1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10),
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: don't use GFP_* in generic Lustre code.
18 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
22 Description: obdclass is limited by single OBD_ALLOC(idarray)
23 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
27 Description: warnings with build patchless client with vanila 2.6.19 and up
28 Details : change old ctl_table style and replace ctl_table/ctl_table_header
29 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
33 Frequency : Only for SLES
34 Description: sanity_quota fail test_1
35 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
36 file, which makes TSTID[2] inunique.
40 Frequency : Only for SLES9
41 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
42 Details : Update target/ChangeLog/which_patch .
46 Frequency : Only for test_10 in sanity-quota.sh
47 Description: a bug in quota test code
48 Details : checking a test flag in a wrong place
52 Frequency : Only for test_13 in sanity-quota.sh
53 Description: a bug in quota test script
54 Details : shouldn't check overfull allocated quota
58 Description: jbd statistics
59 Details : Port older jbd statistics patch for sles10
61 Severity : enhancement
63 Description: Add printing inode info into message about error in writepage.
68 Description: reply_lock_interpret crash due to race with it and lock cancel.
69 Details : Do not replay locks that are being cancelled. Do not reference
70 locks by their address during replay, just by their handle.
74 Frequency : When flocks are used.
75 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
76 Details : Flock locks might destroy just granted lock if it could be merged
77 with another existing flock, this is done in completion handler,
78 so teach ldlm_cli_enquque_fini that this is a valid case for
84 Description: Oops in read and write path when failing to allocate lock.
85 Details : Check if lock allocation failed and return error back.
89 Description: lstripe command fails for valid OST index
90 Details : The stripe offset is compared to lov->desc.ld_tgt_count
91 instead of lov->desc.ld_active_tgt_count.
95 Description: sanity.sh failed test 103
96 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
101 Description: df doesn't work properly if diskfs blocksize != 4K
102 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
106 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
107 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
111 Description: obd_health_check_timeout too short
112 Details : set obd_health_check_timeout as 1.5x of obd_timeout
116 Description: llapi_file_create() does not allow some changes
117 Details : add llapi_file_open() that allows specifying the mode and
118 open flags, and also returns an open file handle.
122 Description: merge and cleanup kernel patches.
123 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
127 Description: Add journal checksum support.(Kernel part)
128 Details : The journal checksum feature adds two new flags i.e
129 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
130 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
131 indicates that the commit block contains the checksum for
132 the blocks described by the descriptor blocks. Now commit
133 record can be sent to disk without waiting for descriptor
134 blocks to be written to disk. This behavior is controlled
135 using JBD2_FEATURE_ASYNC_COMMIT flag.
137 --------------------------------------------------------------------------------
139 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
141 * Support for kernels:
142 2.6.5-7.283 (SLES 9),
143 2.6.9-55.EL (RHEL 4),
144 2.6.16.46-0.14 (SLES 10),
145 2.6.18.8 vanilla (kernel.org)
146 * Client support for unpatched kernels:
147 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
148 2.6.16 - 2.6.22 vanilla (kernel.org)
149 * Due to recently discovered recovery problems, we do not recommend
150 using patchless RHEL 4 clients with this or any earlier release.
151 * Recommended e2fsprogs version: 1.39.cfs8
152 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
153 * Starting with this release, the ldiskfs backing filesystem required
154 by Lustre is now in its own package, lustre-ldiskfs. This package
155 should be installed. It is versioned separately from Lustre and
156 may be released separately in future.
160 Description: block reactivating mgc import until all deactivates complete
161 Details : Fix race when failing back MDT/MGS to itself (testing)
163 Severity : enhancement
165 Description: add optional extra BUILD_VERSION info
166 Details : add a new environment variable (namely LUSTRE_VERS) which allows
167 to override the lustre version.
170 Frequency : 2.6.18 servers only
172 Description: ll_kern_mount() doesn't release the module reference
173 Details : The ldiskfs module reference count never drops down to 0
174 because ll_kern_mount() doesn't release the module reference.
179 Description: server LBUG when using old ost_num_threads parameter
180 Details : Accept the old ost_num_threads parameter but warn that it
181 is deprecated, and fix an off-by-one error that caused an LBUG.
186 Description: Transient SCSI error results in persistent IO issue
187 Details : iobuf->dr_error is not reinitialized to 0 between two
191 Frequency : sometimes when underlying device returns I/O errors
193 Description: OSTs not going read-only during write failures
194 Details : OSTs are not remounted read-only when the journal commit threads
195 get I/O errors because fsfilt_ext3 calls journal_start/stop()
196 instead of the ext3 wrappers.
200 Description: poor connect scaling with increasing client count
201 Details : Don't run filter_grant_sanity_check for more than 100 exports
202 to improve scaling for large numbers of clients.
205 Frequency : SLES10 only
207 Description: sanity-quota.sh quotacheck failed: rc = -22
208 Details : Quotas cannot be enabled on SLES10.
211 Frequency : liblustre clients only
213 Description: getdirentries does not give error when run on compute nodes
214 Details : getdirentries does not fail when the size specified as an argument
215 is too small to contain at least one entry
217 Severity : enhancement
219 Description: Add LNET router traceability for debug purposes
220 Details : If a checksum failure occurs with a router as part of the
221 IO path, the NID of the last router that forwarded the bulk data
222 is printed so it can be identified.
227 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
228 Details : Due to a race condition, liblustre clients were occasionally
231 Severity : enhancement
233 Description: lfs setstripe use optional parameters instead of postional
236 Severity : enhancement
238 Description: Nanosecond timestamp support for ldiskfs
239 Details : The on-disk ldiskfs filesystem has added support for nanosecond
240 resolution timestamps. There is not yet support for this at
241 the Lustre filesystem level.
244 Frequency : during server recovery
246 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
247 Details : request with rq_no_resend flag not awake l_wait_event if they get a
251 Frequency : nfs export on patchless client
253 Description: connectathon hang when test nfs export over patchless client
254 Details : Disconnected dentry cannot be found with lookup, so we do not need
255 to unhash it or make it invalid
258 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
259 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
260 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
265 Description: open req refcounting wrong on reconnect
266 Details : If reconnect happened between getting open reply from server and
267 call to mdc_set_replay_data in ll_file_open, we will schedule
268 replay for unreferenced request that we are about to free.
269 Subsequent close will crash in variety of ways.
270 Check that request is still eligible for replay in
271 mdc_set_replay_data().
276 Description: disable writes to filesystem when reading health_check file
277 Details : the default for reading the health_check proc file has changed
278 to NOT do a journal transaction and write to disk, because this
279 can cause reads of the /proc file to hang and block HA state
280 checking on a healthy but otherwise heavily loaded system. It
281 is possible to return to the previous behaviour during configure
282 with --enable-health-write.
284 Severity : enhancement
286 Description: 64-bit inode version
287 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
288 to the inode. This will be required for version-based recovery.
293 Description: MDS fails to start if a duplicate client export is detected
294 Details : in some rare cases it was possible for a client to connect to
295 an MDS multiple times. Upon recovery the MDS would detect this
296 and fail during startup. Handle this more gracefully.
298 Severity : enhancement
300 Description: Add -o localflock option to simulate old noflock
302 Details : This will achieve local-only flock/fcntl locks
308 Description: log_commit_thread vs filter_destroy race leads to crash
309 Details : Take import reference before releasing llog record semaphore
314 Description: Wrong request locking in request set processing
315 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
316 imp_delayed_list, in this place should be used imp_lock.
319 Frequency : when reconnection
321 Description: Grant Leak when osc reconnect to OST
322 Details : When osc reconnect ost, OST(filter) should check whether it
323 should grant more space to client by comparing fed_grant and
324 cl_avail_grant, and return the granted space to client instead
325 of "new granted" space, because client will call osc_init_grant
326 to update the client grant space info.
329 Frequency : when client reconnect to OST
331 Description: Grant Leak when osc do resend and replay bulk write
332 Details : When osc reconnect to OST, OST(filter)should clear grant info of
333 bulk write request, because the grant info while be sync between
334 OSC and OST when reconnect, and we should ignore the grant info
335 these of resend/replay write req.
340 Description: Grant space more than avaiable left space sometimes.
341 Details : When then OST is about to be full, if two bulk writing from
342 different clients came to OST. Accord the avaliable space of the
343 OST, the first req should be permitted, and the second one
344 should be denied by ENOSPC. But if the seconde arrived before
345 the first one is commited. The OST might wrongly permit second
346 writing, which will cause grant space > avaiable space.
349 Frequency : when client is evicted
351 Description: Grant might be wrongly erased when osc is evicted by OST
352 Details : when the import is evicted by server, it will fork another
353 thread ptlrpc_invalidate_import_thread to invalidate the
354 import, where the grant will be set to 0. While the original
355 thread will update the grant it got when connecting. So if
356 the former happened latter, the grant will be wrongly errased
357 because of this race.
362 Description: Checking Stale with correct fid
363 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
364 to check whether it is stale, because sometimes, we want the
365 enqueue happened anyway, and op_data.fid2 will not be initialized.
367 Severity : enhancement
369 Description: update patchless client
370 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
373 Frequency : only with 2.4 kernel
375 Description: random memory corruption
376 Details : size of struct ll_inode_info is to big for union inode.u and this
377 can be cause of random memory corruption.
382 Description: Memory leak in recovery
383 Details : Lov_mds_md was not free in an error handler in mds_create_object.
384 It should also check obd_fail before fsfilt_start, otherwise if
385 fsfilt_start return -EROFS,(failover mds during mds recovery).
386 then the req will return with repmsg->transno = 0 and rc = EROFS.
387 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
388 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
389 be freed no matter whether fsfilt_commit success or not.
392 Frequency : only with huge count clients
394 Description: Prevents from taking the superblock lock in llap_from_page for
396 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
397 for llap_from_page call in ll_removepage prevents from taking the
398 superblock lock for a soon died page.
403 Description: Not check open intent error before release open handle
404 Details : in some rare cases, the open intent error is not checked before
405 release open handle, which may cause
406 ASSERTION(open_req->rq_transno != 0), because it tries to release
407 the failed open handle.
412 Description: Set cat log bitmap only after create log success.
413 Details : in some rare cases, the cat log bitmap is set too early. and it
414 should be set only after create log success.
418 Description: Accessing a block bevice can re-enable I/O when Lustre is
419 tearing down a device.
420 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
425 Description: service threads may hog cpus when there are a lot of requests
427 Details : Insert cond_resched to give other threads a chance to use some of
433 Description: the cat log was not initialized in recovery
434 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
435 unlink log on mds will not be initialized until mds post
436 recovery. And also in mds post recovery, the unlink log will
437 initialization will be done asynchronausly, so there will be race
438 between add unlink log and unlink log initialization.
442 Description: brw_stats were being printed incorrectly
443 Details : brw_stats were being printed as log2 but all of them were not
444 recorded as log2. Also remove some code duplication arising from
445 filter_tally_{read,write}.
449 Frequency : rare, only in recovery.
450 Description: ASSERTION(req->rq_type != LI_POISON) failed
451 Details : imp_lock should be held while iterating over imp_sending_list for
452 prevent destroy request after get timeout in ptlrpc_queue_wait.
456 Description: replay-single.sh test 52 fails
457 Details : A lock's skiplist need to be cleanup when it being unlinked
458 from its resource list.
462 Description: Short directio read returns full requested size rather than
464 Details : Direct I/O operations should return actual amount of bytes
465 transferred rather than requested size.
467 Severity : enhancement
469 Description: metadata RPC reduction (e.g. for rm performance)
470 Details : decrease the amount of synchronous RPC between clients and servers
471 by canceling conflicing lock before the operation on the client side
472 and packing thier handles into the main operation RPC to server.
474 --------------------------------------------------------------------------------
476 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
481 Frequency : on some architectures
483 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
484 Details : Uninitialized flags sometimes cause configuration commands to
488 Frequency : patchless clients only
490 Description: missing __iget() symbol export
491 Details : The __iget() symbol export is missing. To avoid the need for
492 this on patchless clients the deathrow inode reaper is turned
493 off, and we depend on the VM to clean up old inodes. This
494 dependency was during via the fix for bug 12181.
496 --------------------------------------------------------------------------------
498 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
500 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
501 INTEROPERATE with older versions automatically. In many cases a
502 special upgrade step is needed. Please read the
503 user documentation before upgrading any part of a live system.
504 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
505 version will not interoperate with 1.6 betas before beta5 (1.5.95).
506 * WARNING: Lustre configuration and startup changes are required with
507 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
512 Severity : enhancement
514 Description: MountConf
515 Details : Lustre configuration is now managed via mkfs and mount
516 commands instead of lmc and lconf. New obd types (MGS, MGC)
517 are added for dynamic configuration management. See
518 https://mail.clusterfs.com/wikis/lustre/MountConf for
521 Severity : enhancement
523 Description: dynamic OST addition
524 Details : OSTs can now be added to a live filesystem
526 Severity : enhancement
528 Description: startup order invariance
529 Details : MDTs and OSTs can be started in any order. Clients only
530 require the MDT to complete startup.
532 Severity : enhancement
534 Description: parallel, asynchronous orphan cleanup
535 Details : orphan cleanup is now performed in separate threads for each
536 OST, allowing parallel non-blocking operation.
538 Severity : enhancement
540 Description: optimized stripe assignment
541 Details : stripe assignments are now made based on ost space available,
542 ost previous usage, and OSS previous usage, in order to try
543 to optimize storage space and networking resources.
545 Severity : enhancement
547 Description: Permanently set tunables
548 Details : All writable /proc/fs/lustre tunables can now be permanently
549 set on a per-server basis, at mkfs time or on a live
552 Severity : enhancement
554 Description: Lustre message v2
555 Details : Add lustre message format v2.
557 Severity : enhancement
559 Description: client OST exclusion list
560 Details : Clients can be started with a list of OSTs that should be
561 declared "inactive" for known non-responsive OSTs.
565 Description: SPEC SFS validation failure on NFS v2 over lustre.
566 Details : Changes the blocksize for regular files to be 2x RPC size,
567 and not depend on stripe size.
569 Severity : enhancement
571 Description: Multiple MD RPCs in flight.
572 Details : Further unserialise some read-only MDS RPCs - learn about intents.
573 To avoid overly-overloading MDS, introduce a limit on number of
574 MDS RPCs in flight for a single client and add /proc controls
575 to adjust this limit.
577 Severity : enhancement
579 Description: client read/write statistics
580 Details : Add client read/write call usage stats for performance
581 analysis of user processes.
582 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
583 file access. extents_stats shows chunk size distribution.
584 extents_stats_per_process show chunk size distribution per
587 Severity : enhancement
589 Description: mds statistics
590 Details : Add detailed mds operations statistics in
591 /proc/fs/lustre/mds/*/stats.
595 Description: Failure of copying files with lustre special EAs.
596 Details : Client side always return success for setxattr call for lustre
597 special xattr (currently only "trusted.lov").
602 Description: Refcount LNET uuids
603 Details : The global LNET uuid list grew linearly with every startup;
604 refcount repeated list entries instead of always adding to
607 Severity : enhancement
609 Description: Dynamic service threads
610 Details : Within a small range, start extra service threads
611 automatically when the request queue builds up.
614 Frequency : mixed-endian client/server environments
616 Description: mixed-endian crashes
617 Details : The new msg_v2 system had some failures in mixed-endian
620 Severity : enhancement
622 Description: Easy OST removal
623 Details : OSTs can be permanently deactivated with e.g. 'lctl
624 conf_param lustre-OST0001.osc.active=0'
626 Severity : enhancement
628 Description: MGS proc entries
629 Details : Added basic proc entries for the MGS showing what filesystems
632 Severity : enhancement
634 Description: provide MGS failover
635 Details : Added config lock reacquisition after MGS server failover.
637 Severity : enhancement
639 Description: add Linux 2.4 support
640 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
644 Description: a large application tries to do I/O to the same resource and dies
646 Details : Check the req->rq_arrival time after the call to
647 ost_brw_lock_get(), but before we do anything about
648 processing it & sending the BULK transfer request. This
649 should help move old stale pending locks off the queue as
650 quickly as obd_timeout.
653 Frequency : when an incorrect nid is specified during startup
655 Description: ptlrpc connect to non-existant node causes kernel crash
656 Details : LNET can't be re-entered from an event callback, which
657 happened when we expire a message after the export has been
658 cleaned up. Instead, hand the zombie cleanup off to another
661 Severity : enhancement
663 Description: plain/inodebits lock performance improvement
664 Details : Grouping plain/inodebits in granted list by their request modes
665 and bits policy, thus improving the performance of search through
669 Frequency : only if OST filesystem is corrupted
671 Description: client incorrectly hits assertion in ptlrpc_replay_req()
672 Details : for a short time RPCs with bulk IO are in the replay list,
673 but replay of bulk IOs is unimplemented. If the OST filesystem
674 is corrupted due to disk cache incoherency and then replay is
675 started it is possible to trip an assertion. Avoid putting
676 committed RPCs into the replay list at all to avoid this issue.
679 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
681 Description: System hang on startup
682 Details : This bug allowed the liblustre (e.g. catamount) client to
683 return to the app before handling all startup RPCs. This
684 could leave the node unresponsive to lustre network traffic
685 and manifested as a server ptllnd timeout.
687 Severity : enhancement
689 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
690 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
691 boolean to control whether to print peer debug info when a
692 client's RPC times out.
695 Frequency : only for kernels with patches from Lustre below 1.4.3
697 Description: Remove old rdonly API
698 Details : Remove old rdonly API which unused from at least lustre 1.4.3
701 Frequency : only for devices with external journals
703 Description: Set external device read-only also
704 Details : During a commanded failover stop, we set the disk device
705 read-only while the server shuts down. We now also set any
706 external journal device read-only at the same time.
709 Frequency : when upgrading from 1.4 while trying to change parameters
711 Description: The wrong (new) MDC name was used when setting parameters for
712 upgraded MDT's. Also allows changing of OSC (and MDC)
713 parameters if --writeconf is specified at tunefs upgrade time.
716 Frequency : when setting specific ost indicies
718 Description: QOS code breaks on skipped indicies
719 Details : Add checks for missing OST indicies in the QOS code, so OSTs
720 created with --index need not be sequential.
722 Severity : enhancement
724 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
725 Details : The uninit_groups feature works in conjunction with the kernel
726 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
727 up the pass1 processing of e2fsck. This is a read-only feature
728 in ldiskfs2 only, so older kernels and current ldiskfs cannot
729 mount filesystems that have had this feature enabled.
731 Severity : enhancement
733 Description: Improve multi-block allocation algorithm to avoid fragmentation
734 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
735 allocation locality and avoid filesystem fragmentation.
737 ------------------------------------------------------------------------------
739 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
741 * Support for kernels:
742 2.4.21-47.0.1.EL (RHEL 3)
744 2.6.9-42.0.10.EL (RHEL 4)
745 2.6.12.6 vanilla (kernel.org)
746 2.6.16.27-0.9 (SLES 10)
747 * Recommended e2fsprogs version: 1.39.cfs5
749 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
753 Frequency : occasional, depends on client load and configuration
754 Bugzilla : 12181, 12203
755 Description: data loss for recently-modified files
757 Details : In some cases it is possible that recently written or created
758 files may not be written to disk in a timely manner (this should
759 normally be within 30s unless client IO load is very high).
760 The problem appears as zero-length files or files that are a
761 multiple of 1MB in size after a client crash or client eviction
762 that are missing data at the end of the file.
764 This problem is more likely to be hit on clients where files are
765 repeatedly created and unlinked in the same directory, clients
766 have a large amount of RAM, have many CPUs, the filesystem has
767 many OSTs, the clients are rebooted frequently, and/or the files
768 are not accessed by other nodes after being written.
770 The presence of the problem can be detected by looking at
771 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
772 smaller than the second (nr_unused) then dirty files will not
773 be flushed automatically to disk. "sync; sleep 10" should be
774 run several times on the node before unmounting it to update
775 Lustre (this is also safe to run on nodes without this problem).
777 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
778 that can cause this same problem, so customers using that kernel
779 also need to update the kernel in addition to Lustre. In order
780 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
782 It is normal that files written just before a client crash (less
783 than 30s) may not yet have been flushed to disk, even for local
787 Frequency : frequent on thin XT3 nodes
789 Description: UUID collision on thin XT3 Linux nodes
790 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
791 randomly, since we relied on an insufficiently-seeded PRNG.
796 Description: OSS hangs after "All ost request buffers busy"
797 Details : A deadlock between quota and journal operations caused OSS
798 hangs after printing "All ost request buffers busy."
801 Frequency : always on liblustre builds
803 Description: Cleanup compiler warnings on liblustre
806 Frequency : always on liblustre builds on XT3
808 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
812 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
813 > 32000 subdirectories
814 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
815 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
816 subdirectory count crosses 32000. This will aid e2fsck to
817 correctly handle more than 32000 subdirectories.
820 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
822 Description: System hang on startup
823 Details : This bug allowed the liblustre (e.g. catamount) client to
824 return to the app before handling all startup RPCs. This
825 could leave the node unresponsive to lustre network traffic
826 and manifested as a server ptllnd timeout.
828 Severity : enhancement
830 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
831 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
832 boolean to control whether to print peer debug info when a
833 client's RPC times out.
838 Description: make O_SYNC working on 2.6 kernels
839 Details : 2.6 kernels use different method for mark pages for write,
840 so need add a code to lustre for O_SYNC work.
845 Description: Failure to close file and release space on NFS
846 Details : Put inode details into lock acquired in ll_intent_file_open.
847 Use mdc_intent_lock in ll_intent_open to properly
848 detect all kind of errors unhandled by mdc_enqueue.
853 Description: proc file read during shutdown sometimes raced obd removal,
855 Details : Add lock to prevent obd access after proc file removal.
858 Frequency : Only for files larger than 4GB on 32-bit clients.
860 Description: improperly doing page alignment of locks
861 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
862 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
867 Description: RPCs being resent when they shouldn't be
868 Details : Some RPCs that should not be resent are being resent. This
869 can cause inconsistencies in the RPC state machine. Do not
870 resend such requests.
873 Frequency : rare, only with NFS export
875 Description: Crash on NFS re-export node
876 Details : under very unusual load conditions an assertion is hit in
877 ll_intent_file_open()
880 Frequency : only if OST filesystem is corrupted
882 Description: client incorrectly hits assertion in ptlrpc_replay_req()
883 Details : for a short time RPCs with bulk IO are in the replay list,
884 but replay of bulk IOs is unimplemented. If the OST filesystem
885 is corrupted due to disk cache incoherency and then replay is
886 started it is possible to trip an assertion. Avoid putting
887 committed RPCs into the replay list at all to avoid this issue.
892 Description: large O_DIRECT requests fail under memory pressure/fragmentation
893 Details : Large single O_DIRECT read and write calls can fail to allocate
894 a sufficiently large buffer to process the request. In case of
895 allocation failure the allocation is retried with a smaller
896 buffer and broken into smaller requests.
898 Severity : enhancement
900 Description: Add -o localflock option to simulate old noflock behaviour.
901 Details : This will achieve local-only flock/fcntl locks coherentness.
906 Description: versioning check is incomplete
907 Details : Checking the version difference of client vs. server, report
908 error if the gap is too big.
913 Description: add support PG_writeback bit
914 Details : add support for PG_writeback bit for Lustre, for more carefull
915 work with page cache in 2.6 kernel. This also fix some deadlocks
916 and remove hack for work O_SYNC with 2.6 kernel.
918 Severity : enhancement
920 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
921 Details : The uninit_groups feature works in conjunction with the kernel
922 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
923 up the pass1 processing of e2fsck. This is a read-only feature
924 in ldiskfs2 only, so older kernels and current ldiskfs cannot
925 mount filesystems that have had this feature enabled.
927 Severity : enhancement
929 Description: Improve multi-block allocation algorithm to avoid fragmentation
930 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
931 allocation locality and avoid filesystem fragmentation.
933 ------------------------------------------------------------------------------
935 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
937 * Support for kernels:
938 2.6.9-42.0.3.EL (RHEL 4)
940 2.4.21-47.0.1.EL (RHEL 3)
941 2.6.12.6 vanilla (kernel.org)
942 2.6.16.21-0.8 (SLES10)
943 * Recommended e2fsprogs version: 1.39.cfs2-0
945 * The backwards-compatible /proc/sys/portals symlink has been removed
946 in this release. Before upgrading, please ensure that you change
947 any configuration scripts or /etc/sysctl.conf files that access
948 /proc/sys/portals/* or sysctl portals.* to use the corresponding
949 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
950 in advance of the upgrade on any system running Lustre 1.4.6 or
951 newer, since /proc/sys/lnet was added in that version.
952 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
956 Frequency : only when quota is used
958 Description: avoid scanning export list for quota master
959 Details : Change the algorithms to avoid scanning export list in order
960 to improve the efficiency.
963 Frequency : MDS failover only, very rarely
965 Description: "went back in time" messages on mds failover
966 Details : The greatest transno may be lost when the current operation
967 finishes with an error (transno==0) and the client's last_rcvd
968 record is over-written. Save the greatest transno in the
969 mds_last_transno for this case.
972 Frequency : always for specific kernels and striping counts
974 Description: client may get "Matching packet too big" without ACL support
975 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
976 when trying to access files in certain configurations. The
977 clients should in fact be denied when mounting because they do
981 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
983 Description: many clients connecting with IO in progress causes connect timeouts
984 Details : Avoid synchronous journal commits to avoid delays caused by many
985 clients connecting/disconnecting when bulk IO is in progress.
986 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
987 OST_IO_PORTAL to avoid delays behind potentially many pending
991 Frequency : occasionally with multiple writers to a single file
993 Description: shared writes to file may result in wrong size reported by stat()
994 Details : Allow growing of kms when extent lock is cancelled
997 Frequency : always with random mmap IO to multi-striped file
999 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1000 Details : Only if the hole is at the end of OST object so that kms is too
1001 small. Fix is to increase kms accordingly in ll_nopage.
1004 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1006 Description: writes to a missing object would leak memory on the OST
1007 Details : If there is an inconsistency between the MDS and OST filesystems,
1008 such that the MDS references an object that doesn't exist, writes
1009 to that object will leak memory due to incorrect cleanup in the
1010 error handling path, eventually running out of memory on the OST.
1015 Description: Creating too long symlink causes lustre errors
1016 Details : Check symlink and name lengths before sending requests to MDS.
1019 Frequency : only if flock is enabled (not on by default)
1021 Description: posix locks not released on fd closure on 2.6.9+
1022 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1023 kernels, this caused such locks not to be released on fd close and
1024 then assertions on fs unmount about still used locks.
1027 Frequency : MDS failover only, very rarely
1029 Description: clients may get ASSERTION(granted_lock != NULL)
1030 Details : When request was taking a long time, and a client was resending
1031 a getattr by name lock request. The were multiple lock requests
1032 with the same client lock handle and
1033 mds_getattr_name->fixup_handle_for_resent_request found one of the
1034 lock handles but later failed with ASSERTION(granted_lock != NULL).
1039 Description: handle->h_buffer_credits > 0, assertion failure
1040 Details : h_buffer_credits is zero after truncate, causing assertion
1041 failure. This patch extends the transaction or creates a new
1045 Frequency : NFS re-export or patchless client
1046 Bugzilla : 11179, 10796
1047 Description: Crash on NFS re-export node (__d_move)
1048 Details : We do not want to hash the dentry if we don't have a lock.
1049 But if this dentry is later used in d_move, we'd hit uninitialised
1050 list head d_hash, so we just do this to init d_hash field but
1051 leave dentry unhashed.
1054 Frequency : NFS re-export or patchless client
1056 Description: NFS exports has problem with symbolic link
1057 Details : lustre client didn't properly install dentry when re-exported
1058 to NFS or running patchless client.
1061 Frequency : NFS re-export or patchless client
1063 Description: Various nfs/patchless fixes.
1064 Details : fixes reuse disconected alias for lookup process - this fixes
1065 warning "find_exported_dentry: npd != pd",
1066 fix permission error with open files at nfs.
1067 fix apply umask when do revalidate.
1070 Frequency : occasional
1072 Description: Crash on NFS re-export node
1073 Details : calling clear_page() on the wrong pointer triggered oops in
1074 generic_mapping_read().
1077 Frequency : rarely, using O_DIRECT IO
1079 Description: unaligned directio crashes client with LASSERT
1080 Details : check for unaligned buffers before trying any requests.
1083 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1085 Description: stale data returned from RAID cache
1086 Details : If only a small amount of IO is done to the RAID device before
1087 reading it again it is possible to get stale data from the RAID
1088 cache instead of reading it from disk.
1091 Frequency : always for sles10 kernel
1093 Description: sles10 support
1094 Details : ll_follow_link: compile fixes and using of nd_set_link
1095 under newer kernels.
1098 Frequency : depends on arch, kernel and compiler version, always on sles10
1101 Description: recursive or deep enough symlinks cause stack overflow
1102 Details : getting rid of large stack-allocated variable in
1106 Frequency : depends on hardware
1108 Description: lustre write performance loss in the SLES10 kernel
1109 Details : the performance loss is caused by using of write barriers in the
1110 ext3 code. The SLES10 kernel turns barrier support on by
1111 default. The fix is to undo that change for ldiskfs.
1113 ------------------------------------------------------------------------------
1115 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1117 * Support for kernels:
1118 2.6.9-42.0.3EL (RHEL 4)
1119 2.6.5-7.276 (SLES 9)
1120 2.4.21-47.0.1.EL (RHEL 3)
1121 2.6.12.6 vanilla (kernel.org)
1125 Frequency : quota enabled and large files being deleted
1127 Description: releasing more than 4GB of quota at once hangs OST
1128 Details : If a user deletes more than 4GB of files on a single OST it
1129 will cause the OST to spin in an infinite loop. Release
1130 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1135 Description: statfs data retrieved from /proc may be stale or zero
1136 Details : When reading per-device statfs data from /proc, in the
1137 {kbytes,files}_{total,free,avail} files, it may appear
1138 as zero or be out of date.
1141 Frequency : systems with MD RAID1 external journal devices
1143 Description: lconf's call to blkid is confused by RAID1 journal devices
1144 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1145 of returning all block devices that match the journal UUID.
1148 Frequency : always, for aggregate stripe size over 4GB
1150 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1151 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1152 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1155 Frequency : always if "lfs find" used on a local file/directory
1157 Description: "lfs find" segfaults if used on a local file/directory
1158 Details : The case where a directory component was not specified wasn't
1159 handled correctly. Handle this properly.
1162 Frequency : always on ppc64
1164 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1165 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1166 perform endianness conversion.
1169 Frequency : rarely (truncate to non-zero file size after write under load)
1170 Bugzilla : 10730, 10687
1171 Description: Files padded with zeros to next 4K multiple
1172 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1173 it is possible that files that are truncated to a non-zero size
1174 immediately after being written are filled with zero bytes beyond
1175 the truncated size. No file data is lost.
1177 Severity : enhancement
1179 Description: Allow recovery/failover for liblustre clients.
1180 Details : liblustre clients were unaware of failover configurations until
1183 Severity : enhancement
1185 Description: user file locks should fail when not mounting with flock option
1186 Details : Set up an error-returning stub in ll_file_operations.lock field
1187 to prevent incorrect behaviour when client is mounted without
1188 flock option. Also, set up properly f_op->flock field for
1192 Frequency : always on ia64
1194 Description: "lfs df" loops on printing out MDS statfs information
1195 Details : The obd_ioctl_data was not initialized and in some systems
1196 this caused a failure during the ioctl that did not return
1197 an error. Initialize the struct and return an error on failure.
1200 Frequency : SLES 9 only
1202 Description: Error of copying files with lustre special EAs as root
1203 Details : Client side always return success for setxattr call for lustre
1204 special xattr (currently only "trusted.lov").
1207 Frequency : rarely on clusters with both ia64+i386 clients
1209 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1210 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1211 clients are doing concurrent writes to the same file it is
1212 possible that the ia64 clients may LASSERT because the OST
1213 extent locks are not PAGE_SIZE aligned. Ensure that grown
1214 locks are always aligned on the request boundary.
1217 Frequency : specific use, occasional
1219 Description: Overwriting in use executable truncates on-disk binary image
1220 Details : If one node attempts to overwrite an executable in use by
1221 another node, we now correctly return ETXTBSY instead of
1222 truncating the file.
1224 Severity : enhancement
1226 Description: Async OSC create to avoid the blocking unnecessarily.
1227 Details : If a OST has no remain object, system will block on the creating
1228 when need to create a new object on this OST. Now, ways use
1229 pre-created objects when available, instead of blocking on an
1230 empty osc while others are not empty. If we must block, we block
1231 for the shortest possible period of time.
1236 Description: chmod on Lustre root is propagated to other clients
1237 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1238 invalid by the follow_mount time.
1241 Frequency : liblustre clients only
1243 Description: Race in 'instant cancel' lock handling could lead to such locks
1244 never to be granted in case of SMP MDS
1245 Details : Do not destroy not yet granted but cbpending locks in
1249 Frequency : replay/resend of open
1251 Description: non null lock assetion failure in mds_intent_policy
1252 Details : Trying to replay/resend lockless open requests resulted in
1253 mds_open() returning 0 with no lock. Now it sets a flag if
1254 a lock is going to be returned.
1256 Severity : enhancement
1258 Description: Checksum enhancements
1259 Details : New checksum enhancements allow for resending RPCs that failed
1262 Severity : enhancement
1264 Description: Tunables on number of dirty pages in cacche
1265 Details : Allow to set limit on number of dirty pages cached.
1270 Description: client crash on unmount - lock still has references
1271 Details : In some error handling cases it was possible to leak a lock
1272 reference on a client while accessing a file. This was not
1273 harmful to the client during operation, but would cause the
1274 client to crash when the filesystem is unmounted.
1277 Frequency : specific case, rare
1279 Description: ETXTBSY on mds though file not in use by client
1280 Details : ETXTBSY is no longer incorrectly returned when attempting to
1281 chmod or chown a directory that the user previously tried to
1282 execute or a currently-executing binary.
1285 Frequency : extremely rare except on liblustre-based clients
1287 Description: Lustre space not freed when files are deleted
1288 Details : Clean up open-unlinked files after client eviction. Previously
1289 the unlink was skipped and the files remained as orphans.
1294 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1295 Details : In some cases it was possible to send a blocking callback to
1296 a client doing a glimpse, even though that client didn't get
1297 a lock granted. When the glimpse lock is cancelled on the OST
1298 the freed lock is left on the waiting list and corrupted the list.
1301 Frequency : all core dumps
1303 Description: Broke core dumps to lustre
1304 Details : Negative dentry may be unhashed if parent does not have UPDATE
1305 lock, but some callers, e.g. do_coredump, expect dentry to be
1306 hashed after successful create, hash it in ll_create_it.
1308 ------------------------------------------------------------------------------
1310 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1312 * Support for kernels:
1313 2.6.9-42.0.2.EL (RHEL 4)
1314 2.6.5-7.276 (SLES 9)
1315 2.4.21-40.EL (RHEL 3)
1316 2.6.12.6 vanilla (kernel.org)
1320 Frequency : always on RHEL 3
1322 Description: Number of open files grows over time
1323 Details : The number of open files grows over time, whether or not
1324 Lustre is started. This was due to a filp leak introduced
1325 by one of our kernel patches.
1327 ------------------------------------------------------------------------------
1329 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1331 * Support for kernels:
1332 2.6.9-42.EL (RHEL 4)
1333 2.6.5-7.276 (SLES 9)
1334 2.4.21-40.EL (RHEL 3)
1335 2.6.12.6 vanilla (kernel.org)
1340 Bugzilla : 5719, 9635, 9792, 9684
1341 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1342 Details : If a server is under heavy load and cannot reply to new
1343 connection requests before the client resends the (re)connect,
1344 the connection handling code can behave badly if two service
1345 threads are concurrently handing separate (re)connections from
1346 the same client. Add better locking to the connection handling
1347 code, and ensure that only a single connection will be processed
1348 for a given client UUID, even if the lock is dropped.
1350 Severity : enhancement
1352 Description: add TCP zero-copy support to kernel
1353 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1354 sends if the hardware supports scatter-gather and checksumming.
1355 This allows socklnd to do client-write and server-read more
1356 efficiently and reduce CPU utilization from skbuf copying.
1359 Frequency : only if NFS exporting from client
1361 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1362 Details : If setting attributes on a file created under NFS that had
1363 never been opened it would be possible to oops the client
1364 if the file had no objects.
1367 Frequency : always for liblustre
1369 Description: liblustre client does MDS+OSTs setattr RPC for each write
1370 Details : When doing a write from a liblustre client, the client
1371 incorrectly issued an RPC to the MDS and each OST the file was
1372 striped over in order to update the timestamps. When writing
1373 with small chunks and many clients this could overwhelm the MDS
1374 with RPCs. In all cases it would slow down the write because
1375 these RPCs are unnecessary.
1377 Severity : enhancement
1379 Description: allow number of MDS service threads to be changed at module load
1380 Details : It is now possible to change the number of MDS service threads
1381 running. Adding "options mds mds_num_threads={N}" to the MDS's
1382 /etc/modprobe.conf will set the number of threads for the next
1383 time Lustre is restarted (assuming the "mds" module is also
1384 reloaded at that time). The default number of threads will
1385 stay the same, 32 for most systems.
1390 Description: OST crash if filesystem is unformatted or corrupt
1391 Details : If an OST is started on a device that has never been formatted
1392 or if the filesystem is corrupt and cannot even mount then the
1393 error handling cleanup routines would dereference a NULL pointer.
1398 Description: NULL pointer deref in llap_from_page.
1399 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1400 incorrect) mapping if the page was truncated/reclaimed while it was
1401 searched for. Check for this condition and skip such pages when
1402 doing readahead. Introduce extra check to llap_from_page() to
1403 verify page->mapping->host is non-NULL (so page is not anonymous).
1406 Frequency : Sometimes when using sys_sendfile
1408 Description: "page not covered by a lock" warnings from ll_readpage
1409 Details : sendfile called ll_readpage without right page locks present.
1410 Now we introduced ll_file_sendfile that does necessary locking
1411 around call to generic_file_sendfile() much like we do in
1415 Frequency : with certain MDS communication failures at client mount time
1417 Description: NULL pointer deref after failed client mount
1418 Details : a client connection request may delayed by the network layer
1419 and not be sent until after the PTLRPC layer has timed out the
1420 request. If the client fails the mount immediately it will try
1421 to clean up before the network times out the request. Add a
1422 reference from the request import to the obd device and delay
1423 the cleanup until the network drops the request.
1426 Frequency : occasionally during client (re)connect
1428 Description: assertion failure during client (re)connect
1429 Details : processing a client connection request may be delayed by the
1430 client or server longer than the client connect timeout. This
1431 causes the client to resend the connection request. If the
1432 original connection request is replied in this interval, the
1433 client may trip an assertion failure in ptlrpc_connect_interpret()
1434 which thought it would be the only running connect process.
1437 Frequency : only with obd_echo servers and clients that are rebooted
1439 Description: kernel BUG accessing uninitialized data structure
1440 Details : When running an obd_echo server it did not start the ping_evictor
1441 thread, and when a client was evicted an uninitialized data
1442 structure was accessed. Start the ping_evictor in the RPC
1443 service startup instead of the OBD startup.
1445 Severity : enhancement
1446 Bugzilla : 10193 (patchless)
1447 Description: Remove dependency on various unexported kernel interfaces.
1448 Details : No longer need reparent_to_init, exit_mm, exit_files,
1449 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1452 Frequency : rare (only users of deprecated and unsupported LDAP config)
1454 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1455 Details : LDAP apparently contains 'lustreName' attributes instead of
1456 'name'. A simple remapping of the name is sufficient.
1459 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1461 Description: waiting_locks_callback trips kernel BUG if client is evicted
1462 Details : Running with the dump_on_timeout debug flag turned on makes
1463 it possible that the waiting_locks_callback() can try to dump
1464 the Lustre kernel debug logs from an interrupt handler. Defer
1465 this log dumping to the expired_lock_main() thread.
1467 Severity : enhancement
1469 Description: Support NFS exporting on 2.6 kernels.
1470 Details : Implement non-rawops metadata methods for NFS server to use without
1471 changing NFS server code.
1474 Frequency : very rare (synthetic metadata workload only)
1476 Description: two racing renames might cause an MDS thread to deadlock
1477 Details : Running the "racer" program may cause one MDS thread to rename
1478 a file from being the source of a rename to being the target of
1479 a rename at exactly the same time that another thread is doing
1480 so, and the second thread has already enqueued these locks after
1481 doing a lookup of the target and is trying to relock them in
1482 order. Ensure that we don't try to re-lock the same resource.
1485 Frequency : only very large systems with liblustre clients
1487 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1488 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1489 all OSTs in parallel. This allows the eviction of stale liblustre
1490 clients to proceed much faster than if they were done in series,
1491 and also offers similar improvements for other set_info RPCs.
1496 Description: excessive CPU usage during initial read phase on client
1497 Details : During the initial read phase on a client, it would agressively
1498 retry readahead on the file, consuming too much CPU and impacting
1499 performance (since 1.4.5.8). Improve the readahead algorithm
1500 to avoid this, and also improve some other common cases (read
1501 of small files in particular, where "small" is files smaller than
1502 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1507 Description: MDS crash when receiving packet with unknown intent.
1508 Details : Do not LBUG in unknown intent case, just return -EFAULT
1510 Severity : enhancement
1511 Bugzilla : 9293, 9385
1512 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1513 Details : Do not serialize getattr (non-intent version) and statfs.
1516 Frequency : occasional, when OST network is overloaded/intermittent
1518 Description: client evicted by OST after bulk IO timeout
1519 Details : If a client sends a bulk IO request (read or write) the OST
1520 may evict the client if it is unresposive to its data GET/PUT
1521 request. This is incorrect if the network is overloaded (takes
1522 too long to transfer the RPC data) or dropped the OST GET/PUT
1523 request. There is no need to evict the client at all, since
1524 the pinger and/or lock callbacks will handle this, and the
1525 client can restart the bulk request.
1528 Frequency : Always when mmapping file with no objects
1530 Description: client crashes when mmapping file with no objects
1531 Details : Check that we actually have objects in a file before doing any
1532 operations on objects in ll_vm_open, ll_vm_close and
1538 Description: Request leak when working with deleted CWD
1539 Details : Introduce advanced request refcount tracking for requests
1540 referenced from lustre intent.
1542 Severity : Enhancement
1544 Description: Cache open file handles on client.
1545 Details : MDS now will return special lock along with openhandle, if
1546 requested and client is allowed to hold openhandle, even if unused,
1547 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1548 closing files for every read/write openration.
1550 Severity : Enhancement
1552 Description: Cache open negative dentries on client when possible.
1553 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1554 negative dentries on lock revocation.
1559 Description: Remounting a client read-only wasn't possible with a zconf mount
1560 Details : It wasn't possible to remount a client read-only with llmount.
1562 Severity : enhancement
1563 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1564 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1565 MPICH is included to add Lustre-specific ADIO interfaces.
1566 This is based closely on the UFS ADIO layer and only differs
1567 in file creation, in order to allow the OST striping to be set.
1568 This is user-contributed code and not supported by CFS.
1573 Description: extended inode attributes (immutable, append-only) work improperly
1574 when 2.4 and 2.6 kernels are used on client/server or vice versa
1575 Details : Introduce kernel-independent values for these flags.
1577 Severity : enhancement
1580 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1581 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1582 granularity. Now, Fractional MB granularity is be supported,
1583 this is very useful for low memory system.
1585 Severity : enhancement
1587 Description: Getattr by fid
1588 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1589 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1590 have locks and avoids getattr rpc requests for it on every stat.
1593 Frequency : Always, for filesystems larger than 2TB
1595 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1596 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1597 filesystems larger than 2TB to be mounted with mballoc enabled.
1600 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1602 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1603 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1604 filesystems larger than 2TB is mounted then the kernel will
1605 silently corrupt the start of the filesystem. CONFIG_LBD is
1606 enabled for all CFS-supported kernels, but the possibility of
1607 this happening with a modified kernel config exists.
1609 Severity : enhancement
1611 Description: add client O_DIRECT support for 2.6 kernels
1612 Details : It is now possible to do O_DIRECT reads and writes to files
1613 in the Lustre client mountpoint on 2.6 kernel clients.
1615 Severity : enhancement
1617 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1618 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1619 parallel, not waiting for response from every OST before sending
1620 a rpc to the next OST.
1625 Description: setattr vs write race when updating file timestamps
1626 Details : Client processes that update a file timestamp into the past
1627 right after writing to the file (e.g. tar) it is possible that
1628 the updated file modification time can be reset to the current
1629 time due to a race between processing the setattr and write RPC.
1631 Severity : enhancement
1633 Description: Bring 'lfs find' closer in line with regular Linux find.
1634 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1635 -print0 options and obtains all the needed info through the lustre
1638 Severity : enhancement
1640 Description: support up to 1024 configured devices on one node
1641 Details : change obd_dev array from statically allocated to dynamically
1642 allocated structs as they are first used to reduce memory usage
1647 Description: Flush dirty partially truncated pages during truncate
1648 Details : Immediatelly flush partially truncated pages in filter_setattr,
1649 this way we completely avoid having any pages in page cache on OST
1650 and can retire ugly workarounds during writes to flush such pages.
1655 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1656 Details : protect inode from truncation within vfs_unlink() context
1657 just take a reference before calling vfs_unlink() and release it
1658 when parent's i_sem is free.
1663 Description: last_id value checked outside lock on OST caused LASSERT failure
1664 Details : If there were multiple MDS->OST object precreate requests in
1665 flight, it was possible that the OST's last object id was checked
1666 outside a lock and incorrectly tripped an assertion. Move checks
1667 inside locks, and discard old precreate requests.
1670 Frequency : always, if extents are used on OSTs
1672 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1673 Details : OSTs using the extents format would not zero the high 16 bits of
1674 the index physical block number. This is not a problem for any
1675 OST filesystems smaller than 16TB, and no kernels support ext3
1676 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1677 new/modified files) and can be fixed for existing filesystems
1678 with e2fsprogs-1.39-cfs1.
1683 Description: import connection selection may be incorrect if timer wraps
1684 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1685 import connections to be ignored if the 32-bit jiffies counter
1686 wraps. Use a 64-bit jiffies counter.
1689 Frequency : very large clusters immediately after boot
1691 Description: LNET request buffers exhausted under heavy short-term load
1692 Details : If a large number of client requests are generated on a service
1693 that has previously never seen so many requests it is possible
1694 that the request buffer growth cannot keep up with the spike in
1695 demand. Instead of dropping incoming requests, they are held in
1696 the LND until the RPC service can accept more requests.
1699 Frequency : Sometimes during replay
1701 Description: Assertion failure in ll_local_open after replay.
1702 Details : If replay happened on an open request reply before we were able
1703 to set replay handler, reply will become not swabbed tripping the
1704 assertion in ll_local_open. Now we set the handler right after
1705 recognising of open request
1708 Frequency : very rare
1710 Description: kernel reports "badness in vsnprintf"
1711 Details : Reading from the "recovery_status" /proc file in small chunks
1712 may cause a negative length in lprocfs_obd_rd_recovery_status()
1713 call to vsnprintf() (which is otherwise harmless). Exit early
1714 if there is no more space in the output buffer.
1716 Severity : enhancement
1718 Description: clear OBD RPC statistics by writing to them
1719 Details : It is now possible to clear the OBD RPC statistics by writing
1720 to the "stats" file.
1725 Description: Client mtime is not the same on different clients after utimes
1726 Details : In some cases, the client was using the utimes() syscall on
1727 a file cached on another node. The clients now validate the
1728 ctime from the MDS + OSTs to determine which one is right.
1733 Description: Inability to activate failout mode
1734 Details : lconf script incorrectly assumed that in pythong string's numeric
1735 value is used in comparisons.
1738 Frequency : always with multiple stripes per file
1740 Description: Inefficient object allocation for mutli-stripe files
1741 Details : When selecting which OSTs to stripe files over, for files with
1742 a stripe count that divides evenly into the number of OSTs,
1743 the MDS is always picking the same starting OST for each file.
1744 Return the OST selection heuristic to the original design.
1749 Description: mount failures may take full timeout to return an error
1750 Details : Under some heavy load conditions it is possible that a
1751 failed mount can wait for the full obd_timeout interval,
1752 possibly several minutes, before reporting an error.
1753 Instead return an error as soon as the status is known.
1755 Frequency : quota enabled and large files being deleted
1757 Description: releasing more than 4GB of quota at once hangs OST
1758 Details : If a user deletes more than 4GB of files on a single OST it
1759 will cause the OST to spin in an infinite loop. Release
1760 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1765 Description: statfs data retrieved from /proc may be stale or zero
1766 Details : When reading per-device statfs data from /proc, in the
1767 {kbytes,files}_{total,free,avail} files, it may appear
1768 as zero or be out of date.
1771 Frequency : systems with MD RAID1 external journal devices
1773 Description: lconf's call to blkid is confused by RAID1 journal devices
1774 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1775 of returning all block devices that match the journal UUID.
1778 Frequency : always, for aggregate stripe size over 4GB
1780 Description: assertion fails when trying to use 4GB stripe size
1781 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1782 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1785 Frequency : always on ppc64
1787 Description: the first write on an ext3 filesystem with mballoc got stuck
1788 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1789 perform endianness conversion.
1791 ------------------------------------------------------------------------------
1793 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1795 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1796 INTEROPERATE with older versions automatically. Please read the
1797 user documentation before upgrading any part of a live system.
1798 * WARNING: Lustre networking configuration changes are required with
1799 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1802 * Support for kernels:
1803 2.6.9-22.0.2.EL (RHEL 4)
1804 2.6.5-7.244 (SLES 9)
1805 2.6.12.6 vanilla (kernel.org)
1808 Severity : enhancement
1809 Bugzilla : 7981/8208
1810 Description: Introduced Lustre Networking (LNET)
1811 Details : LNET is new networking infrastructure for Lustre, it includes
1812 a reorganized network configuration mode (see the user
1813 documentation for full details) as well as support for routing
1814 between different network fabrics. Lustre Networking Devices
1815 (LNDs) for the supported network fabrics have also been
1816 created for this new infrastructure.
1818 Severity : enhancement
1819 Description: Introduced Access control lists
1820 Details : clients can set ACLs on files and directories in order to have
1821 more fine-grained permissions than the standard Unix UGO+RWX.
1822 The MDS must be started with the "-o acl" mount option.
1824 Severity : enhancement
1825 Description: Introduced filesystem quotas
1826 Details : Administrators may now establish per-user quotas on the
1829 Severity : enhancement
1831 Description: Configuration change for the XT3
1832 The PTLLND is now used to run Lustre over Portals on the XT3
1833 The configure option(s) --with-cray-portals are no longer used.
1834 Rather --with-portals=<path-to-portals-includes> is used to
1835 enable building on the XT3. In addition to enable XT3 specific
1836 features the option --enable-cray-xt3 must be used.
1841 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1842 Details : A race between allocating a new llcd and re-getting the llcd_lock
1843 allowed another thread to grab newly-allocated llcd.
1845 Severity : enhancement
1847 Description: 2.6 OST async journal commit and locking fix to improve performance
1848 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1849 2.6 kernels are now async as they already were in 2.4 kernels so
1850 that they can commit concurrently with the network bulk transfer.
1851 For block-allocated files the filter allocation semaphore is held
1852 to avoid filesystem fragmentation during allocation. BKL lock
1853 removed for 2.6 xattr operations where it is no longer needed.
1858 Description: lconf incorrectly determined whether two IP networks could talk
1859 Details : In some more complicated routing and multiple-network
1860 configurations, lconf will avoid trying to make a network
1861 connection to a disjoint part of the IP space. It was doing the
1862 math incorrectly for one set of cases.
1867 Description: Fix for potential infinite loop processing records in an llog.
1868 Details : If an llog record is corrupted/zeroed, it is possible to loop
1869 forever in llog_process(). Validate the llog record length
1870 and skip the remainder of the block on error.
1873 Frequency : occasional (liblustre only)
1875 Description: liblustre could not open files whose last component is a symlink
1876 Details : sysio_path_walk() would incorrectly pass the open intent to
1877 intermediate path components.
1880 Frequency : rare (liblustre only with non-standard tuning)
1881 Bugzilla : 7201 (7350)
1882 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1883 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1884 after it has been released, only do this for referenced locks
1886 Severity : enhancement
1888 Description: specifying an (invalid) directory default stripe_size of -1
1889 would reset the directory default striping
1890 Details : stripe_size -1 was used internally to signal directory stripe
1891 removal, now use "all default" to signal dir stripe removal
1892 as a directory striping of "all default" is not useful
1895 Frequency : common for large clusters running liblustre clients
1897 Description: doing an ls when liblustre clients are running is slow
1898 Details : sending a glimpse AST to a liblustre client waits for every AST
1899 to time out, as liblustre clients will not respond. Since they
1900 cannot cache data we refresh the OST lock LVB from disk instead.
1902 Severity : enhancement
1904 Description: doing an ls at the same time as file IO can be slow
1905 Details : enqueue and other "small" requests can be blocked behind many
1906 large IO requests. Create a new OST IO portal for non-IO
1907 requests so they can be processed faster.
1910 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1912 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1913 created file with mode 000
1914 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1915 allow an ftruncate() as a non-root user to a file with mode 000.
1916 Lustre now respects this flag to disable mode checks when
1917 truncating a file owned by the user
1920 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1923 Description: Revoking locks from clients that went dead or catatonic might take
1925 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1926 cancellation of such locks instant on servers without waiting for
1927 any reply from clients. Clients drops these locks when cancel
1928 notification from server is received without replying.
1931 Frequency : liblustre-only, when liblustre client dies or becomes busy
1933 Description: Doing ls on Linux clients can take a long time with active
1935 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1936 granting such locks to it in the first place if possible. Locks
1937 are taken by proxy on the OST during the read or write and
1938 dropped immediately afterward. Add connect flags handling, do
1939 not grant locks to liblustre clients for glimpse ASTs.
1941 Severity : enhancement
1943 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1944 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1945 sequential reads and behaves badly with applications doing random
1946 reads. Improve it by reading ahead at least read region, and
1947 avoiding excessive large RPC for small reads.
1949 Severity : enhancement
1951 Description: Creating more than 1000 files for a single job may cause a load
1952 imbalance on the OSTs if there are also a large number of OSTs.
1953 Details : qos_prep_create() uses an OST index reseed value that is an
1954 even multiple of the number of available OSTs so that if the
1955 reseed happens in the middle of the object allocation it will
1956 still utilize the OSTs as uniformly as possible.
1961 Description: OST or MDS may oops in ping_evictor_main()
1962 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1963 but doesn't restart at beginning of obd_exports_timed list
1966 Severity : enhancement
1968 Description: improve by-nid export eviction on the MDS and OST
1969 Details : allow multiple exports with the same NID to be evicted at one
1970 time without re-searching the exports list.
1973 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1975 Description: MDS may oops in groups_free()
1976 Details : in rare race conditions a newly allocated group_info struct is
1977 freed again, and this can be NULL. The 2.4 compatibility code
1978 for groups_free() checked for a NULL pointer, but 2.6 did not.
1981 Frequency : common for liblustre clients doing little filesystem IO
1982 Bugzilla : 9352, 7313
1983 Description: server may evict liblustre clients accessing contended locks
1984 Details : if a client is granted a lock or receives a completion AST
1985 with a blocking AST already set it would not reply to the AST
1986 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1989 Frequency : lfs setstripe, only systems with more than 160 OSTs
1991 Description: unable to set striping with a starting offset beyond OST 160
1992 Details : llapi_create_file() incorrectly limited the starting stripe
1993 index to the maximum single-file stripe count.
1996 Frequency : LDAP users only
1998 Description: lconf did not handle in-kernel recovery with LDAP properly
1999 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2001 Severity : enhancement
2003 Description: bind OST threads to NUMA nodes to improve performance
2004 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2005 node and do their allocations there to localize memory access
2007 Severity : enhancement
2009 Description: llmount can determine client NID directly from Myrinet (GM)
2010 Details : the client NID code from gmnalnid was moved directly into
2011 llmount, removing the need to use this or specifying the
2012 client NID explicitly when mounting GM clients with zeroconf
2015 Frequency : if client is started with down MDS
2017 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2018 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2019 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2020 ptlrpc_import_delayed_req() because we didn't check if the
2021 request was interrupted, and we also didn't break out of the
2022 event loop if there was no timeout
2027 Description: data loss during non-page-aligned writes to a single file from
2028 both multiple nodes and multiple threads on one node at same time
2029 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2030 inconsistency led to false short-reads, that were cached and later
2031 used by ->prepare_write() to fill in partially written page,
2032 leading to data loss.
2035 Frequency : always, if lconf --abort_recovery used
2037 Description: lconf --abort_recovery fails with 'Operation not supported'
2038 Details : lconf was attempting to abort recovery on the MDT device and not
2041 Severity : enhancement
2043 Description: remove cleanup logs
2044 Details : replace lconf-generated cleanup logs with lustre internal
2045 cleanup routines. Eliminates the need for client-cleanup and
2048 Severity : enhancement
2050 Description: add support for EAs (user and system) on lustre filesystems
2051 Details : it is now possible to store extended attributes in the Lustre
2052 client filesystem, and with the user_xattr mount option it
2053 is possible to allow users to store EAs on their files also
2055 Severity : enhancement
2057 Description: Add possibility (config option) to show minimal available OST free
2059 Details : When compiled with --enable-mindf configure option, statfs(2)
2060 (and so, df) will return least minimal free space available from
2061 all OSTs as amount of free space on FS, instead of summary of
2062 free spaces of all OSTs.
2064 Severity : enhancement
2066 Description: do not expand extent locks acquired on OST-side
2067 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2068 by server: they are not cached anyway.
2071 Frequency : when mmap is used/binaries executed from Lustre
2073 Description: Unmmap pages before throwing them away from read cache.
2074 Details : llap_shrink cache now attempts to unmap pages before discarding
2075 them (if unmapping failed - do not discard). SLES9 kernel has
2076 extra checks that trigger if this unmapping is not done first.
2081 Description: lconf didn't resolve symlinks before checking to see whether a
2082 given mountpoint was already in use
2085 Frequency : when migrating failover services
2086 Bugzilla : 6395, 9514
2087 Description: When migrating a subset of services from a node (e.g. failback
2088 from a failover service node) the remaining services would
2089 time out and evict clients.
2090 Details : lconf --force (implied by --failover) sets the global obd_timeout
2091 to 5 seconds in order to quickly disconnect, but this caused
2092 other RPCs to time out too quickly. Do not change the global
2093 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2095 Severity : enhancement
2096 Frequency : if MDS is started with down OST
2097 Bugzilla : 9439,5706
2098 Description: Allow startup/shutdown of an MDS without depending on the
2099 availability of the OSTs.
2100 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2101 Add appropriate locking and lov-osc refcounts for safe
2102 cleaning. Add osc abort_inflight calls in case the
2103 synchronize never started.
2106 Frequency : occasional (Cray XT3 only)
2108 Description: root not authorized to access files in CRAY_PORTALS environment
2109 Details : The client process capabilities were not honoured on the MDS in
2110 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2111 been accessed by an authorized user then root was able to access
2112 the file on the local client also. The root user capabilities
2113 are now allowed on the MDS, as this environment has secure UID.
2116 Frequency : occasional
2118 Description: ldiskfs "too long searching" message happens too often
2119 Details : A debugging message (otherwise harmless) prints too often on
2120 the OST console. This has been reduced to only happen when
2121 there are fragmentation problems on the filesystem.
2126 Description: Division by zero in statfs when all OSCs are inactive
2127 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2128 lov_statfs divides by value returned by lov_get_stripecnt().
2132 Bugzilla : 9489, 3273
2133 Description: First write from each client to each OST was only 4kB in size,
2134 to initialize client writeback cache, which caused sub-optimal
2135 RPCs and poor layout on disk for the first writen file.
2136 Details : Clients now request an initial cache grant at (re)connect time
2137 and so that they can start streaming writes to the cache right
2138 away and always do full-sized RPCs if there is enough data.
2139 If the OST is rebooted the client also re-establishes its grant
2140 so that client cached writes will be honoured under the grant.
2145 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2146 Details : Now I/O RPCs go to different portal number and (presumably) fast
2147 lock requests (and glimses) and other RPCs get their own service
2148 threads pool that should be able to service those RPCs
2151 Severity : enhancement
2153 Description: Ability to exchange lustre version between client and servers and
2154 issue warnings at client side if client is too old. Also for
2155 liblustre clients there is ability to refuse connection of too old
2157 Details : New 'version' field is added to connect data structure that is
2158 filled with version info. That info is later checked by server and
2162 Frequency : rare, liblustre only.
2163 Bugzilla : 9296, 9581
2164 Description: Two simultaneous writes from liblustre at offset within same page
2165 might proceed at the same time overwriting eachother with stale
2167 Details : I/O lock withing llu_file_prwv was released too early, before data
2168 actually was hitting the wire. Extended lock-holding time until
2169 server acknowledges receiving data.
2172 Frequency : extremely rare. Never observed in practice.
2174 Description: avoid generating lustre_handle cookie of 0.
2175 Details : class_handle_hash() generates handle cookies by incrementing
2176 global counter, and can hit 0 occasionaly (this is unlikely, but
2177 not impossible, because initial value of cookie counter is
2178 selected randonly). Value of 0 is used as a sentinel meaning
2179 "unassigned handle" --- avoid it. Also coalesce two critical
2180 sections in this function into one.
2182 Severity : enhancement
2184 Description: allow liblustre clients to delegate truncate locking to OST
2185 Details : To avoid overhead of locking, liblustre client instructs OST to
2186 take extent lock in ost_punch() on client's behalf. New connection
2187 flag is added to handle backward compatibility.
2189 Severity : enhancement
2190 Bugzilla : 4928, 7341, 9758
2191 Description: allow number of OST service threads to be specified
2192 Details : a module parameter allows the number of OST service threads
2193 to be specified via "options ost ost_num_threads={N}" in the
2194 OSS's /etc/modules.conf or /etc/modprobe.conf.
2198 Bugzilla : 6146, 9635, 9895
2199 Description: servers crash with bad pointer in target_handle_connect()
2200 Details : In rare cases when a client is reconnecting it was possible that
2201 the connection request was the last reference for that export.
2202 We would temporarily drop the export reference and get a new
2203 one, but this may have been the last reference and the export
2204 was just destroyed. Get new reference before dropping old one.
2206 Severity : enhancement
2207 Frequency : if client is started with failover MDS
2209 Description: Allow multiple MDS hostnames in the mount command
2210 Details : Try to read the configuration from all specified MDS
2211 hostnames during a client mount in case the "primary"
2214 Severity : enhancement
2216 Description: Stop sending data to evicted clients as soon as possible.
2217 Details : Check if the client we are about to send or are sending data to
2218 was evicted already. (Check is done every second of waiting,
2219 for which l_wait_event interface was extended to allow checking
2220 of exit condition at specified intervals).
2223 Frequency : rare, normally only when NFS exporting is done from client
2225 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2227 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2230 Frequency : always, if srand() is called before liblustre initialization
2232 Description: Liblustre uses system PRNG disturbing its usage by user application
2233 Details : Introduce internal to lustre fast and high-quality PRNG for
2234 lustre usage and make liblustre and some other places in generic
2235 lustre code to use it.
2237 Severity : enhancement
2238 Bugzilla : 9477, 9557, 9870
2239 Description: Verify that the MDS configuration logs are updated when xml is
2240 Details : Check if the .xml configuration logs are newer than the config
2241 logs stored on the MDS and report an error if this is the case.
2242 Request --write-conf, or allow starting with --old_conf.
2244 Severity : enhancement
2246 Description: Handle symlinks in the path when checking if Lustre is mounted.
2247 Details : Resolve intermediate symlinks when checking if a client has
2248 mounted a filesystem to avoid duplicate client mounts.
2253 Description: lconf can hit an error exception but still return success.
2254 Details : The lconf command catches the Command error exception at the top
2255 level script context and will exit with the associated exit
2256 status, but doesn't ensure that this exit status is non-zero.
2261 Description: failure of ptlrpc thread startup can cause oops
2262 Details : Starting a ptlrpc service thread can fail if there are a large
2263 number of threads or the server memory is very fragmented.
2264 Handle this without oopsing.
2267 Frequency : always, only if liblustre and non-default acceptor port was used
2269 Description: liblustre cannot connect to servers with non-default acceptor port
2270 Details : tcpnal_set_default_params() was not called and was therefore
2271 ignoring the environment varaible TCPNAL_PORT, as well as other
2272 TCPNAL_ environment variables
2277 Description: two objects could be created on the same OST for a single file
2278 Details : If an OST is down, in some cases it was possible to create two
2279 objects on a single OST for a single file. No problems other
2280 than potential performance impact and spurious error messages.
2284 Bugzilla : 5681, 9562
2285 Description: Client may oops in ll_unhash_aliases
2286 Details : Client dcache may become inconsistent in race condition.
2287 In some cases "getcwd" can fail if the current directory is
2293 Description: Inode refcounting problems in NFS export code
2294 Details : link_raw functions used to call d_instantiate without obtaining
2295 extra inode reference first.
2299 Bugzilla : 9942, 9903
2300 Description: Referencing freed requests leading to crash, memleaks with NFS.
2301 Details : We used to require that call to ll_revalidate_it was always
2302 followed by ll_lookup_it. Also with revalidate_special() it is
2303 possible to call ll_revalidate_it() twice for the same dentry
2304 even if first occurence returned success. This fix changes semantic
2305 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2306 reference on a request referred from the intent.
2307 ll_intent_release() then releases such a request.
2310 Frequency : rare, normally benchmark loads only
2312 Description: unlinked inodes were kept in memory on the client
2313 Details : If a client is repeatedly creating and unlinking files it
2314 can accumulate a lot of stale inodes in the inode slab cache.
2315 If there is no other client load running this can cause the
2316 client node to run out of memory. Instead flush old inodes
2317 from client cache that have the same inode number as a new inode.
2320 Frequency : SLES9 2.6.5 kernel and long filenames only
2321 Bugzilla : 9969, 10379
2322 Description: utime reports stale NFS file handle
2323 Details : SLES9 uses out-of-dentry names in some cases, which confused
2324 the lustre dentry revalidation. Change it to always use the
2328 Frequency : rare, unless heavy write-truncate concurrency is continuous
2329 Bugzilla : 4180, 6984, 7171, 9963, 9331
2330 Description: OST becomes very slow and/or deadlocked during object unlink
2331 Details : filter_destroy() was holding onto the parent directory lock
2332 while truncating+unlinking objects. For very large objects this
2333 may block other threads for a long time and slow overall OST
2334 responsiveness. It may also be possible to get a lock ordering
2335 deadlock in this case, or run out of journal credits because of
2336 the combined truncate+unlink. Solution is to do object truncate
2337 first in one transaction without parent lock, and then do the
2338 final unlink in a new transaction with the parent lock. This
2339 reduces the lock hold time dramatically.
2342 Frequency : rare, 2.4 kernels only
2344 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2345 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2346 before unmounting their filesystem in order to avoid deadlock.
2347 The kernel_locked() function in 2.4 kernels only checks whether
2348 the kernel lock is held, not whether it is this process that is
2349 holding it as 2.6 kernels do.
2354 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2355 Details : The client ptlrpc code may be trying to reconnect to a down
2356 server before a previous connection attempt has timed out.
2357 Increase the reconnect interval to be longer than the connection
2358 timeout interval to avoid sending duplicate connections to
2362 Frequency : echo_client brw_test command
2364 Description: fix echo_client to work with OST preallocated code
2365 Details : OST preallocation code (5137) didn't take echo_client IO path
2366 into account: echo_client calls filter methods outside of any
2367 OST thread and, hence, there is no per-thread preallocated
2368 pages and buffers to use. Solution: hijack pga pages for IO. As
2369 a byproduct, this avoids unnecessary data copying.
2373 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2374 Description: Client can oops in mdc_commit_close() after open replay
2375 Details : It was possible for the MDS to return an open request with no
2376 transaction number in mds_finish_transno() if the client was
2377 evicted, but without actually returning an error. Clients
2378 would later try to replay that open and may trip an assertion
2379 Simplify the client close codepath, and always return an error
2380 from the MDS in case the open is not successful.
2383 Frequency : rare, 2.6 OSTs only
2385 Description: OST may deadlock under high load on fragmented files
2386 Details : If there was a heavy load and highly-fragmented OST filesystems
2387 it was possible to have all the OST threads deadlock waiting on
2388 allocation of biovecs, because the biovecs were not released
2389 until the entire RPC IO was completed. Instead, release biovecs
2390 as soon as they are complete to ensure forward IO progress.
2392 Severity : enhancement
2394 Description: Support for specifying external journal device at mount
2395 Details : If an OST or MDS device is formatted with an external journal
2396 device, this device major/minor is stored in the ext3 superblock
2397 and may not be valid for failover. Allow detecting and
2398 specifying the external journal at mount time.
2403 Description: Mounting an MDS with pending unlinked files may cause oops
2404 Details : target_finish_recovery() calls mds_postrecov() which returned
2405 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2406 considers this an error and immediately begins cleaning up the
2407 lov, just after starting the mds_lov process
2409 Severity : enhancement
2411 Description: Implement 'lfs df' to report actual free space on per-OST basis
2412 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2413 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2414 usage of objects; '-h' to report in human readable format.
2416 ------------------------------------------------------------------------------
2418 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2425 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2426 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2427 at startup during orphan unlinking. Instead, skip buddy update
2428 before setup, as it will be regenerated after recovery is complete.
2433 Description: If an OST is inactive, its locks might reference stale inodes.
2434 Details : lov_change_cbdata() must iterate over all namespaces, even if
2435 they are inactive to clear inode references from the lock.
2437 Severity : enhancement
2438 Frequency : occasional, if non-standard max_dirty_mb used
2440 Description: Client will block write RPCs if not enough grant
2441 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2442 then the client will block writes while waiting for another RPC
2443 to complete instead of consuming its dirty limit. With change
2444 we get improved performance when max_dirty_mb is small.
2446 Severity : enhancement
2447 Bugzilla : 3389, 6253
2448 Description: Add support for supplementary groups on the MDS.
2449 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2450 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2451 lookups for user supplementary groups into a cache.
2455 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2456 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2457 for liblustre O_EXCL is enforced by MDS.
2461 Description: readdir never returns NULL in liblustre.
2462 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2463 dentry in struct dirent.
2467 Frequency : liblustre only, depends on application IO pattern
2468 Description: liblustre clients evicted if not contacting servers
2469 Details : Don't put liblustre clients into the ping_evictor list, so
2470 they will not be evicted by the pinger ever.
2472 Severity : enhancement
2474 Description: Add ability to evict clients by NID from MDS.
2475 Details : By echoing "nid:$NID" string into
2476 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2477 $NID would be instantly evicted from this MDS and from all active
2478 OSTs connected to it.
2482 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2483 Details : lookup_it_finish() used to query file size from OSTs that was not
2488 Description: service threads change working directory to that of init
2489 Details : Starting lustre service threads may pin the working directory
2490 of the parent thread, making that filesystem busy. Threads
2491 now change to the working directory of init to avoid this.
2495 Frequency : during shutdown only
2496 Description: shutdown with a failed MDS or OST can cause unmount to hang
2497 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2502 Frequency : 2.6 only
2503 Description: chmod/chown may include an extra supplementary group
2504 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2505 supplementary group and if none is specified this is used.
2508 Bugzilla : 5479 (6816)
2510 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2511 Details : If lookup is in progress on a file that is unlinked we might try
2512 to revalidate the inode and fail in revalidate after lookup is
2513 complete and ll_file_open() enqueues the open again but
2514 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2517 Frequency : always, if lconf --abort_recovery used
2519 Description: lconf --abort_recovery fails with 'Operation not supported'
2520 Details : lconf was attempting to abort recovery on the MDT device and not
2523 ------------------------------------------------------------------------------
2525 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2530 Frequency : rare (only unsupported configurations with a node running as an
2532 Bugzilla : 6514, 5137
2533 Description: Mounting a Lustre file system on a node running as an OST could
2535 Details : OSTs now preallocates memory needed to write out data at
2536 startup, instead of when needed, to avoid having to
2537 allocate memory in possibly low memory situations.
2538 Specifically, if the file system is mounted on on OST,
2539 memory pressure could force it to try to write out data,
2540 which it needed to allocate memory to do. Due to the low
2541 memory, it would be unable to do so and the node would
2542 become unresponsive.
2544 Severity : enhancement
2546 Description: Addition of lconf --service command line option
2547 Details : lconf now accepts a '--service <arg>' option, which is
2548 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2550 Severity : enhancement
2552 Description: Failover mode is now the default for OSTs.
2553 Details : By default, OSTs will now run in failover mode. To return to
2554 the old behaviour, add '--failout' to the lmc line for OSTs.
2556 Severity : enhancement
2558 Description: Health checks are now provided for MDS and OSTs
2559 Details : Additional detailed health check information on MSD and OSTs
2560 is now provided through the procfs health_check value.
2563 Frequency : occasional, depends on IO load
2565 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2566 after numerous create/delete cycles
2567 Details : The ext3 inode allocation policy would not allocate new inodes
2568 very well on the OSTs because there are no new directories
2569 being created. Instead we look for groups with free space if
2570 the parent directories are nearly full.
2575 Description: Network or server problems during mount may cause partially
2576 mounted clients instead of returning an error.
2577 Details : The config llog parsing code may overwrite the error return
2578 code during mount error handling, returning success instead
2584 Description: MDS can fail to allocate large reply buffers
2585 Details : After long uptimes the MDS can fail to allocate large reply
2586 buffers (e.g. zconf client mount config records) due to memory
2587 fragmentation or consumption by the buffer cache. Preallocate
2588 some large reply buffers so that these replies can be sent even
2589 under memory pressure.
2593 Frequency : rare (liblustre)
2594 Description: fsx running with liblustre complained that using truncate() to
2595 extend the file doesn't work. This patch corrects that issue.
2596 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2597 ATTR_SIZE and lsm use in llu_setattr_raw.
2601 Frequency : rare, only 2.6 kernels
2602 Description: Unusual file access patterns on the MDS may result in inode
2603 data being lost in very rare circumstances.
2604 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2605 code in the 2.6 kernel caused the inode and/or EA data not to
2606 be read from disk, causing single-file corruption.
2610 Frequency : rare, only 2.6 filesystems using extents
2611 Description: Heavy concurrent write and delete load may cause data corruption.
2612 Details : It was possible under high-load situations to have an extent
2613 metadata block in the block device cache from a just-unlinked
2614 file overwrite a newly-allocated data block. We now unmap any
2615 metadata buffers that alias just-allocated data blocks.
2619 Frequency : filesystems with default stripe_count larger than 77
2620 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2621 Details : lconf specifies an inode size of 4096 bytes when the default
2622 stripe_count is larger than 77. This conflicts with the default
2623 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2624 this case to avoid pinning too much memory for large EAs.
2626 ------------------------------------------------------------------------------
2628 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2633 Frequency : rare (extremely heavy IO load with hundreds of clients)
2635 Description: Client is evicted, gets IO error writing to file
2636 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2637 caused clients to be evicted for AST timeouts. The fixes in
2638 bug 5192 mean we no longer need to have such short AST timeouts
2639 so ldlm_timeout has been increased.
2642 Frequency : occasional during --force or --failover shutdown under load
2643 Bugzilla : 5949, 4834
2644 Description: Server oops/LBUG if stopped with --force or --failover under load
2645 Details : a collection of import/export refcount and cleanup ordering
2646 issues fixed for safer force cleanup
2649 Frequency : only filesystems larger than 120 OSTs
2650 Bugzilla : 5990, 6223
2651 Description: lfs getstripe would oops on a very large filesystem
2652 Details : lov_getconfig used kfree on vmalloc'd memory
2655 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2656 Bugzilla : 6242, 6243
2657 Description: reading from files that had been truncated to a non-zero size
2658 but never opened returned no data
2659 Details : ll_file_read() reads zeros from no-object files to EOF
2664 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2665 Details : There's little harm in aborting MDS/OSS recovery and letting it
2666 try again, so I removed the LASSERT and return an error instead.
2668 Severity : enhancement
2670 Description: New debugging infrastructure for tracking down data corruption
2671 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2672 (b) cover more of the client-side code path, and (c) try to narrow
2673 down where problems occurred
2677 Bugzilla : 3819, 4364, 4397, 6313
2678 Description: Racing close and eviction MDS could cause assertion in mds_close
2679 Details : It was possible to get multiple mfd references during close and
2680 client eviction, leading to one thread referencing a freed mfd.
2682 Severity: : enhancement
2683 Bugzilla : 3262, 6359
2684 Description: Attempts to reconnect to servers are now more aggressive.
2685 Details : This builds on the enhanced upcall-less recovery that was added
2686 in 1.4.2. When trying to reconnect to servers, clients will
2687 now try each server in the failover group every 10 seconds. By
2688 default, clients would previously try one server every 25 seconds.
2693 Description: After recovery, certain operations trigger a failed
2694 assertion on a client.
2695 Details : Failing over an mds, using lconf -d --failover, while a
2696 client was doing a readdir() call would cause the client to
2697 LBUG after recovery completed and the readdir() was resent.
2699 Severity : enhancement
2701 Description: Default groups are now added by lconf
2702 Details : You can now run lconf --group <servicename> without having to
2703 manually add groups with lmc.
2706 Frequency : occasional
2708 Description: Nodes with an elan id of 0 trigger a failed assertion
2711 Frequency : always when accessing e.g. tty/console device nodes
2713 Description: tty and some other devices nodes cannot be used on lustre
2714 Details : file's private_data field is used by device data and lustre
2715 values in there got lost. New field was added to struct file to
2716 store fs-specific private data.
2719 Frequency : when exporting Lustre via NFS
2721 Description: NFSD failed occasionally when looking up a path component
2722 Details : NFSD is looking up ".." which was broken in ext3 directories
2723 that had grown large enough to become hashed.
2726 Frequency : Clusters with multiple interfaces not on the same subnet
2728 Description: Nodes will repeatedly try to reconnect to an interface which it
2729 cannot reach and report an error to the log.
2730 Details : Extra peer list entries will be created by lconf with some peers
2731 unreachable. lconf now validates the peer before adding it.
2734 Frequency : Only if a default stripe is set on the filesystem root.
2736 Description: Setting a default stripe on the filesystem root prevented the
2737 filesystem from being remounted.
2738 Details : The client was sending extra request flags in the root getattr
2739 request and did not allocate a reply buffer for the dir EA.
2742 Frequency : occasional, higher if lots of files are accessed by one client
2743 Bugzilla : 6159, 6097
2744 Description: Client trips assertion regarding lsm mismatch/magic
2745 Details : While revalidating inodes the VFS looks up inodes with ifind()
2746 and in rare cases can find an inode that is being freed.
2747 The ll_test_inode() code will free the lsm during ifind()
2748 when it finds an existing inode and then the VFS later attaches
2749 this free lsm to a new inode.
2753 Bugzilla : 6422, 7030
2754 Description: MDS deadlock between mkdir and client eviction
2755 Details : Creating a new file via mkdir or mknod (starting a transaction
2756 and getting the ns lock) can deadlock with client eviction
2757 (gets ns lock and trying to finish a synchronous transaction).
2760 Frequency : occasional
2761 Description: While starting a server, the fsfilt_ext3 module could not be
2763 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2764 kernels. Previously, lconf would still use the ext3 name
2765 when trying to load modules. Now, it will correctly use
2766 ext3 on 2.4 and ldiskfs on 2.6.
2768 Severity : enhancement
2769 Description: The default stripe count has been changed to 1
2770 Details : The interpretation of the default stripe count (0, to lfs
2771 or lmc) has been changed to mean striping across a single
2772 OST, rather than all available. For general usage we have
2773 found a stripe count of 1 or 2 works best.
2775 Severity : enhancement
2776 Description: Add support for compiling against Cray portals.
2777 Details : Conditional compiling for some areas that are different
2781 Frequency : occasional
2782 Bugzilla : 6409, 6834
2783 Description: Creating files with an explicit stripe count may lead to
2784 a failed assertion on the MDS
2785 Details : If some OSTs are full or unavailable, creating files may
2786 trigger a failed assertion on the MDS. Now, Lustre will
2787 try to use other servers or return an error to the
2791 Frequency : occasional
2793 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2794 caused lock timeout during readahead (since 1.4.2).
2795 Details : Processes doing readahead might match a lock that hasn't been
2796 granted yet if there are overlapping and conflicting lock
2797 requests. The readahead process waits on ungranted lock
2798 (original lock is CBPENDING), while OST waits for that process
2799 to cancel CBPENDING read lock and eventually evicts client.
2801 Severity : enhancement
2803 Description: Initial enabling of flock support for clients
2804 Details : Implements fcntl advisory locking and file status functions.
2805 This feature is provided as an optional mount flag (default
2806 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2807 locking are implemented yet, and those that are are not guaranteed
2808 to be completely correct in production environments.
2809 mount -t lustre -o [flock|noflock] ...
2812 Frequency : occasional
2814 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2815 assertion in the ext3 JBD (journaling) layer.
2816 Details : The b_committed_data struct is protected by the big kernel lock
2817 in 2.4 kernels, serializing journal_commit_transaction() and
2818 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2819 access to this struct. In 2.6 kernels there is finer grained
2820 locking to improve SMP performance of the JBD layer.
2824 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2829 Description: merge and cleanup kernel patches.
2832 -----------------------------------------------------------------------------
2834 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2836 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2837 versions of Lustre. Please update all servers and clients to
2838 version 1.4.2 or later at the same time. You must also run
2839 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2840 to update the configuration logs.
2842 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2843 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2844 - swab llog records of type '0' so we get proper header size/idx (5861)
2845 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2846 - fix rename of one directory over another leaking an inode (5953)
2847 - avoid SetPageDirty on 2.6 (5981)
2848 - don't re-add just-being-destroyed locks to the waiting list (5653)
2849 - when creating new directories, inherit the parent's custom
2850 striping settings if present parent (3048)
2851 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2852 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2853 - don't hold client locks on temporary worklist from l_lru (5666)
2854 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2855 - automatically evict dead clients (5921)
2856 - Update file size properly in create+truncate+fstat case (6196)
2857 - Do not unhash mountpoint dentries, do not allow removal of
2859 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2860 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2861 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2862 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2863 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2864 - Correctly handle reads of files with no objects (6243)
2865 - lctl recover will also mark a device active if deactivate used (5933)
2867 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2868 - allow --write-conf on an MDS with different nettype than client (5619)
2869 - don't write config llogs to MDS for mounts not from that MDS (5617)
2870 - lconf should create multiple TCP connections from a client (5201)
2871 - init scripts are now turned off by default; run chkconfig --on
2872 lustre and chkconfig --on lustrefs to use them
2873 - upcalls are no longer needed for clients to recover to failover
2875 - add --abort-recovery option to lconf to abort recovery on device
2877 - add support for an arbitrary number of OSTs (3026)
2878 - Quota support protocol changes.
2879 - forward compatibility changes to wire structs (6007)
2880 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2881 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2882 - improved "lustre status" script
2883 - initialize blocksize for non-regular files (6062)
2884 - added --disable-server and --disable-client configure options (5782)
2885 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2886 - Vanilla 2.4.29 support
2887 - increase maximum number of obd devices to 520 (6242)
2888 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2889 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2890 SLES 9 2.6.5 kernels
2891 - limit stripes per file to 160 (the maximum EA size) (6093)
2893 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2896 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2897 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2898 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2899 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2900 - small ext3 extents cleanups and fixes (5733)
2901 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2902 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2903 - avoid needless client->OST connect, fix handle mismatch (5317)
2904 - fix DLM error path that led to out-of-sync client, long delays (5779)
2905 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2906 - fix several locking issues related to i_size (5492,5624,5654,5672)
2907 - don't move pending lock onto export if it is already evicted (5683)
2908 - fix kernel oops when creating .foo in unlinked directory (5548)
2909 - fix deadlock in obdfilter statistics vs. object create (5811)
2910 - use time_{before,after} to avoid timer jiffies wrap (5882)
2911 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2912 - Do not leak request if server was not able to process it (5154)
2913 - If mds_open unable to find parent dir, make that negative lookup(5154)
2914 - don't create new directories with extent-mapping (5909, 5936)
2916 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2917 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2918 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2919 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2920 - support for RHEL4 kernel on i686 (5773)
2921 - provide error messages when incompatible logs are encountered (5898)
2923 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2924 * version 1.4.0.10 (1.4.1 release candidate 1)
2926 - don't keep a lock reference when lock is not granted (4238)
2927 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2928 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2929 - drop import inflight refcount on signal_completed_replay error (5255)
2930 - unlock page after async write error during send (3677)
2931 - handle missing objects in filter_preprw_read properly (5265)
2932 - no transno return for symlink open, don't save no-trasno open (3440)
2933 - don't try to complete elan receive that already failed (4012)
2934 - free RPC server reply state on error (5406)
2935 - clean up thread from ptlrpc_start_thread() on error (5160)
2936 - readahead could read extra page into cache that wasn't ejected (5388)
2937 - prevent races in class_attach/setup/cleanup/detach (5260)
2938 - don't dereference de->d_inode after l_dput of de (5458)
2939 - use "int" for stripe value returned from lock_to_stripe (5544)
2940 - mballoc allocation and error-checking fixes in 2.6 (5504)
2941 - block device patches to fix I/O request sizes in 2.6 (5482)
2942 - look up hostnames for IB nals (5602)
2943 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2944 - don't start multiple acceptors for the same port (5277)
2945 - fix incorrect LASSERT in mds_getattr_name (5635)
2946 - export a proc file for general "ping" checking (5628)
2947 - fix "lfs check" to not block when the MDS is down (5628)
2949 - service request history (4965)
2950 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2951 - create an "evict_client" /proc entry on OSTs, like the MDS has
2952 - fix mount usage message, return errors per mount(8) (5168)
2953 - change grep [] to grep "[]" in tests so they work in more UMLs
2954 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2955 - remove ancient LOV_MAGIC_V0 EA support (5047)
2956 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2957 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2958 - allow readahead allocations to fail when low on memory (5383)
2959 - mmap locking landed again, after considerable improvement (2828)
2960 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2962 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2965 - send OST transaction number in read/write reply to free req (4966)
2966 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2967 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2968 - the watchdog thread now runs as interruptible (5246)
2969 - flock/lockf fixes (but it's still disabled, pending 5135)
2970 - don't use EXT3 constants in llite code (5094)
2971 - memory shortage at startup could cause assertion (5176)
2973 - reorganization of lov code
2974 - single portals codebase
2976 - add extents/mballoc support (5025)
2977 - direct I/O reads in the obdfilter (4048)
2978 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2980 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2983 - send OST transaction number in read/write reply to free req (4966)
2984 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2985 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2986 - the watchdog thread now runs as interruptible (5246)
2987 - handle missing objects in filter_preprw_read properly (5265)
2988 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2989 - drop import inflight refcount on signal_completed_replay error (5255)
2990 - unlock page after async write error during send (3677)
2991 - return original error code on reconstructed replies (3761)
2992 - no transno return for symlink open, don't save no-trasno open (3440)
2994 - add pid to ldlm debugging output (4922)
2995 - bump the watchdog timeouts -- we can't handle 30sec yet
2996 - extra debugging for orphan dentry/inode bug (5259)
2998 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3001 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3002 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3003 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3004 - fix NULL dereference in /proc/sys/portals/routes (4827)
3005 - allow failed mdc_close() operations to be interrupted (4561)
3006 - stop precreate on OST before MDS would time out on it (4778)
3007 - don't send partial-page writes before EOF from client (4410)
3008 - discard client grant for sub-page writes on large-page clients (4520)
3009 - don't free dentries not owned by NFS code, check generation (4806)
3010 - fix lsm leak if mds_create_objects() fails (4801)
3011 - limit debug_daemon file size, always print CERROR messages (4789)
3012 - use transno after validating reply (3892)
3013 - process timed out requests if import state changes (3754)
3014 - update mtime on OST during writes, return in glimpse (4829)
3015 - add mkfsoptions to LDAP (4679)
3016 - use ->max_readahead method instead of zapping global ra (5039)
3017 - don't interrupt __l_wait_event() during strace
3019 - add software watchdogs to catch hung threads quickly (4941)
3020 - make lustrefs init script start after nfs is mounted
3021 - fix CWARN/ERROR duplication (4930)
3022 - return async write errors to application if possible (2248)
3023 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3024 - print NAL number in %x format (4645)
3025 - update barely-supported suse-2.4.21-171 series (4842)
3026 - support for sles 9 %post scripts
3027 - support for building 2.6 kernel-source packages
3028 - support for sles km_* packages
3030 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3033 - ignore -ENOENT errors in osc_destroy (3639)
3034 - notify osc create thread that OSC is being cleaned up (4600)
3035 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3036 - reconstruct ost_handle() like mds_handle() (4657)
3037 - create a new thread to do import eviction to avoid deadlock (3969)
3038 - let lconf resolve symlinked-to devices (4629)
3039 - don't unlink "objects" from directory with default EA (4554)
3040 - hold socknal file ref over connect in case target is down (4394)
3041 - allow more than 32000 subdirectories in a single directory (3244)
3042 - fix blocks count for O_DIRECT writes (3751)
3043 - OST returns ENOSPC from object create when no space left (4539)
3044 - don't send truncate RPC if file size isn't changing (4410)
3045 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3046 - bind to privileged port in socknal and tcpnal (3689)
3048 - rate limit CERROR/CWARN console message to avoid overload (4519)
3049 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3050 - basic mmap support (3918)
3051 - kernel patch series update from b1_4 (4711)
3053 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3056 - avoid crash during MDS cleanup with OST shut down (2775)
3057 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3058 - don't use bad inodes on the MDS (3744)
3059 - dynamic object preallocation to improve recovery speed (4236)
3060 - don't hold spinlock over lock dumping or change debug flags (4401)
3061 - don't zero obd_dev when it is force cleaned (3651)
3062 - print grants to console if they go negative (4431)
3063 - "lctl deactivate" will stop automatic recovery attempts (3406)
3064 - look for existing locks in ldlm_handle_enqueue() (3764)
3065 - don't resolve lock handle twice in recovery avoiding race (4401)
3066 - revalidate should check working dir is a directory (4134)
3068 - don't always mark "slow" obdfilter messages as errors (4418)
3070 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3073 - don't close LustreDB during write_conf until it is done (3860)
3074 - fix typo in lconf for_each_profile (3821)
3075 - allow dumping logs from multiple threads at one time (3820)
3076 - don't allow multiple threads in OSC recovery (3812)
3077 - fix debug_size parameters (3864)
3078 - fix mds_postrecov to initialize import for llog ctxt (3121)
3079 - replace config semaphore with spinlock (3306)
3080 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3081 - don't allow enqueue to complete on a destroyed export (3822)
3082 - down write_lock before checking llog header bitmap (3825)
3083 - recover from lock replay timeout (3764)
3084 - up llog sem before sending rpc (3652)
3085 - reduce ns lock hold times when setting kms (3267)
3086 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3087 - fix NULL deref and obd_dev leak on setup error (3312)
3088 - replace some LBUG about llog ops with error handling (3841)
3089 - don't match INVALID dentries from d_lookup and spin (3784)
3090 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3091 - fix invalid assertion in ptlrpc_set_wait (3880)
3093 - add libwrap support for the TCP acceptor (3996)
3094 - add /proc/sys/portals/routes for non-root route listing (3994)
3095 - allow setting MDS UUID in .xml (2580)
3096 - print the stack of a process that LBUGs (4228)
3098 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3101 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3102 - make sure to unset replay flag from failed open requests (3440)
3103 - if default stripe count is 0, use OST count for inode size (3636)
3104 - update parent mtime/ctime on client for create/unlink (2611)
3105 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3106 - free recovery state on server during a forced cleanup (3571)
3107 - unregister_reply for resent reqs (3063)
3108 - loop back devices mounting and status check on 2.6 (3563)
3109 - fix resource-creation race that can provoke i_size == 0 (3513)
3110 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3111 - more debugging for page-accounting assertion (3746)
3112 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3113 - avoid deadlock after precreation failure (3758)
3114 - fix race and lock order deadlock in orphan handling (3450, 3750)
3115 - add validity checks when grabbing inodes from l_ast_data (3599)
3117 - add /proc/.../recovery_status to obdfilter (3428)
3118 - lightweight CDEBUG infrastructure, debug daemon (3668)
3119 - change default OSC RPC parameters to be better on small clusters
3120 - turn off OST read cache for files smaller than 32MB
3121 - install man pages and include them in rpms (3100)
3122 - add new init script for (un)mounting lustre filesystems (2593)
3123 - run chkconfig in %post for init scripts (3701)
3124 - drop scimac NAL (unmaintained)
3126 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3129 - clean kiobufs before and after use (3485)
3130 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3131 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3132 - change init script to fail more clearly if not run as root (1528)
3133 - allow clients to reconnect during replay (1742)
3134 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3135 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3136 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3137 - fix cond_resched (3554)
3138 - only evict unfinished clients after recovery (3515)
3139 - allow bulk resend, prevent data loss (3570)
3140 - dynamic ptlrpc request buffer allocation (2102)
3141 - don't allow unlinking open directory if it isn't empty (2904)
3142 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3143 - remove extraneous obd dereference causing LASSERT failure (3334)
3144 - don't use get_cycles() when creating temp. files on the mds (3156)
3145 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3146 - handle EEXIST for set-stripe, set proper directory name (3336)
3148 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3149 - fix ksocknal_fmb_callback() error messages (2918)
3151 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3154 - don't copy lvb into (possibly NULL) reply on error (2983)
3155 - don't deref dentry after dput, don't free lvb on error (2922)
3156 - use the kms to determine writeback rpc length (2947)
3157 - increment oti_logcookies when osc is inactive (2948)
3158 - update client's i_blocks count via lvb messages (2543)
3159 - handle intent open/close of special files properly (1557)
3160 - mount MDS with errors=remount-ro, like obdfilter (2009)
3161 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3162 - don't use cancelling-locks' kms values (2947)
3163 - use highest lock extent for kms, not last one (2925)
3164 - don't dereference ERR_PTR() dentry in error handling path (3107)
3165 - fix thread race in portals_debug_dumplog() (3122)
3166 - create lprocfs device entries at setup instead of at attach (1519)
3167 - common AST error handler, don't evict client on completion race (3145)
3168 - zero nameidata in detach_mnt in 2.6 (3118)
3169 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3170 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3171 - initialize RPC timeout timer earlier for 2.6 (3219)
3172 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3173 - print nal/nid for unknown nid (3258)
3174 - additional checks for oscc recovery before doing precreate (3284)
3175 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3176 - don't crash in mdc_close for bad permissions on open (3285)
3177 - zero i_rdev for non-device files (3147)
3178 - clear page->private before handing to FS, better assertion (3119)
3179 - tune the read pipeline (3236)
3180 - fix incorrect decref of invalidated dentry (2350)
3181 - provide read-ahead stats and refine rpc in flight stats (3328)
3182 - don't hold journal transaction open across create RPC (3313)
3183 - update atime on MDS at close time (3265)
3184 - close LDAP connection when recovering to avoid server load (3315)
3185 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3186 - don't leak open file on MDS after open resend (3325)
3187 - serialize filter_precreate and filter_destroy_precreated (3329)
3188 - loop device shouldn't call sync_dev() for nul device (3092)
3189 - clear page cache after eviction (2766)
3190 - resynchronize MDS->OST in background (2824)
3191 - refuse to mount the same filesystem twice on same mountpoint (3394)
3192 - allow llmount to create routes for mounting behind routers (3320)
3193 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3194 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3195 - fix rare problem with rename on htree directories (3417)
3197 - allow default OST striping configuration per directory (1414)
3198 - fix compilation for qswnal for 2.6 kernels (3125)
3199 - increase maximum number of MDS request buffers for large systems
3200 - change liblustreapi to be useful for external progs like lfsck (3098)
3201 - increase local configuration timeout for slow disks (3353)
3202 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3204 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3207 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3208 - don't overwrite extent policy data in reply if lock was blocked (2901)
3209 - drop filter export grants atomically with removal from device (2663)
3210 - del obd_self_export from work_list in class_disconnect_exports (2908)
3211 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3212 - swab reply message in mdc_close, other PPC fixes (2464)
3213 - fix destroying of named logs (2325)
3214 - overwrite old logs when running lconf --write_conf (2264)
3215 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3216 - fix race in target_handle_connect (2898)
3217 - mds_reint_create() should take same inode create lock (2926)
3218 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3219 - don't close files for self_export to avoid uninitialized obd (2936)
3220 - allow MDS with the same name as client node (2939)
3221 - hold dentry reference for closed log files for unlink (2325)
3222 - reserve space for all logs during transactions (2059)
3223 - don't evict page beyond end of stripe extent (2925)
3224 - don't oops on a deleted current working directory (2399)
3225 - handle hard links to targets without a parent properly (2517)
3226 - don't dereference NULL lock when racing during eviction (2867)
3227 - don't grow lock extents when lots of conflicting locks (2919)
3229 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3232 - account for cache space usage on clients to avoid data loss (974)
3233 - lfsck support in lustre kernel code (2349)
3234 - reduce journal credits needed for BRW writes (2370)
3235 - orphan handling to avoid losing space on client/server crashes
3236 - ptlrpcd can be blocked, stopping ALL progress (2477)
3237 - use lock value blocks to assist in proper KMS, faster stat (1021)
3238 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3239 - recovery for initial connections (2355)
3240 - fixes for mds_cleanup_orphans (1934)
3241 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3242 - block all file creations until orphan recovery completes (1901)
3243 - client remove rq_connection from request struct (2423)
3244 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3245 - recovery timer race (2670)
3246 - mdc_close recovey bug (2532)
3247 - ptlrpc cleanup bug (2710)
3248 - mds timeout on local locks (2588)
3249 - namespace lock held during RPCs (2431)
3250 - handle interrupted sync write properly (2503)
3251 - don't try to handle a message that hasn't been replied to (2699)
3252 - client assert failure during cleanup after abort recovery (2701)
3253 - leak mdc device after failed mount (2712)
3254 - ptlrpc_check_set allows timedout requests to complete (2714)
3255 - wait for inflight reqs when ptlrpcd finishes (2710)
3256 - make sure unregistered services are removed from the srv_list
3257 - reset bulk XID's when resending them (caught by 1138 test)
3258 - unregister_bulk after timeout
3259 - fix lconf error (2694)
3260 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3261 - readahead locks pages, leaves pending causing memory pressure (2673)
3262 - increase OST request buffers to 4096 on large machines (2729)
3263 - fix up permission of existing directories in simple_mkdir (2661)
3264 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3265 - don't assign transno to errored transactions (2742)
3266 - don't delete objects on OST if given a bogus objid from MDS (2751)
3267 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3268 - if rq_no_resend, then timeout request after recovery (2432)
3269 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3270 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3271 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3272 - MDS recovery shouldn't delete objects that it has given out (2730)
3273 - if enqueue arrives after completion, don't clobber LVB (2819)
3274 - don't unlock pages twice when trigger_group_io returns error (2814)
3275 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3276 - don't write pages to disk if there was an error (1450)
3277 - don't ping imports that have recovery disabled (2676)
3278 - take buffered bytes into account when balancing socknal conn (2817)
3279 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3280 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3281 - remove little-endian swabbing of llog records (1987)
3282 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3283 - retry reposting request buffers if they fail (1191)
3284 - grow extent at grant time to avoid granting a revoked lock (2809)
3285 - lock revoke doesn't evict page if covered by a second lock (2765)
3286 - disable VM readahead to avoid reading outside lock extents (2805)
3288 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3289 - updated kernel patches for hp-2.4.20 kernel (2681)
3291 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3294 - fix truncated write corruption (2366)
3295 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3297 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3298 - ensure len(uuid) < 37 in lmc (1171)
3299 - fix ia64 OOPS in llog_test (2255)
3300 - zero end of page at obdfilter for partial page writes (2648)
3301 - don't leave stale dentries around after renames (bug 2428)
3302 - fix timeouts when evicting a client with a single lock held (2642)
3303 - set deadline for the initial HELLO message to drain (2634)
3304 - print out dotted-quad IP addresses in the socknal (2302)
3306 - additional debugging for MDS client eviction problem (2443)
3307 - fix mkfsoptions support for osts (2603, 2604)
3309 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3312 - add series for the vanilla 2.6.0 kernel
3313 - add series for the vanilla 2.4.24 kernel
3314 - add series for a cray x86/64 UL kernel drop
3315 - fix xattr patches for the vanilla 2.4.19 series
3317 - generate true UUIDs in lmc (1171)
3318 - have portals stack dumping break in UML (2466)
3319 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3320 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3321 - wake callback waiting threads on client eviction (2460)
3322 - Add --ptldebug and --subsystem to lmc (1719)
3323 - update assertion to allow safe interrupt allocation
3324 - set rq_no_resend for cancel requests (2432)
3325 - recalculate ptlrpcd timeout after resend (2494)
3326 - call vfs_rmdir when removing pending directories (2368)
3327 - fix renaming a file to itself (2429)
3328 - lmc creates a default one-stripe lov (2454)
3329 - expand procfs space to handle large clusters (2326)
3330 - increase UML stack to avoid overflow
3331 - update lconf's list of debug and subsystem masks
3332 - fix lfs find --obd (2510)
3333 - /proc tunable for disabling filter read caching (2591)
3334 - stop rpm packages from altering slapd.conf (2301)
3335 - disable nagle in the socknal under 0conf (2578)
3336 - choose mds inode size based on stripe count (2572)
3337 - fix kernel-source rpm problems (2516)
3339 - add --disable-doc to avoid pdf generation (2421)
3340 - update documentation, tests, type-os, comments
3341 - avoid format warnings on ia64
3342 - remove the TOE NAL
3343 - tiny code cleanups by removing unused fields
3345 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3348 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3349 - remove the most blatant lies from BUILDING file (2371)
3350 - change default debug level to reasonable production setting
3351 - reduce client side cache size to reduce cache flush time
3352 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3353 - make TCP zerocopy and pinger support enabled by default (2476)
3354 - sync writes completed after process exits caused crashes (2319)
3355 - maintain correct mount count on the MDS (2356)
3356 - backout 1557, because 2316 wasn't really fixed
3357 - better file I/O statistics gathering in /proc
3358 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3359 - another kernel patch to fix zero-copy TCP function export
3360 - don't take duplicate lock when processing re-sent getattr (2420)
3361 - lctl uses obd_self_export instead of creating new conn (2353)
3362 - MDS/OST recovery case which requires object creation asserted (2425)
3363 - move lfs from /usr/sbin to /usr/bin in packages
3364 - fix race between mds_client_add and mds_client_free (2417)
3365 - use kmalloc instead of slabs in portals (2430)
3366 - don't create duplicate records when a failover MDS is present (2442)
3367 - remove unnecessary mount age check (2332)
3368 - don't remove directory inodes from locks prematurely (2451)
3369 - don't break if MDS service name is the same as hostname (2103)
3370 - fix races in client write RPC generation when cache full (2482)
3372 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3375 - remove now-unused request->rq_obd (278)
3376 - if an allocation fails, print out how much memory we've used (1933)
3377 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3378 - add the "configurable stack size" patch to most series files (1256)
3379 - ability to write large log records, for 100+ OST configs (2306)
3380 - fix NULL deref when filter_prep fails (2314)
3381 - fix operator precedence error in filter_sync
3382 - dynamic allocation of socknal TX descriptors (2315)
3383 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3384 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3385 - ensure that CWARN messages in Portals always get to the syslog
3386 - __init/__exit are not for prototype decls (ldlm_init/exit)
3387 - x86-64 compile warning fixes
3388 - fix gateway LMC keyword conflict (2318)
3389 - fix MDS lock inversions in getattr/reint paths (1844)
3390 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3391 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3392 - DLM race condition prevented some lock evictions (2328)
3393 - ENOMEM detection and retry on socknal sends (2230)
3394 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3395 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3396 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3397 - better error messages when a client is rejected during recovery (1505)
3398 - avoid cancelling locks which were never granted, after failure (2330)
3399 - fix i_sem/journal inversion in mds_client_add (2333)
3400 - fix truncate/getattr lock cycle deadlock (2334)
3401 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3402 - fix two rare exit paths which could leak an l_lock() ref (2321)
3403 - fencepost error in MDS/OST orphan recovery (2226)
3404 - make log record alignment 8 bytes (1988)
3405 - lstripe now fails when requested offset > ost_count (2237)
3406 - ensure that all kernel series have a complete list.h (1607)
3407 - fix crashes in special-file operations (2316)
3408 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3410 - allow configurable automake binary, for testing new versions
3411 - small update to the lfs documentation
3413 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3415 * fix negative export reference count in fsfilt_sync (2312)
3417 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3418 * release candidate 0.9.1
3420 - orphans are moved into the PENDING directory for possible recovery
3421 - replayed opens now open by fid for orphan/rename safety (1042)
3422 - last close of an orphan inode generates a transno (683)
3423 - chdir() and mount() now pin the directory entry (1020)
3424 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3425 - discard very old requests without processing them (1502)
3426 - remove some common, well-understood CERRORs (1505)
3427 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3428 - clear "grant" flags in OST replies until OST grant code lands (1644)
3429 - fix read performance by not clobbering i_blksize on client (1598)
3430 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3431 - make LRU size a /proc tunable, clears locks when reduced (707)
3432 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3433 - server locks take references on exports now (1558)
3434 - build fixes for 2.4.20-rh trees (1663)
3435 - return an error from lov_create if all OSCs are inactive (1751)
3436 - fix import levels when a reconnect happens without a timeout (1597)
3437 - exit early from mds_open if we get a lookup error (1749)
3438 - partial page read at EOF wouldn't wait for disk before sending (1642)
3439 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3440 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3441 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3442 - write/truncate lock inversion (1639)
3443 - Don't auto-load obdclass, portals modules during cleanup (1495)
3444 - fix timestamps from jumping to "now" (1763)
3445 - extra journal assertions (1648)
3446 - add an extra multiunlink test (1771)
3447 - fix read_record/write_record API (1776)
3448 - fix leak of offset_extent, possible incorrect i_size later (1772)
3449 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3450 - Debugging for the kqswnal_get_idle_tx problems (1820)
3451 - Allow recovery to be attempted multiple times (1536)
3452 - Write out MDS last_rcvd file after it is first created (1600)
3453 - Fix tx_descriptor leak in failed transmit situations (1827)
3454 - ext3 journaling fixes for assertion failure after IO error (1871)
3455 - class_export_put() on freed export after completion AST error (1896)
3456 - Fix revalidate looping in VFS (1322)
3457 - Don't access a freed export during MDS_REINT timeout (1521)
3458 - Add open-unlink recovery support on the MDS (1673,1764)
3459 - Return an error if no MDS data was read from last_rcvd (1946)
3460 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3461 - Don't setup a disk device that doesn't match exported UUID (317)
3462 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3463 - avoid committing NULL handle in force close
3464 - local.sh is now a one-stripe LOV configuration
3465 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3466 - fix ext3 htree duplicate directory entry corruption (1516)
3467 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3468 - update to new LOV EA format (2097)
3469 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3471 2003-06-15 Phil Schwan <phil@clusterfs.com>
3474 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3475 - per-import/export recovery handling (958, 931, 959)
3476 - multiple last-rcvd slots, for serving multiple FSes (949)
3477 - connections are again shared between multiple imp/exports (963, 964)
3478 - "umount -f" would hang if any requests needed to be sent (393, 978)
3479 - avoid pinning large req buffer by copying for queued messages (989)
3480 - add "uuid" to "lctl device" command to help upcalls (991)
3481 - "open" RPCs with transnos would confuse recovery counters (1037)
3482 - do proper endian conversion of all wire messages (288, 340, 891)
3483 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3484 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3485 - fix ptlrpc_connection leak in target_handle_connect (1174)
3486 - fix import refcounting bug in OST and MDS cleanup (1134)
3487 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3488 - fix very unlikely obd_types race condition (501)
3489 - remove osc_open hack for echo_client (1187)
3490 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3491 - a failure in read_inode2 leads to deadlock (1139)
3492 - cancel ack-locks as soon as transaction is committed (1072)
3493 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3494 - make sure to commitrw after any preprw to avoid deadlock (1162)
3495 - failing to execute a file in a lustre FS would lock inode (1203)
3496 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3497 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3498 - obd_uuid2tgt would walk off the end of the list (1255)
3499 - on IA64 the osc would give portals incorrect bulk size (1258)
3500 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3501 - fix lock inversion caused by new llite matching code (1282)
3502 - limit the number of dirty pages on a client to 10MB (1286)
3503 - timed out locks were not being corrected cancelled (1289)
3504 - fix O_DIRECT above 4GB on IA-32 (1292)
3505 * major user-visible changes
3506 - fail out/fail over policy now controlled by the upcall (993)
3508 - add OBD_PING to check server availability and failure (954)
3509 - lustre messages are now sent in sending host order (288, 340, 891)
3510 - add eadatalen to MDS getattr reply (340)
3511 - OST read replies may contain second buffer, with per-page status (593)
3513 2003-03-11 Phil Schwan <phil@clusterfs.com>
3516 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3517 - failed open()s could cause deadlock; fixed (867, 869)
3518 - stop cancelling OST locks when files are closed (481)
3519 - overlapping XID spaces caused network corruption (851, 853)
3520 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3521 - setattr_raw added, to do single-RPC, server-side setattrs
3522 - lmc/lconf syntax change for OST UUIDs
3523 - fix crashy race condition between ptlrpc_free_req and osc_close
3524 - don't use request in mdc_enqueue if we hit a timeout (889)
3525 - don't set the inode i_size for regular files from the MDS (896)
3526 - handle out of order completion AST (842)
3527 - don't LBUG if a lock request times out after receiving AST (913)
3528 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3529 - if a bad lock AST arrives, send an error instead of dropping entirely
3530 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3531 - fix leak in bulk IO when only partially completed (899, 900, 926)
3532 - fix O_DIRECT for ia64 (55)
3533 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3534 - C-z after timeout could hang a process forever; fixed (977)
3536 - client-side I/O cache (678, 924, 929, 941, 970)
3538 - READPAGE and SETATTRs which don't take server-side locks get
3541 2003-02-11 Phil Schwan <phil@clusterfs.com>
3544 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3545 - Fix fsfilt_extN_readpage() to read a full page of directory
3546 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3547 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3548 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3549 - Most intent operations were removed, in favour of a new RPC mode
3550 that does a single RPC to the server and bypasses most of the VFS
3551 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3552 - Aggressively cancel local locks on DLM servers
3553 - mds_reint_unlink sends EA to the client if it's the last nlink.
3554 client uses that EA to unlink OST objects.
3555 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3556 - recursive symlinks were fixed (439)
3557 - fixed NULL deref in DEBUG_REQ
3558 - filter_update_lastobjid no longer calls sync, which annoyed extN
3559 - fixed multi-client small-writes to a single file problem (445)
3560 - fixed mtime updates during file writes (607)
3561 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3562 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3563 - fixed Linux OST size reporting problem (444, 656)
3564 - OST now updates object mtime with writes or setattr (607, 619)
3565 - client verifies file size before zeroing page past EOF (445)
3566 - OST now writes last allocated objid to disk with allocation (108)
3567 - LOV on echo now works (409)
3569 - mds_reint_unlink sends a new buffer, with the EA included. this
3570 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3571 set if a regular file was being unlinked, and it was the last link
3572 - use PtlGet from the target for bulk writes (315)
3573 - OST now updates object mtime with writes or setattr (607, 619)
3574 - LDLM now has a grant-time callback to revalidate locked items, if
3576 - Many MDS operations were reorganized to combat race conditions
3578 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3579 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3581 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3584 - Fully reactivate OST imports after reconnection (512, others)
3585 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3586 - More graceful error handling for truncating on dead OST (515)
3587 - Don't error out unless we're actually accessing dead stripes (474)
3588 - Fix garbage sizes when stripes are missing (410)
3589 - LRU counters were broken, causing constant lock purge (433, 432)
3590 - garbage on read from stripes with failed OSTs (441)
3591 - mark OSCs as active before reconnecting during recovery (438)
3592 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3593 - lfind did not preserve OST order in output (443)
3594 - symlinks cause hung clients, incorrect data (439)
3595 - stop dereferencing request after dropping refcount (457)
3596 - don't LASSERT(spin_is_locked) on non-SMP (455)
3597 - fixes for many rename() bugs
3598 - fstat didn't correctly synchronize attributes (399)
3599 - server must handle lock cancellation during blocking AST prep (487)
3600 - bulk descriptors were free()d too soon (511)
3601 - fix paths in lconf, which would load incorrect modules (451, 507)
3602 - fix confusing lconf 'host not found' error message (386)
3603 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3604 - fix race condition in mdc_blocking_ast() for inode access (526)
3605 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3606 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3607 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3608 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3609 - fix lstripe in several ways (kernel side) (527)
3610 - fix request leak in ldlm_cli_enqueue (262)
3611 - incorrect OSC was marked inactive after OST failure
3612 - call mds_fs_cleanup before unmounting filesystem (524)
3613 - fix races between taking ns_lock and ldlm_lock_change_resource
3614 - fix races updating LOV export open file list
3615 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3616 - fix recovery NULL deref in ldlm_cli_cancel_unused
3617 - fix some DLM races by using new hash table for lock handles (419)
3618 - permit the client to specify desired inodes, at replay
3619 - duplicate requests when we queue them for replay reintegration
3620 - fix last_rcvd offset calculation
3621 - sync after each recovered transaction, so we always make progress
3622 - never, not always, ERESTART requests without transnos
3623 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3625 - skip replay if the MDS says that the client is already connected
3626 - don't check for a recovery-enabled export to match lctl's UUID
3627 - don't INC_USE_COUNT for phantom exports
3628 - don't crash when cleaning up phantom exports (567)
3629 - don't double-finish or set replay data for errored mdc_open requests
3630 - abort requests when they time out, so we don't get old replies
3631 - send/receive replies for AST messages again
3632 - if the client says that it doesn't have the lock, cancel it on the
3634 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3635 mark it as destroyed, it will all work out when decref is called
3636 - fix module use counts (22, 581)
3638 - ASTs now expect a reply (server cancels lock on error reply)
3640 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3643 - fix many simultaneous client startup (392)
3644 - fix dentry->d_it clobbering
3645 - credentials weren't being shipped for readdir/getattr operations
3646 - remove invalid assertions triggered during some concurrent MD
3648 - proper Lustre versions added (336, 389)
3649 - fix memory leak for create error case (398)
3650 - fix LOV locking bug that would get cli/srv out of sync
3651 - fix echo client over LOV (409)
3652 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3653 - fix double-O_EXCL intent crash (424)
3654 - avoid sending multiple lock CANCELs (352)
3656 - MDS can do multi-client recovery (modulo bugs in new code)
3658 - many updates, edits, cleanups
3660 2002-11-18 Phil Schwan <phil@clusterfs.com>
3663 - fix null d_it dereference (346)
3664 - fix full OST/dbench hang (333)
3665 - fix permission problem with file removal (286)
3666 - fix removal of OSCs from LOV when they fail
3667 - fix NULL deref during bulk timeout (214)
3668 - fix problems related to multiple filesystems on one MDS (241)
3669 - fixed serious subtle metadata locking bugs
3670 - free locks on clients when inodes are removed due to memory
3672 - fix inode pointer in lock data (285)
3673 - partial support for multiple MDS on a single host (241)
3674 - data locks weren't cancelled at clear_inode time (290, 311)
3675 - intent locks could lead to unbounded lock growth (205)
3676 - added a maximum lock count, an LRU list, and a flusher
3677 - fix multiple rename (365)
3678 - properly abstracted the echo client
3679 - OSC locked 1 byte too many; fixed
3680 - rewrote brw callback code:
3681 - fixed recovery bugs related to LOVs (306)
3682 - fixed too-many-pages-in-one-write crash (191)
3683 - fixed (again) crash in sync_io_timeout (214)
3684 - probably fixed callback-related race (385)
3686 - Add capability to MDS protocol
3687 - LDLM cancellations and callbacks on different portals
3689 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3692 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3693 - defer open object destruction to close time (601981, 138)
3694 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3695 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3696 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3697 - don't call obd_setattr in ll_file_release for destroyed objects
3698 * protocol change to lustre_msg: move |version| and add |flags|
3699 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3700 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3701 * added replay of create, unlink, link and rename operations during
3702 MDS failover; recovery should be much more robust now
3703 * remove failed OSCs from LOVs (only lov_create uses this so far)
3704 * the lustre-HOWTO was brought (more) up to date (582544)
3706 2002-10-23 Phil Schwan <phil@clusterfs.com>
3709 - in-use dentries weren't being reused properly (617851)
3710 - prevent multiple LDLM setup (599178)
3711 - fix LOV size calculations for truncate (617853)
3712 - fix client handling of MDS intent errors (POSIX)
3713 - fix permission bug in lovstripe.c test (624321)
3714 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3715 - truncate past end of file could corrupt data
3716 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3717 - a race in recovery could return ETIMEDOUT to apps (623947)
3718 - building outside the source directory was fixed
3719 * the lustre-HOWTO was brought (more) up to date (582544)
3720 * major progress was made on recovery functionality
3722 2002-10-10 Phil Schwan <phil@clusterfs.com>
3725 - recovery deadlock fix
3726 - rm -rf causes LBUG fix (617817)
3727 - file open by multiple tasks fix (618962)
3728 - directory permissions bugs (602707 and 620007)
3729 - journal_stop fixed with locking (611313)
3730 - O_APPEND failures resolved (618273, perhaps 614459)
3731 - lconf PATH fix (619770)
3732 - IA64 build fix (621450)
3733 - RPC buffer sizes scale with amount of memory
3735 2002-10-01 Phil Schwan <phil@clusterfs.com>
3738 - locks would be cancelled without throwing away data pages,
3739 resulting in inconsistent data (605627)
3740 - inode attributes were not always being refreshed (605627, 612449)
3741 - lconf now continues to cleanup after lctl reports an error
3742 - MDS now enforces user permissions (602707)
3743 - lprocfs cleanup fixed, but not yet enabled (614157)
3744 - fixed infinite server hang, should a client not respond to an AST
3745 - avoid going into recovery if user calls readlink() with a buffer
3746 that's too small (613941)
3747 - AST RPCs no longer require replies (614867) -- this may be changed
3748 - don't crash server if client sends an IOV that's too big (611336)
3749 - fixed lock conversion deadlock (611892)
3750 - fixed the following of symlinks (614622)
3751 * recovery: the server can remove locks from a client that dies, other
3752 clients can make progress
3753 * more extN patch fixes
3754 * compile-time configurable ptlrpc buffer allocations
3756 - collaborative read cache document
3757 - Lustre Lite Performance CDR document-in-progress
3759 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3762 - fix typo in patch-2.4.18
3764 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3767 - clear ptlrpc request each time in handle_incoming_request()
3768 - unlink of files now destroys the object on the OST
3770 2002-09-19 Peter Braam <braam@clusterfs.com>
3772 * add hard link support
3773 * change obdfile creation method
3774 * kernel patch changed
3776 2002-09-19 Peter Braam <braam@clusterfs.com>
3779 - stack overflow bug in extN fixed
3781 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3783 * documentation updates
3784 - add man pages for config tools
3785 - update tests/README to describe testing with new config tools
3786 - finish metadata API descriptions
3787 * bug fixes and cleanups
3788 - statfs workaround for 16TB limit
3789 - LOV stripe allocation improved, can stripe on subset of OSTs
3790 - LOV file size/IO offset was wrong for files > 4GB in size
3791 - object EA data was being dropped, caused files to be unreadable
3792 - memory overflow with non-LOV OST caused memory corruption
3793 - fixed regression tests to work with new config tools, obdfilter
3794 - fixed bug when directory size became larger than 1 block
3795 - fixed bug (for single client case) when PWD was deleted
3796 - invalidate local directory pages when doing intent-based ops
3797 - avoid LDLM oops when lock callback contained bad data
3799 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3801 * documentation updates
3802 * bug fixes and cleanups
3803 - configuration tools
3806 - 64-bit compile warnings
3807 - 64-bit internal statfs data
3809 * test_brw on persistent OST devices
3811 * lprocfs (disabled)
3813 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3815 * documentation updates
3816 * bug fixes and cleanups
3817 * configuration tools
3819 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3822 * many small fixes to 0_5_4
3823 * io/network handling
3824 * thinkos in MDS operations
3826 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3829 * crucial basic fixes to 0.5.3
3830 * IOR, Iozone work over Elan
3833 2002-08-07 Phil Schwan <phil@clusterfs.com>
3834 * version 0_5_3, our first alpha
3835 * we use the new Portals iovs
3836 * documentation updates
3837 * bug fixes and cleanups
3838 * small changes in the DLM wire protocol
3840 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3841 * version 0_5_1 with some initial stability,
3842 * locking on MD and file I/O.
3843 * documentation updates
3844 * several bug fixes since 0.5.0
3845 * small changes in wire protocol
3847 2002-07-18 Phil Schwan <phil@clusterfs.com>
3849 * delivered as Lustre Light Alpha
3850 * fixed a crash after handling invalid MDS requests
3851 * fixed directory pages for architectures with non-4k pages sizes
3853 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3854 * release version v0_4_4
3855 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3857 * Fixes a number of interruption problems with OST operations.
3858 * Update documentation for portals header changes
3859 * Move all wire protocol structs/defines to lustre_idl.h
3860 * Fixes symlink length bug.
3861 * Add tcpdump to repository.
3863 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3864 * release version v0_4_3
3865 * Fixes statfs for inodes on extN.
3866 * Fixes bug in runtests which would delete /etc/hosts.
3867 * Use 64-bit object IDs wherever possible (not into VFS though)
3868 Remove ost_get_info, which is unused by lustre, and out of date.
3870 2002-07-03 Peter Braam <braam@clusterfs.com>
3871 * release version v0_4_2 Fixes a lookup error (type not passed)
3872 * move forward to head of Portals
3873 * move forward to latest Lustre kernel
3875 2002-06-25 Peter Braam <braam@clusterfs.com>
3876 * release version v0_4_1. Hopefully stable on single node use.