 2.6.18 servers only
177 Description: ll_kern_mount() doesn't release the module reference
178 Details : The ldiskfs module reference count never drops down to 0
179 because ll_kern_mount() doesn't release the module reference.
184 Description: server LBUG when using old ost_num_threads parameter
185 Details : Accept the old ost_num_threads parameter but warn that it
186 is deprecated, and fix an off-by-one error that caused an LBUG.
191 Description: Transient SCSI error results in persistent IO issue
192 Details : iobuf->dr_error is not reinitialized to 0 between two
196 Frequency : sometimes when underlying device returns I/O errors
198 Description: OSTs not going read-only during write failures
199 Details : OSTs are not remounted read-only when the journal commit threads
200 get I/O errors because fsfilt_ext3 calls journal_start/stop()
201 instead of the ext3 wrappers.
205 Description: poor connect scaling with increasing client count
206 Details : Don't run filter_grant_sanity_check for more than 100 exports
207 to improve scaling for large numbers of clients.
210 Frequency : SLES10 only
212 Description: sanity-quota.sh quotacheck failed: rc = -22
213 Details : Quotas cannot be enabled on SLES10.
216 Frequency : liblustre clients only
218 Description: getdirentries does not give error when run on compute nodes
219 Details : getdirentries does not fail when the size specified as an argument
220 is too small to contain at least one entry
222 Severity : enhancement
224 Description: Add LNET router traceability for debug purposes
225 Details : If a checksum failure occurs with a router as part of the
226 IO path, the NID of the last router that forwarded the bulk data
227 is printed so it can be identified.
232 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
233 Details : Due to a race condition, liblustre clients were occasionally
236 Severity : enhancement
238 Description: lfs setstripe use optional parameters instead of postional
241 Severity : enhancement
243 Description: Nanosecond timestamp support for ldiskfs
244 Details : The on-disk ldiskfs filesystem has added support for nanosecond
245 resolution timestamps. There is not yet support for this at
246 the Lustre filesystem level.
249 Frequency : during server recovery
251 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
252 Details : request with rq_no_resend flag not awake l_wait_event if they get a
256 Frequency : nfs export on patchless client
258 Description: connectathon hang when test nfs export over patchless client
259 Details : Disconnected dentry cannot be found with lookup, so we do not need
260 to unhash it or make it invalid
263 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
264 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
265 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
270 Description: open req refcounting wrong on reconnect
271 Details : If reconnect happened between getting open reply from server and
272 call to mdc_set_replay_data in ll_file_open, we will schedule
273 replay for unreferenced request that we are about to free.
274 Subsequent close will crash in variety of ways.
275 Check that request is still eligible for replay in
276 mdc_set_replay_data().
281 Description: disable writes to filesystem when reading health_check file
282 Details : the default for reading the health_check proc file has changed
283 to NOT do a journal transaction and write to disk, because this
284 can cause reads of the /proc file to hang and block HA state
285 checking on a healthy but otherwise heavily loaded system. It
286 is possible to return to the previous behaviour during configure
287 with --enable-health-write.
289 Severity : enhancement
291 Description: 64-bit inode version
292 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
293 to the inode. This will be required for version-based recovery.
298 Description: MDS fails to start if a duplicate client export is detected
299 Details : in some rare cases it was possible for a client to connect to
300 an MDS multiple times. Upon recovery the MDS would detect this
301 and fail during startup. Handle this more gracefully.
303 Severity : enhancement
305 Description: Add -o localflock option to simulate old noflock
307 Details : This will achieve local-only flock/fcntl locks
313 Description: log_commit_thread vs filter_destroy race leads to crash
314 Details : Take import reference before releasing llog record semaphore
319 Description: Wrong request locking in request set processing
320 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
321 imp_delayed_list, in this place should be used imp_lock.
324 Frequency : when reconnection
326 Description: Grant Leak when osc reconnect to OST
327 Details : When osc reconnect ost, OST(filter) should check whether it
328 should grant more space to client by comparing fed_grant and
329 cl_avail_grant, and return the granted space to client instead
330 of "new granted" space, because client will call osc_init_grant
331 to update the client grant space info.
334 Frequency : when client reconnect to OST
336 Description: Grant Leak when osc do resend and replay bulk write
337 Details : When osc reconnect to OST, OST(filter)should clear grant info of
338 bulk write request, because the grant info while be sync between
339 OSC and OST when reconnect, and we should ignore the grant info
340 these of resend/replay write req.
345 Description: Grant space more than avaiable left space sometimes.
346 Details : When then OST is about to be full, if two bulk writing from
347 different clients came to OST. Accord the avaliable space of the
348 OST, the first req should be permitted, and the second one
349 should be denied by ENOSPC. But if the seconde arrived before
350 the first one is commited. The OST might wrongly permit second
351 writing, which will cause grant space > avaiable space.
354 Frequency : when client is evicted
356 Description: Grant might be wrongly erased when osc is evicted by OST
357 Details : when the import is evicted by server, it will fork another
358 thread ptlrpc_invalidate_import_thread to invalidate the
359 import, where the grant will be set to 0. While the original
360 thread will update the grant it got when connecting. So if
361 the former happened latter, the grant will be wrongly errased
362 because of this race.
367 Description: Checking Stale with correct fid
368 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
369 to check whether it is stale, because sometimes, we want the
370 enqueue happened anyway, and op_data.fid2 will not be initialized.
372 Severity : enhancement
374 Description: update patchless client
375 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
378 Frequency : only with 2.4 kernel
380 Description: random memory corruption
381 Details : size of struct ll_inode_info is to big for union inode.u and this
382 can be cause of random memory corruption.
387 Description: Memory leak in recovery
388 Details : Lov_mds_md was not free in an error handler in mds_create_object.
389 It should also check obd_fail before fsfilt_start, otherwise if
390 fsfilt_start return -EROFS,(failover mds during mds recovery).
391 then the req will return with repmsg->transno = 0 and rc = EROFS.
392 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
393 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
394 be freed no matter whether fsfilt_commit success or not.
397 Frequency : only with huge count clients
399 Description: Prevents from taking the superblock lock in llap_from_page for
401 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
402 for llap_from_page call in ll_removepage prevents from taking the
403 superblock lock for a soon died page.
408 Description: Not check open intent error before release open handle
409 Details : in some rare cases, the open intent error is not checked before
410 release open handle, which may cause
411 ASSERTION(open_req->rq_transno != 0), because it tries to release
412 the failed open handle.
417 Description: Set cat log bitmap only after create log success.
418 Details : in some rare cases, the cat log bitmap is set too early. and it
419 should be set only after create log success.
423 Description: Accessing a block bevice can re-enable I/O when Lustre is
424 tearing down a device.
425 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
430 Description: service threads may hog cpus when there are a lot of requests
432 Details : Insert cond_resched to give other threads a chance to use some of
438 Description: the cat log was not initialized in recovery
439 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
440 unlink log on mds will not be initialized until mds post
441 recovery. And also in mds post recovery, the unlink log will
442 initialization will be done asynchronausly, so there will be race
443 between add unlink log and unlink log initialization.
447 Description: brw_stats were being printed incorrectly
448 Details : brw_stats were being printed as log2 but all of them were not
449 recorded as log2. Also remove some code duplication arising from
450 filter_tally_{read,write}.
454 Frequency : rare, only in recovery.
455 Description: ASSERTION(req->rq_type != LI_POISON) failed
456 Details : imp_lock should be held while iterating over imp_sending_list for
457 prevent destroy request after get timeout in ptlrpc_queue_wait.
461 Description: replay-single.sh test 52 fails
462 Details : A lock's skiplist need to be cleanup when it being unlinked
463 from its resource list.
467 Description: Short directio read returns full requested size rather than
469 Details : Direct I/O operations should return actual amount of bytes
470 transferred rather than requested size.
472 Severity : enhancement
474 Description: metadata RPC reduction (e.g. for rm performance)
475 Details : decrease the amount of synchronous RPC between clients and servers
476 by canceling conflicing lock before the operation on the client side
477 and packing thier handles into the main operation RPC to server.
479 Severity : enhancement
481 Description: Async OSC create to avoid the blocking unnecessarily.
482 Details : If a OST has no remain object, system will block on the creating
483 when need to create a new object on this OST. Now, ways use
484 pre-created objects when available, instead of blocking on an
485 empty osc while others are not empty. If we must block, we block
486 for the shortest possible period of time.
488 --------------------------------------------------------------------------------
490 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
495 Frequency : on some architectures
497 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
498 Details : Uninitialized flags sometimes cause configuration commands to
502 Frequency : patchless clients only
504 Description: missing __iget() symbol export
505 Details : The __iget() symbol export is missing. To avoid the need for
506 this on patchless clients the deathrow inode reaper is turned
507 off, and we depend on the VM to clean up old inodes. This
508 dependency was during via the fix for bug 12181.
510 --------------------------------------------------------------------------------
512 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
514 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
515 INTEROPERATE with older versions automatically. In many cases a
516 special upgrade step is needed. Please read the
517 user documentation before upgrading any part of a live system.
518 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
519 version will not interoperate with 1.6 betas before beta5 (1.5.95).
520 * WARNING: Lustre configuration and startup changes are required with
521 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
527 Severity : enhancement
529 Description: MountConf
530 Details : Lustre configuration is now managed via mkfs and mount
531 commands instead of lmc and lconf. New obd types (MGS, MGC)
532 are added for dynamic configuration management. See
533 https://mail.clusterfs.com/wikis/lustre/MountConf for
536 Severity : enhancement
538 Description: dynamic OST addition
539 Details : OSTs can now be added to a live filesystem
541 Severity : enhancement
543 Description: startup order invariance
544 Details : MDTs and OSTs can be started in any order. Clients only
545 require the MDT to complete startup.
547 Severity : enhancement
549 Description: parallel, asynchronous orphan cleanup
550 Details : orphan cleanup is now performed in separate threads for each
551 OST, allowing parallel non-blocking operation.
553 Severity : enhancement
555 Description: optimized stripe assignment
556 Details : stripe assignments are now made based on ost space available,
557 ost previous usage, and OSS previous usage, in order to try
558 to optimize storage space and networking resources.
560 Severity : enhancement
562 Description: Permanently set tunables
563 Details : All writable /proc/fs/lustre tunables can now be permanently
564 set on a per-server basis, at mkfs time or on a live
567 Severity : enhancement
569 Description: Lustre message v2
570 Details : Add lustre message format v2.
572 Severity : enhancement
574 Description: client OST exclusion list
575 Details : Clients can be started with a list of OSTs that should be
576 declared "inactive" for known non-responsive OSTs.
580 Description: SPEC SFS validation failure on NFS v2 over lustre.
581 Details : Changes the blocksize for regular files to be 2x RPC size,
582 and not depend on stripe size.
584 Severity : enhancement
586 Description: Multiple MD RPCs in flight.
587 Details : Further unserialise some read-only MDS RPCs - learn about intents.
588 To avoid overly-overloading MDS, introduce a limit on number of
589 MDS RPCs in flight for a single client and add /proc controls
590 to adjust this limit.
592 Severity : enhancement
594 Description: client read/write statistics
595 Details : Add client read/write call usage stats for performance
596 analysis of user processes.
597 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
598 file access. extents_stats shows chunk size distribution.
599 extents_stats_per_process show chunk size distribution per
602 Severity : enhancement
604 Description: mds statistics
605 Details : Add detailed mds operations statistics in
606 /proc/fs/lustre/mds/*/stats.
610 Description: Failure of copying files with lustre special EAs.
611 Details : Client side always return success for setxattr call for lustre
612 special xattr (currently only "trusted.lov").
617 Description: Refcount LNET uuids
618 Details : The global LNET uuid list grew linearly with every startup;
619 refcount repeated list entries instead of always adding to
622 Severity : enhancement
624 Description: Dynamic service threads
625 Details : Within a small range, start extra service threads
626 automatically when the request queue builds up.
629 Frequency : mixed-endian client/server environments
631 Description: mixed-endian crashes
632 Details : The new msg_v2 system had some failures in mixed-endian
635 Severity : enhancement
637 Description: Easy OST removal
638 Details : OSTs can be permanently deactivated with e.g. 'lctl
639 conf_param lustre-OST0001.osc.active=0'
641 Severity : enhancement
643 Description: MGS proc entries
644 Details : Added basic proc entries for the MGS showing what filesystems
647 Severity : enhancement
649 Description: provide MGS failover
650 Details : Added config lock reacquisition after MGS server failover.
652 Severity : enhancement
654 Description: add Linux 2.4 support
655 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
659 Description: a large application tries to do I/O to the same resource and dies
661 Details : Check the req->rq_arrival time after the call to
662 ost_brw_lock_get(), but before we do anything about
663 processing it & sending the BULK transfer request. This
664 should help move old stale pending locks off the queue as
665 quickly as obd_timeout.
668 Frequency : when an incorrect nid is specified during startup
670 Description: ptlrpc connect to non-existant node causes kernel crash
671 Details : LNET can't be re-entered from an event callback, which
672 happened when we expire a message after the export has been
673 cleaned up. Instead, hand the zombie cleanup off to another
676 Severity : enhancement
678 Description: plain/inodebits lock performance improvement
679 Details : Grouping plain/inodebits in granted list by their request modes
680 and bits policy, thus improving the performance of search through
684 Frequency : only if OST filesystem is corrupted
686 Description: client incorrectly hits assertion in ptlrpc_replay_req()
687 Details : for a short time RPCs with bulk IO are in the replay list,
688 but replay of bulk IOs is unimplemented. If the OST filesystem
689 is corrupted due to disk cache incoherency and then replay is
690 started it is possible to trip an assertion. Avoid putting
691 committed RPCs into the replay list at all to avoid this issue.
694 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
696 Description: System hang on startup
697 Details : This bug allowed the liblustre (e.g. catamount) client to
698 return to the app before handling all startup RPCs. This
699 could leave the node unresponsive to lustre network traffic
700 and manifested as a server ptllnd timeout.
702 Severity : enhancement
704 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
705 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
706 boolean to control whether to print peer debug info when a
707 client's RPC times out.
710 Frequency : only for kernels with patches from Lustre below 1.4.3
712 Description: Remove old rdonly API
713 Details : Remove old rdonly API which unused from at least lustre 1.4.3
716 Frequency : only for devices with external journals
718 Description: Set external device read-only also
719 Details : During a commanded failover stop, we set the disk device
720 read-only while the server shuts down. We now also set any
721 external journal device read-only at the same time.
724 Frequency : when upgrading from 1.4 while trying to change parameters
726 Description: The wrong (new) MDC name was used when setting parameters for
727 upgraded MDT's. Also allows changing of OSC (and MDC)
728 parameters if --writeconf is specified at tunefs upgrade time.
731 Frequency : when setting specific ost indicies
733 Description: QOS code breaks on skipped indicies
734 Details : Add checks for missing OST indicies in the QOS code, so OSTs
735 created with --index need not be sequential.
737 Severity : enhancement
739 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
740 Details : The uninit_groups feature works in conjunction with the kernel
741 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
742 up the pass1 processing of e2fsck. This is a read-only feature
743 in ldiskfs2 only, so older kernels and current ldiskfs cannot
744 mount filesystems that have had this feature enabled.
746 Severity : enhancement
748 Description: Improve multi-block allocation algorithm to avoid fragmentation
749 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
750 allocation locality and avoid filesystem fragmentation.
752 ------------------------------------------------------------------------------
754 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
756 * Support for kernels:
757 2.4.21-47.0.1.EL (RHEL 3)
759 2.6.9-42.0.10.EL (RHEL 4)
760 2.6.12.6 vanilla (kernel.org)
761 2.6.16.27-0.9 (SLES 10)
762 * Recommended e2fsprogs version: 1.39.cfs5
764 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
768 Frequency : occasional, depends on client load and configuration
769 Bugzilla : 12181, 12203
770 Description: data loss for recently-modified files
772 Details : In some cases it is possible that recently written or created
773 files may not be written to disk in a timely manner (this should
774 normally be within 30s unless client IO load is very high).
775 The problem appears as zero-length files or files that are a
776 multiple of 1MB in size after a client crash or client eviction
777 that are missing data at the end of the file.
779 This problem is more likely to be hit on clients where files are
780 repeatedly created and unlinked in the same directory, clients
781 have a large amount of RAM, have many CPUs, the filesystem has
782 many OSTs, the clients are rebooted frequently, and/or the files
783 are not accessed by other nodes after being written.
785 The presence of the problem can be detected by looking at
786 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
787 smaller than the second (nr_unused) then dirty files will not
788 be flushed automatically to disk. "sync; sleep 10" should be
789 run several times on the node before unmounting it to update
790 Lustre (this is also safe to run on nodes without this problem).
792 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
793 that can cause this same problem, so customers using that kernel
794 also need to update the kernel in addition to Lustre. In order
795 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
797 It is normal that files written just before a client crash (less
798 than 30s) may not yet have been flushed to disk, even for local
802 Frequency : frequent on thin XT3 nodes
804 Description: UUID collision on thin XT3 Linux nodes
805 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
806 randomly, since we relied on an insufficiently-seeded PRNG.
811 Description: OSS hangs after "All ost request buffers busy"
812 Details : A deadlock between quota and journal operations caused OSS
813 hangs after printing "All ost request buffers busy."
816 Frequency : always on liblustre builds
818 Description: Cleanup compiler warnings on liblustre
821 Frequency : always on liblustre builds on XT3
823 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
827 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
828 > 32000 subdirectories
829 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
830 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
831 subdirectory count crosses 32000. This will aid e2fsck to
832 correctly handle more than 32000 subdirectories.
835 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
837 Description: System hang on startup
838 Details : This bug allowed the liblustre (e.g. catamount) client to
839 return to the app before handling all startup RPCs. This
840 could leave the node unresponsive to lustre network traffic
841 and manifested as a server ptllnd timeout.
843 Severity : enhancement
845 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
846 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
847 boolean to control whether to print peer debug info when a
848 client's RPC times out.
853 Description: make O_SYNC working on 2.6 kernels
854 Details : 2.6 kernels use different method for mark pages for write,
855 so need add a code to lustre for O_SYNC work.
860 Description: Failure to close file and release space on NFS
861 Details : Put inode details into lock acquired in ll_intent_file_open.
862 Use mdc_intent_lock in ll_intent_open to properly
863 detect all kind of errors unhandled by mdc_enqueue.
868 Description: proc file read during shutdown sometimes raced obd removal,
870 Details : Add lock to prevent obd access after proc file removal.
873 Frequency : Only for files larger than 4GB on 32-bit clients.
875 Description: improperly doing page alignment of locks
876 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
877 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
882 Description: RPCs being resent when they shouldn't be
883 Details : Some RPCs that should not be resent are being resent. This
884 can cause inconsistencies in the RPC state machine. Do not
885 resend such requests.
888 Frequency : rare, only with NFS export
890 Description: Crash on NFS re-export node
891 Details : under very unusual load conditions an assertion is hit in
892 ll_intent_file_open()
895 Frequency : only if OST filesystem is corrupted
897 Description: client incorrectly hits assertion in ptlrpc_replay_req()
898 Details : for a short time RPCs with bulk IO are in the replay list,
899 but replay of bulk IOs is unimplemented. If the OST filesystem
900 is corrupted due to disk cache incoherency and then replay is
901 started it is possible to trip an assertion. Avoid putting
902 committed RPCs into the replay list at all to avoid this issue.
907 Description: large O_DIRECT requests fail under memory pressure/fragmentation
908 Details : Large single O_DIRECT read and write calls can fail to allocate
909 a sufficiently large buffer to process the request. In case of
910 allocation failure the allocation is retried with a smaller
911 buffer and broken into smaller requests.
913 Severity : enhancement
915 Description: Add -o localflock option to simulate old noflock behaviour.
916 Details : This will achieve local-only flock/fcntl locks coherentness.
921 Description: versioning check is incomplete
922 Details : Checking the version difference of client vs. server, report
923 error if the gap is too big.
928 Description: add support PG_writeback bit
929 Details : add support for PG_writeback bit for Lustre, for more carefull
930 work with page cache in 2.6 kernel. This also fix some deadlocks
931 and remove hack for work O_SYNC with 2.6 kernel.
933 Severity : enhancement
935 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
936 Details : The uninit_groups feature works in conjunction with the kernel
937 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
938 up the pass1 processing of e2fsck. This is a read-only feature
939 in ldiskfs2 only, so older kernels and current ldiskfs cannot
940 mount filesystems that have had this feature enabled.
942 Severity : enhancement
944 Description: Improve multi-block allocation algorithm to avoid fragmentation
945 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
946 allocation locality and avoid filesystem fragmentation.
948 ------------------------------------------------------------------------------
950 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
952 * Support for kernels:
953 2.6.9-42.0.3.EL (RHEL 4)
955 2.4.21-47.0.1.EL (RHEL 3)
956 2.6.12.6 vanilla (kernel.org)
957 2.6.16.21-0.8 (SLES10)
958 * Recommended e2fsprogs version: 1.39.cfs2-0
960 * The backwards-compatible /proc/sys/portals symlink has been removed
961 in this release. Before upgrading, please ensure that you change
962 any configuration scripts or /etc/sysctl.conf files that access
963 /proc/sys/portals/* or sysctl portals.* to use the corresponding
964 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
965 in advance of the upgrade on any system running Lustre 1.4.6 or
966 newer, since /proc/sys/lnet was added in that version.
967 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
971 Frequency : only when quota is used
973 Description: avoid scanning export list for quota master
974 Details : Change the algorithms to avoid scanning export list in order
975 to improve the efficiency.
978 Frequency : MDS failover only, very rarely
980 Description: "went back in time" messages on mds failover
981 Details : The greatest transno may be lost when the current operation
982 finishes with an error (transno==0) and the client's last_rcvd
983 record is over-written. Save the greatest transno in the
984 mds_last_transno for this case.
987 Frequency : always for specific kernels and striping counts
989 Description: client may get "Matching packet too big" without ACL support
990 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
991 when trying to access files in certain configurations. The
992 clients should in fact be denied when mounting because they do
996 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
998 Description: many clients connecting with IO in progress causes connect timeouts
999 Details : Avoid synchronous journal commits to avoid delays caused by many
1000 clients connecting/disconnecting when bulk IO is in progress.
1001 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1002 OST_IO_PORTAL to avoid delays behind potentially many pending
1006 Frequency : occasionally with multiple writers to a single file
1008 Description: shared writes to file may result in wrong size reported by stat()
1009 Details : Allow growing of kms when extent lock is cancelled
1012 Frequency : always with random mmap IO to multi-striped file
1014 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1015 Details : Only if the hole is at the end of OST object so that kms is too
1016 small. Fix is to increase kms accordingly in ll_nopage.
1019 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1021 Description: writes to a missing object would leak memory on the OST
1022 Details : If there is an inconsistency between the MDS and OST filesystems,
1023 such that the MDS references an object that doesn't exist, writes
1024 to that object will leak memory due to incorrect cleanup in the
1025 error handling path, eventually running out of memory on the OST.
1030 Description: Creating too long symlink causes lustre errors
1031 Details : Check symlink and name lengths before sending requests to MDS.
1034 Frequency : only if flock is enabled (not on by default)
1036 Description: posix locks not released on fd closure on 2.6.9+
1037 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1038 kernels, this caused such locks not to be released on fd close and
1039 then assertions on fs unmount about still used locks.
1042 Frequency : MDS failover only, very rarely
1044 Description: clients may get ASSERTION(granted_lock != NULL)
1045 Details : When request was taking a long time, and a client was resending
1046 a getattr by name lock request. The were multiple lock requests
1047 with the same client lock handle and
1048 mds_getattr_name->fixup_handle_for_resent_request found one of the
1049 lock handles but later failed with ASSERTION(granted_lock != NULL).
1054 Description: handle->h_buffer_credits > 0, assertion failure
1055 Details : h_buffer_credits is zero after truncate, causing assertion
1056 failure. This patch extends the transaction or creates a new
1060 Frequency : NFS re-export or patchless client
1061 Bugzilla : 11179, 10796
1062 Description: Crash on NFS re-export node (__d_move)
1063 Details : We do not want to hash the dentry if we don't have a lock.
1064 But if this dentry is later used in d_move, we'd hit uninitialised
1065 list head d_hash, so we just do this to init d_hash field but
1066 leave dentry unhashed.
1069 Frequency : NFS re-export or patchless client
1071 Description: NFS exports has problem with symbolic link
1072 Details : lustre client didn't properly install dentry when re-exported
1073 to NFS or running patchless client.
1076 Frequency : NFS re-export or patchless client
1078 Description: Various nfs/patchless fixes.
1079 Details : fixes reuse disconected alias for lookup process - this fixes
1080 warning "find_exported_dentry: npd != pd",
1081 fix permission error with open files at nfs.
1082 fix apply umask when do revalidate.
1085 Frequency : occasional
1087 Description: Crash on NFS re-export node
1088 Details : calling clear_page() on the wrong pointer triggered oops in
1089 generic_mapping_read().
1092 Frequency : rarely, using O_DIRECT IO
1094 Description: unaligned directio crashes client with LASSERT
1095 Details : check for unaligned buffers before trying any requests.
1098 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1100 Description: stale data returned from RAID cache
1101 Details : If only a small amount of IO is done to the RAID device before
1102 reading it again it is possible to get stale data from the RAID
1103 cache instead of reading it from disk.
1106 Frequency : always for sles10 kernel
1108 Description: sles10 support
1109 Details : ll_follow_link: compile fixes and using of nd_set_link
1110 under newer kernels.
1113 Frequency : depends on arch, kernel and compiler version, always on sles10
1116 Description: recursive or deep enough symlinks cause stack overflow
1117 Details : getting rid of large stack-allocated variable in
1121 Frequency : depends on hardware
1123 Description: lustre write performance loss in the SLES10 kernel
1124 Details : the performance loss is caused by using of write barriers in the
1125 ext3 code. The SLES10 kernel turns barrier support on by
1126 default. The fix is to undo that change for ldiskfs.
1128 ------------------------------------------------------------------------------
1130 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1132 * Support for kernels:
1133 2.6.9-42.0.3EL (RHEL 4)
1134 2.6.5-7.276 (SLES 9)
1135 2.4.21-47.0.1.EL (RHEL 3)
1136 2.6.12.6 vanilla (kernel.org)
1140 Frequency : quota enabled and large files being deleted
1142 Description: releasing more than 4GB of quota at once hangs OST
1143 Details : If a user deletes more than 4GB of files on a single OST it
1144 will cause the OST to spin in an infinite loop. Release
1145 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1150 Description: statfs data retrieved from /proc may be stale or zero
1151 Details : When reading per-device statfs data from /proc, in the
1152 {kbytes,files}_{total,free,avail} files, it may appear
1153 as zero or be out of date.
1156 Frequency : systems with MD RAID1 external journal devices
1158 Description: lconf's call to blkid is confused by RAID1 journal devices
1159 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1160 of returning all block devices that match the journal UUID.
1163 Frequency : always, for aggregate stripe size over 4GB
1165 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1166 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1167 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1170 Frequency : always if "lfs find" used on a local file/directory
1172 Description: "lfs find" segfaults if used on a local file/directory
1173 Details : The case where a directory component was not specified wasn't
1174 handled correctly. Handle this properly.
1177 Frequency : always on ppc64
1179 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1180 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1181 perform endianness conversion.
1184 Frequency : rarely (truncate to non-zero file size after write under load)
1185 Bugzilla : 10730, 10687
1186 Description: Files padded with zeros to next 4K multiple
1187 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1188 it is possible that files that are truncated to a non-zero size
1189 immediately after being written are filled with zero bytes beyond
1190 the truncated size. No file data is lost.
1192 Severity : enhancement
1194 Description: Allow recovery/failover for liblustre clients.
1195 Details : liblustre clients were unaware of failover configurations until
1198 Severity : enhancement
1200 Description: user file locks should fail when not mounting with flock option
1201 Details : Set up an error-returning stub in ll_file_operations.lock field
1202 to prevent incorrect behaviour when client is mounted without
1203 flock option. Also, set up properly f_op->flock field for
1207 Frequency : always on ia64
1209 Description: "lfs df" loops on printing out MDS statfs information
1210 Details : The obd_ioctl_data was not initialized and in some systems
1211 this caused a failure during the ioctl that did not return
1212 an error. Initialize the struct and return an error on failure.
1215 Frequency : SLES 9 only
1217 Description: Error of copying files with lustre special EAs as root
1218 Details : Client side always return success for setxattr call for lustre
1219 special xattr (currently only "trusted.lov").
1222 Frequency : rarely on clusters with both ia64+i386 clients
1224 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1225 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1226 clients are doing concurrent writes to the same file it is
1227 possible that the ia64 clients may LASSERT because the OST
1228 extent locks are not PAGE_SIZE aligned. Ensure that grown
1229 locks are always aligned on the request boundary.
1232 Frequency : specific use, occasional
1234 Description: Overwriting in use executable truncates on-disk binary image
1235 Details : If one node attempts to overwrite an executable in use by
1236 another node, we now correctly return ETXTBSY instead of
1237 truncating the file.
1239 Severity : enhancement
1241 Description: Async OSC create to avoid the blocking unnecessarily.
1242 Details : If a OST has no remain object, system will block on the creating
1243 when need to create a new object on this OST. Now, ways use
1244 pre-created objects when available, instead of blocking on an
1245 empty osc while others are not empty. If we must block, we block
1246 for the shortest possible period of time.
1251 Description: chmod on Lustre root is propagated to other clients
1252 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1253 invalid by the follow_mount time.
1256 Frequency : liblustre clients only
1258 Description: Race in 'instant cancel' lock handling could lead to such locks
1259 never to be granted in case of SMP MDS
1260 Details : Do not destroy not yet granted but cbpending locks in
1264 Frequency : replay/resend of open
1266 Description: non null lock assetion failure in mds_intent_policy
1267 Details : Trying to replay/resend lockless open requests resulted in
1268 mds_open() returning 0 with no lock. Now it sets a flag if
1269 a lock is going to be returned.
1271 Severity : enhancement
1273 Description: Checksum enhancements
1274 Details : New checksum enhancements allow for resending RPCs that failed
1277 Severity : enhancement
1279 Description: Tunables on number of dirty pages in cacche
1280 Details : Allow to set limit on number of dirty pages cached.
1285 Description: client crash on unmount - lock still has references
1286 Details : In some error handling cases it was possible to leak a lock
1287 reference on a client while accessing a file. This was not
1288 harmful to the client during operation, but would cause the
1289 client to crash when the filesystem is unmounted.
1292 Frequency : specific case, rare
1294 Description: ETXTBSY on mds though file not in use by client
1295 Details : ETXTBSY is no longer incorrectly returned when attempting to
1296 chmod or chown a directory that the user previously tried to
1297 execute or a currently-executing binary.
1300 Frequency : extremely rare except on liblustre-based clients
1302 Description: Lustre space not freed when files are deleted
1303 Details : Clean up open-unlinked files after client eviction. Previously
1304 the unlink was skipped and the files remained as orphans.
1309 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1310 Details : In some cases it was possible to send a blocking callback to
1311 a client doing a glimpse, even though that client didn't get
1312 a lock granted. When the glimpse lock is cancelled on the OST
1313 the freed lock is left on the waiting list and corrupted the list.
1316 Frequency : all core dumps
1318 Description: Broke core dumps to lustre
1319 Details : Negative dentry may be unhashed if parent does not have UPDATE
1320 lock, but some callers, e.g. do_coredump, expect dentry to be
1321 hashed after successful create, hash it in ll_create_it.
1323 ------------------------------------------------------------------------------
1325 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1327 * Support for kernels:
1328 2.6.9-42.0.2.EL (RHEL 4)
1329 2.6.5-7.276 (SLES 9)
1330 2.4.21-40.EL (RHEL 3)
1331 2.6.12.6 vanilla (kernel.org)
1335 Frequency : always on RHEL 3
1337 Description: Number of open files grows over time
1338 Details : The number of open files grows over time, whether or not
1339 Lustre is started. This was due to a filp leak introduced
1340 by one of our kernel patches.
1342 ------------------------------------------------------------------------------
1344 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1346 * Support for kernels:
1347 2.6.9-42.EL (RHEL 4)
1348 2.6.5-7.276 (SLES 9)
1349 2.4.21-40.EL (RHEL 3)
1350 2.6.12.6 vanilla (kernel.org)
1355 Bugzilla : 5719, 9635, 9792, 9684
1356 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1357 Details : If a server is under heavy load and cannot reply to new
1358 connection requests before the client resends the (re)connect,
1359 the connection handling code can behave badly if two service
1360 threads are concurrently handing separate (re)connections from
1361 the same client. Add better locking to the connection handling
1362 code, and ensure that only a single connection will be processed
1363 for a given client UUID, even if the lock is dropped.
1365 Severity : enhancement
1367 Description: add TCP zero-copy support to kernel
1368 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1369 sends if the hardware supports scatter-gather and checksumming.
1370 This allows socklnd to do client-write and server-read more
1371 efficiently and reduce CPU utilization from skbuf copying.
1374 Frequency : only if NFS exporting from client
1376 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1377 Details : If setting attributes on a file created under NFS that had
1378 never been opened it would be possible to oops the client
1379 if the file had no objects.
1382 Frequency : always for liblustre
1384 Description: liblustre client does MDS+OSTs setattr RPC for each write
1385 Details : When doing a write from a liblustre client, the client
1386 incorrectly issued an RPC to the MDS and each OST the file was
1387 striped over in order to update the timestamps. When writing
1388 with small chunks and many clients this could overwhelm the MDS
1389 with RPCs. In all cases it would slow down the write because
1390 these RPCs are unnecessary.
1392 Severity : enhancement
1394 Description: allow number of MDS service threads to be changed at module load
1395 Details : It is now possible to change the number of MDS service threads
1396 running. Adding "options mds mds_num_threads={N}" to the MDS's
1397 /etc/modprobe.conf will set the number of threads for the next
1398 time Lustre is restarted (assuming the "mds" module is also
1399 reloaded at that time). The default number of threads will
1400 stay the same, 32 for most systems.
1405 Description: OST crash if filesystem is unformatted or corrupt
1406 Details : If an OST is started on a device that has never been formatted
1407 or if the filesystem is corrupt and cannot even mount then the
1408 error handling cleanup routines would dereference a NULL pointer.
1413 Description: NULL pointer deref in llap_from_page.
1414 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1415 incorrect) mapping if the page was truncated/reclaimed while it was
1416 searched for. Check for this condition and skip such pages when
1417 doing readahead. Introduce extra check to llap_from_page() to
1418 verify page->mapping->host is non-NULL (so page is not anonymous).
1421 Frequency : Sometimes when using sys_sendfile
1423 Description: "page not covered by a lock" warnings from ll_readpage
1424 Details : sendfile called ll_readpage without right page locks present.
1425 Now we introduced ll_file_sendfile that does necessary locking
1426 around call to generic_file_sendfile() much like we do in
1430 Frequency : with certain MDS communication failures at client mount time
1432 Description: NULL pointer deref after failed client mount
1433 Details : a client connection request may delayed by the network layer
1434 and not be sent until after the PTLRPC layer has timed out the
1435 request. If the client fails the mount immediately it will try
1436 to clean up before the network times out the request. Add a
1437 reference from the request import to the obd device and delay
1438 the cleanup until the network drops the request.
1441 Frequency : occasionally during client (re)connect
1443 Description: assertion failure during client (re)connect
1444 Details : processing a client connection request may be delayed by the
1445 client or server longer than the client connect timeout. This
1446 causes the client to resend the connection request. If the
1447 original connection request is replied in this interval, the
1448 client may trip an assertion failure in ptlrpc_connect_interpret()
1449 which thought it would be the only running connect process.
1452 Frequency : only with obd_echo servers and clients that are rebooted
1454 Description: kernel BUG accessing uninitialized data structure
1455 Details : When running an obd_echo server it did not start the ping_evictor
1456 thread, and when a client was evicted an uninitialized data
1457 structure was accessed. Start the ping_evictor in the RPC
1458 service startup instead of the OBD startup.
1460 Severity : enhancement
1461 Bugzilla : 10193 (patchless)
1462 Description: Remove dependency on various unexported kernel interfaces.
1463 Details : No longer need reparent_to_init, exit_mm, exit_files,
1464 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1467 Frequency : rare (only users of deprecated and unsupported LDAP config)
1469 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1470 Details : LDAP apparently contains 'lustreName' attributes instead of
1471 'name'. A simple remapping of the name is sufficient.
1474 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1476 Description: waiting_locks_callback trips kernel BUG if client is evicted
1477 Details : Running with the dump_on_timeout debug flag turned on makes
1478 it possible that the waiting_locks_callback() can try to dump
1479 the Lustre kernel debug logs from an interrupt handler. Defer
1480 this log dumping to the expired_lock_main() thread.
1482 Severity : enhancement
1484 Description: Support NFS exporting on 2.6 kernels.
1485 Details : Implement non-rawops metadata methods for NFS server to use without
1486 changing NFS server code.
1489 Frequency : very rare (synthetic metadata workload only)
1491 Description: two racing renames might cause an MDS thread to deadlock
1492 Details : Running the "racer" program may cause one MDS thread to rename
1493 a file from being the source of a rename to being the target of
1494 a rename at exactly the same time that another thread is doing
1495 so, and the second thread has already enqueued these locks after
1496 doing a lookup of the target and is trying to relock them in
1497 order. Ensure that we don't try to re-lock the same resource.
1500 Frequency : only very large systems with liblustre clients
1502 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1503 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1504 all OSTs in parallel. This allows the eviction of stale liblustre
1505 clients to proceed much faster than if they were done in series,
1506 and also offers similar improvements for other set_info RPCs.
1511 Description: excessive CPU usage during initial read phase on client
1512 Details : During the initial read phase on a client, it would agressively
1513 retry readahead on the file, consuming too much CPU and impacting
1514 performance (since 1.4.5.8). Improve the readahead algorithm
1515 to avoid this, and also improve some other common cases (read
1516 of small files in particular, where "small" is files smaller than
1517 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1522 Description: MDS crash when receiving packet with unknown intent.
1523 Details : Do not LBUG in unknown intent case, just return -EFAULT
1525 Severity : enhancement
1526 Bugzilla : 9293, 9385
1527 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1528 Details : Do not serialize getattr (non-intent version) and statfs.
1531 Frequency : occasional, when OST network is overloaded/intermittent
1533 Description: client evicted by OST after bulk IO timeout
1534 Details : If a client sends a bulk IO request (read or write) the OST
1535 may evict the client if it is unresposive to its data GET/PUT
1536 request. This is incorrect if the network is overloaded (takes
1537 too long to transfer the RPC data) or dropped the OST GET/PUT
1538 request. There is no need to evict the client at all, since
1539 the pinger and/or lock callbacks will handle this, and the
1540 client can restart the bulk request.
1543 Frequency : Always when mmapping file with no objects
1545 Description: client crashes when mmapping file with no objects
1546 Details : Check that we actually have objects in a file before doing any
1547 operations on objects in ll_vm_open, ll_vm_close and
1553 Description: Request leak when working with deleted CWD
1554 Details : Introduce advanced request refcount tracking for requests
1555 referenced from lustre intent.
1557 Severity : Enhancement
1559 Description: Cache open file handles on client.
1560 Details : MDS now will return special lock along with openhandle, if
1561 requested and client is allowed to hold openhandle, even if unused,
1562 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1563 closing files for every read/write openration.
1565 Severity : Enhancement
1567 Description: Cache open negative dentries on client when possible.
1568 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1569 negative dentries on lock revocation.
1574 Description: Remounting a client read-only wasn't possible with a zconf mount
1575 Details : It wasn't possible to remount a client read-only with llmount.
1577 Severity : enhancement
1578 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1579 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1580 MPICH is included to add Lustre-specific ADIO interfaces.
1581 This is based closely on the UFS ADIO layer and only differs
1582 in file creation, in order to allow the OST striping to be set.
1583 This is user-contributed code and not supported by CFS.
1588 Description: extended inode attributes (immutable, append-only) work improperly
1589 when 2.4 and 2.6 kernels are used on client/server or vice versa
1590 Details : Introduce kernel-independent values for these flags.
1592 Severity : enhancement
1595 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1596 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1597 granularity. Now, Fractional MB granularity is be supported,
1598 this is very useful for low memory system.
1600 Severity : enhancement
1602 Description: Getattr by fid
1603 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1604 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1605 have locks and avoids getattr rpc requests for it on every stat.
1608 Frequency : Always, for filesystems larger than 2TB
1610 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1611 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1612 filesystems larger than 2TB to be mounted with mballoc enabled.
1615 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1617 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1618 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1619 filesystems larger than 2TB is mounted then the kernel will
1620 silently corrupt the start of the filesystem. CONFIG_LBD is
1621 enabled for all CFS-supported kernels, but the possibility of
1622 this happening with a modified kernel config exists.
1624 Severity : enhancement
1626 Description: add client O_DIRECT support for 2.6 kernels
1627 Details : It is now possible to do O_DIRECT reads and writes to files
1628 in the Lustre client mountpoint on 2.6 kernel clients.
1630 Severity : enhancement
1632 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1633 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1634 parallel, not waiting for response from every OST before sending
1635 a rpc to the next OST.
1640 Description: setattr vs write race when updating file timestamps
1641 Details : Client processes that update a file timestamp into the past
1642 right after writing to the file (e.g. tar) it is possible that
1643 the updated file modification time can be reset to the current
1644 time due to a race between processing the setattr and write RPC.
1646 Severity : enhancement
1648 Description: Bring 'lfs find' closer in line with regular Linux find.
1649 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1650 -print0 options and obtains all the needed info through the lustre
1653 Severity : enhancement
1655 Description: support up to 1024 configured devices on one node
1656 Details : change obd_dev array from statically allocated to dynamically
1657 allocated structs as they are first used to reduce memory usage
1662 Description: Flush dirty partially truncated pages during truncate
1663 Details : Immediatelly flush partially truncated pages in filter_setattr,
1664 this way we completely avoid having any pages in page cache on OST
1665 and can retire ugly workarounds during writes to flush such pages.
1670 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1671 Details : protect inode from truncation within vfs_unlink() context
1672 just take a reference before calling vfs_unlink() and release it
1673 when parent's i_sem is free.
1678 Description: last_id value checked outside lock on OST caused LASSERT failure
1679 Details : If there were multiple MDS->OST object precreate requests in
1680 flight, it was possible that the OST's last object id was checked
1681 outside a lock and incorrectly tripped an assertion. Move checks
1682 inside locks, and discard old precreate requests.
1685 Frequency : always, if extents are used on OSTs
1687 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1688 Details : OSTs using the extents format would not zero the high 16 bits of
1689 the index physical block number. This is not a problem for any
1690 OST filesystems smaller than 16TB, and no kernels support ext3
1691 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1692 new/modified files) and can be fixed for existing filesystems
1693 with e2fsprogs-1.39-cfs1.
1698 Description: import connection selection may be incorrect if timer wraps
1699 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1700 import connections to be ignored if the 32-bit jiffies counter
1701 wraps. Use a 64-bit jiffies counter.
1704 Frequency : very large clusters immediately after boot
1706 Description: LNET request buffers exhausted under heavy short-term load
1707 Details : If a large number of client requests are generated on a service
1708 that has previously never seen so many requests it is possible
1709 that the request buffer growth cannot keep up with the spike in
1710 demand. Instead of dropping incoming requests, they are held in
1711 the LND until the RPC service can accept more requests.
1714 Frequency : Sometimes during replay
1716 Description: Assertion failure in ll_local_open after replay.
1717 Details : If replay happened on an open request reply before we were able
1718 to set replay handler, reply will become not swabbed tripping the
1719 assertion in ll_local_open. Now we set the handler right after
1720 recognising of open request
1723 Frequency : very rare
1725 Description: kernel reports "badness in vsnprintf"
1726 Details : Reading from the "recovery_status" /proc file in small chunks
1727 may cause a negative length in lprocfs_obd_rd_recovery_status()
1728 call to vsnprintf() (which is otherwise harmless). Exit early
1729 if there is no more space in the output buffer.
1731 Severity : enhancement
1733 Description: clear OBD RPC statistics by writing to them
1734 Details : It is now possible to clear the OBD RPC statistics by writing
1735 to the "stats" file.
1740 Description: Client mtime is not the same on different clients after utimes
1741 Details : In some cases, the client was using the utimes() syscall on
1742 a file cached on another node. The clients now validate the
1743 ctime from the MDS + OSTs to determine which one is right.
1748 Description: Inability to activate failout mode
1749 Details : lconf script incorrectly assumed that in pythong string's numeric
1750 value is used in comparisons.
1753 Frequency : always with multiple stripes per file
1755 Description: Inefficient object allocation for mutli-stripe files
1756 Details : When selecting which OSTs to stripe files over, for files with
1757 a stripe count that divides evenly into the number of OSTs,
1758 the MDS is always picking the same starting OST for each file.
1759 Return the OST selection heuristic to the original design.
1764 Description: mount failures may take full timeout to return an error
1765 Details : Under some heavy load conditions it is possible that a
1766 failed mount can wait for the full obd_timeout interval,
1767 possibly several minutes, before reporting an error.
1768 Instead return an error as soon as the status is known.
1770 Frequency : quota enabled and large files being deleted
1772 Description: releasing more than 4GB of quota at once hangs OST
1773 Details : If a user deletes more than 4GB of files on a single OST it
1774 will cause the OST to spin in an infinite loop. Release
1775 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1780 Description: statfs data retrieved from /proc may be stale or zero
1781 Details : When reading per-device statfs data from /proc, in the
1782 {kbytes,files}_{total,free,avail} files, it may appear
1783 as zero or be out of date.
1786 Frequency : systems with MD RAID1 external journal devices
1788 Description: lconf's call to blkid is confused by RAID1 journal devices
1789 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1790 of returning all block devices that match the journal UUID.
1793 Frequency : always, for aggregate stripe size over 4GB
1795 Description: assertion fails when trying to use 4GB stripe size
1796 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1797 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1800 Frequency : always on ppc64
1802 Description: the first write on an ext3 filesystem with mballoc got stuck
1803 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1804 perform endianness conversion.
1806 ------------------------------------------------------------------------------
1808 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1810 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1811 INTEROPERATE with older versions automatically. Please read the
1812 user documentation before upgrading any part of a live system.
1813 * WARNING: Lustre networking configuration changes are required with
1814 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1817 * Support for kernels:
1818 2.6.9-22.0.2.EL (RHEL 4)
1819 2.6.5-7.244 (SLES 9)
1820 2.6.12.6 vanilla (kernel.org)
1823 Severity : enhancement
1824 Bugzilla : 7981/8208
1825 Description: Introduced Lustre Networking (LNET)
1826 Details : LNET is new networking infrastructure for Lustre, it includes
1827 a reorganized network configuration mode (see the user
1828 documentation for full details) as well as support for routing
1829 between different network fabrics. Lustre Networking Devices
1830 (LNDs) for the supported network fabrics have also been
1831 created for this new infrastructure.
1833 Severity : enhancement
1834 Description: Introduced Access control lists
1835 Details : clients can set ACLs on files and directories in order to have
1836 more fine-grained permissions than the standard Unix UGO+RWX.
1837 The MDS must be started with the "-o acl" mount option.
1839 Severity : enhancement
1840 Description: Introduced filesystem quotas
1841 Details : Administrators may now establish per-user quotas on the
1844 Severity : enhancement
1846 Description: Configuration change for the XT3
1847 The PTLLND is now used to run Lustre over Portals on the XT3
1848 The configure option(s) --with-cray-portals are no longer used.
1849 Rather --with-portals=<path-to-portals-includes> is used to
1850 enable building on the XT3. In addition to enable XT3 specific
1851 features the option --enable-cray-xt3 must be used.
1856 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1857 Details : A race between allocating a new llcd and re-getting the llcd_lock
1858 allowed another thread to grab newly-allocated llcd.
1860 Severity : enhancement
1862 Description: 2.6 OST async journal commit and locking fix to improve performance
1863 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1864 2.6 kernels are now async as they already were in 2.4 kernels so
1865 that they can commit concurrently with the network bulk transfer.
1866 For block-allocated files the filter allocation semaphore is held
1867 to avoid filesystem fragmentation during allocation. BKL lock
1868 removed for 2.6 xattr operations where it is no longer needed.
1873 Description: lconf incorrectly determined whether two IP networks could talk
1874 Details : In some more complicated routing and multiple-network
1875 configurations, lconf will avoid trying to make a network
1876 connection to a disjoint part of the IP space. It was doing the
1877 math incorrectly for one set of cases.
1882 Description: Fix for potential infinite loop processing records in an llog.
1883 Details : If an llog record is corrupted/zeroed, it is possible to loop
1884 forever in llog_process(). Validate the llog record length
1885 and skip the remainder of the block on error.
1888 Frequency : occasional (liblustre only)
1890 Description: liblustre could not open files whose last component is a symlink
1891 Details : sysio_path_walk() would incorrectly pass the open intent to
1892 intermediate path components.
1895 Frequency : rare (liblustre only with non-standard tuning)
1896 Bugzilla : 7201 (7350)
1897 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1898 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1899 after it has been released, only do this for referenced locks
1901 Severity : enhancement
1903 Description: specifying an (invalid) directory default stripe_size of -1
1904 would reset the directory default striping
1905 Details : stripe_size -1 was used internally to signal directory stripe
1906 removal, now use "all default" to signal dir stripe removal
1907 as a directory striping of "all default" is not useful
1910 Frequency : common for large clusters running liblustre clients
1912 Description: doing an ls when liblustre clients are running is slow
1913 Details : sending a glimpse AST to a liblustre client waits for every AST
1914 to time out, as liblustre clients will not respond. Since they
1915 cannot cache data we refresh the OST lock LVB from disk instead.
1917 Severity : enhancement
1919 Description: doing an ls at the same time as file IO can be slow
1920 Details : enqueue and other "small" requests can be blocked behind many
1921 large IO requests. Create a new OST IO portal for non-IO
1922 requests so they can be processed faster.
1925 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1927 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1928 created file with mode 000
1929 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1930 allow an ftruncate() as a non-root user to a file with mode 000.
1931 Lustre now respects this flag to disable mode checks when
1932 truncating a file owned by the user
1935 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1938 Description: Revoking locks from clients that went dead or catatonic might take
1940 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1941 cancellation of such locks instant on servers without waiting for
1942 any reply from clients. Clients drops these locks when cancel
1943 notification from server is received without replying.
1946 Frequency : liblustre-only, when liblustre client dies or becomes busy
1948 Description: Doing ls on Linux clients can take a long time with active
1950 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1951 granting such locks to it in the first place if possible. Locks
1952 are taken by proxy on the OST during the read or write and
1953 dropped immediately afterward. Add connect flags handling, do
1954 not grant locks to liblustre clients for glimpse ASTs.
1956 Severity : enhancement
1958 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1959 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1960 sequential reads and behaves badly with applications doing random
1961 reads. Improve it by reading ahead at least read region, and
1962 avoiding excessive large RPC for small reads.
1964 Severity : enhancement
1966 Description: Creating more than 1000 files for a single job may cause a load
1967 imbalance on the OSTs if there are also a large number of OSTs.
1968 Details : qos_prep_create() uses an OST index reseed value that is an
1969 even multiple of the number of available OSTs so that if the
1970 reseed happens in the middle of the object allocation it will
1971 still utilize the OSTs as uniformly as possible.
1976 Description: OST or MDS may oops in ping_evictor_main()
1977 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1978 but doesn't restart at beginning of obd_exports_timed list
1981 Severity : enhancement
1983 Description: improve by-nid export eviction on the MDS and OST
1984 Details : allow multiple exports with the same NID to be evicted at one
1985 time without re-searching the exports list.
1988 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1990 Description: MDS may oops in groups_free()
1991 Details : in rare race conditions a newly allocated group_info struct is
1992 freed again, and this can be NULL. The 2.4 compatibility code
1993 for groups_free() checked for a NULL pointer, but 2.6 did not.
1996 Frequency : common for liblustre clients doing little filesystem IO
1997 Bugzilla : 9352, 7313
1998 Description: server may evict liblustre clients accessing contended locks
1999 Details : if a client is granted a lock or receives a completion AST
2000 with a blocking AST already set it would not reply to the AST
2001 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2004 Frequency : lfs setstripe, only systems with more than 160 OSTs
2006 Description: unable to set striping with a starting offset beyond OST 160
2007 Details : llapi_create_file() incorrectly limited the starting stripe
2008 index to the maximum single-file stripe count.
2011 Frequency : LDAP users only
2013 Description: lconf did not handle in-kernel recovery with LDAP properly
2014 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2016 Severity : enhancement
2018 Description: bind OST threads to NUMA nodes to improve performance
2019 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2020 node and do their allocations there to localize memory access
2022 Severity : enhancement
2024 Description: llmount can determine client NID directly from Myrinet (GM)
2025 Details : the client NID code from gmnalnid was moved directly into
2026 llmount, removing the need to use this or specifying the
2027 client NID explicitly when mounting GM clients with zeroconf
2030 Frequency : if client is started with down MDS
2032 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2033 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2034 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2035 ptlrpc_import_delayed_req() because we didn't check if the
2036 request was interrupted, and we also didn't break out of the
2037 event loop if there was no timeout
2042 Description: data loss during non-page-aligned writes to a single file from
2043 both multiple nodes and multiple threads on one node at same time
2044 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2045 inconsistency led to false short-reads, that were cached and later
2046 used by ->prepare_write() to fill in partially written page,
2047 leading to data loss.
2050 Frequency : always, if lconf --abort_recovery used
2052 Description: lconf --abort_recovery fails with 'Operation not supported'
2053 Details : lconf was attempting to abort recovery on the MDT device and not
2056 Severity : enhancement
2058 Description: remove cleanup logs
2059 Details : replace lconf-generated cleanup logs with lustre internal
2060 cleanup routines. Eliminates the need for client-cleanup and
2063 Severity : enhancement
2065 Description: add support for EAs (user and system) on lustre filesystems
2066 Details : it is now possible to store extended attributes in the Lustre
2067 client filesystem, and with the user_xattr mount option it
2068 is possible to allow users to store EAs on their files also
2070 Severity : enhancement
2072 Description: Add possibility (config option) to show minimal available OST free
2074 Details : When compiled with --enable-mindf configure option, statfs(2)
2075 (and so, df) will return least minimal free space available from
2076 all OSTs as amount of free space on FS, instead of summary of
2077 free spaces of all OSTs.
2079 Severity : enhancement
2081 Description: do not expand extent locks acquired on OST-side
2082 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2083 by server: they are not cached anyway.
2086 Frequency : when mmap is used/binaries executed from Lustre
2088 Description: Unmmap pages before throwing them away from read cache.
2089 Details : llap_shrink cache now attempts to unmap pages before discarding
2090 them (if unmapping failed - do not discard). SLES9 kernel has
2091 extra checks that trigger if this unmapping is not done first.
2096 Description: lconf didn't resolve symlinks before checking to see whether a
2097 given mountpoint was already in use
2100 Frequency : when migrating failover services
2101 Bugzilla : 6395, 9514
2102 Description: When migrating a subset of services from a node (e.g. failback
2103 from a failover service node) the remaining services would
2104 time out and evict clients.
2105 Details : lconf --force (implied by --failover) sets the global obd_timeout
2106 to 5 seconds in order to quickly disconnect, but this caused
2107 other RPCs to time out too quickly. Do not change the global
2108 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2110 Severity : enhancement
2111 Frequency : if MDS is started with down OST
2112 Bugzilla : 9439,5706
2113 Description: Allow startup/shutdown of an MDS without depending on the
2114 availability of the OSTs.
2115 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2116 Add appropriate locking and lov-osc refcounts for safe
2117 cleaning. Add osc abort_inflight calls in case the
2118 synchronize never started.
2121 Frequency : occasional (Cray XT3 only)
2123 Description: root not authorized to access files in CRAY_PORTALS environment
2124 Details : The client process capabilities were not honoured on the MDS in
2125 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2126 been accessed by an authorized user then root was able to access
2127 the file on the local client also. The root user capabilities
2128 are now allowed on the MDS, as this environment has secure UID.
2131 Frequency : occasional
2133 Description: ldiskfs "too long searching" message happens too often
2134 Details : A debugging message (otherwise harmless) prints too often on
2135 the OST console. This has been reduced to only happen when
2136 there are fragmentation problems on the filesystem.
2141 Description: Division by zero in statfs when all OSCs are inactive
2142 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2143 lov_statfs divides by value returned by lov_get_stripecnt().
2147 Bugzilla : 9489, 3273
2148 Description: First write from each client to each OST was only 4kB in size,
2149 to initialize client writeback cache, which caused sub-optimal
2150 RPCs and poor layout on disk for the first writen file.
2151 Details : Clients now request an initial cache grant at (re)connect time
2152 and so that they can start streaming writes to the cache right
2153 away and always do full-sized RPCs if there is enough data.
2154 If the OST is rebooted the client also re-establishes its grant
2155 so that client cached writes will be honoured under the grant.
2160 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2161 Details : Now I/O RPCs go to different portal number and (presumably) fast
2162 lock requests (and glimses) and other RPCs get their own service
2163 threads pool that should be able to service those RPCs
2166 Severity : enhancement
2168 Description: Ability to exchange lustre version between client and servers and
2169 issue warnings at client side if client is too old. Also for
2170 liblustre clients there is ability to refuse connection of too old
2172 Details : New 'version' field is added to connect data structure that is
2173 filled with version info. That info is later checked by server and
2177 Frequency : rare, liblustre only.
2178 Bugzilla : 9296, 9581
2179 Description: Two simultaneous writes from liblustre at offset within same page
2180 might proceed at the same time overwriting eachother with stale
2182 Details : I/O lock withing llu_file_prwv was released too early, before data
2183 actually was hitting the wire. Extended lock-holding time until
2184 server acknowledges receiving data.
2187 Frequency : extremely rare. Never observed in practice.
2189 Description: avoid generating lustre_handle cookie of 0.
2190 Details : class_handle_hash() generates handle cookies by incrementing
2191 global counter, and can hit 0 occasionaly (this is unlikely, but
2192 not impossible, because initial value of cookie counter is
2193 selected randonly). Value of 0 is used as a sentinel meaning
2194 "unassigned handle" --- avoid it. Also coalesce two critical
2195 sections in this function into one.
2197 Severity : enhancement
2199 Description: allow liblustre clients to delegate truncate locking to OST
2200 Details : To avoid overhead of locking, liblustre client instructs OST to
2201 take extent lock in ost_punch() on client's behalf. New connection
2202 flag is added to handle backward compatibility.
2204 Severity : enhancement
2205 Bugzilla : 4928, 7341, 9758
2206 Description: allow number of OST service threads to be specified
2207 Details : a module parameter allows the number of OST service threads
2208 to be specified via "options ost ost_num_threads={N}" in the
2209 OSS's /etc/modules.conf or /etc/modprobe.conf.
2213 Bugzilla : 6146, 9635, 9895
2214 Description: servers crash with bad pointer in target_handle_connect()
2215 Details : In rare cases when a client is reconnecting it was possible that
2216 the connection request was the last reference for that export.
2217 We would temporarily drop the export reference and get a new
2218 one, but this may have been the last reference and the export
2219 was just destroyed. Get new reference before dropping old one.
2221 Severity : enhancement
2222 Frequency : if client is started with failover MDS
2224 Description: Allow multiple MDS hostnames in the mount command
2225 Details : Try to read the configuration from all specified MDS
2226 hostnames during a client mount in case the "primary"
2229 Severity : enhancement
2231 Description: Stop sending data to evicted clients as soon as possible.
2232 Details : Check if the client we are about to send or are sending data to
2233 was evicted already. (Check is done every second of waiting,
2234 for which l_wait_event interface was extended to allow checking
2235 of exit condition at specified intervals).
2238 Frequency : rare, normally only when NFS exporting is done from client
2240 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2242 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2245 Frequency : always, if srand() is called before liblustre initialization
2247 Description: Liblustre uses system PRNG disturbing its usage by user application
2248 Details : Introduce internal to lustre fast and high-quality PRNG for
2249 lustre usage and make liblustre and some other places in generic
2250 lustre code to use it.
2252 Severity : enhancement
2253 Bugzilla : 9477, 9557, 9870
2254 Description: Verify that the MDS configuration logs are updated when xml is
2255 Details : Check if the .xml configuration logs are newer than the config
2256 logs stored on the MDS and report an error if this is the case.
2257 Request --write-conf, or allow starting with --old_conf.
2259 Severity : enhancement
2261 Description: Handle symlinks in the path when checking if Lustre is mounted.
2262 Details : Resolve intermediate symlinks when checking if a client has
2263 mounted a filesystem to avoid duplicate client mounts.
2268 Description: lconf can hit an error exception but still return success.
2269 Details : The lconf command catches the Command error exception at the top
2270 level script context and will exit with the associated exit
2271 status, but doesn't ensure that this exit status is non-zero.
2276 Description: failure of ptlrpc thread startup can cause oops
2277 Details : Starting a ptlrpc service thread can fail if there are a large
2278 number of threads or the server memory is very fragmented.
2279 Handle this without oopsing.
2282 Frequency : always, only if liblustre and non-default acceptor port was used
2284 Description: liblustre cannot connect to servers with non-default acceptor port
2285 Details : tcpnal_set_default_params() was not called and was therefore
2286 ignoring the environment varaible TCPNAL_PORT, as well as other
2287 TCPNAL_ environment variables
2292 Description: two objects could be created on the same OST for a single file
2293 Details : If an OST is down, in some cases it was possible to create two
2294 objects on a single OST for a single file. No problems other
2295 than potential performance impact and spurious error messages.
2299 Bugzilla : 5681, 9562
2300 Description: Client may oops in ll_unhash_aliases
2301 Details : Client dcache may become inconsistent in race condition.
2302 In some cases "getcwd" can fail if the current directory is
2308 Description: Inode refcounting problems in NFS export code
2309 Details : link_raw functions used to call d_instantiate without obtaining
2310 extra inode reference first.
2314 Bugzilla : 9942, 9903
2315 Description: Referencing freed requests leading to crash, memleaks with NFS.
2316 Details : We used to require that call to ll_revalidate_it was always
2317 followed by ll_lookup_it. Also with revalidate_special() it is
2318 possible to call ll_revalidate_it() twice for the same dentry
2319 even if first occurence returned success. This fix changes semantic
2320 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2321 reference on a request referred from the intent.
2322 ll_intent_release() then releases such a request.
2325 Frequency : rare, normally benchmark loads only
2327 Description: unlinked inodes were kept in memory on the client
2328 Details : If a client is repeatedly creating and unlinking files it
2329 can accumulate a lot of stale inodes in the inode slab cache.
2330 If there is no other client load running this can cause the
2331 client node to run out of memory. Instead flush old inodes
2332 from client cache that have the same inode number as a new inode.
2335 Frequency : SLES9 2.6.5 kernel and long filenames only
2336 Bugzilla : 9969, 10379
2337 Description: utime reports stale NFS file handle
2338 Details : SLES9 uses out-of-dentry names in some cases, which confused
2339 the lustre dentry revalidation. Change it to always use the
2343 Frequency : rare, unless heavy write-truncate concurrency is continuous
2344 Bugzilla : 4180, 6984, 7171, 9963, 9331
2345 Description: OST becomes very slow and/or deadlocked during object unlink
2346 Details : filter_destroy() was holding onto the parent directory lock
2347 while truncating+unlinking objects. For very large objects this
2348 may block other threads for a long time and slow overall OST
2349 responsiveness. It may also be possible to get a lock ordering
2350 deadlock in this case, or run out of journal credits because of
2351 the combined truncate+unlink. Solution is to do object truncate
2352 first in one transaction without parent lock, and then do the
2353 final unlink in a new transaction with the parent lock. This
2354 reduces the lock hold time dramatically.
2357 Frequency : rare, 2.4 kernels only
2359 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2360 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2361 before unmounting their filesystem in order to avoid deadlock.
2362 The kernel_locked() function in 2.4 kernels only checks whether
2363 the kernel lock is held, not whether it is this process that is
2364 holding it as 2.6 kernels do.
2369 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2370 Details : The client ptlrpc code may be trying to reconnect to a down
2371 server before a previous connection attempt has timed out.
2372 Increase the reconnect interval to be longer than the connection
2373 timeout interval to avoid sending duplicate connections to
2377 Frequency : echo_client brw_test command
2379 Description: fix echo_client to work with OST preallocated code
2380 Details : OST preallocation code (5137) didn't take echo_client IO path
2381 into account: echo_client calls filter methods outside of any
2382 OST thread and, hence, there is no per-thread preallocated
2383 pages and buffers to use. Solution: hijack pga pages for IO. As
2384 a byproduct, this avoids unnecessary data copying.
2388 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2389 Description: Client can oops in mdc_commit_close() after open replay
2390 Details : It was possible for the MDS to return an open request with no
2391 transaction number in mds_finish_transno() if the client was
2392 evicted, but without actually returning an error. Clients
2393 would later try to replay that open and may trip an assertion
2394 Simplify the client close codepath, and always return an error
2395 from the MDS in case the open is not successful.
2398 Frequency : rare, 2.6 OSTs only
2400 Description: OST may deadlock under high load on fragmented files
2401 Details : If there was a heavy load and highly-fragmented OST filesystems
2402 it was possible to have all the OST threads deadlock waiting on
2403 allocation of biovecs, because the biovecs were not released
2404 until the entire RPC IO was completed. Instead, release biovecs
2405 as soon as they are complete to ensure forward IO progress.
2407 Severity : enhancement
2409 Description: Support for specifying external journal device at mount
2410 Details : If an OST or MDS device is formatted with an external journal
2411 device, this device major/minor is stored in the ext3 superblock
2412 and may not be valid for failover. Allow detecting and
2413 specifying the external journal at mount time.
2418 Description: Mounting an MDS with pending unlinked files may cause oops
2419 Details : target_finish_recovery() calls mds_postrecov() which returned
2420 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2421 considers this an error and immediately begins cleaning up the
2422 lov, just after starting the mds_lov process
2424 Severity : enhancement
2426 Description: Implement 'lfs df' to report actual free space on per-OST basis
2427 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2428 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2429 usage of objects; '-h' to report in human readable format.
2431 ------------------------------------------------------------------------------
2433 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2440 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2441 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2442 at startup during orphan unlinking. Instead, skip buddy update
2443 before setup, as it will be regenerated after recovery is complete.
2448 Description: If an OST is inactive, its locks might reference stale inodes.
2449 Details : lov_change_cbdata() must iterate over all namespaces, even if
2450 they are inactive to clear inode references from the lock.
2452 Severity : enhancement
2453 Frequency : occasional, if non-standard max_dirty_mb used
2455 Description: Client will block write RPCs if not enough grant
2456 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2457 then the client will block writes while waiting for another RPC
2458 to complete instead of consuming its dirty limit. With change
2459 we get improved performance when max_dirty_mb is small.
2461 Severity : enhancement
2462 Bugzilla : 3389, 6253
2463 Description: Add support for supplementary groups on the MDS.
2464 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2465 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2466 lookups for user supplementary groups into a cache.
2470 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2471 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2472 for liblustre O_EXCL is enforced by MDS.
2476 Description: readdir never returns NULL in liblustre.
2477 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2478 dentry in struct dirent.
2482 Frequency : liblustre only, depends on application IO pattern
2483 Description: liblustre clients evicted if not contacting servers
2484 Details : Don't put liblustre clients into the ping_evictor list, so
2485 they will not be evicted by the pinger ever.
2487 Severity : enhancement
2489 Description: Add ability to evict clients by NID from MDS.
2490 Details : By echoing "nid:$NID" string into
2491 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2492 $NID would be instantly evicted from this MDS and from all active
2493 OSTs connected to it.
2497 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2498 Details : lookup_it_finish() used to query file size from OSTs that was not
2503 Description: service threads change working directory to that of init
2504 Details : Starting lustre service threads may pin the working directory
2505 of the parent thread, making that filesystem busy. Threads
2506 now change to the working directory of init to avoid this.
2510 Frequency : during shutdown only
2511 Description: shutdown with a failed MDS or OST can cause unmount to hang
2512 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2517 Frequency : 2.6 only
2518 Description: chmod/chown may include an extra supplementary group
2519 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2520 supplementary group and if none is specified this is used.
2523 Bugzilla : 5479 (6816)
2525 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2526 Details : If lookup is in progress on a file that is unlinked we might try
2527 to revalidate the inode and fail in revalidate after lookup is
2528 complete and ll_file_open() enqueues the open again but
2529 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2532 Frequency : always, if lconf --abort_recovery used
2534 Description: lconf --abort_recovery fails with 'Operation not supported'
2535 Details : lconf was attempting to abort recovery on the MDT device and not
2538 ------------------------------------------------------------------------------
2540 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2545 Frequency : rare (only unsupported configurations with a node running as an
2547 Bugzilla : 6514, 5137
2548 Description: Mounting a Lustre file system on a node running as an OST could
2550 Details : OSTs now preallocates memory needed to write out data at
2551 startup, instead of when needed, to avoid having to
2552 allocate memory in possibly low memory situations.
2553 Specifically, if the file system is mounted on on OST,
2554 memory pressure could force it to try to write out data,
2555 which it needed to allocate memory to do. Due to the low
2556 memory, it would be unable to do so and the node would
2557 become unresponsive.
2559 Severity : enhancement
2561 Description: Addition of lconf --service command line option
2562 Details : lconf now accepts a '--service <arg>' option, which is
2563 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2565 Severity : enhancement
2567 Description: Failover mode is now the default for OSTs.
2568 Details : By default, OSTs will now run in failover mode. To return to
2569 the old behaviour, add '--failout' to the lmc line for OSTs.
2571 Severity : enhancement
2573 Description: Health checks are now provided for MDS and OSTs
2574 Details : Additional detailed health check information on MSD and OSTs
2575 is now provided through the procfs health_check value.
2578 Frequency : occasional, depends on IO load
2580 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2581 after numerous create/delete cycles
2582 Details : The ext3 inode allocation policy would not allocate new inodes
2583 very well on the OSTs because there are no new directories
2584 being created. Instead we look for groups with free space if
2585 the parent directories are nearly full.
2590 Description: Network or server problems during mount may cause partially
2591 mounted clients instead of returning an error.
2592 Details : The config llog parsing code may overwrite the error return
2593 code during mount error handling, returning success instead
2599 Description: MDS can fail to allocate large reply buffers
2600 Details : After long uptimes the MDS can fail to allocate large reply
2601 buffers (e.g. zconf client mount config records) due to memory
2602 fragmentation or consumption by the buffer cache. Preallocate
2603 some large reply buffers so that these replies can be sent even
2604 under memory pressure.
2608 Frequency : rare (liblustre)
2609 Description: fsx running with liblustre complained that using truncate() to
2610 extend the file doesn't work. This patch corrects that issue.
2611 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2612 ATTR_SIZE and lsm use in llu_setattr_raw.
2616 Frequency : rare, only 2.6 kernels
2617 Description: Unusual file access patterns on the MDS may result in inode
2618 data being lost in very rare circumstances.
2619 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2620 code in the 2.6 kernel caused the inode and/or EA data not to
2621 be read from disk, causing single-file corruption.
2625 Frequency : rare, only 2.6 filesystems using extents
2626 Description: Heavy concurrent write and delete load may cause data corruption.
2627 Details : It was possible under high-load situations to have an extent
2628 metadata block in the block device cache from a just-unlinked
2629 file overwrite a newly-allocated data block. We now unmap any
2630 metadata buffers that alias just-allocated data blocks.
2634 Frequency : filesystems with default stripe_count larger than 77
2635 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2636 Details : lconf specifies an inode size of 4096 bytes when the default
2637 stripe_count is larger than 77. This conflicts with the default
2638 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2639 this case to avoid pinning too much memory for large EAs.
2641 ------------------------------------------------------------------------------
2643 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2648 Frequency : rare (extremely heavy IO load with hundreds of clients)
2650 Description: Client is evicted, gets IO error writing to file
2651 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2652 caused clients to be evicted for AST timeouts. The fixes in
2653 bug 5192 mean we no longer need to have such short AST timeouts
2654 so ldlm_timeout has been increased.
2657 Frequency : occasional during --force or --failover shutdown under load
2658 Bugzilla : 5949, 4834
2659 Description: Server oops/LBUG if stopped with --force or --failover under load
2660 Details : a collection of import/export refcount and cleanup ordering
2661 issues fixed for safer force cleanup
2664 Frequency : only filesystems larger than 120 OSTs
2665 Bugzilla : 5990, 6223
2666 Description: lfs getstripe would oops on a very large filesystem
2667 Details : lov_getconfig used kfree on vmalloc'd memory
2670 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2671 Bugzilla : 6242, 6243
2672 Description: reading from files that had been truncated to a non-zero size
2673 but never opened returned no data
2674 Details : ll_file_read() reads zeros from no-object files to EOF
2679 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2680 Details : There's little harm in aborting MDS/OSS recovery and letting it
2681 try again, so I removed the LASSERT and return an error instead.
2683 Severity : enhancement
2685 Description: New debugging infrastructure for tracking down data corruption
2686 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2687 (b) cover more of the client-side code path, and (c) try to narrow
2688 down where problems occurred
2692 Bugzilla : 3819, 4364, 4397, 6313
2693 Description: Racing close and eviction MDS could cause assertion in mds_close
2694 Details : It was possible to get multiple mfd references during close and
2695 client eviction, leading to one thread referencing a freed mfd.
2697 Severity: : enhancement
2698 Bugzilla : 3262, 6359
2699 Description: Attempts to reconnect to servers are now more aggressive.
2700 Details : This builds on the enhanced upcall-less recovery that was added
2701 in 1.4.2. When trying to reconnect to servers, clients will
2702 now try each server in the failover group every 10 seconds. By
2703 default, clients would previously try one server every 25 seconds.
2708 Description: After recovery, certain operations trigger a failed
2709 assertion on a client.
2710 Details : Failing over an mds, using lconf -d --failover, while a
2711 client was doing a readdir() call would cause the client to
2712 LBUG after recovery completed and the readdir() was resent.
2714 Severity : enhancement
2716 Description: Default groups are now added by lconf
2717 Details : You can now run lconf --group <servicename> without having to
2718 manually add groups with lmc.
2721 Frequency : occasional
2723 Description: Nodes with an elan id of 0 trigger a failed assertion
2726 Frequency : always when accessing e.g. tty/console device nodes
2728 Description: tty and some other devices nodes cannot be used on lustre
2729 Details : file's private_data field is used by device data and lustre
2730 values in there got lost. New field was added to struct file to
2731 store fs-specific private data.
2734 Frequency : when exporting Lustre via NFS
2736 Description: NFSD failed occasionally when looking up a path component
2737 Details : NFSD is looking up ".." which was broken in ext3 directories
2738 that had grown large enough to become hashed.
2741 Frequency : Clusters with multiple interfaces not on the same subnet
2743 Description: Nodes will repeatedly try to reconnect to an interface which it
2744 cannot reach and report an error to the log.
2745 Details : Extra peer list entries will be created by lconf with some peers
2746 unreachable. lconf now validates the peer before adding it.
2749 Frequency : Only if a default stripe is set on the filesystem root.
2751 Description: Setting a default stripe on the filesystem root prevented the
2752 filesystem from being remounted.
2753 Details : The client was sending extra request flags in the root getattr
2754 request and did not allocate a reply buffer for the dir EA.
2757 Frequency : occasional, higher if lots of files are accessed by one client
2758 Bugzilla : 6159, 6097
2759 Description: Client trips assertion regarding lsm mismatch/magic
2760 Details : While revalidating inodes the VFS looks up inodes with ifind()
2761 and in rare cases can find an inode that is being freed.
2762 The ll_test_inode() code will free the lsm during ifind()
2763 when it finds an existing inode and then the VFS later attaches
2764 this free lsm to a new inode.
2768 Bugzilla : 6422, 7030
2769 Description: MDS deadlock between mkdir and client eviction
2770 Details : Creating a new file via mkdir or mknod (starting a transaction
2771 and getting the ns lock) can deadlock with client eviction
2772 (gets ns lock and trying to finish a synchronous transaction).
2775 Frequency : occasional
2776 Description: While starting a server, the fsfilt_ext3 module could not be
2778 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2779 kernels. Previously, lconf would still use the ext3 name
2780 when trying to load modules. Now, it will correctly use
2781 ext3 on 2.4 and ldiskfs on 2.6.
2783 Severity : enhancement
2784 Description: The default stripe count has been changed to 1
2785 Details : The interpretation of the default stripe count (0, to lfs
2786 or lmc) has been changed to mean striping across a single
2787 OST, rather than all available. For general usage we have
2788 found a stripe count of 1 or 2 works best.
2790 Severity : enhancement
2791 Description: Add support for compiling against Cray portals.
2792 Details : Conditional compiling for some areas that are different
2796 Frequency : occasional
2797 Bugzilla : 6409, 6834
2798 Description: Creating files with an explicit stripe count may lead to
2799 a failed assertion on the MDS
2800 Details : If some OSTs are full or unavailable, creating files may
2801 trigger a failed assertion on the MDS. Now, Lustre will
2802 try to use other servers or return an error to the
2806 Frequency : occasional
2808 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2809 caused lock timeout during readahead (since 1.4.2).
2810 Details : Processes doing readahead might match a lock that hasn't been
2811 granted yet if there are overlapping and conflicting lock
2812 requests. The readahead process waits on ungranted lock
2813 (original lock is CBPENDING), while OST waits for that process
2814 to cancel CBPENDING read lock and eventually evicts client.
2816 Severity : enhancement
2818 Description: Initial enabling of flock support for clients
2819 Details : Implements fcntl advisory locking and file status functions.
2820 This feature is provided as an optional mount flag (default
2821 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2822 locking are implemented yet, and those that are are not guaranteed
2823 to be completely correct in production environments.
2824 mount -t lustre -o [flock|noflock] ...
2827 Frequency : occasional
2829 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2830 assertion in the ext3 JBD (journaling) layer.
2831 Details : The b_committed_data struct is protected by the big kernel lock
2832 in 2.4 kernels, serializing journal_commit_transaction() and
2833 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2834 access to this struct. In 2.6 kernels there is finer grained
2835 locking to improve SMP performance of the JBD layer.
2839 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2844 Description: merge and cleanup kernel patches.
2847 -----------------------------------------------------------------------------
2849 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2851 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2852 versions of Lustre. Please update all servers and clients to
2853 version 1.4.2 or later at the same time. You must also run
2854 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2855 to update the configuration logs.
2857 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2858 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2859 - swab llog records of type '0' so we get proper header size/idx (5861)
2860 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2861 - fix rename of one directory over another leaking an inode (5953)
2862 - avoid SetPageDirty on 2.6 (5981)
2863 - don't re-add just-being-destroyed locks to the waiting list (5653)
2864 - when creating new directories, inherit the parent's custom
2865 striping settings if present parent (3048)
2866 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2867 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2868 - don't hold client locks on temporary worklist from l_lru (5666)
2869 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2870 - automatically evict dead clients (5921)
2871 - Update file size properly in create+truncate+fstat case (6196)
2872 - Do not unhash mountpoint dentries, do not allow removal of
2874 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2875 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2876 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2877 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2878 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2879 - Correctly handle reads of files with no objects (6243)
2880 - lctl recover will also mark a device active if deactivate used (5933)
2882 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2883 - allow --write-conf on an MDS with different nettype than client (5619)
2884 - don't write config llogs to MDS for mounts not from that MDS (5617)
2885 - lconf should create multiple TCP connections from a client (5201)
2886 - init scripts are now turned off by default; run chkconfig --on
2887 lustre and chkconfig --on lustrefs to use them
2888 - upcalls are no longer needed for clients to recover to failover
2890 - add --abort-recovery option to lconf to abort recovery on device
2892 - add support for an arbitrary number of OSTs (3026)
2893 - Quota support protocol changes.
2894 - forward compatibility changes to wire structs (6007)
2895 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2896 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2897 - improved "lustre status" script
2898 - initialize blocksize for non-regular files (6062)
2899 - added --disable-server and --disable-client configure options (5782)
2900 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2901 - Vanilla 2.4.29 support
2902 - increase maximum number of obd devices to 520 (6242)
2903 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2904 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2905 SLES 9 2.6.5 kernels
2906 - limit stripes per file to 160 (the maximum EA size) (6093)
2908 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2911 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2912 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2913 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2914 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2915 - small ext3 extents cleanups and fixes (5733)
2916 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2917 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2918 - avoid needless client->OST connect, fix handle mismatch (5317)
2919 - fix DLM error path that led to out-of-sync client, long delays (5779)
2920 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2921 - fix several locking issues related to i_size (5492,5624,5654,5672)
2922 - don't move pending lock onto export if it is already evicted (5683)
2923 - fix kernel oops when creating .foo in unlinked directory (5548)
2924 - fix deadlock in obdfilter statistics vs. object create (5811)
2925 - use time_{before,after} to avoid timer jiffies wrap (5882)
2926 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2927 - Do not leak request if server was not able to process it (5154)
2928 - If mds_open unable to find parent dir, make that negative lookup(5154)
2929 - don't create new directories with extent-mapping (5909, 5936)
2931 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2932 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2933 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2934 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2935 - support for RHEL4 kernel on i686 (5773)
2936 - provide error messages when incompatible logs are encountered (5898)
2938 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2939 * version 1.4.0.10 (1.4.1 release candidate 1)
2941 - don't keep a lock reference when lock is not granted (4238)
2942 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2943 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2944 - drop import inflight refcount on signal_completed_replay error (5255)
2945 - unlock page after async write error during send (3677)
2946 - handle missing objects in filter_preprw_read properly (5265)
2947 - no transno return for symlink open, don't save no-trasno open (3440)
2948 - don't try to complete elan receive that already failed (4012)
2949 - free RPC server reply state on error (5406)
2950 - clean up thread from ptlrpc_start_thread() on error (5160)
2951 - readahead could read extra page into cache that wasn't ejected (5388)
2952 - prevent races in class_attach/setup/cleanup/detach (5260)
2953 - don't dereference de->d_inode after l_dput of de (5458)
2954 - use "int" for stripe value returned from lock_to_stripe (5544)
2955 - mballoc allocation and error-checking fixes in 2.6 (5504)
2956 - block device patches to fix I/O request sizes in 2.6 (5482)
2957 - look up hostnames for IB nals (5602)
2958 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2959 - don't start multiple acceptors for the same port (5277)
2960 - fix incorrect LASSERT in mds_getattr_name (5635)
2961 - export a proc file for general "ping" checking (5628)
2962 - fix "lfs check" to not block when the MDS is down (5628)
2964 - service request history (4965)
2965 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2966 - create an "evict_client" /proc entry on OSTs, like the MDS has
2967 - fix mount usage message, return errors per mount(8) (5168)
2968 - change grep [] to grep "[]" in tests so they work in more UMLs
2969 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2970 - remove ancient LOV_MAGIC_V0 EA support (5047)
2971 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2972 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2973 - allow readahead allocations to fail when low on memory (5383)
2974 - mmap locking landed again, after considerable improvement (2828)
2975 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2977 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2980 - send OST transaction number in read/write reply to free req (4966)
2981 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2982 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2983 - the watchdog thread now runs as interruptible (5246)
2984 - flock/lockf fixes (but it's still disabled, pending 5135)
2985 - don't use EXT3 constants in llite code (5094)
2986 - memory shortage at startup could cause assertion (5176)
2988 - reorganization of lov code
2989 - single portals codebase
2991 - add extents/mballoc support (5025)
2992 - direct I/O reads in the obdfilter (4048)
2993 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2995 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2998 - send OST transaction number in read/write reply to free req (4966)
2999 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3000 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3001 - the watchdog thread now runs as interruptible (5246)
3002 - handle missing objects in filter_preprw_read properly (5265)
3003 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3004 - drop import inflight refcount on signal_completed_replay error (5255)
3005 - unlock page after async write error during send (3677)
3006 - return original error code on reconstructed replies (3761)
3007 - no transno return for symlink open, don't save no-trasno open (3440)
3009 - add pid to ldlm debugging output (4922)
3010 - bump the watchdog timeouts -- we can't handle 30sec yet
3011 - extra debugging for orphan dentry/inode bug (5259)
3013 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3016 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3017 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3018 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3019 - fix NULL dereference in /proc/sys/portals/routes (4827)
3020 - allow failed mdc_close() operations to be interrupted (4561)
3021 - stop precreate on OST before MDS would time out on it (4778)
3022 - don't send partial-page writes before EOF from client (4410)
3023 - discard client grant for sub-page writes on large-page clients (4520)
3024 - don't free dentries not owned by NFS code, check generation (4806)
3025 - fix lsm leak if mds_create_objects() fails (4801)
3026 - limit debug_daemon file size, always print CERROR messages (4789)
3027 - use transno after validating reply (3892)
3028 - process timed out requests if import state changes (3754)
3029 - update mtime on OST during writes, return in glimpse (4829)
3030 - add mkfsoptions to LDAP (4679)
3031 - use ->max_readahead method instead of zapping global ra (5039)
3032 - don't interrupt __l_wait_event() during strace
3034 - add software watchdogs to catch hung threads quickly (4941)
3035 - make lustrefs init script start after nfs is mounted
3036 - fix CWARN/ERROR duplication (4930)
3037 - return async write errors to application if possible (2248)
3038 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3039 - print NAL number in %x format (4645)
3040 - update barely-supported suse-2.4.21-171 series (4842)
3041 - support for sles 9 %post scripts
3042 - support for building 2.6 kernel-source packages
3043 - support for sles km_* packages
3045 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3048 - ignore -ENOENT errors in osc_destroy (3639)
3049 - notify osc create thread that OSC is being cleaned up (4600)
3050 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3051 - reconstruct ost_handle() like mds_handle() (4657)
3052 - create a new thread to do import eviction to avoid deadlock (3969)
3053 - let lconf resolve symlinked-to devices (4629)
3054 - don't unlink "objects" from directory with default EA (4554)
3055 - hold socknal file ref over connect in case target is down (4394)
3056 - allow more than 32000 subdirectories in a single directory (3244)
3057 - fix blocks count for O_DIRECT writes (3751)
3058 - OST returns ENOSPC from object create when no space left (4539)
3059 - don't send truncate RPC if file size isn't changing (4410)
3060 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3061 - bind to privileged port in socknal and tcpnal (3689)
3063 - rate limit CERROR/CWARN console message to avoid overload (4519)
3064 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3065 - basic mmap support (3918)
3066 - kernel patch series update from b1_4 (4711)
3068 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3071 - avoid crash during MDS cleanup with OST shut down (2775)
3072 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3073 - don't use bad inodes on the MDS (3744)
3074 - dynamic object preallocation to improve recovery speed (4236)
3075 - don't hold spinlock over lock dumping or change debug flags (4401)
3076 - don't zero obd_dev when it is force cleaned (3651)
3077 - print grants to console if they go negative (4431)
3078 - "lctl deactivate" will stop automatic recovery attempts (3406)
3079 - look for existing locks in ldlm_handle_enqueue() (3764)
3080 - don't resolve lock handle twice in recovery avoiding race (4401)
3081 - revalidate should check working dir is a directory (4134)
3083 - don't always mark "slow" obdfilter messages as errors (4418)
3085 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3088 - don't close LustreDB during write_conf until it is done (3860)
3089 - fix typo in lconf for_each_profile (3821)
3090 - allow dumping logs from multiple threads at one time (3820)
3091 - don't allow multiple threads in OSC recovery (3812)
3092 - fix debug_size parameters (3864)
3093 - fix mds_postrecov to initialize import for llog ctxt (3121)
3094 - replace config semaphore with spinlock (3306)
3095 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3096 - don't allow enqueue to complete on a destroyed export (3822)
3097 - down write_lock before checking llog header bitmap (3825)
3098 - recover from lock replay timeout (3764)
3099 - up llog sem before sending rpc (3652)
3100 - reduce ns lock hold times when setting kms (3267)
3101 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3102 - fix NULL deref and obd_dev leak on setup error (3312)
3103 - replace some LBUG about llog ops with error handling (3841)
3104 - don't match INVALID dentries from d_lookup and spin (3784)
3105 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3106 - fix invalid assertion in ptlrpc_set_wait (3880)
3108 - add libwrap support for the TCP acceptor (3996)
3109 - add /proc/sys/portals/routes for non-root route listing (3994)
3110 - allow setting MDS UUID in .xml (2580)
3111 - print the stack of a process that LBUGs (4228)
3113 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3116 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3117 - make sure to unset replay flag from failed open requests (3440)
3118 - if default stripe count is 0, use OST count for inode size (3636)
3119 - update parent mtime/ctime on client for create/unlink (2611)
3120 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3121 - free recovery state on server during a forced cleanup (3571)
3122 - unregister_reply for resent reqs (3063)
3123 - loop back devices mounting and status check on 2.6 (3563)
3124 - fix resource-creation race that can provoke i_size == 0 (3513)
3125 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3126 - more debugging for page-accounting assertion (3746)
3127 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3128 - avoid deadlock after precreation failure (3758)
3129 - fix race and lock order deadlock in orphan handling (3450, 3750)
3130 - add validity checks when grabbing inodes from l_ast_data (3599)
3132 - add /proc/.../recovery_status to obdfilter (3428)
3133 - lightweight CDEBUG infrastructure, debug daemon (3668)
3134 - change default OSC RPC parameters to be better on small clusters
3135 - turn off OST read cache for files smaller than 32MB
3136 - install man pages and include them in rpms (3100)
3137 - add new init script for (un)mounting lustre filesystems (2593)
3138 - run chkconfig in %post for init scripts (3701)
3139 - drop scimac NAL (unmaintained)
3141 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3144 - clean kiobufs before and after use (3485)
3145 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3146 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3147 - change init script to fail more clearly if not run as root (1528)
3148 - allow clients to reconnect during replay (1742)
3149 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3150 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3151 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3152 - fix cond_resched (3554)
3153 - only evict unfinished clients after recovery (3515)
3154 - allow bulk resend, prevent data loss (3570)
3155 - dynamic ptlrpc request buffer allocation (2102)
3156 - don't allow unlinking open directory if it isn't empty (2904)
3157 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3158 - remove extraneous obd dereference causing LASSERT failure (3334)
3159 - don't use get_cycles() when creating temp. files on the mds (3156)
3160 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3161 - handle EEXIST for set-stripe, set proper directory name (3336)
3163 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3164 - fix ksocknal_fmb_callback() error messages (2918)
3166 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3169 - don't copy lvb into (possibly NULL) reply on error (2983)
3170 - don't deref dentry after dput, don't free lvb on error (2922)
3171 - use the kms to determine writeback rpc length (2947)
3172 - increment oti_logcookies when osc is inactive (2948)
3173 - update client's i_blocks count via lvb messages (2543)
3174 - handle intent open/close of special files properly (1557)
3175 - mount MDS with errors=remount-ro, like obdfilter (2009)
3176 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3177 - don't use cancelling-locks' kms values (2947)
3178 - use highest lock extent for kms, not last one (2925)
3179 - don't dereference ERR_PTR() dentry in error handling path (3107)
3180 - fix thread race in portals_debug_dumplog() (3122)
3181 - create lprocfs device entries at setup instead of at attach (1519)
3182 - common AST error handler, don't evict client on completion race (3145)
3183 - zero nameidata in detach_mnt in 2.6 (3118)
3184 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3185 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3186 - initialize RPC timeout timer earlier for 2.6 (3219)
3187 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3188 - print nal/nid for unknown nid (3258)
3189 - additional checks for oscc recovery before doing precreate (3284)
3190 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3191 - don't crash in mdc_close for bad permissions on open (3285)
3192 - zero i_rdev for non-device files (3147)
3193 - clear page->private before handing to FS, better assertion (3119)
3194 - tune the read pipeline (3236)
3195 - fix incorrect decref of invalidated dentry (2350)
3196 - provide read-ahead stats and refine rpc in flight stats (3328)
3197 - don't hold journal transaction open across create RPC (3313)
3198 - update atime on MDS at close time (3265)
3199 - close LDAP connection when recovering to avoid server load (3315)
3200 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3201 - don't leak open file on MDS after open resend (3325)
3202 - serialize filter_precreate and filter_destroy_precreated (3329)
3203 - loop device shouldn't call sync_dev() for nul device (3092)
3204 - clear page cache after eviction (2766)
3205 - resynchronize MDS->OST in background (2824)
3206 - refuse to mount the same filesystem twice on same mountpoint (3394)
3207 - allow llmount to create routes for mounting behind routers (3320)
3208 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3209 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3210 - fix rare problem with rename on htree directories (3417)
3212 - allow default OST striping configuration per directory (1414)
3213 - fix compilation for qswnal for 2.6 kernels (3125)
3214 - increase maximum number of MDS request buffers for large systems
3215 - change liblustreapi to be useful for external progs like lfsck (3098)
3216 - increase local configuration timeout for slow disks (3353)
3217 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3219 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3222 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3223 - don't overwrite extent policy data in reply if lock was blocked (2901)
3224 - drop filter export grants atomically with removal from device (2663)
3225 - del obd_self_export from work_list in class_disconnect_exports (2908)
3226 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3227 - swab reply message in mdc_close, other PPC fixes (2464)
3228 - fix destroying of named logs (2325)
3229 - overwrite old logs when running lconf --write_conf (2264)
3230 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3231 - fix race in target_handle_connect (2898)
3232 - mds_reint_create() should take same inode create lock (2926)
3233 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3234 - don't close files for self_export to avoid uninitialized obd (2936)
3235 - allow MDS with the same name as client node (2939)
3236 - hold dentry reference for closed log files for unlink (2325)
3237 - reserve space for all logs during transactions (2059)
3238 - don't evict page beyond end of stripe extent (2925)
3239 - don't oops on a deleted current working directory (2399)
3240 - handle hard links to targets without a parent properly (2517)
3241 - don't dereference NULL lock when racing during eviction (2867)
3242 - don't grow lock extents when lots of conflicting locks (2919)
3244 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3247 - account for cache space usage on clients to avoid data loss (974)
3248 - lfsck support in lustre kernel code (2349)
3249 - reduce journal credits needed for BRW writes (2370)
3250 - orphan handling to avoid losing space on client/server crashes
3251 - ptlrpcd can be blocked, stopping ALL progress (2477)
3252 - use lock value blocks to assist in proper KMS, faster stat (1021)
3253 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3254 - recovery for initial connections (2355)
3255 - fixes for mds_cleanup_orphans (1934)
3256 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3257 - block all file creations until orphan recovery completes (1901)
3258 - client remove rq_connection from request struct (2423)
3259 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3260 - recovery timer race (2670)
3261 - mdc_close recovey bug (2532)
3262 - ptlrpc cleanup bug (2710)
3263 - mds timeout on local locks (2588)
3264 - namespace lock held during RPCs (2431)
3265 - handle interrupted sync write properly (2503)
3266 - don't try to handle a message that hasn't been replied to (2699)
3267 - client assert failure during cleanup after abort recovery (2701)
3268 - leak mdc device after failed mount (2712)
3269 - ptlrpc_check_set allows timedout requests to complete (2714)
3270 - wait for inflight reqs when ptlrpcd finishes (2710)
3271 - make sure unregistered services are removed from the srv_list
3272 - reset bulk XID's when resending them (caught by 1138 test)
3273 - unregister_bulk after timeout
3274 - fix lconf error (2694)
3275 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3276 - readahead locks pages, leaves pending causing memory pressure (2673)
3277 - increase OST request buffers to 4096 on large machines (2729)
3278 - fix up permission of existing directories in simple_mkdir (2661)
3279 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3280 - don't assign transno to errored transactions (2742)
3281 - don't delete objects on OST if given a bogus objid from MDS (2751)
3282 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3283 - if rq_no_resend, then timeout request after recovery (2432)
3284 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3285 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3286 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3287 - MDS recovery shouldn't delete objects that it has given out (2730)
3288 - if enqueue arrives after completion, don't clobber LVB (2819)
3289 - don't unlock pages twice when trigger_group_io returns error (2814)
3290 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3291 - don't write pages to disk if there was an error (1450)
3292 - don't ping imports that have recovery disabled (2676)
3293 - take buffered bytes into account when balancing socknal conn (2817)
3294 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3295 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3296 - remove little-endian swabbing of llog records (1987)
3297 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3298 - retry reposting request buffers if they fail (1191)
3299 - grow extent at grant time to avoid granting a revoked lock (2809)
3300 - lock revoke doesn't evict page if covered by a second lock (2765)
3301 - disable VM readahead to avoid reading outside lock extents (2805)
3303 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3304 - updated kernel patches for hp-2.4.20 kernel (2681)
3306 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3309 - fix truncated write corruption (2366)
3310 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3312 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3313 - ensure len(uuid) < 37 in lmc (1171)
3314 - fix ia64 OOPS in llog_test (2255)
3315 - zero end of page at obdfilter for partial page writes (2648)
3316 - don't leave stale dentries around after renames (bug 2428)
3317 - fix timeouts when evicting a client with a single lock held (2642)
3318 - set deadline for the initial HELLO message to drain (2634)
3319 - print out dotted-quad IP addresses in the socknal (2302)
3321 - additional debugging for MDS client eviction problem (2443)
3322 - fix mkfsoptions support for osts (2603, 2604)
3324 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3327 - add series for the vanilla 2.6.0 kernel
3328 - add series for the vanilla 2.4.24 kernel
3329 - add series for a cray x86/64 UL kernel drop
3330 - fix xattr patches for the vanilla 2.4.19 series
3332 - generate true UUIDs in lmc (1171)
3333 - have portals stack dumping break in UML (2466)
3334 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3335 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3336 - wake callback waiting threads on client eviction (2460)
3337 - Add --ptldebug and --subsystem to lmc (1719)
3338 - update assertion to allow safe interrupt allocation
3339 - set rq_no_resend for cancel requests (2432)
3340 - recalculate ptlrpcd timeout after resend (2494)
3341 - call vfs_rmdir when removing pending directories (2368)
3342 - fix renaming a file to itself (2429)
3343 - lmc creates a default one-stripe lov (2454)
3344 - expand procfs space to handle large clusters (2326)
3345 - increase UML stack to avoid overflow
3346 - update lconf's list of debug and subsystem masks
3347 - fix lfs find --obd (2510)
3348 - /proc tunable for disabling filter read caching (2591)
3349 - stop rpm packages from altering slapd.conf (2301)
3350 - disable nagle in the socknal under 0conf (2578)
3351 - choose mds inode size based on stripe count (2572)
3352 - fix kernel-source rpm problems (2516)
3354 - add --disable-doc to avoid pdf generation (2421)
3355 - update documentation, tests, type-os, comments
3356 - avoid format warnings on ia64
3357 - remove the TOE NAL
3358 - tiny code cleanups by removing unused fields
3360 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3363 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3364 - remove the most blatant lies from BUILDING file (2371)
3365 - change default debug level to reasonable production setting
3366 - reduce client side cache size to reduce cache flush time
3367 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3368 - make TCP zerocopy and pinger support enabled by default (2476)
3369 - sync writes completed after process exits caused crashes (2319)
3370 - maintain correct mount count on the MDS (2356)
3371 - backout 1557, because 2316 wasn't really fixed
3372 - better file I/O statistics gathering in /proc
3373 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3374 - another kernel patch to fix zero-copy TCP function export
3375 - don't take duplicate lock when processing re-sent getattr (2420)
3376 - lctl uses obd_self_export instead of creating new conn (2353)
3377 - MDS/OST recovery case which requires object creation asserted (2425)
3378 - move lfs from /usr/sbin to /usr/bin in packages
3379 - fix race between mds_client_add and mds_client_free (2417)
3380 - use kmalloc instead of slabs in portals (2430)
3381 - don't create duplicate records when a failover MDS is present (2442)
3382 - remove unnecessary mount age check (2332)
3383 - don't remove directory inodes from locks prematurely (2451)
3384 - don't break if MDS service name is the same as hostname (2103)
3385 - fix races in client write RPC generation when cache full (2482)
3387 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3390 - remove now-unused request->rq_obd (278)
3391 - if an allocation fails, print out how much memory we've used (1933)
3392 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3393 - add the "configurable stack size" patch to most series files (1256)
3394 - ability to write large log records, for 100+ OST configs (2306)
3395 - fix NULL deref when filter_prep fails (2314)
3396 - fix operator precedence error in filter_sync
3397 - dynamic allocation of socknal TX descriptors (2315)
3398 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3399 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3400 - ensure that CWARN messages in Portals always get to the syslog
3401 - __init/__exit are not for prototype decls (ldlm_init/exit)
3402 - x86-64 compile warning fixes
3403 - fix gateway LMC keyword conflict (2318)
3404 - fix MDS lock inversions in getattr/reint paths (1844)
3405 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3406 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3407 - DLM race condition prevented some lock evictions (2328)
3408 - ENOMEM detection and retry on socknal sends (2230)
3409 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3410 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3411 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3412 - better error messages when a client is rejected during recovery (1505)
3413 - avoid cancelling locks which were never granted, after failure (2330)
3414 - fix i_sem/journal inversion in mds_client_add (2333)
3415 - fix truncate/getattr lock cycle deadlock (2334)
3416 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3417 - fix two rare exit paths which could leak an l_lock() ref (2321)
3418 - fencepost error in MDS/OST orphan recovery (2226)
3419 - make log record alignment 8 bytes (1988)
3420 - lstripe now fails when requested offset > ost_count (2237)
3421 - ensure that all kernel series have a complete list.h (1607)
3422 - fix crashes in special-file operations (2316)
3423 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3425 - allow configurable automake binary, for testing new versions
3426 - small update to the lfs documentation
3428 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3430 * fix negative export reference count in fsfilt_sync (2312)
3432 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3433 * release candidate 0.9.1
3435 - orphans are moved into the PENDING directory for possible recovery
3436 - replayed opens now open by fid for orphan/rename safety (1042)
3437 - last close of an orphan inode generates a transno (683)
3438 - chdir() and mount() now pin the directory entry (1020)
3439 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3440 - discard very old requests without processing them (1502)
3441 - remove some common, well-understood CERRORs (1505)
3442 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3443 - clear "grant" flags in OST replies until OST grant code lands (1644)
3444 - fix read performance by not clobbering i_blksize on client (1598)
3445 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3446 - make LRU size a /proc tunable, clears locks when reduced (707)
3447 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3448 - server locks take references on exports now (1558)
3449 - build fixes for 2.4.20-rh trees (1663)
3450 - return an error from lov_create if all OSCs are inactive (1751)
3451 - fix import levels when a reconnect happens without a timeout (1597)
3452 - exit early from mds_open if we get a lookup error (1749)
3453 - partial page read at EOF wouldn't wait for disk before sending (1642)
3454 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3455 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3456 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3457 - write/truncate lock inversion (1639)
3458 - Don't auto-load obdclass, portals modules during cleanup (1495)
3459 - fix timestamps from jumping to "now" (1763)
3460 - extra journal assertions (1648)
3461 - add an extra multiunlink test (1771)
3462 - fix read_record/write_record API (1776)
3463 - fix leak of offset_extent, possible incorrect i_size later (1772)
3464 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3465 - Debugging for the kqswnal_get_idle_tx problems (1820)
3466 - Allow recovery to be attempted multiple times (1536)
3467 - Write out MDS last_rcvd file after it is first created (1600)
3468 - Fix tx_descriptor leak in failed transmit situations (1827)
3469 - ext3 journaling fixes for assertion failure after IO error (1871)
3470 - class_export_put() on freed export after completion AST error (1896)
3471 - Fix revalidate looping in VFS (1322)
3472 - Don't access a freed export during MDS_REINT timeout (1521)
3473 - Add open-unlink recovery support on the MDS (1673,1764)
3474 - Return an error if no MDS data was read from last_rcvd (1946)
3475 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3476 - Don't setup a disk device that doesn't match exported UUID (317)
3477 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3478 - avoid committing NULL handle in force close
3479 - local.sh is now a one-stripe LOV configuration
3480 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3481 - fix ext3 htree duplicate directory entry corruption (1516)
3482 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3483 - update to new LOV EA format (2097)
3484 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3486 2003-06-15 Phil Schwan <phil@clusterfs.com>
3489 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3490 - per-import/export recovery handling (958, 931, 959)
3491 - multiple last-rcvd slots, for serving multiple FSes (949)
3492 - connections are again shared between multiple imp/exports (963, 964)
3493 - "umount -f" would hang if any requests needed to be sent (393, 978)
3494 - avoid pinning large req buffer by copying for queued messages (989)
3495 - add "uuid" to "lctl device" command to help upcalls (991)
3496 - "open" RPCs with transnos would confuse recovery counters (1037)
3497 - do proper endian conversion of all wire messages (288, 340, 891)
3498 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3499 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3500 - fix ptlrpc_connection leak in target_handle_connect (1174)
3501 - fix import refcounting bug in OST and MDS cleanup (1134)
3502 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3503 - fix very unlikely obd_types race condition (501)
3504 - remove osc_open hack for echo_client (1187)
3505 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3506 - a failure in read_inode2 leads to deadlock (1139)
3507 - cancel ack-locks as soon as transaction is committed (1072)
3508 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3509 - make sure to commitrw after any preprw to avoid deadlock (1162)
3510 - failing to execute a file in a lustre FS would lock inode (1203)
3511 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3512 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3513 - obd_uuid2tgt would walk off the end of the list (1255)
3514 - on IA64 the osc would give portals incorrect bulk size (1258)
3515 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3516 - fix lock inversion caused by new llite matching code (1282)
3517 - limit the number of dirty pages on a client to 10MB (1286)
3518 - timed out locks were not being corrected cancelled (1289)
3519 - fix O_DIRECT above 4GB on IA-32 (1292)
3520 * major user-visible changes
3521 - fail out/fail over policy now controlled by the upcall (993)
3523 - add OBD_PING to check server availability and failure (954)
3524 - lustre messages are now sent in sending host order (288, 340, 891)
3525 - add eadatalen to MDS getattr reply (340)
3526 - OST read replies may contain second buffer, with per-page status (593)
3528 2003-03-11 Phil Schwan <phil@clusterfs.com>
3531 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3532 - failed open()s could cause deadlock; fixed (867, 869)
3533 - stop cancelling OST locks when files are closed (481)
3534 - overlapping XID spaces caused network corruption (851, 853)
3535 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3536 - setattr_raw added, to do single-RPC, server-side setattrs
3537 - lmc/lconf syntax change for OST UUIDs
3538 - fix crashy race condition between ptlrpc_free_req and osc_close
3539 - don't use request in mdc_enqueue if we hit a timeout (889)
3540 - don't set the inode i_size for regular files from the MDS (896)
3541 - handle out of order completion AST (842)
3542 - don't LBUG if a lock request times out after receiving AST (913)
3543 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3544 - if a bad lock AST arrives, send an error instead of dropping entirely
3545 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3546 - fix leak in bulk IO when only partially completed (899, 900, 926)
3547 - fix O_DIRECT for ia64 (55)
3548 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3549 - C-z after timeout could hang a process forever; fixed (977)
3551 - client-side I/O cache (678, 924, 929, 941, 970)
3553 - READPAGE and SETATTRs which don't take server-side locks get
3556 2003-02-11 Phil Schwan <phil@clusterfs.com>
3559 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3560 - Fix fsfilt_extN_readpage() to read a full page of directory
3561 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3562 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3563 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3564 - Most intent operations were removed, in favour of a new RPC mode
3565 that does a single RPC to the server and bypasses most of the VFS
3566 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3567 - Aggressively cancel local locks on DLM servers
3568 - mds_reint_unlink sends EA to the client if it's the last nlink.
3569 client uses that EA to unlink OST objects.
3570 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3571 - recursive symlinks were fixed (439)
3572 - fixed NULL deref in DEBUG_REQ
3573 - filter_update_lastobjid no longer calls sync, which annoyed extN
3574 - fixed multi-client small-writes to a single file problem (445)
3575 - fixed mtime updates during file writes (607)
3576 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3577 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3578 - fixed Linux OST size reporting problem (444, 656)
3579 - OST now updates object mtime with writes or setattr (607, 619)
3580 - client verifies file size before zeroing page past EOF (445)
3581 - OST now writes last allocated objid to disk with allocation (108)
3582 - LOV on echo now works (409)
3584 - mds_reint_unlink sends a new buffer, with the EA included. this
3585 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3586 set if a regular file was being unlinked, and it was the last link
3587 - use PtlGet from the target for bulk writes (315)
3588 - OST now updates object mtime with writes or setattr (607, 619)
3589 - LDLM now has a grant-time callback to revalidate locked items, if
3591 - Many MDS operations were reorganized to combat race conditions
3593 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3594 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3596 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3599 - Fully reactivate OST imports after reconnection (512, others)
3600 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3601 - More graceful error handling for truncating on dead OST (515)
3602 - Don't error out unless we're actually accessing dead stripes (474)
3603 - Fix garbage sizes when stripes are missing (410)
3604 - LRU counters were broken, causing constant lock purge (433, 432)
3605 - garbage on read from stripes with failed OSTs (441)
3606 - mark OSCs as active before reconnecting during recovery (438)
3607 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3608 - lfind did not preserve OST order in output (443)
3609 - symlinks cause hung clients, incorrect data (439)
3610 - stop dereferencing request after dropping refcount (457)
3611 - don't LASSERT(spin_is_locked) on non-SMP (455)
3612 - fixes for many rename() bugs
3613 - fstat didn't correctly synchronize attributes (399)
3614 - server must handle lock cancellation during blocking AST prep (487)
3615 - bulk descriptors were free()d too soon (511)
3616 - fix paths in lconf, which would load incorrect modules (451, 507)
3617 - fix confusing lconf 'host not found' error message (386)
3618 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3619 - fix race condition in mdc_blocking_ast() for inode access (526)
3620 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3621 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3622 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3623 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3624 - fix lstripe in several ways (kernel side) (527)
3625 - fix request leak in ldlm_cli_enqueue (262)
3626 - incorrect OSC was marked inactive after OST failure
3627 - call mds_fs_cleanup before unmounting filesystem (524)
3628 - fix races between taking ns_lock and ldlm_lock_change_resource
3629 - fix races updating LOV export open file list
3630 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3631 - fix recovery NULL deref in ldlm_cli_cancel_unused
3632 - fix some DLM races by using new hash table for lock handles (419)
3633 - permit the client to specify desired inodes, at replay
3634 - duplicate requests when we queue them for replay reintegration
3635 - fix last_rcvd offset calculation
3636 - sync after each recovered transaction, so we always make progress
3637 - never, not always, ERESTART requests without transnos
3638 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3640 - skip replay if the MDS says that the client is already connected
3641 - don't check for a recovery-enabled export to match lctl's UUID
3642 - don't INC_USE_COUNT for phantom exports
3643 - don't crash when cleaning up phantom exports (567)
3644 - don't double-finish or set replay data for errored mdc_open requests
3645 - abort requests when they time out, so we don't get old replies
3646 - send/receive replies for AST messages again
3647 - if the client says that it doesn't have the lock, cancel it on the
3649 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3650 mark it as destroyed, it will all work out when decref is called
3651 - fix module use counts (22, 581)
3653 - ASTs now expect a reply (server cancels lock on error reply)
3655 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3658 - fix many simultaneous client startup (392)
3659 - fix dentry->d_it clobbering
3660 - credentials weren't being shipped for readdir/getattr operations
3661 - remove invalid assertions triggered during some concurrent MD
3663 - proper Lustre versions added (336, 389)
3664 - fix memory leak for create error case (398)
3665 - fix LOV locking bug that would get cli/srv out of sync
3666 - fix echo client over LOV (409)
3667 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3668 - fix double-O_EXCL intent crash (424)
3669 - avoid sending multiple lock CANCELs (352)
3671 - MDS can do multi-client recovery (modulo bugs in new code)
3673 - many updates, edits, cleanups
3675 2002-11-18 Phil Schwan <phil@clusterfs.com>
3678 - fix null d_it dereference (346)
3679 - fix full OST/dbench hang (333)
3680 - fix permission problem with file removal (286)
3681 - fix removal of OSCs from LOV when they fail
3682 - fix NULL deref during bulk timeout (214)
3683 - fix problems related to multiple filesystems on one MDS (241)
3684 - fixed serious subtle metadata locking bugs
3685 - free locks on clients when inodes are removed due to memory
3687 - fix inode pointer in lock data (285)
3688 - partial support for multiple MDS on a single host (241)
3689 - data locks weren't cancelled at clear_inode time (290, 311)
3690 - intent locks could lead to unbounded lock growth (205)
3691 - added a maximum lock count, an LRU list, and a flusher
3692 - fix multiple rename (365)
3693 - properly abstracted the echo client
3694 - OSC locked 1 byte too many; fixed
3695 - rewrote brw callback code:
3696 - fixed recovery bugs related to LOVs (306)
3697 - fixed too-many-pages-in-one-write crash (191)
3698 - fixed (again) crash in sync_io_timeout (214)
3699 - probably fixed callback-related race (385)
3701 - Add capability to MDS protocol
3702 - LDLM cancellations and callbacks on different portals
3704 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3707 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3708 - defer open object destruction to close time (601981, 138)
3709 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3710 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3711 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3712 - don't call obd_setattr in ll_file_release for destroyed objects
3713 * protocol change to lustre_msg: move |version| and add |flags|
3714 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3715 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3716 * added replay of create, unlink, link and rename operations during
3717 MDS failover; recovery should be much more robust now
3718 * remove failed OSCs from LOVs (only lov_create uses this so far)
3719 * the lustre-HOWTO was brought (more) up to date (582544)
3721 2002-10-23 Phil Schwan <phil@clusterfs.com>
3724 - in-use dentries weren't being reused properly (617851)
3725 - prevent multiple LDLM setup (599178)
3726 - fix LOV size calculations for truncate (617853)
3727 - fix client handling of MDS intent errors (POSIX)
3728 - fix permission bug in lovstripe.c test (624321)
3729 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3730 - truncate past end of file could corrupt data
3731 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3732 - a race in recovery could return ETIMEDOUT to apps (623947)
3733 - building outside the source directory was fixed
3734 * the lustre-HOWTO was brought (more) up to date (582544)
3735 * major progress was made on recovery functionality
3737 2002-10-10 Phil Schwan <phil@clusterfs.com>
3740 - recovery deadlock fix
3741 - rm -rf causes LBUG fix (617817)
3742 - file open by multiple tasks fix (618962)
3743 - directory permissions bugs (602707 and 620007)
3744 - journal_stop fixed with locking (611313)
3745 - O_APPEND failures resolved (618273, perhaps 614459)
3746 - lconf PATH fix (619770)
3747 - IA64 build fix (621450)
3748 - RPC buffer sizes scale with amount of memory
3750 2002-10-01 Phil Schwan <phil@clusterfs.com>
3753 - locks would be cancelled without throwing away data pages,
3754 resulting in inconsistent data (605627)
3755 - inode attributes were not always being refreshed (605627, 612449)
3756 - lconf now continues to cleanup after lctl reports an error
3757 - MDS now enforces user permissions (602707)
3758 - lprocfs cleanup fixed, but not yet enabled (614157)
3759 - fixed infinite server hang, should a client not respond to an AST
3760 - avoid going into recovery if user calls readlink() with a buffer
3761 that's too small (613941)
3762 - AST RPCs no longer require replies (614867) -- this may be changed
3763 - don't crash server if client sends an IOV that's too big (611336)
3764 - fixed lock conversion deadlock (611892)
3765 - fixed the following of symlinks (614622)
3766 * recovery: the server can remove locks from a client that dies, other
3767 clients can make progress
3768 * more extN patch fixes
3769 * compile-time configurable ptlrpc buffer allocations
3771 - collaborative read cache document
3772 - Lustre Lite Performance CDR document-in-progress
3774 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3777 - fix typo in patch-2.4.18
3779 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3782 - clear ptlrpc request each time in handle_incoming_request()
3783 - unlink of files now destroys the object on the OST
3785 2002-09-19 Peter Braam <braam@clusterfs.com>
3787 * add hard link support
3788 * change obdfile creation method
3789 * kernel patch changed
3791 2002-09-19 Peter Braam <braam@clusterfs.com>
3794 - stack overflow bug in extN fixed
3796 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3798 * documentation updates
3799 - add man pages for config tools
3800 - update tests/README to describe testing with new config tools
3801 - finish metadata API descriptions
3802 * bug fixes and cleanups
3803 - statfs workaround for 16TB limit
3804 - LOV stripe allocation improved, can stripe on subset of OSTs
3805 - LOV file size/IO offset was wrong for files > 4GB in size
3806 - object EA data was being dropped, caused files to be unreadable
3807 - memory overflow with non-LOV OST caused memory corruption
3808 - fixed regression tests to work with new config tools, obdfilter
3809 - fixed bug when directory size became larger than 1 block
3810 - fixed bug (for single client case) when PWD was deleted
3811 - invalidate local directory pages when doing intent-based ops
3812 - avoid LDLM oops when lock callback contained bad data
3814 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3816 * documentation updates
3817 * bug fixes and cleanups
3818 - configuration tools
3821 - 64-bit compile warnings
3822 - 64-bit internal statfs data
3824 * test_brw on persistent OST devices
3826 * lprocfs (disabled)
3828 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3830 * documentation updates
3831 * bug fixes and cleanups
3832 * configuration tools
3834 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3837 * many small fixes to 0_5_4
3838 * io/network handling
3839 * thinkos in MDS operations
3841 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3844 * crucial basic fixes to 0.5.3
3845 * IOR, Iozone work over Elan
3848 2002-08-07 Phil Schwan <phil@clusterfs.com>
3849 * version 0_5_3, our first alpha
3850 * we use the new Portals iovs
3851 * documentation updates
3852 * bug fixes and cleanups
3853 * small changes in the DLM wire protocol
3855 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3856 * version 0_5_1 with some initial stability,
3857 * locking on MD and file I/O.
3858 * documentation updates
3859 * several bug fixes since 0.5.0
3860 * small changes in wire protocol
3862 2002-07-18 Phil Schwan <phil@clusterfs.com>
3864 * delivered as Lustre Light Alpha
3865 * fixed a crash after handling invalid MDS requests
3866 * fixed directory pages for architectures with non-4k pages sizes
3868 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3869 * release version v0_4_4
3870 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3872 * Fixes a number of interruption problems with OST operations.
3873 * Update documentation for portals header changes
3874 * Move all wire protocol structs/defines to lustre_idl.h
3875 * Fixes symlink length bug.
3876 * Add tcpdump to repository.
3878 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3879 * release version v0_4_3
3880 * Fixes statfs for inodes on extN.
3881 * Fixes bug in runtests which would delete /etc/hosts.
3882 * Use 64-bit object IDs wherever possible (not into VFS though)
3883 Remove ost_get_info, which is unused by lustre, and out of date.
3885 2002-07-03 Peter Braam <braam@clusterfs.com>
3886 * release version v0_4_2 Fixes a lookup error (type not passed)
3887 * move forward to head of Portals
3888 * move forward to latest Lustre kernel
3890 2002-06-25 Peter Braam <braam@clusterfs.com>
3891 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob