1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10),
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.39.cfs8
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: don't use GFP_* in generic Lustre code.
18 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
22 Description: obdclass is limited by single OBD_ALLOC(idarray)
23 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
27 Description: warnings with build patchless client with vanila 2.6.19 and up
28 Details : change old ctl_table style and replace ctl_table/ctl_table_header
29 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
33 Frequency : Only for SLES
34 Description: sanity_quota fail test_1
35 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
36 file, which makes TSTID[2] inunique.
40 Frequency : Only for SLES9
41 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
42 Details : Update target/ChangeLog/which_patch .
46 Frequency : Only for test_10 in sanity-quota.sh
47 Description: a bug in quota test code
48 Details : checking a test flag in a wrong place
52 Frequency : Only for test_13 in sanity-quota.sh
53 Description: a bug in quota test script
54 Details : shouldn't check overfull allocated quota
58 Description: jbd statistics
59 Details : Port older jbd statistics patch for sles10
61 Severity : enhancement
63 Description: Add printing inode info into message about error in writepage.
68 Description: reply_lock_interpret crash due to race with it and lock cancel.
69 Details : Do not replay locks that are being cancelled. Do not reference
70 locks by their address during replay, just by their handle.
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.
120 --------------------------------------------------------------------------------
122 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
124 * Support for kernels:
125 2.6.5-7.283 (SLES 9),
126 2.6.9-55.EL (RHEL 4),
127 2.6.16.46-0.14 (SLES 10),
128 2.6.18.8 vanilla (kernel.org)
129 * Client support for unpatched kernels:
130 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
131 2.6.16 - 2.6.22 vanilla (kernel.org)
132 * Due to recently discovered recovery problems, we do not recommend
133 using patchless RHEL 4 clients with this or any earlier release.
134 * Recommended e2fsprogs version: 1.39.cfs8
135 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
136 * Starting with this release, the ldiskfs backing filesystem required
137 by Lustre is now in its own package, lustre-ldiskfs. This package
138 should be installed. It is versioned separately from Lustre and
139 may be released separately in future.
143 Description: block reactivating mgc import until all deactivates complete
144 Details : Fix race when failing back MDT/MGS to itself (testing)
146 Severity : enhancement
148 Description: add optional extra BUILD_VERSION info
149 Details : add a new environment variable (namely LUSTRE_VERS) which allows
150 to override the lustre version.
153 Frequency : 2.6.18 servers only
155 Description: ll_kern_mount() doesn't release the module reference
156 Details : The ldiskfs module reference count never drops down to 0
157 because ll_kern_mount() doesn't release the module reference.
162 Description: server LBUG when using old ost_num_threads parameter
163 Details : Accept the old ost_num_threads parameter but warn that it
164 is deprecated, and fix an off-by-one error that caused an LBUG.
169 Description: Transient SCSI error results in persistent IO issue
170 Details : iobuf->dr_error is not reinitialized to 0 between two
174 Frequency : sometimes when underlying device returns I/O errors
176 Description: OSTs not going read-only during write failures
177 Details : OSTs are not remounted read-only when the journal commit threads
178 get I/O errors because fsfilt_ext3 calls journal_start/stop()
179 instead of the ext3 wrappers.
183 Description: poor connect scaling with increasing client count
184 Details : Don't run filter_grant_sanity_check for more than 100 exports
185 to improve scaling for large numbers of clients.
188 Frequency : SLES10 only
190 Description: sanity-quota.sh quotacheck failed: rc = -22
191 Details : Quotas cannot be enabled on SLES10.
194 Frequency : liblustre clients only
196 Description: getdirentries does not give error when run on compute nodes
197 Details : getdirentries does not fail when the size specified as an argument
198 is too small to contain at least one entry
200 Severity : enhancement
202 Description: Add LNET router traceability for debug purposes
203 Details : If a checksum failure occurs with a router as part of the
204 IO path, the NID of the last router that forwarded the bulk data
205 is printed so it can be identified.
210 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
211 Details : Due to a race condition, liblustre clients were occasionally
214 Severity : enhancement
216 Description: lfs setstripe use optional parameters instead of postional
219 Severity : enhancement
221 Description: Nanosecond timestamp support for ldiskfs
222 Details : The on-disk ldiskfs filesystem has added support for nanosecond
223 resolution timestamps. There is not yet support for this at
224 the Lustre filesystem level.
227 Frequency : during server recovery
229 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
230 Details : request with rq_no_resend flag not awake l_wait_event if they get a
234 Frequency : nfs export on patchless client
236 Description: connectathon hang when test nfs export over patchless client
237 Details : Disconnected dentry cannot be found with lookup, so we do not need
238 to unhash it or make it invalid
241 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
242 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
243 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
248 Description: open req refcounting wrong on reconnect
249 Details : If reconnect happened between getting open reply from server and
250 call to mdc_set_replay_data in ll_file_open, we will schedule
251 replay for unreferenced request that we are about to free.
252 Subsequent close will crash in variety of ways.
253 Check that request is still eligible for replay in
254 mdc_set_replay_data().
259 Description: disable writes to filesystem when reading health_check file
260 Details : the default for reading the health_check proc file has changed
261 to NOT do a journal transaction and write to disk, because this
262 can cause reads of the /proc file to hang and block HA state
263 checking on a healthy but otherwise heavily loaded system. It
264 is possible to return to the previous behaviour during configure
265 with --enable-health-write.
267 Severity : enhancement
269 Description: 64-bit inode version
270 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
271 to the inode. This will be required for version-based recovery.
276 Description: MDS fails to start if a duplicate client export is detected
277 Details : in some rare cases it was possible for a client to connect to
278 an MDS multiple times. Upon recovery the MDS would detect this
279 and fail during startup. Handle this more gracefully.
281 Severity : enhancement
283 Description: Add -o localflock option to simulate old noflock
285 Details : This will achieve local-only flock/fcntl locks
291 Description: log_commit_thread vs filter_destroy race leads to crash
292 Details : Take import reference before releasing llog record semaphore
297 Description: Wrong request locking in request set processing
298 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
299 imp_delayed_list, in this place should be used imp_lock.
302 Frequency : when reconnection
304 Description: Grant Leak when osc reconnect to OST
305 Details : When osc reconnect ost, OST(filter) should check whether it
306 should grant more space to client by comparing fed_grant and
307 cl_avail_grant, and return the granted space to client instead
308 of "new granted" space, because client will call osc_init_grant
309 to update the client grant space info.
312 Frequency : when client reconnect to OST
314 Description: Grant Leak when osc do resend and replay bulk write
315 Details : When osc reconnect to OST, OST(filter)should clear grant info of
316 bulk write request, because the grant info while be sync between
317 OSC and OST when reconnect, and we should ignore the grant info
318 these of resend/replay write req.
323 Description: Grant space more than avaiable left space sometimes.
324 Details : When then OST is about to be full, if two bulk writing from
325 different clients came to OST. Accord the avaliable space of the
326 OST, the first req should be permitted, and the second one
327 should be denied by ENOSPC. But if the seconde arrived before
328 the first one is commited. The OST might wrongly permit second
329 writing, which will cause grant space > avaiable space.
332 Frequency : when client is evicted
334 Description: Grant might be wrongly erased when osc is evicted by OST
335 Details : when the import is evicted by server, it will fork another
336 thread ptlrpc_invalidate_import_thread to invalidate the
337 import, where the grant will be set to 0. While the original
338 thread will update the grant it got when connecting. So if
339 the former happened latter, the grant will be wrongly errased
340 because of this race.
345 Description: Checking Stale with correct fid
346 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
347 to check whether it is stale, because sometimes, we want the
348 enqueue happened anyway, and op_data.fid2 will not be initialized.
350 Severity : enhancement
352 Description: update patchless client
353 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
356 Frequency : only with 2.4 kernel
358 Description: random memory corruption
359 Details : size of struct ll_inode_info is to big for union inode.u and this
360 can be cause of random memory corruption.
365 Description: Memory leak in recovery
366 Details : Lov_mds_md was not free in an error handler in mds_create_object.
367 It should also check obd_fail before fsfilt_start, otherwise if
368 fsfilt_start return -EROFS,(failover mds during mds recovery).
369 then the req will return with repmsg->transno = 0 and rc = EROFS.
370 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
371 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
372 be freed no matter whether fsfilt_commit success or not.
375 Frequency : only with huge count clients
377 Description: Prevents from taking the superblock lock in llap_from_page for
379 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
380 for llap_from_page call in ll_removepage prevents from taking the
381 superblock lock for a soon died page.
386 Description: Not check open intent error before release open handle
387 Details : in some rare cases, the open intent error is not checked before
388 release open handle, which may cause
389 ASSERTION(open_req->rq_transno != 0), because it tries to release
390 the failed open handle.
395 Description: Set cat log bitmap only after create log success.
396 Details : in some rare cases, the cat log bitmap is set too early. and it
397 should be set only after create log success.
401 Description: Accessing a block bevice can re-enable I/O when Lustre is
402 tearing down a device.
403 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
408 Description: service threads may hog cpus when there are a lot of requests
410 Details : Insert cond_resched to give other threads a chance to use some of
416 Description: the cat log was not initialized in recovery
417 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
418 unlink log on mds will not be initialized until mds post
419 recovery. And also in mds post recovery, the unlink log will
420 initialization will be done asynchronausly, so there will be race
421 between add unlink log and unlink log initialization.
425 Description: brw_stats were being printed incorrectly
426 Details : brw_stats were being printed as log2 but all of them were not
427 recorded as log2. Also remove some code duplication arising from
428 filter_tally_{read,write}.
432 Frequency : rare, only in recovery.
433 Description: ASSERTION(req->rq_type != LI_POISON) failed
434 Details : imp_lock should be held while iterating over imp_sending_list for
435 prevent destroy request after get timeout in ptlrpc_queue_wait.
439 Description: replay-single.sh test 52 fails
440 Details : A lock's skiplist need to be cleanup when it being unlinked
441 from its resource list.
445 Description: Short directio read returns full requested size rather than
447 Details : Direct I/O operations should return actual amount of bytes
448 transferred rather than requested size.
450 Severity : enhancement
452 Description: metadata RPC reduction (e.g. for rm performance)
453 Details : decrease the amount of synchronous RPC between clients and servers
454 by canceling conflicing lock before the operation on the client side
455 and packing thier handles into the main operation RPC to server.
457 --------------------------------------------------------------------------------
459 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
464 Frequency : on some architectures
466 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
467 Details : Uninitialized flags sometimes cause configuration commands to
471 Frequency : patchless clients only
473 Description: missing __iget() symbol export
474 Details : The __iget() symbol export is missing. To avoid the need for
475 this on patchless clients the deathrow inode reaper is turned
476 off, and we depend on the VM to clean up old inodes. This
477 dependency was during via the fix for bug 12181.
479 --------------------------------------------------------------------------------
481 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
483 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
484 INTEROPERATE with older versions automatically. In many cases a
485 special upgrade step is needed. Please read the
486 user documentation before upgrading any part of a live system.
487 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
488 version will not interoperate with 1.6 betas before beta5 (1.5.95).
489 * WARNING: Lustre configuration and startup changes are required with
490 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
495 Severity : enhancement
497 Description: MountConf
498 Details : Lustre configuration is now managed via mkfs and mount
499 commands instead of lmc and lconf. New obd types (MGS, MGC)
500 are added for dynamic configuration management. See
501 https://mail.clusterfs.com/wikis/lustre/MountConf for
504 Severity : enhancement
506 Description: dynamic OST addition
507 Details : OSTs can now be added to a live filesystem
509 Severity : enhancement
511 Description: startup order invariance
512 Details : MDTs and OSTs can be started in any order. Clients only
513 require the MDT to complete startup.
515 Severity : enhancement
517 Description: parallel, asynchronous orphan cleanup
518 Details : orphan cleanup is now performed in separate threads for each
519 OST, allowing parallel non-blocking operation.
521 Severity : enhancement
523 Description: optimized stripe assignment
524 Details : stripe assignments are now made based on ost space available,
525 ost previous usage, and OSS previous usage, in order to try
526 to optimize storage space and networking resources.
528 Severity : enhancement
530 Description: Permanently set tunables
531 Details : All writable /proc/fs/lustre tunables can now be permanently
532 set on a per-server basis, at mkfs time or on a live
535 Severity : enhancement
537 Description: Lustre message v2
538 Details : Add lustre message format v2.
540 Severity : enhancement
542 Description: client OST exclusion list
543 Details : Clients can be started with a list of OSTs that should be
544 declared "inactive" for known non-responsive OSTs.
548 Description: SPEC SFS validation failure on NFS v2 over lustre.
549 Details : Changes the blocksize for regular files to be 2x RPC size,
550 and not depend on stripe size.
552 Severity : enhancement
554 Description: Multiple MD RPCs in flight.
555 Details : Further unserialise some read-only MDS RPCs - learn about intents.
556 To avoid overly-overloading MDS, introduce a limit on number of
557 MDS RPCs in flight for a single client and add /proc controls
558 to adjust this limit.
560 Severity : enhancement
562 Description: client read/write statistics
563 Details : Add client read/write call usage stats for performance
564 analysis of user processes.
565 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
566 file access. extents_stats shows chunk size distribution.
567 extents_stats_per_process show chunk size distribution per
570 Severity : enhancement
572 Description: mds statistics
573 Details : Add detailed mds operations statistics in
574 /proc/fs/lustre/mds/*/stats.
578 Description: Failure of copying files with lustre special EAs.
579 Details : Client side always return success for setxattr call for lustre
580 special xattr (currently only "trusted.lov").
585 Description: Refcount LNET uuids
586 Details : The global LNET uuid list grew linearly with every startup;
587 refcount repeated list entries instead of always adding to
590 Severity : enhancement
592 Description: Dynamic service threads
593 Details : Within a small range, start extra service threads
594 automatically when the request queue builds up.
597 Frequency : mixed-endian client/server environments
599 Description: mixed-endian crashes
600 Details : The new msg_v2 system had some failures in mixed-endian
603 Severity : enhancement
605 Description: Easy OST removal
606 Details : OSTs can be permanently deactivated with e.g. 'lctl
607 conf_param lustre-OST0001.osc.active=0'
609 Severity : enhancement
611 Description: MGS proc entries
612 Details : Added basic proc entries for the MGS showing what filesystems
615 Severity : enhancement
617 Description: provide MGS failover
618 Details : Added config lock reacquisition after MGS server failover.
620 Severity : enhancement
622 Description: add Linux 2.4 support
623 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
627 Description: a large application tries to do I/O to the same resource and dies
629 Details : Check the req->rq_arrival time after the call to
630 ost_brw_lock_get(), but before we do anything about
631 processing it & sending the BULK transfer request. This
632 should help move old stale pending locks off the queue as
633 quickly as obd_timeout.
636 Frequency : when an incorrect nid is specified during startup
638 Description: ptlrpc connect to non-existant node causes kernel crash
639 Details : LNET can't be re-entered from an event callback, which
640 happened when we expire a message after the export has been
641 cleaned up. Instead, hand the zombie cleanup off to another
644 Severity : enhancement
646 Description: plain/inodebits lock performance improvement
647 Details : Grouping plain/inodebits in granted list by their request modes
648 and bits policy, thus improving the performance of search through
652 Frequency : only if OST filesystem is corrupted
654 Description: client incorrectly hits assertion in ptlrpc_replay_req()
655 Details : for a short time RPCs with bulk IO are in the replay list,
656 but replay of bulk IOs is unimplemented. If the OST filesystem
657 is corrupted due to disk cache incoherency and then replay is
658 started it is possible to trip an assertion. Avoid putting
659 committed RPCs into the replay list at all to avoid this issue.
662 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
664 Description: System hang on startup
665 Details : This bug allowed the liblustre (e.g. catamount) client to
666 return to the app before handling all startup RPCs. This
667 could leave the node unresponsive to lustre network traffic
668 and manifested as a server ptllnd timeout.
670 Severity : enhancement
672 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
673 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
674 boolean to control whether to print peer debug info when a
675 client's RPC times out.
678 Frequency : only for kernels with patches from Lustre below 1.4.3
680 Description: Remove old rdonly API
681 Details : Remove old rdonly API which unused from at least lustre 1.4.3
684 Frequency : only for devices with external journals
686 Description: Set external device read-only also
687 Details : During a commanded failover stop, we set the disk device
688 read-only while the server shuts down. We now also set any
689 external journal device read-only at the same time.
692 Frequency : when upgrading from 1.4 while trying to change parameters
694 Description: The wrong (new) MDC name was used when setting parameters for
695 upgraded MDT's. Also allows changing of OSC (and MDC)
696 parameters if --writeconf is specified at tunefs upgrade time.
699 Frequency : when setting specific ost indicies
701 Description: QOS code breaks on skipped indicies
702 Details : Add checks for missing OST indicies in the QOS code, so OSTs
703 created with --index need not be sequential.
705 Severity : enhancement
707 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
708 Details : The uninit_groups feature works in conjunction with the kernel
709 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
710 up the pass1 processing of e2fsck. This is a read-only feature
711 in ldiskfs2 only, so older kernels and current ldiskfs cannot
712 mount filesystems that have had this feature enabled.
714 Severity : enhancement
716 Description: Improve multi-block allocation algorithm to avoid fragmentation
717 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
718 allocation locality and avoid filesystem fragmentation.
720 ------------------------------------------------------------------------------
722 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
724 * Support for kernels:
725 2.4.21-47.0.1.EL (RHEL 3)
727 2.6.9-42.0.10.EL (RHEL 4)
728 2.6.12.6 vanilla (kernel.org)
729 2.6.16.27-0.9 (SLES 10)
730 * Recommended e2fsprogs version: 1.39.cfs5
732 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
736 Frequency : occasional, depends on client load and configuration
737 Bugzilla : 12181, 12203
738 Description: data loss for recently-modified files
740 Details : In some cases it is possible that recently written or created
741 files may not be written to disk in a timely manner (this should
742 normally be within 30s unless client IO load is very high).
743 The problem appears as zero-length files or files that are a
744 multiple of 1MB in size after a client crash or client eviction
745 that are missing data at the end of the file.
747 This problem is more likely to be hit on clients where files are
748 repeatedly created and unlinked in the same directory, clients
749 have a large amount of RAM, have many CPUs, the filesystem has
750 many OSTs, the clients are rebooted frequently, and/or the files
751 are not accessed by other nodes after being written.
753 The presence of the problem can be detected by looking at
754 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
755 smaller than the second (nr_unused) then dirty files will not
756 be flushed automatically to disk. "sync; sleep 10" should be
757 run several times on the node before unmounting it to update
758 Lustre (this is also safe to run on nodes without this problem).
760 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
761 that can cause this same problem, so customers using that kernel
762 also need to update the kernel in addition to Lustre. In order
763 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
765 It is normal that files written just before a client crash (less
766 than 30s) may not yet have been flushed to disk, even for local
770 Frequency : frequent on thin XT3 nodes
772 Description: UUID collision on thin XT3 Linux nodes
773 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
774 randomly, since we relied on an insufficiently-seeded PRNG.
779 Description: OSS hangs after "All ost request buffers busy"
780 Details : A deadlock between quota and journal operations caused OSS
781 hangs after printing "All ost request buffers busy."
784 Frequency : always on liblustre builds
786 Description: Cleanup compiler warnings on liblustre
789 Frequency : always on liblustre builds on XT3
791 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
795 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
796 > 32000 subdirectories
797 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
798 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
799 subdirectory count crosses 32000. This will aid e2fsck to
800 correctly handle more than 32000 subdirectories.
803 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
805 Description: System hang on startup
806 Details : This bug allowed the liblustre (e.g. catamount) client to
807 return to the app before handling all startup RPCs. This
808 could leave the node unresponsive to lustre network traffic
809 and manifested as a server ptllnd timeout.
811 Severity : enhancement
813 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
814 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
815 boolean to control whether to print peer debug info when a
816 client's RPC times out.
821 Description: make O_SYNC working on 2.6 kernels
822 Details : 2.6 kernels use different method for mark pages for write,
823 so need add a code to lustre for O_SYNC work.
828 Description: Failure to close file and release space on NFS
829 Details : Put inode details into lock acquired in ll_intent_file_open.
830 Use mdc_intent_lock in ll_intent_open to properly
831 detect all kind of errors unhandled by mdc_enqueue.
836 Description: proc file read during shutdown sometimes raced obd removal,
838 Details : Add lock to prevent obd access after proc file removal.
841 Frequency : Only for files larger than 4GB on 32-bit clients.
843 Description: improperly doing page alignment of locks
844 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
845 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
850 Description: RPCs being resent when they shouldn't be
851 Details : Some RPCs that should not be resent are being resent. This
852 can cause inconsistencies in the RPC state machine. Do not
853 resend such requests.
856 Frequency : rare, only with NFS export
858 Description: Crash on NFS re-export node
859 Details : under very unusual load conditions an assertion is hit in
860 ll_intent_file_open()
863 Frequency : only if OST filesystem is corrupted
865 Description: client incorrectly hits assertion in ptlrpc_replay_req()
866 Details : for a short time RPCs with bulk IO are in the replay list,
867 but replay of bulk IOs is unimplemented. If the OST filesystem
868 is corrupted due to disk cache incoherency and then replay is
869 started it is possible to trip an assertion. Avoid putting
870 committed RPCs into the replay list at all to avoid this issue.
875 Description: large O_DIRECT requests fail under memory pressure/fragmentation
876 Details : Large single O_DIRECT read and write calls can fail to allocate
877 a sufficiently large buffer to process the request. In case of
878 allocation failure the allocation is retried with a smaller
879 buffer and broken into smaller requests.
881 Severity : enhancement
883 Description: Add -o localflock option to simulate old noflock behaviour.
884 Details : This will achieve local-only flock/fcntl locks coherentness.
889 Description: versioning check is incomplete
890 Details : Checking the version difference of client vs. server, report
891 error if the gap is too big.
896 Description: add support PG_writeback bit
897 Details : add support for PG_writeback bit for Lustre, for more carefull
898 work with page cache in 2.6 kernel. This also fix some deadlocks
899 and remove hack for work O_SYNC with 2.6 kernel.
901 Severity : enhancement
903 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
904 Details : The uninit_groups feature works in conjunction with the kernel
905 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
906 up the pass1 processing of e2fsck. This is a read-only feature
907 in ldiskfs2 only, so older kernels and current ldiskfs cannot
908 mount filesystems that have had this feature enabled.
910 Severity : enhancement
912 Description: Improve multi-block allocation algorithm to avoid fragmentation
913 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
914 allocation locality and avoid filesystem fragmentation.
916 ------------------------------------------------------------------------------
918 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
920 * Support for kernels:
921 2.6.9-42.0.3.EL (RHEL 4)
923 2.4.21-47.0.1.EL (RHEL 3)
924 2.6.12.6 vanilla (kernel.org)
925 2.6.16.21-0.8 (SLES10)
926 * Recommended e2fsprogs version: 1.39.cfs2-0
928 * The backwards-compatible /proc/sys/portals symlink has been removed
929 in this release. Before upgrading, please ensure that you change
930 any configuration scripts or /etc/sysctl.conf files that access
931 /proc/sys/portals/* or sysctl portals.* to use the corresponding
932 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
933 in advance of the upgrade on any system running Lustre 1.4.6 or
934 newer, since /proc/sys/lnet was added in that version.
935 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
939 Frequency : only when quota is used
941 Description: avoid scanning export list for quota master
942 Details : Change the algorithms to avoid scanning export list in order
943 to improve the efficiency.
946 Frequency : MDS failover only, very rarely
948 Description: "went back in time" messages on mds failover
949 Details : The greatest transno may be lost when the current operation
950 finishes with an error (transno==0) and the client's last_rcvd
951 record is over-written. Save the greatest transno in the
952 mds_last_transno for this case.
955 Frequency : always for specific kernels and striping counts
957 Description: client may get "Matching packet too big" without ACL support
958 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
959 when trying to access files in certain configurations. The
960 clients should in fact be denied when mounting because they do
964 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
966 Description: many clients connecting with IO in progress causes connect timeouts
967 Details : Avoid synchronous journal commits to avoid delays caused by many
968 clients connecting/disconnecting when bulk IO is in progress.
969 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
970 OST_IO_PORTAL to avoid delays behind potentially many pending
974 Frequency : occasionally with multiple writers to a single file
976 Description: shared writes to file may result in wrong size reported by stat()
977 Details : Allow growing of kms when extent lock is cancelled
980 Frequency : always with random mmap IO to multi-striped file
982 Description: mmap write might be lost if we are writing to a 'hole' in stripe
983 Details : Only if the hole is at the end of OST object so that kms is too
984 small. Fix is to increase kms accordingly in ll_nopage.
987 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
989 Description: writes to a missing object would leak memory on the OST
990 Details : If there is an inconsistency between the MDS and OST filesystems,
991 such that the MDS references an object that doesn't exist, writes
992 to that object will leak memory due to incorrect cleanup in the
993 error handling path, eventually running out of memory on the OST.
998 Description: Creating too long symlink causes lustre errors
999 Details : Check symlink and name lengths before sending requests to MDS.
1002 Frequency : only if flock is enabled (not on by default)
1004 Description: posix locks not released on fd closure on 2.6.9+
1005 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1006 kernels, this caused such locks not to be released on fd close and
1007 then assertions on fs unmount about still used locks.
1010 Frequency : MDS failover only, very rarely
1012 Description: clients may get ASSERTION(granted_lock != NULL)
1013 Details : When request was taking a long time, and a client was resending
1014 a getattr by name lock request. The were multiple lock requests
1015 with the same client lock handle and
1016 mds_getattr_name->fixup_handle_for_resent_request found one of the
1017 lock handles but later failed with ASSERTION(granted_lock != NULL).
1022 Description: handle->h_buffer_credits > 0, assertion failure
1023 Details : h_buffer_credits is zero after truncate, causing assertion
1024 failure. This patch extends the transaction or creates a new
1028 Frequency : NFS re-export or patchless client
1029 Bugzilla : 11179, 10796
1030 Description: Crash on NFS re-export node (__d_move)
1031 Details : We do not want to hash the dentry if we don't have a lock.
1032 But if this dentry is later used in d_move, we'd hit uninitialised
1033 list head d_hash, so we just do this to init d_hash field but
1034 leave dentry unhashed.
1037 Frequency : NFS re-export or patchless client
1039 Description: NFS exports has problem with symbolic link
1040 Details : lustre client didn't properly install dentry when re-exported
1041 to NFS or running patchless client.
1044 Frequency : NFS re-export or patchless client
1046 Description: Various nfs/patchless fixes.
1047 Details : fixes reuse disconected alias for lookup process - this fixes
1048 warning "find_exported_dentry: npd != pd",
1049 fix permission error with open files at nfs.
1050 fix apply umask when do revalidate.
1053 Frequency : occasional
1055 Description: Crash on NFS re-export node
1056 Details : calling clear_page() on the wrong pointer triggered oops in
1057 generic_mapping_read().
1060 Frequency : rarely, using O_DIRECT IO
1062 Description: unaligned directio crashes client with LASSERT
1063 Details : check for unaligned buffers before trying any requests.
1066 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1068 Description: stale data returned from RAID cache
1069 Details : If only a small amount of IO is done to the RAID device before
1070 reading it again it is possible to get stale data from the RAID
1071 cache instead of reading it from disk.
1074 Frequency : always for sles10 kernel
1076 Description: sles10 support
1077 Details : ll_follow_link: compile fixes and using of nd_set_link
1078 under newer kernels.
1081 Frequency : depends on arch, kernel and compiler version, always on sles10
1084 Description: recursive or deep enough symlinks cause stack overflow
1085 Details : getting rid of large stack-allocated variable in
1089 Frequency : depends on hardware
1091 Description: lustre write performance loss in the SLES10 kernel
1092 Details : the performance loss is caused by using of write barriers in the
1093 ext3 code. The SLES10 kernel turns barrier support on by
1094 default. The fix is to undo that change for ldiskfs.
1096 ------------------------------------------------------------------------------
1098 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1100 * Support for kernels:
1101 2.6.9-42.0.3EL (RHEL 4)
1102 2.6.5-7.276 (SLES 9)
1103 2.4.21-47.0.1.EL (RHEL 3)
1104 2.6.12.6 vanilla (kernel.org)
1108 Frequency : quota enabled and large files being deleted
1110 Description: releasing more than 4GB of quota at once hangs OST
1111 Details : If a user deletes more than 4GB of files on a single OST it
1112 will cause the OST to spin in an infinite loop. Release
1113 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1118 Description: statfs data retrieved from /proc may be stale or zero
1119 Details : When reading per-device statfs data from /proc, in the
1120 {kbytes,files}_{total,free,avail} files, it may appear
1121 as zero or be out of date.
1124 Frequency : systems with MD RAID1 external journal devices
1126 Description: lconf's call to blkid is confused by RAID1 journal devices
1127 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1128 of returning all block devices that match the journal UUID.
1131 Frequency : always, for aggregate stripe size over 4GB
1133 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1134 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1135 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1138 Frequency : always if "lfs find" used on a local file/directory
1140 Description: "lfs find" segfaults if used on a local file/directory
1141 Details : The case where a directory component was not specified wasn't
1142 handled correctly. Handle this properly.
1145 Frequency : always on ppc64
1147 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1148 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1149 perform endianness conversion.
1152 Frequency : rarely (truncate to non-zero file size after write under load)
1153 Bugzilla : 10730, 10687
1154 Description: Files padded with zeros to next 4K multiple
1155 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1156 it is possible that files that are truncated to a non-zero size
1157 immediately after being written are filled with zero bytes beyond
1158 the truncated size. No file data is lost.
1160 Severity : enhancement
1162 Description: Allow recovery/failover for liblustre clients.
1163 Details : liblustre clients were unaware of failover configurations until
1166 Severity : enhancement
1168 Description: user file locks should fail when not mounting with flock option
1169 Details : Set up an error-returning stub in ll_file_operations.lock field
1170 to prevent incorrect behaviour when client is mounted without
1171 flock option. Also, set up properly f_op->flock field for
1175 Frequency : always on ia64
1177 Description: "lfs df" loops on printing out MDS statfs information
1178 Details : The obd_ioctl_data was not initialized and in some systems
1179 this caused a failure during the ioctl that did not return
1180 an error. Initialize the struct and return an error on failure.
1183 Frequency : SLES 9 only
1185 Description: Error of copying files with lustre special EAs as root
1186 Details : Client side always return success for setxattr call for lustre
1187 special xattr (currently only "trusted.lov").
1190 Frequency : rarely on clusters with both ia64+i386 clients
1192 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1193 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1194 clients are doing concurrent writes to the same file it is
1195 possible that the ia64 clients may LASSERT because the OST
1196 extent locks are not PAGE_SIZE aligned. Ensure that grown
1197 locks are always aligned on the request boundary.
1200 Frequency : specific use, occasional
1202 Description: Overwriting in use executable truncates on-disk binary image
1203 Details : If one node attempts to overwrite an executable in use by
1204 another node, we now correctly return ETXTBSY instead of
1205 truncating the file.
1207 Severity : enhancement
1209 Description: Async OSC create to avoid the blocking unnecessarily.
1210 Details : If a OST has no remain object, system will block on the creating
1211 when need to create a new object on this OST. Now, ways use
1212 pre-created objects when available, instead of blocking on an
1213 empty osc while others are not empty. If we must block, we block
1214 for the shortest possible period of time.
1219 Description: chmod on Lustre root is propagated to other clients
1220 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1221 invalid by the follow_mount time.
1224 Frequency : liblustre clients only
1226 Description: Race in 'instant cancel' lock handling could lead to such locks
1227 never to be granted in case of SMP MDS
1228 Details : Do not destroy not yet granted but cbpending locks in
1232 Frequency : replay/resend of open
1234 Description: non null lock assetion failure in mds_intent_policy
1235 Details : Trying to replay/resend lockless open requests resulted in
1236 mds_open() returning 0 with no lock. Now it sets a flag if
1237 a lock is going to be returned.
1239 Severity : enhancement
1241 Description: Checksum enhancements
1242 Details : New checksum enhancements allow for resending RPCs that failed
1245 Severity : enhancement
1247 Description: Tunables on number of dirty pages in cacche
1248 Details : Allow to set limit on number of dirty pages cached.
1253 Description: client crash on unmount - lock still has references
1254 Details : In some error handling cases it was possible to leak a lock
1255 reference on a client while accessing a file. This was not
1256 harmful to the client during operation, but would cause the
1257 client to crash when the filesystem is unmounted.
1260 Frequency : specific case, rare
1262 Description: ETXTBSY on mds though file not in use by client
1263 Details : ETXTBSY is no longer incorrectly returned when attempting to
1264 chmod or chown a directory that the user previously tried to
1265 execute or a currently-executing binary.
1268 Frequency : extremely rare except on liblustre-based clients
1270 Description: Lustre space not freed when files are deleted
1271 Details : Clean up open-unlinked files after client eviction. Previously
1272 the unlink was skipped and the files remained as orphans.
1277 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1278 Details : In some cases it was possible to send a blocking callback to
1279 a client doing a glimpse, even though that client didn't get
1280 a lock granted. When the glimpse lock is cancelled on the OST
1281 the freed lock is left on the waiting list and corrupted the list.
1284 Frequency : all core dumps
1286 Description: Broke core dumps to lustre
1287 Details : Negative dentry may be unhashed if parent does not have UPDATE
1288 lock, but some callers, e.g. do_coredump, expect dentry to be
1289 hashed after successful create, hash it in ll_create_it.
1291 ------------------------------------------------------------------------------
1293 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1295 * Support for kernels:
1296 2.6.9-42.0.2.EL (RHEL 4)
1297 2.6.5-7.276 (SLES 9)
1298 2.4.21-40.EL (RHEL 3)
1299 2.6.12.6 vanilla (kernel.org)
1303 Frequency : always on RHEL 3
1305 Description: Number of open files grows over time
1306 Details : The number of open files grows over time, whether or not
1307 Lustre is started. This was due to a filp leak introduced
1308 by one of our kernel patches.
1310 ------------------------------------------------------------------------------
1312 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1314 * Support for kernels:
1315 2.6.9-42.EL (RHEL 4)
1316 2.6.5-7.276 (SLES 9)
1317 2.4.21-40.EL (RHEL 3)
1318 2.6.12.6 vanilla (kernel.org)
1323 Bugzilla : 5719, 9635, 9792, 9684
1324 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1325 Details : If a server is under heavy load and cannot reply to new
1326 connection requests before the client resends the (re)connect,
1327 the connection handling code can behave badly if two service
1328 threads are concurrently handing separate (re)connections from
1329 the same client. Add better locking to the connection handling
1330 code, and ensure that only a single connection will be processed
1331 for a given client UUID, even if the lock is dropped.
1333 Severity : enhancement
1335 Description: add TCP zero-copy support to kernel
1336 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1337 sends if the hardware supports scatter-gather and checksumming.
1338 This allows socklnd to do client-write and server-read more
1339 efficiently and reduce CPU utilization from skbuf copying.
1342 Frequency : only if NFS exporting from client
1344 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1345 Details : If setting attributes on a file created under NFS that had
1346 never been opened it would be possible to oops the client
1347 if the file had no objects.
1350 Frequency : always for liblustre
1352 Description: liblustre client does MDS+OSTs setattr RPC for each write
1353 Details : When doing a write from a liblustre client, the client
1354 incorrectly issued an RPC to the MDS and each OST the file was
1355 striped over in order to update the timestamps. When writing
1356 with small chunks and many clients this could overwhelm the MDS
1357 with RPCs. In all cases it would slow down the write because
1358 these RPCs are unnecessary.
1360 Severity : enhancement
1362 Description: allow number of MDS service threads to be changed at module load
1363 Details : It is now possible to change the number of MDS service threads
1364 running. Adding "options mds mds_num_threads={N}" to the MDS's
1365 /etc/modprobe.conf will set the number of threads for the next
1366 time Lustre is restarted (assuming the "mds" module is also
1367 reloaded at that time). The default number of threads will
1368 stay the same, 32 for most systems.
1373 Description: OST crash if filesystem is unformatted or corrupt
1374 Details : If an OST is started on a device that has never been formatted
1375 or if the filesystem is corrupt and cannot even mount then the
1376 error handling cleanup routines would dereference a NULL pointer.
1381 Description: NULL pointer deref in llap_from_page.
1382 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1383 incorrect) mapping if the page was truncated/reclaimed while it was
1384 searched for. Check for this condition and skip such pages when
1385 doing readahead. Introduce extra check to llap_from_page() to
1386 verify page->mapping->host is non-NULL (so page is not anonymous).
1389 Frequency : Sometimes when using sys_sendfile
1391 Description: "page not covered by a lock" warnings from ll_readpage
1392 Details : sendfile called ll_readpage without right page locks present.
1393 Now we introduced ll_file_sendfile that does necessary locking
1394 around call to generic_file_sendfile() much like we do in
1398 Frequency : with certain MDS communication failures at client mount time
1400 Description: NULL pointer deref after failed client mount
1401 Details : a client connection request may delayed by the network layer
1402 and not be sent until after the PTLRPC layer has timed out the
1403 request. If the client fails the mount immediately it will try
1404 to clean up before the network times out the request. Add a
1405 reference from the request import to the obd device and delay
1406 the cleanup until the network drops the request.
1409 Frequency : occasionally during client (re)connect
1411 Description: assertion failure during client (re)connect
1412 Details : processing a client connection request may be delayed by the
1413 client or server longer than the client connect timeout. This
1414 causes the client to resend the connection request. If the
1415 original connection request is replied in this interval, the
1416 client may trip an assertion failure in ptlrpc_connect_interpret()
1417 which thought it would be the only running connect process.
1420 Frequency : only with obd_echo servers and clients that are rebooted
1422 Description: kernel BUG accessing uninitialized data structure
1423 Details : When running an obd_echo server it did not start the ping_evictor
1424 thread, and when a client was evicted an uninitialized data
1425 structure was accessed. Start the ping_evictor in the RPC
1426 service startup instead of the OBD startup.
1428 Severity : enhancement
1429 Bugzilla : 10193 (patchless)
1430 Description: Remove dependency on various unexported kernel interfaces.
1431 Details : No longer need reparent_to_init, exit_mm, exit_files,
1432 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1435 Frequency : rare (only users of deprecated and unsupported LDAP config)
1437 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1438 Details : LDAP apparently contains 'lustreName' attributes instead of
1439 'name'. A simple remapping of the name is sufficient.
1442 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1444 Description: waiting_locks_callback trips kernel BUG if client is evicted
1445 Details : Running with the dump_on_timeout debug flag turned on makes
1446 it possible that the waiting_locks_callback() can try to dump
1447 the Lustre kernel debug logs from an interrupt handler. Defer
1448 this log dumping to the expired_lock_main() thread.
1450 Severity : enhancement
1452 Description: Support NFS exporting on 2.6 kernels.
1453 Details : Implement non-rawops metadata methods for NFS server to use without
1454 changing NFS server code.
1457 Frequency : very rare (synthetic metadata workload only)
1459 Description: two racing renames might cause an MDS thread to deadlock
1460 Details : Running the "racer" program may cause one MDS thread to rename
1461 a file from being the source of a rename to being the target of
1462 a rename at exactly the same time that another thread is doing
1463 so, and the second thread has already enqueued these locks after
1464 doing a lookup of the target and is trying to relock them in
1465 order. Ensure that we don't try to re-lock the same resource.
1468 Frequency : only very large systems with liblustre clients
1470 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1471 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1472 all OSTs in parallel. This allows the eviction of stale liblustre
1473 clients to proceed much faster than if they were done in series,
1474 and also offers similar improvements for other set_info RPCs.
1479 Description: excessive CPU usage during initial read phase on client
1480 Details : During the initial read phase on a client, it would agressively
1481 retry readahead on the file, consuming too much CPU and impacting
1482 performance (since 1.4.5.8). Improve the readahead algorithm
1483 to avoid this, and also improve some other common cases (read
1484 of small files in particular, where "small" is files smaller than
1485 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1490 Description: MDS crash when receiving packet with unknown intent.
1491 Details : Do not LBUG in unknown intent case, just return -EFAULT
1493 Severity : enhancement
1494 Bugzilla : 9293, 9385
1495 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1496 Details : Do not serialize getattr (non-intent version) and statfs.
1499 Frequency : occasional, when OST network is overloaded/intermittent
1501 Description: client evicted by OST after bulk IO timeout
1502 Details : If a client sends a bulk IO request (read or write) the OST
1503 may evict the client if it is unresposive to its data GET/PUT
1504 request. This is incorrect if the network is overloaded (takes
1505 too long to transfer the RPC data) or dropped the OST GET/PUT
1506 request. There is no need to evict the client at all, since
1507 the pinger and/or lock callbacks will handle this, and the
1508 client can restart the bulk request.
1511 Frequency : Always when mmapping file with no objects
1513 Description: client crashes when mmapping file with no objects
1514 Details : Check that we actually have objects in a file before doing any
1515 operations on objects in ll_vm_open, ll_vm_close and
1521 Description: Request leak when working with deleted CWD
1522 Details : Introduce advanced request refcount tracking for requests
1523 referenced from lustre intent.
1525 Severity : Enhancement
1527 Description: Cache open file handles on client.
1528 Details : MDS now will return special lock along with openhandle, if
1529 requested and client is allowed to hold openhandle, even if unused,
1530 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1531 closing files for every read/write openration.
1533 Severity : Enhancement
1535 Description: Cache open negative dentries on client when possible.
1536 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1537 negative dentries on lock revocation.
1542 Description: Remounting a client read-only wasn't possible with a zconf mount
1543 Details : It wasn't possible to remount a client read-only with llmount.
1545 Severity : enhancement
1546 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1547 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1548 MPICH is included to add Lustre-specific ADIO interfaces.
1549 This is based closely on the UFS ADIO layer and only differs
1550 in file creation, in order to allow the OST striping to be set.
1551 This is user-contributed code and not supported by CFS.
1556 Description: extended inode attributes (immutable, append-only) work improperly
1557 when 2.4 and 2.6 kernels are used on client/server or vice versa
1558 Details : Introduce kernel-independent values for these flags.
1560 Severity : enhancement
1563 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1564 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1565 granularity. Now, Fractional MB granularity is be supported,
1566 this is very useful for low memory system.
1568 Severity : enhancement
1570 Description: Getattr by fid
1571 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1572 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1573 have locks and avoids getattr rpc requests for it on every stat.
1576 Frequency : Always, for filesystems larger than 2TB
1578 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1579 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1580 filesystems larger than 2TB to be mounted with mballoc enabled.
1583 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1585 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1586 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1587 filesystems larger than 2TB is mounted then the kernel will
1588 silently corrupt the start of the filesystem. CONFIG_LBD is
1589 enabled for all CFS-supported kernels, but the possibility of
1590 this happening with a modified kernel config exists.
1592 Severity : enhancement
1594 Description: add client O_DIRECT support for 2.6 kernels
1595 Details : It is now possible to do O_DIRECT reads and writes to files
1596 in the Lustre client mountpoint on 2.6 kernel clients.
1598 Severity : enhancement
1600 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1601 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1602 parallel, not waiting for response from every OST before sending
1603 a rpc to the next OST.
1608 Description: setattr vs write race when updating file timestamps
1609 Details : Client processes that update a file timestamp into the past
1610 right after writing to the file (e.g. tar) it is possible that
1611 the updated file modification time can be reset to the current
1612 time due to a race between processing the setattr and write RPC.
1614 Severity : enhancement
1616 Description: Bring 'lfs find' closer in line with regular Linux find.
1617 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1618 -print0 options and obtains all the needed info through the lustre
1621 Severity : enhancement
1623 Description: support up to 1024 configured devices on one node
1624 Details : change obd_dev array from statically allocated to dynamically
1625 allocated structs as they are first used to reduce memory usage
1630 Description: Flush dirty partially truncated pages during truncate
1631 Details : Immediatelly flush partially truncated pages in filter_setattr,
1632 this way we completely avoid having any pages in page cache on OST
1633 and can retire ugly workarounds during writes to flush such pages.
1638 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1639 Details : protect inode from truncation within vfs_unlink() context
1640 just take a reference before calling vfs_unlink() and release it
1641 when parent's i_sem is free.
1646 Description: last_id value checked outside lock on OST caused LASSERT failure
1647 Details : If there were multiple MDS->OST object precreate requests in
1648 flight, it was possible that the OST's last object id was checked
1649 outside a lock and incorrectly tripped an assertion. Move checks
1650 inside locks, and discard old precreate requests.
1653 Frequency : always, if extents are used on OSTs
1655 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1656 Details : OSTs using the extents format would not zero the high 16 bits of
1657 the index physical block number. This is not a problem for any
1658 OST filesystems smaller than 16TB, and no kernels support ext3
1659 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1660 new/modified files) and can be fixed for existing filesystems
1661 with e2fsprogs-1.39-cfs1.
1666 Description: import connection selection may be incorrect if timer wraps
1667 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1668 import connections to be ignored if the 32-bit jiffies counter
1669 wraps. Use a 64-bit jiffies counter.
1672 Frequency : very large clusters immediately after boot
1674 Description: LNET request buffers exhausted under heavy short-term load
1675 Details : If a large number of client requests are generated on a service
1676 that has previously never seen so many requests it is possible
1677 that the request buffer growth cannot keep up with the spike in
1678 demand. Instead of dropping incoming requests, they are held in
1679 the LND until the RPC service can accept more requests.
1682 Frequency : Sometimes during replay
1684 Description: Assertion failure in ll_local_open after replay.
1685 Details : If replay happened on an open request reply before we were able
1686 to set replay handler, reply will become not swabbed tripping the
1687 assertion in ll_local_open. Now we set the handler right after
1688 recognising of open request
1691 Frequency : very rare
1693 Description: kernel reports "badness in vsnprintf"
1694 Details : Reading from the "recovery_status" /proc file in small chunks
1695 may cause a negative length in lprocfs_obd_rd_recovery_status()
1696 call to vsnprintf() (which is otherwise harmless). Exit early
1697 if there is no more space in the output buffer.
1699 Severity : enhancement
1701 Description: clear OBD RPC statistics by writing to them
1702 Details : It is now possible to clear the OBD RPC statistics by writing
1703 to the "stats" file.
1708 Description: Client mtime is not the same on different clients after utimes
1709 Details : In some cases, the client was using the utimes() syscall on
1710 a file cached on another node. The clients now validate the
1711 ctime from the MDS + OSTs to determine which one is right.
1716 Description: Inability to activate failout mode
1717 Details : lconf script incorrectly assumed that in pythong string's numeric
1718 value is used in comparisons.
1721 Frequency : always with multiple stripes per file
1723 Description: Inefficient object allocation for mutli-stripe files
1724 Details : When selecting which OSTs to stripe files over, for files with
1725 a stripe count that divides evenly into the number of OSTs,
1726 the MDS is always picking the same starting OST for each file.
1727 Return the OST selection heuristic to the original design.
1732 Description: mount failures may take full timeout to return an error
1733 Details : Under some heavy load conditions it is possible that a
1734 failed mount can wait for the full obd_timeout interval,
1735 possibly several minutes, before reporting an error.
1736 Instead return an error as soon as the status is known.
1738 Frequency : quota enabled and large files being deleted
1740 Description: releasing more than 4GB of quota at once hangs OST
1741 Details : If a user deletes more than 4GB of files on a single OST it
1742 will cause the OST to spin in an infinite loop. Release
1743 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1748 Description: statfs data retrieved from /proc may be stale or zero
1749 Details : When reading per-device statfs data from /proc, in the
1750 {kbytes,files}_{total,free,avail} files, it may appear
1751 as zero or be out of date.
1754 Frequency : systems with MD RAID1 external journal devices
1756 Description: lconf's call to blkid is confused by RAID1 journal devices
1757 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1758 of returning all block devices that match the journal UUID.
1761 Frequency : always, for aggregate stripe size over 4GB
1763 Description: assertion fails when trying to use 4GB stripe size
1764 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1765 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1768 Frequency : always on ppc64
1770 Description: the first write on an ext3 filesystem with mballoc got stuck
1771 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1772 perform endianness conversion.
1774 ------------------------------------------------------------------------------
1776 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1778 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1779 INTEROPERATE with older versions automatically. Please read the
1780 user documentation before upgrading any part of a live system.
1781 * WARNING: Lustre networking configuration changes are required with
1782 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1785 * Support for kernels:
1786 2.6.9-22.0.2.EL (RHEL 4)
1787 2.6.5-7.244 (SLES 9)
1788 2.6.12.6 vanilla (kernel.org)
1791 Severity : enhancement
1792 Bugzilla : 7981/8208
1793 Description: Introduced Lustre Networking (LNET)
1794 Details : LNET is new networking infrastructure for Lustre, it includes
1795 a reorganized network configuration mode (see the user
1796 documentation for full details) as well as support for routing
1797 between different network fabrics. Lustre Networking Devices
1798 (LNDs) for the supported network fabrics have also been
1799 created for this new infrastructure.
1801 Severity : enhancement
1802 Description: Introduced Access control lists
1803 Details : clients can set ACLs on files and directories in order to have
1804 more fine-grained permissions than the standard Unix UGO+RWX.
1805 The MDS must be started with the "-o acl" mount option.
1807 Severity : enhancement
1808 Description: Introduced filesystem quotas
1809 Details : Administrators may now establish per-user quotas on the
1812 Severity : enhancement
1814 Description: Configuration change for the XT3
1815 The PTLLND is now used to run Lustre over Portals on the XT3
1816 The configure option(s) --with-cray-portals are no longer used.
1817 Rather --with-portals=<path-to-portals-includes> is used to
1818 enable building on the XT3. In addition to enable XT3 specific
1819 features the option --enable-cray-xt3 must be used.
1824 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1825 Details : A race between allocating a new llcd and re-getting the llcd_lock
1826 allowed another thread to grab newly-allocated llcd.
1828 Severity : enhancement
1830 Description: 2.6 OST async journal commit and locking fix to improve performance
1831 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1832 2.6 kernels are now async as they already were in 2.4 kernels so
1833 that they can commit concurrently with the network bulk transfer.
1834 For block-allocated files the filter allocation semaphore is held
1835 to avoid filesystem fragmentation during allocation. BKL lock
1836 removed for 2.6 xattr operations where it is no longer needed.
1841 Description: lconf incorrectly determined whether two IP networks could talk
1842 Details : In some more complicated routing and multiple-network
1843 configurations, lconf will avoid trying to make a network
1844 connection to a disjoint part of the IP space. It was doing the
1845 math incorrectly for one set of cases.
1850 Description: Fix for potential infinite loop processing records in an llog.
1851 Details : If an llog record is corrupted/zeroed, it is possible to loop
1852 forever in llog_process(). Validate the llog record length
1853 and skip the remainder of the block on error.
1856 Frequency : occasional (liblustre only)
1858 Description: liblustre could not open files whose last component is a symlink
1859 Details : sysio_path_walk() would incorrectly pass the open intent to
1860 intermediate path components.
1863 Frequency : rare (liblustre only with non-standard tuning)
1864 Bugzilla : 7201 (7350)
1865 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1866 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1867 after it has been released, only do this for referenced locks
1869 Severity : enhancement
1871 Description: specifying an (invalid) directory default stripe_size of -1
1872 would reset the directory default striping
1873 Details : stripe_size -1 was used internally to signal directory stripe
1874 removal, now use "all default" to signal dir stripe removal
1875 as a directory striping of "all default" is not useful
1878 Frequency : common for large clusters running liblustre clients
1880 Description: doing an ls when liblustre clients are running is slow
1881 Details : sending a glimpse AST to a liblustre client waits for every AST
1882 to time out, as liblustre clients will not respond. Since they
1883 cannot cache data we refresh the OST lock LVB from disk instead.
1885 Severity : enhancement
1887 Description: doing an ls at the same time as file IO can be slow
1888 Details : enqueue and other "small" requests can be blocked behind many
1889 large IO requests. Create a new OST IO portal for non-IO
1890 requests so they can be processed faster.
1893 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1895 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1896 created file with mode 000
1897 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1898 allow an ftruncate() as a non-root user to a file with mode 000.
1899 Lustre now respects this flag to disable mode checks when
1900 truncating a file owned by the user
1903 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1906 Description: Revoking locks from clients that went dead or catatonic might take
1908 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1909 cancellation of such locks instant on servers without waiting for
1910 any reply from clients. Clients drops these locks when cancel
1911 notification from server is received without replying.
1914 Frequency : liblustre-only, when liblustre client dies or becomes busy
1916 Description: Doing ls on Linux clients can take a long time with active
1918 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1919 granting such locks to it in the first place if possible. Locks
1920 are taken by proxy on the OST during the read or write and
1921 dropped immediately afterward. Add connect flags handling, do
1922 not grant locks to liblustre clients for glimpse ASTs.
1924 Severity : enhancement
1926 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1927 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1928 sequential reads and behaves badly with applications doing random
1929 reads. Improve it by reading ahead at least read region, and
1930 avoiding excessive large RPC for small reads.
1932 Severity : enhancement
1934 Description: Creating more than 1000 files for a single job may cause a load
1935 imbalance on the OSTs if there are also a large number of OSTs.
1936 Details : qos_prep_create() uses an OST index reseed value that is an
1937 even multiple of the number of available OSTs so that if the
1938 reseed happens in the middle of the object allocation it will
1939 still utilize the OSTs as uniformly as possible.
1944 Description: OST or MDS may oops in ping_evictor_main()
1945 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1946 but doesn't restart at beginning of obd_exports_timed list
1949 Severity : enhancement
1951 Description: improve by-nid export eviction on the MDS and OST
1952 Details : allow multiple exports with the same NID to be evicted at one
1953 time without re-searching the exports list.
1956 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1958 Description: MDS may oops in groups_free()
1959 Details : in rare race conditions a newly allocated group_info struct is
1960 freed again, and this can be NULL. The 2.4 compatibility code
1961 for groups_free() checked for a NULL pointer, but 2.6 did not.
1964 Frequency : common for liblustre clients doing little filesystem IO
1965 Bugzilla : 9352, 7313
1966 Description: server may evict liblustre clients accessing contended locks
1967 Details : if a client is granted a lock or receives a completion AST
1968 with a blocking AST already set it would not reply to the AST
1969 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1972 Frequency : lfs setstripe, only systems with more than 160 OSTs
1974 Description: unable to set striping with a starting offset beyond OST 160
1975 Details : llapi_create_file() incorrectly limited the starting stripe
1976 index to the maximum single-file stripe count.
1979 Frequency : LDAP users only
1981 Description: lconf did not handle in-kernel recovery with LDAP properly
1982 Details : lconf/LustreDB get_refs() is searching the wrong namespace
1984 Severity : enhancement
1986 Description: bind OST threads to NUMA nodes to improve performance
1987 Details : all OST threads are uniformly bound to CPUs on a single NUMA
1988 node and do their allocations there to localize memory access
1990 Severity : enhancement
1992 Description: llmount can determine client NID directly from Myrinet (GM)
1993 Details : the client NID code from gmnalnid was moved directly into
1994 llmount, removing the need to use this or specifying the
1995 client NID explicitly when mounting GM clients with zeroconf
1998 Frequency : if client is started with down MDS
2000 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2001 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2002 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2003 ptlrpc_import_delayed_req() because we didn't check if the
2004 request was interrupted, and we also didn't break out of the
2005 event loop if there was no timeout
2010 Description: data loss during non-page-aligned writes to a single file from
2011 both multiple nodes and multiple threads on one node at same time
2012 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2013 inconsistency led to false short-reads, that were cached and later
2014 used by ->prepare_write() to fill in partially written page,
2015 leading to data loss.
2018 Frequency : always, if lconf --abort_recovery used
2020 Description: lconf --abort_recovery fails with 'Operation not supported'
2021 Details : lconf was attempting to abort recovery on the MDT device and not
2024 Severity : enhancement
2026 Description: remove cleanup logs
2027 Details : replace lconf-generated cleanup logs with lustre internal
2028 cleanup routines. Eliminates the need for client-cleanup and
2031 Severity : enhancement
2033 Description: add support for EAs (user and system) on lustre filesystems
2034 Details : it is now possible to store extended attributes in the Lustre
2035 client filesystem, and with the user_xattr mount option it
2036 is possible to allow users to store EAs on their files also
2038 Severity : enhancement
2040 Description: Add possibility (config option) to show minimal available OST free
2042 Details : When compiled with --enable-mindf configure option, statfs(2)
2043 (and so, df) will return least minimal free space available from
2044 all OSTs as amount of free space on FS, instead of summary of
2045 free spaces of all OSTs.
2047 Severity : enhancement
2049 Description: do not expand extent locks acquired on OST-side
2050 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2051 by server: they are not cached anyway.
2054 Frequency : when mmap is used/binaries executed from Lustre
2056 Description: Unmmap pages before throwing them away from read cache.
2057 Details : llap_shrink cache now attempts to unmap pages before discarding
2058 them (if unmapping failed - do not discard). SLES9 kernel has
2059 extra checks that trigger if this unmapping is not done first.
2064 Description: lconf didn't resolve symlinks before checking to see whether a
2065 given mountpoint was already in use
2068 Frequency : when migrating failover services
2069 Bugzilla : 6395, 9514
2070 Description: When migrating a subset of services from a node (e.g. failback
2071 from a failover service node) the remaining services would
2072 time out and evict clients.
2073 Details : lconf --force (implied by --failover) sets the global obd_timeout
2074 to 5 seconds in order to quickly disconnect, but this caused
2075 other RPCs to time out too quickly. Do not change the global
2076 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2078 Severity : enhancement
2079 Frequency : if MDS is started with down OST
2080 Bugzilla : 9439,5706
2081 Description: Allow startup/shutdown of an MDS without depending on the
2082 availability of the OSTs.
2083 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2084 Add appropriate locking and lov-osc refcounts for safe
2085 cleaning. Add osc abort_inflight calls in case the
2086 synchronize never started.
2089 Frequency : occasional (Cray XT3 only)
2091 Description: root not authorized to access files in CRAY_PORTALS environment
2092 Details : The client process capabilities were not honoured on the MDS in
2093 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2094 been accessed by an authorized user then root was able to access
2095 the file on the local client also. The root user capabilities
2096 are now allowed on the MDS, as this environment has secure UID.
2099 Frequency : occasional
2101 Description: ldiskfs "too long searching" message happens too often
2102 Details : A debugging message (otherwise harmless) prints too often on
2103 the OST console. This has been reduced to only happen when
2104 there are fragmentation problems on the filesystem.
2109 Description: Division by zero in statfs when all OSCs are inactive
2110 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2111 lov_statfs divides by value returned by lov_get_stripecnt().
2115 Bugzilla : 9489, 3273
2116 Description: First write from each client to each OST was only 4kB in size,
2117 to initialize client writeback cache, which caused sub-optimal
2118 RPCs and poor layout on disk for the first writen file.
2119 Details : Clients now request an initial cache grant at (re)connect time
2120 and so that they can start streaming writes to the cache right
2121 away and always do full-sized RPCs if there is enough data.
2122 If the OST is rebooted the client also re-establishes its grant
2123 so that client cached writes will be honoured under the grant.
2128 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2129 Details : Now I/O RPCs go to different portal number and (presumably) fast
2130 lock requests (and glimses) and other RPCs get their own service
2131 threads pool that should be able to service those RPCs
2134 Severity : enhancement
2136 Description: Ability to exchange lustre version between client and servers and
2137 issue warnings at client side if client is too old. Also for
2138 liblustre clients there is ability to refuse connection of too old
2140 Details : New 'version' field is added to connect data structure that is
2141 filled with version info. That info is later checked by server and
2145 Frequency : rare, liblustre only.
2146 Bugzilla : 9296, 9581
2147 Description: Two simultaneous writes from liblustre at offset within same page
2148 might proceed at the same time overwriting eachother with stale
2150 Details : I/O lock withing llu_file_prwv was released too early, before data
2151 actually was hitting the wire. Extended lock-holding time until
2152 server acknowledges receiving data.
2155 Frequency : extremely rare. Never observed in practice.
2157 Description: avoid generating lustre_handle cookie of 0.
2158 Details : class_handle_hash() generates handle cookies by incrementing
2159 global counter, and can hit 0 occasionaly (this is unlikely, but
2160 not impossible, because initial value of cookie counter is
2161 selected randonly). Value of 0 is used as a sentinel meaning
2162 "unassigned handle" --- avoid it. Also coalesce two critical
2163 sections in this function into one.
2165 Severity : enhancement
2167 Description: allow liblustre clients to delegate truncate locking to OST
2168 Details : To avoid overhead of locking, liblustre client instructs OST to
2169 take extent lock in ost_punch() on client's behalf. New connection
2170 flag is added to handle backward compatibility.
2172 Severity : enhancement
2173 Bugzilla : 4928, 7341, 9758
2174 Description: allow number of OST service threads to be specified
2175 Details : a module parameter allows the number of OST service threads
2176 to be specified via "options ost ost_num_threads={N}" in the
2177 OSS's /etc/modules.conf or /etc/modprobe.conf.
2181 Bugzilla : 6146, 9635, 9895
2182 Description: servers crash with bad pointer in target_handle_connect()
2183 Details : In rare cases when a client is reconnecting it was possible that
2184 the connection request was the last reference for that export.
2185 We would temporarily drop the export reference and get a new
2186 one, but this may have been the last reference and the export
2187 was just destroyed. Get new reference before dropping old one.
2189 Severity : enhancement
2190 Frequency : if client is started with failover MDS
2192 Description: Allow multiple MDS hostnames in the mount command
2193 Details : Try to read the configuration from all specified MDS
2194 hostnames during a client mount in case the "primary"
2197 Severity : enhancement
2199 Description: Stop sending data to evicted clients as soon as possible.
2200 Details : Check if the client we are about to send or are sending data to
2201 was evicted already. (Check is done every second of waiting,
2202 for which l_wait_event interface was extended to allow checking
2203 of exit condition at specified intervals).
2206 Frequency : rare, normally only when NFS exporting is done from client
2208 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2210 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2213 Frequency : always, if srand() is called before liblustre initialization
2215 Description: Liblustre uses system PRNG disturbing its usage by user application
2216 Details : Introduce internal to lustre fast and high-quality PRNG for
2217 lustre usage and make liblustre and some other places in generic
2218 lustre code to use it.
2220 Severity : enhancement
2221 Bugzilla : 9477, 9557, 9870
2222 Description: Verify that the MDS configuration logs are updated when xml is
2223 Details : Check if the .xml configuration logs are newer than the config
2224 logs stored on the MDS and report an error if this is the case.
2225 Request --write-conf, or allow starting with --old_conf.
2227 Severity : enhancement
2229 Description: Handle symlinks in the path when checking if Lustre is mounted.
2230 Details : Resolve intermediate symlinks when checking if a client has
2231 mounted a filesystem to avoid duplicate client mounts.
2236 Description: lconf can hit an error exception but still return success.
2237 Details : The lconf command catches the Command error exception at the top
2238 level script context and will exit with the associated exit
2239 status, but doesn't ensure that this exit status is non-zero.
2244 Description: failure of ptlrpc thread startup can cause oops
2245 Details : Starting a ptlrpc service thread can fail if there are a large
2246 number of threads or the server memory is very fragmented.
2247 Handle this without oopsing.
2250 Frequency : always, only if liblustre and non-default acceptor port was used
2252 Description: liblustre cannot connect to servers with non-default acceptor port
2253 Details : tcpnal_set_default_params() was not called and was therefore
2254 ignoring the environment varaible TCPNAL_PORT, as well as other
2255 TCPNAL_ environment variables
2260 Description: two objects could be created on the same OST for a single file
2261 Details : If an OST is down, in some cases it was possible to create two
2262 objects on a single OST for a single file. No problems other
2263 than potential performance impact and spurious error messages.
2267 Bugzilla : 5681, 9562
2268 Description: Client may oops in ll_unhash_aliases
2269 Details : Client dcache may become inconsistent in race condition.
2270 In some cases "getcwd" can fail if the current directory is
2276 Description: Inode refcounting problems in NFS export code
2277 Details : link_raw functions used to call d_instantiate without obtaining
2278 extra inode reference first.
2282 Bugzilla : 9942, 9903
2283 Description: Referencing freed requests leading to crash, memleaks with NFS.
2284 Details : We used to require that call to ll_revalidate_it was always
2285 followed by ll_lookup_it. Also with revalidate_special() it is
2286 possible to call ll_revalidate_it() twice for the same dentry
2287 even if first occurence returned success. This fix changes semantic
2288 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2289 reference on a request referred from the intent.
2290 ll_intent_release() then releases such a request.
2293 Frequency : rare, normally benchmark loads only
2295 Description: unlinked inodes were kept in memory on the client
2296 Details : If a client is repeatedly creating and unlinking files it
2297 can accumulate a lot of stale inodes in the inode slab cache.
2298 If there is no other client load running this can cause the
2299 client node to run out of memory. Instead flush old inodes
2300 from client cache that have the same inode number as a new inode.
2303 Frequency : SLES9 2.6.5 kernel and long filenames only
2304 Bugzilla : 9969, 10379
2305 Description: utime reports stale NFS file handle
2306 Details : SLES9 uses out-of-dentry names in some cases, which confused
2307 the lustre dentry revalidation. Change it to always use the
2311 Frequency : rare, unless heavy write-truncate concurrency is continuous
2312 Bugzilla : 4180, 6984, 7171, 9963, 9331
2313 Description: OST becomes very slow and/or deadlocked during object unlink
2314 Details : filter_destroy() was holding onto the parent directory lock
2315 while truncating+unlinking objects. For very large objects this
2316 may block other threads for a long time and slow overall OST
2317 responsiveness. It may also be possible to get a lock ordering
2318 deadlock in this case, or run out of journal credits because of
2319 the combined truncate+unlink. Solution is to do object truncate
2320 first in one transaction without parent lock, and then do the
2321 final unlink in a new transaction with the parent lock. This
2322 reduces the lock hold time dramatically.
2325 Frequency : rare, 2.4 kernels only
2327 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2328 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2329 before unmounting their filesystem in order to avoid deadlock.
2330 The kernel_locked() function in 2.4 kernels only checks whether
2331 the kernel lock is held, not whether it is this process that is
2332 holding it as 2.6 kernels do.
2337 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2338 Details : The client ptlrpc code may be trying to reconnect to a down
2339 server before a previous connection attempt has timed out.
2340 Increase the reconnect interval to be longer than the connection
2341 timeout interval to avoid sending duplicate connections to
2345 Frequency : echo_client brw_test command
2347 Description: fix echo_client to work with OST preallocated code
2348 Details : OST preallocation code (5137) didn't take echo_client IO path
2349 into account: echo_client calls filter methods outside of any
2350 OST thread and, hence, there is no per-thread preallocated
2351 pages and buffers to use. Solution: hijack pga pages for IO. As
2352 a byproduct, this avoids unnecessary data copying.
2356 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2357 Description: Client can oops in mdc_commit_close() after open replay
2358 Details : It was possible for the MDS to return an open request with no
2359 transaction number in mds_finish_transno() if the client was
2360 evicted, but without actually returning an error. Clients
2361 would later try to replay that open and may trip an assertion
2362 Simplify the client close codepath, and always return an error
2363 from the MDS in case the open is not successful.
2366 Frequency : rare, 2.6 OSTs only
2368 Description: OST may deadlock under high load on fragmented files
2369 Details : If there was a heavy load and highly-fragmented OST filesystems
2370 it was possible to have all the OST threads deadlock waiting on
2371 allocation of biovecs, because the biovecs were not released
2372 until the entire RPC IO was completed. Instead, release biovecs
2373 as soon as they are complete to ensure forward IO progress.
2375 Severity : enhancement
2377 Description: Support for specifying external journal device at mount
2378 Details : If an OST or MDS device is formatted with an external journal
2379 device, this device major/minor is stored in the ext3 superblock
2380 and may not be valid for failover. Allow detecting and
2381 specifying the external journal at mount time.
2386 Description: Mounting an MDS with pending unlinked files may cause oops
2387 Details : target_finish_recovery() calls mds_postrecov() which returned
2388 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2389 considers this an error and immediately begins cleaning up the
2390 lov, just after starting the mds_lov process
2392 Severity : enhancement
2394 Description: Implement 'lfs df' to report actual free space on per-OST basis
2395 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2396 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2397 usage of objects; '-h' to report in human readable format.
2399 ------------------------------------------------------------------------------
2401 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2408 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2409 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2410 at startup during orphan unlinking. Instead, skip buddy update
2411 before setup, as it will be regenerated after recovery is complete.
2416 Description: If an OST is inactive, its locks might reference stale inodes.
2417 Details : lov_change_cbdata() must iterate over all namespaces, even if
2418 they are inactive to clear inode references from the lock.
2420 Severity : enhancement
2421 Frequency : occasional, if non-standard max_dirty_mb used
2423 Description: Client will block write RPCs if not enough grant
2424 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2425 then the client will block writes while waiting for another RPC
2426 to complete instead of consuming its dirty limit. With change
2427 we get improved performance when max_dirty_mb is small.
2429 Severity : enhancement
2430 Bugzilla : 3389, 6253
2431 Description: Add support for supplementary groups on the MDS.
2432 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2433 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2434 lookups for user supplementary groups into a cache.
2438 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2439 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2440 for liblustre O_EXCL is enforced by MDS.
2444 Description: readdir never returns NULL in liblustre.
2445 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2446 dentry in struct dirent.
2450 Frequency : liblustre only, depends on application IO pattern
2451 Description: liblustre clients evicted if not contacting servers
2452 Details : Don't put liblustre clients into the ping_evictor list, so
2453 they will not be evicted by the pinger ever.
2455 Severity : enhancement
2457 Description: Add ability to evict clients by NID from MDS.
2458 Details : By echoing "nid:$NID" string into
2459 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2460 $NID would be instantly evicted from this MDS and from all active
2461 OSTs connected to it.
2465 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2466 Details : lookup_it_finish() used to query file size from OSTs that was not
2471 Description: service threads change working directory to that of init
2472 Details : Starting lustre service threads may pin the working directory
2473 of the parent thread, making that filesystem busy. Threads
2474 now change to the working directory of init to avoid this.
2478 Frequency : during shutdown only
2479 Description: shutdown with a failed MDS or OST can cause unmount to hang
2480 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2485 Frequency : 2.6 only
2486 Description: chmod/chown may include an extra supplementary group
2487 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2488 supplementary group and if none is specified this is used.
2491 Bugzilla : 5479 (6816)
2493 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2494 Details : If lookup is in progress on a file that is unlinked we might try
2495 to revalidate the inode and fail in revalidate after lookup is
2496 complete and ll_file_open() enqueues the open again but
2497 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2500 Frequency : always, if lconf --abort_recovery used
2502 Description: lconf --abort_recovery fails with 'Operation not supported'
2503 Details : lconf was attempting to abort recovery on the MDT device and not
2506 ------------------------------------------------------------------------------
2508 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2513 Frequency : rare (only unsupported configurations with a node running as an
2515 Bugzilla : 6514, 5137
2516 Description: Mounting a Lustre file system on a node running as an OST could
2518 Details : OSTs now preallocates memory needed to write out data at
2519 startup, instead of when needed, to avoid having to
2520 allocate memory in possibly low memory situations.
2521 Specifically, if the file system is mounted on on OST,
2522 memory pressure could force it to try to write out data,
2523 which it needed to allocate memory to do. Due to the low
2524 memory, it would be unable to do so and the node would
2525 become unresponsive.
2527 Severity : enhancement
2529 Description: Addition of lconf --service command line option
2530 Details : lconf now accepts a '--service <arg>' option, which is
2531 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2533 Severity : enhancement
2535 Description: Failover mode is now the default for OSTs.
2536 Details : By default, OSTs will now run in failover mode. To return to
2537 the old behaviour, add '--failout' to the lmc line for OSTs.
2539 Severity : enhancement
2541 Description: Health checks are now provided for MDS and OSTs
2542 Details : Additional detailed health check information on MSD and OSTs
2543 is now provided through the procfs health_check value.
2546 Frequency : occasional, depends on IO load
2548 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2549 after numerous create/delete cycles
2550 Details : The ext3 inode allocation policy would not allocate new inodes
2551 very well on the OSTs because there are no new directories
2552 being created. Instead we look for groups with free space if
2553 the parent directories are nearly full.
2558 Description: Network or server problems during mount may cause partially
2559 mounted clients instead of returning an error.
2560 Details : The config llog parsing code may overwrite the error return
2561 code during mount error handling, returning success instead
2567 Description: MDS can fail to allocate large reply buffers
2568 Details : After long uptimes the MDS can fail to allocate large reply
2569 buffers (e.g. zconf client mount config records) due to memory
2570 fragmentation or consumption by the buffer cache. Preallocate
2571 some large reply buffers so that these replies can be sent even
2572 under memory pressure.
2576 Frequency : rare (liblustre)
2577 Description: fsx running with liblustre complained that using truncate() to
2578 extend the file doesn't work. This patch corrects that issue.
2579 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2580 ATTR_SIZE and lsm use in llu_setattr_raw.
2584 Frequency : rare, only 2.6 kernels
2585 Description: Unusual file access patterns on the MDS may result in inode
2586 data being lost in very rare circumstances.
2587 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2588 code in the 2.6 kernel caused the inode and/or EA data not to
2589 be read from disk, causing single-file corruption.
2593 Frequency : rare, only 2.6 filesystems using extents
2594 Description: Heavy concurrent write and delete load may cause data corruption.
2595 Details : It was possible under high-load situations to have an extent
2596 metadata block in the block device cache from a just-unlinked
2597 file overwrite a newly-allocated data block. We now unmap any
2598 metadata buffers that alias just-allocated data blocks.
2602 Frequency : filesystems with default stripe_count larger than 77
2603 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2604 Details : lconf specifies an inode size of 4096 bytes when the default
2605 stripe_count is larger than 77. This conflicts with the default
2606 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2607 this case to avoid pinning too much memory for large EAs.
2609 ------------------------------------------------------------------------------
2611 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2616 Frequency : rare (extremely heavy IO load with hundreds of clients)
2618 Description: Client is evicted, gets IO error writing to file
2619 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2620 caused clients to be evicted for AST timeouts. The fixes in
2621 bug 5192 mean we no longer need to have such short AST timeouts
2622 so ldlm_timeout has been increased.
2625 Frequency : occasional during --force or --failover shutdown under load
2626 Bugzilla : 5949, 4834
2627 Description: Server oops/LBUG if stopped with --force or --failover under load
2628 Details : a collection of import/export refcount and cleanup ordering
2629 issues fixed for safer force cleanup
2632 Frequency : only filesystems larger than 120 OSTs
2633 Bugzilla : 5990, 6223
2634 Description: lfs getstripe would oops on a very large filesystem
2635 Details : lov_getconfig used kfree on vmalloc'd memory
2638 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2639 Bugzilla : 6242, 6243
2640 Description: reading from files that had been truncated to a non-zero size
2641 but never opened returned no data
2642 Details : ll_file_read() reads zeros from no-object files to EOF
2647 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2648 Details : There's little harm in aborting MDS/OSS recovery and letting it
2649 try again, so I removed the LASSERT and return an error instead.
2651 Severity : enhancement
2653 Description: New debugging infrastructure for tracking down data corruption
2654 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2655 (b) cover more of the client-side code path, and (c) try to narrow
2656 down where problems occurred
2660 Bugzilla : 3819, 4364, 4397, 6313
2661 Description: Racing close and eviction MDS could cause assertion in mds_close
2662 Details : It was possible to get multiple mfd references during close and
2663 client eviction, leading to one thread referencing a freed mfd.
2665 Severity: : enhancement
2666 Bugzilla : 3262, 6359
2667 Description: Attempts to reconnect to servers are now more aggressive.
2668 Details : This builds on the enhanced upcall-less recovery that was added
2669 in 1.4.2. When trying to reconnect to servers, clients will
2670 now try each server in the failover group every 10 seconds. By
2671 default, clients would previously try one server every 25 seconds.
2676 Description: After recovery, certain operations trigger a failed
2677 assertion on a client.
2678 Details : Failing over an mds, using lconf -d --failover, while a
2679 client was doing a readdir() call would cause the client to
2680 LBUG after recovery completed and the readdir() was resent.
2682 Severity : enhancement
2684 Description: Default groups are now added by lconf
2685 Details : You can now run lconf --group <servicename> without having to
2686 manually add groups with lmc.
2689 Frequency : occasional
2691 Description: Nodes with an elan id of 0 trigger a failed assertion
2694 Frequency : always when accessing e.g. tty/console device nodes
2696 Description: tty and some other devices nodes cannot be used on lustre
2697 Details : file's private_data field is used by device data and lustre
2698 values in there got lost. New field was added to struct file to
2699 store fs-specific private data.
2702 Frequency : when exporting Lustre via NFS
2704 Description: NFSD failed occasionally when looking up a path component
2705 Details : NFSD is looking up ".." which was broken in ext3 directories
2706 that had grown large enough to become hashed.
2709 Frequency : Clusters with multiple interfaces not on the same subnet
2711 Description: Nodes will repeatedly try to reconnect to an interface which it
2712 cannot reach and report an error to the log.
2713 Details : Extra peer list entries will be created by lconf with some peers
2714 unreachable. lconf now validates the peer before adding it.
2717 Frequency : Only if a default stripe is set on the filesystem root.
2719 Description: Setting a default stripe on the filesystem root prevented the
2720 filesystem from being remounted.
2721 Details : The client was sending extra request flags in the root getattr
2722 request and did not allocate a reply buffer for the dir EA.
2725 Frequency : occasional, higher if lots of files are accessed by one client
2726 Bugzilla : 6159, 6097
2727 Description: Client trips assertion regarding lsm mismatch/magic
2728 Details : While revalidating inodes the VFS looks up inodes with ifind()
2729 and in rare cases can find an inode that is being freed.
2730 The ll_test_inode() code will free the lsm during ifind()
2731 when it finds an existing inode and then the VFS later attaches
2732 this free lsm to a new inode.
2736 Bugzilla : 6422, 7030
2737 Description: MDS deadlock between mkdir and client eviction
2738 Details : Creating a new file via mkdir or mknod (starting a transaction
2739 and getting the ns lock) can deadlock with client eviction
2740 (gets ns lock and trying to finish a synchronous transaction).
2743 Frequency : occasional
2744 Description: While starting a server, the fsfilt_ext3 module could not be
2746 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2747 kernels. Previously, lconf would still use the ext3 name
2748 when trying to load modules. Now, it will correctly use
2749 ext3 on 2.4 and ldiskfs on 2.6.
2751 Severity : enhancement
2752 Description: The default stripe count has been changed to 1
2753 Details : The interpretation of the default stripe count (0, to lfs
2754 or lmc) has been changed to mean striping across a single
2755 OST, rather than all available. For general usage we have
2756 found a stripe count of 1 or 2 works best.
2758 Severity : enhancement
2759 Description: Add support for compiling against Cray portals.
2760 Details : Conditional compiling for some areas that are different
2764 Frequency : occasional
2765 Bugzilla : 6409, 6834
2766 Description: Creating files with an explicit stripe count may lead to
2767 a failed assertion on the MDS
2768 Details : If some OSTs are full or unavailable, creating files may
2769 trigger a failed assertion on the MDS. Now, Lustre will
2770 try to use other servers or return an error to the
2774 Frequency : occasional
2776 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2777 caused lock timeout during readahead (since 1.4.2).
2778 Details : Processes doing readahead might match a lock that hasn't been
2779 granted yet if there are overlapping and conflicting lock
2780 requests. The readahead process waits on ungranted lock
2781 (original lock is CBPENDING), while OST waits for that process
2782 to cancel CBPENDING read lock and eventually evicts client.
2784 Severity : enhancement
2786 Description: Initial enabling of flock support for clients
2787 Details : Implements fcntl advisory locking and file status functions.
2788 This feature is provided as an optional mount flag (default
2789 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2790 locking are implemented yet, and those that are are not guaranteed
2791 to be completely correct in production environments.
2792 mount -t lustre -o [flock|noflock] ...
2795 Frequency : occasional
2797 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2798 assertion in the ext3 JBD (journaling) layer.
2799 Details : The b_committed_data struct is protected by the big kernel lock
2800 in 2.4 kernels, serializing journal_commit_transaction() and
2801 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2802 access to this struct. In 2.6 kernels there is finer grained
2803 locking to improve SMP performance of the JBD layer.
2807 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2809 -----------------------------------------------------------------------------
2811 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2813 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2814 versions of Lustre. Please update all servers and clients to
2815 version 1.4.2 or later at the same time. You must also run
2816 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2817 to update the configuration logs.
2819 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2820 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2821 - swab llog records of type '0' so we get proper header size/idx (5861)
2822 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2823 - fix rename of one directory over another leaking an inode (5953)
2824 - avoid SetPageDirty on 2.6 (5981)
2825 - don't re-add just-being-destroyed locks to the waiting list (5653)
2826 - when creating new directories, inherit the parent's custom
2827 striping settings if present parent (3048)
2828 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2829 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2830 - don't hold client locks on temporary worklist from l_lru (5666)
2831 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2832 - automatically evict dead clients (5921)
2833 - Update file size properly in create+truncate+fstat case (6196)
2834 - Do not unhash mountpoint dentries, do not allow removal of
2836 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2837 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2838 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2839 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2840 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2841 - Correctly handle reads of files with no objects (6243)
2842 - lctl recover will also mark a device active if deactivate used (5933)
2844 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2845 - allow --write-conf on an MDS with different nettype than client (5619)
2846 - don't write config llogs to MDS for mounts not from that MDS (5617)
2847 - lconf should create multiple TCP connections from a client (5201)
2848 - init scripts are now turned off by default; run chkconfig --on
2849 lustre and chkconfig --on lustrefs to use them
2850 - upcalls are no longer needed for clients to recover to failover
2852 - add --abort-recovery option to lconf to abort recovery on device
2854 - add support for an arbitrary number of OSTs (3026)
2855 - Quota support protocol changes.
2856 - forward compatibility changes to wire structs (6007)
2857 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2858 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2859 - improved "lustre status" script
2860 - initialize blocksize for non-regular files (6062)
2861 - added --disable-server and --disable-client configure options (5782)
2862 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2863 - Vanilla 2.4.29 support
2864 - increase maximum number of obd devices to 520 (6242)
2865 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2866 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2867 SLES 9 2.6.5 kernels
2868 - limit stripes per file to 160 (the maximum EA size) (6093)
2870 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2873 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2874 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2875 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2876 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2877 - small ext3 extents cleanups and fixes (5733)
2878 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2879 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2880 - avoid needless client->OST connect, fix handle mismatch (5317)
2881 - fix DLM error path that led to out-of-sync client, long delays (5779)
2882 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2883 - fix several locking issues related to i_size (5492,5624,5654,5672)
2884 - don't move pending lock onto export if it is already evicted (5683)
2885 - fix kernel oops when creating .foo in unlinked directory (5548)
2886 - fix deadlock in obdfilter statistics vs. object create (5811)
2887 - use time_{before,after} to avoid timer jiffies wrap (5882)
2888 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2889 - Do not leak request if server was not able to process it (5154)
2890 - If mds_open unable to find parent dir, make that negative lookup(5154)
2891 - don't create new directories with extent-mapping (5909, 5936)
2893 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2894 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2895 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2896 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2897 - support for RHEL4 kernel on i686 (5773)
2898 - provide error messages when incompatible logs are encountered (5898)
2900 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2901 * version 1.4.0.10 (1.4.1 release candidate 1)
2903 - don't keep a lock reference when lock is not granted (4238)
2904 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2905 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2906 - drop import inflight refcount on signal_completed_replay error (5255)
2907 - unlock page after async write error during send (3677)
2908 - handle missing objects in filter_preprw_read properly (5265)
2909 - no transno return for symlink open, don't save no-trasno open (3440)
2910 - don't try to complete elan receive that already failed (4012)
2911 - free RPC server reply state on error (5406)
2912 - clean up thread from ptlrpc_start_thread() on error (5160)
2913 - readahead could read extra page into cache that wasn't ejected (5388)
2914 - prevent races in class_attach/setup/cleanup/detach (5260)
2915 - don't dereference de->d_inode after l_dput of de (5458)
2916 - use "int" for stripe value returned from lock_to_stripe (5544)
2917 - mballoc allocation and error-checking fixes in 2.6 (5504)
2918 - block device patches to fix I/O request sizes in 2.6 (5482)
2919 - look up hostnames for IB nals (5602)
2920 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2921 - don't start multiple acceptors for the same port (5277)
2922 - fix incorrect LASSERT in mds_getattr_name (5635)
2923 - export a proc file for general "ping" checking (5628)
2924 - fix "lfs check" to not block when the MDS is down (5628)
2926 - service request history (4965)
2927 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2928 - create an "evict_client" /proc entry on OSTs, like the MDS has
2929 - fix mount usage message, return errors per mount(8) (5168)
2930 - change grep [] to grep "[]" in tests so they work in more UMLs
2931 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2932 - remove ancient LOV_MAGIC_V0 EA support (5047)
2933 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2934 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2935 - allow readahead allocations to fail when low on memory (5383)
2936 - mmap locking landed again, after considerable improvement (2828)
2937 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2939 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2942 - send OST transaction number in read/write reply to free req (4966)
2943 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2944 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2945 - the watchdog thread now runs as interruptible (5246)
2946 - flock/lockf fixes (but it's still disabled, pending 5135)
2947 - don't use EXT3 constants in llite code (5094)
2948 - memory shortage at startup could cause assertion (5176)
2950 - reorganization of lov code
2951 - single portals codebase
2953 - add extents/mballoc support (5025)
2954 - direct I/O reads in the obdfilter (4048)
2955 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2957 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2960 - send OST transaction number in read/write reply to free req (4966)
2961 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2962 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2963 - the watchdog thread now runs as interruptible (5246)
2964 - handle missing objects in filter_preprw_read properly (5265)
2965 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2966 - drop import inflight refcount on signal_completed_replay error (5255)
2967 - unlock page after async write error during send (3677)
2968 - return original error code on reconstructed replies (3761)
2969 - no transno return for symlink open, don't save no-trasno open (3440)
2971 - add pid to ldlm debugging output (4922)
2972 - bump the watchdog timeouts -- we can't handle 30sec yet
2973 - extra debugging for orphan dentry/inode bug (5259)
2975 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
2978 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
2979 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
2980 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
2981 - fix NULL dereference in /proc/sys/portals/routes (4827)
2982 - allow failed mdc_close() operations to be interrupted (4561)
2983 - stop precreate on OST before MDS would time out on it (4778)
2984 - don't send partial-page writes before EOF from client (4410)
2985 - discard client grant for sub-page writes on large-page clients (4520)
2986 - don't free dentries not owned by NFS code, check generation (4806)
2987 - fix lsm leak if mds_create_objects() fails (4801)
2988 - limit debug_daemon file size, always print CERROR messages (4789)
2989 - use transno after validating reply (3892)
2990 - process timed out requests if import state changes (3754)
2991 - update mtime on OST during writes, return in glimpse (4829)
2992 - add mkfsoptions to LDAP (4679)
2993 - use ->max_readahead method instead of zapping global ra (5039)
2994 - don't interrupt __l_wait_event() during strace
2996 - add software watchdogs to catch hung threads quickly (4941)
2997 - make lustrefs init script start after nfs is mounted
2998 - fix CWARN/ERROR duplication (4930)
2999 - return async write errors to application if possible (2248)
3000 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3001 - print NAL number in %x format (4645)
3002 - update barely-supported suse-2.4.21-171 series (4842)
3003 - support for sles 9 %post scripts
3004 - support for building 2.6 kernel-source packages
3005 - support for sles km_* packages
3007 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3010 - ignore -ENOENT errors in osc_destroy (3639)
3011 - notify osc create thread that OSC is being cleaned up (4600)
3012 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3013 - reconstruct ost_handle() like mds_handle() (4657)
3014 - create a new thread to do import eviction to avoid deadlock (3969)
3015 - let lconf resolve symlinked-to devices (4629)
3016 - don't unlink "objects" from directory with default EA (4554)
3017 - hold socknal file ref over connect in case target is down (4394)
3018 - allow more than 32000 subdirectories in a single directory (3244)
3019 - fix blocks count for O_DIRECT writes (3751)
3020 - OST returns ENOSPC from object create when no space left (4539)
3021 - don't send truncate RPC if file size isn't changing (4410)
3022 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3023 - bind to privileged port in socknal and tcpnal (3689)
3025 - rate limit CERROR/CWARN console message to avoid overload (4519)
3026 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3027 - basic mmap support (3918)
3028 - kernel patch series update from b1_4 (4711)
3030 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3033 - avoid crash during MDS cleanup with OST shut down (2775)
3034 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3035 - don't use bad inodes on the MDS (3744)
3036 - dynamic object preallocation to improve recovery speed (4236)
3037 - don't hold spinlock over lock dumping or change debug flags (4401)
3038 - don't zero obd_dev when it is force cleaned (3651)
3039 - print grants to console if they go negative (4431)
3040 - "lctl deactivate" will stop automatic recovery attempts (3406)
3041 - look for existing locks in ldlm_handle_enqueue() (3764)
3042 - don't resolve lock handle twice in recovery avoiding race (4401)
3043 - revalidate should check working dir is a directory (4134)
3045 - don't always mark "slow" obdfilter messages as errors (4418)
3047 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3050 - don't close LustreDB during write_conf until it is done (3860)
3051 - fix typo in lconf for_each_profile (3821)
3052 - allow dumping logs from multiple threads at one time (3820)
3053 - don't allow multiple threads in OSC recovery (3812)
3054 - fix debug_size parameters (3864)
3055 - fix mds_postrecov to initialize import for llog ctxt (3121)
3056 - replace config semaphore with spinlock (3306)
3057 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3058 - don't allow enqueue to complete on a destroyed export (3822)
3059 - down write_lock before checking llog header bitmap (3825)
3060 - recover from lock replay timeout (3764)
3061 - up llog sem before sending rpc (3652)
3062 - reduce ns lock hold times when setting kms (3267)
3063 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3064 - fix NULL deref and obd_dev leak on setup error (3312)
3065 - replace some LBUG about llog ops with error handling (3841)
3066 - don't match INVALID dentries from d_lookup and spin (3784)
3067 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3068 - fix invalid assertion in ptlrpc_set_wait (3880)
3070 - add libwrap support for the TCP acceptor (3996)
3071 - add /proc/sys/portals/routes for non-root route listing (3994)
3072 - allow setting MDS UUID in .xml (2580)
3073 - print the stack of a process that LBUGs (4228)
3075 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3078 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3079 - make sure to unset replay flag from failed open requests (3440)
3080 - if default stripe count is 0, use OST count for inode size (3636)
3081 - update parent mtime/ctime on client for create/unlink (2611)
3082 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3083 - free recovery state on server during a forced cleanup (3571)
3084 - unregister_reply for resent reqs (3063)
3085 - loop back devices mounting and status check on 2.6 (3563)
3086 - fix resource-creation race that can provoke i_size == 0 (3513)
3087 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3088 - more debugging for page-accounting assertion (3746)
3089 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3090 - avoid deadlock after precreation failure (3758)
3091 - fix race and lock order deadlock in orphan handling (3450, 3750)
3092 - add validity checks when grabbing inodes from l_ast_data (3599)
3094 - add /proc/.../recovery_status to obdfilter (3428)
3095 - lightweight CDEBUG infrastructure, debug daemon (3668)
3096 - change default OSC RPC parameters to be better on small clusters
3097 - turn off OST read cache for files smaller than 32MB
3098 - install man pages and include them in rpms (3100)
3099 - add new init script for (un)mounting lustre filesystems (2593)
3100 - run chkconfig in %post for init scripts (3701)
3101 - drop scimac NAL (unmaintained)
3103 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3106 - clean kiobufs before and after use (3485)
3107 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3108 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3109 - change init script to fail more clearly if not run as root (1528)
3110 - allow clients to reconnect during replay (1742)
3111 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3112 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3113 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3114 - fix cond_resched (3554)
3115 - only evict unfinished clients after recovery (3515)
3116 - allow bulk resend, prevent data loss (3570)
3117 - dynamic ptlrpc request buffer allocation (2102)
3118 - don't allow unlinking open directory if it isn't empty (2904)
3119 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3120 - remove extraneous obd dereference causing LASSERT failure (3334)
3121 - don't use get_cycles() when creating temp. files on the mds (3156)
3122 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3123 - handle EEXIST for set-stripe, set proper directory name (3336)
3125 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3126 - fix ksocknal_fmb_callback() error messages (2918)
3128 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3131 - don't copy lvb into (possibly NULL) reply on error (2983)
3132 - don't deref dentry after dput, don't free lvb on error (2922)
3133 - use the kms to determine writeback rpc length (2947)
3134 - increment oti_logcookies when osc is inactive (2948)
3135 - update client's i_blocks count via lvb messages (2543)
3136 - handle intent open/close of special files properly (1557)
3137 - mount MDS with errors=remount-ro, like obdfilter (2009)
3138 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3139 - don't use cancelling-locks' kms values (2947)
3140 - use highest lock extent for kms, not last one (2925)
3141 - don't dereference ERR_PTR() dentry in error handling path (3107)
3142 - fix thread race in portals_debug_dumplog() (3122)
3143 - create lprocfs device entries at setup instead of at attach (1519)
3144 - common AST error handler, don't evict client on completion race (3145)
3145 - zero nameidata in detach_mnt in 2.6 (3118)
3146 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3147 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3148 - initialize RPC timeout timer earlier for 2.6 (3219)
3149 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3150 - print nal/nid for unknown nid (3258)
3151 - additional checks for oscc recovery before doing precreate (3284)
3152 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3153 - don't crash in mdc_close for bad permissions on open (3285)
3154 - zero i_rdev for non-device files (3147)
3155 - clear page->private before handing to FS, better assertion (3119)
3156 - tune the read pipeline (3236)
3157 - fix incorrect decref of invalidated dentry (2350)
3158 - provide read-ahead stats and refine rpc in flight stats (3328)
3159 - don't hold journal transaction open across create RPC (3313)
3160 - update atime on MDS at close time (3265)
3161 - close LDAP connection when recovering to avoid server load (3315)
3162 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3163 - don't leak open file on MDS after open resend (3325)
3164 - serialize filter_precreate and filter_destroy_precreated (3329)
3165 - loop device shouldn't call sync_dev() for nul device (3092)
3166 - clear page cache after eviction (2766)
3167 - resynchronize MDS->OST in background (2824)
3168 - refuse to mount the same filesystem twice on same mountpoint (3394)
3169 - allow llmount to create routes for mounting behind routers (3320)
3170 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3171 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3172 - fix rare problem with rename on htree directories (3417)
3174 - allow default OST striping configuration per directory (1414)
3175 - fix compilation for qswnal for 2.6 kernels (3125)
3176 - increase maximum number of MDS request buffers for large systems
3177 - change liblustreapi to be useful for external progs like lfsck (3098)
3178 - increase local configuration timeout for slow disks (3353)
3179 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3181 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3184 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3185 - don't overwrite extent policy data in reply if lock was blocked (2901)
3186 - drop filter export grants atomically with removal from device (2663)
3187 - del obd_self_export from work_list in class_disconnect_exports (2908)
3188 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3189 - swab reply message in mdc_close, other PPC fixes (2464)
3190 - fix destroying of named logs (2325)
3191 - overwrite old logs when running lconf --write_conf (2264)
3192 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3193 - fix race in target_handle_connect (2898)
3194 - mds_reint_create() should take same inode create lock (2926)
3195 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3196 - don't close files for self_export to avoid uninitialized obd (2936)
3197 - allow MDS with the same name as client node (2939)
3198 - hold dentry reference for closed log files for unlink (2325)
3199 - reserve space for all logs during transactions (2059)
3200 - don't evict page beyond end of stripe extent (2925)
3201 - don't oops on a deleted current working directory (2399)
3202 - handle hard links to targets without a parent properly (2517)
3203 - don't dereference NULL lock when racing during eviction (2867)
3204 - don't grow lock extents when lots of conflicting locks (2919)
3206 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3209 - account for cache space usage on clients to avoid data loss (974)
3210 - lfsck support in lustre kernel code (2349)
3211 - reduce journal credits needed for BRW writes (2370)
3212 - orphan handling to avoid losing space on client/server crashes
3213 - ptlrpcd can be blocked, stopping ALL progress (2477)
3214 - use lock value blocks to assist in proper KMS, faster stat (1021)
3215 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3216 - recovery for initial connections (2355)
3217 - fixes for mds_cleanup_orphans (1934)
3218 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3219 - block all file creations until orphan recovery completes (1901)
3220 - client remove rq_connection from request struct (2423)
3221 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3222 - recovery timer race (2670)
3223 - mdc_close recovey bug (2532)
3224 - ptlrpc cleanup bug (2710)
3225 - mds timeout on local locks (2588)
3226 - namespace lock held during RPCs (2431)
3227 - handle interrupted sync write properly (2503)
3228 - don't try to handle a message that hasn't been replied to (2699)
3229 - client assert failure during cleanup after abort recovery (2701)
3230 - leak mdc device after failed mount (2712)
3231 - ptlrpc_check_set allows timedout requests to complete (2714)
3232 - wait for inflight reqs when ptlrpcd finishes (2710)
3233 - make sure unregistered services are removed from the srv_list
3234 - reset bulk XID's when resending them (caught by 1138 test)
3235 - unregister_bulk after timeout
3236 - fix lconf error (2694)
3237 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3238 - readahead locks pages, leaves pending causing memory pressure (2673)
3239 - increase OST request buffers to 4096 on large machines (2729)
3240 - fix up permission of existing directories in simple_mkdir (2661)
3241 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3242 - don't assign transno to errored transactions (2742)
3243 - don't delete objects on OST if given a bogus objid from MDS (2751)
3244 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3245 - if rq_no_resend, then timeout request after recovery (2432)
3246 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3247 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3248 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3249 - MDS recovery shouldn't delete objects that it has given out (2730)
3250 - if enqueue arrives after completion, don't clobber LVB (2819)
3251 - don't unlock pages twice when trigger_group_io returns error (2814)
3252 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3253 - don't write pages to disk if there was an error (1450)
3254 - don't ping imports that have recovery disabled (2676)
3255 - take buffered bytes into account when balancing socknal conn (2817)
3256 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3257 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3258 - remove little-endian swabbing of llog records (1987)
3259 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3260 - retry reposting request buffers if they fail (1191)
3261 - grow extent at grant time to avoid granting a revoked lock (2809)
3262 - lock revoke doesn't evict page if covered by a second lock (2765)
3263 - disable VM readahead to avoid reading outside lock extents (2805)
3265 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3266 - updated kernel patches for hp-2.4.20 kernel (2681)
3268 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3271 - fix truncated write corruption (2366)
3272 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3274 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3275 - ensure len(uuid) < 37 in lmc (1171)
3276 - fix ia64 OOPS in llog_test (2255)
3277 - zero end of page at obdfilter for partial page writes (2648)
3278 - don't leave stale dentries around after renames (bug 2428)
3279 - fix timeouts when evicting a client with a single lock held (2642)
3280 - set deadline for the initial HELLO message to drain (2634)
3281 - print out dotted-quad IP addresses in the socknal (2302)
3283 - additional debugging for MDS client eviction problem (2443)
3284 - fix mkfsoptions support for osts (2603, 2604)
3286 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3289 - add series for the vanilla 2.6.0 kernel
3290 - add series for the vanilla 2.4.24 kernel
3291 - add series for a cray x86/64 UL kernel drop
3292 - fix xattr patches for the vanilla 2.4.19 series
3294 - generate true UUIDs in lmc (1171)
3295 - have portals stack dumping break in UML (2466)
3296 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3297 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3298 - wake callback waiting threads on client eviction (2460)
3299 - Add --ptldebug and --subsystem to lmc (1719)
3300 - update assertion to allow safe interrupt allocation
3301 - set rq_no_resend for cancel requests (2432)
3302 - recalculate ptlrpcd timeout after resend (2494)
3303 - call vfs_rmdir when removing pending directories (2368)
3304 - fix renaming a file to itself (2429)
3305 - lmc creates a default one-stripe lov (2454)
3306 - expand procfs space to handle large clusters (2326)
3307 - increase UML stack to avoid overflow
3308 - update lconf's list of debug and subsystem masks
3309 - fix lfs find --obd (2510)
3310 - /proc tunable for disabling filter read caching (2591)
3311 - stop rpm packages from altering slapd.conf (2301)
3312 - disable nagle in the socknal under 0conf (2578)
3313 - choose mds inode size based on stripe count (2572)
3314 - fix kernel-source rpm problems (2516)
3316 - add --disable-doc to avoid pdf generation (2421)
3317 - update documentation, tests, type-os, comments
3318 - avoid format warnings on ia64
3319 - remove the TOE NAL
3320 - tiny code cleanups by removing unused fields
3322 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3325 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3326 - remove the most blatant lies from BUILDING file (2371)
3327 - change default debug level to reasonable production setting
3328 - reduce client side cache size to reduce cache flush time
3329 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3330 - make TCP zerocopy and pinger support enabled by default (2476)
3331 - sync writes completed after process exits caused crashes (2319)
3332 - maintain correct mount count on the MDS (2356)
3333 - backout 1557, because 2316 wasn't really fixed
3334 - better file I/O statistics gathering in /proc
3335 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3336 - another kernel patch to fix zero-copy TCP function export
3337 - don't take duplicate lock when processing re-sent getattr (2420)
3338 - lctl uses obd_self_export instead of creating new conn (2353)
3339 - MDS/OST recovery case which requires object creation asserted (2425)
3340 - move lfs from /usr/sbin to /usr/bin in packages
3341 - fix race between mds_client_add and mds_client_free (2417)
3342 - use kmalloc instead of slabs in portals (2430)
3343 - don't create duplicate records when a failover MDS is present (2442)
3344 - remove unnecessary mount age check (2332)
3345 - don't remove directory inodes from locks prematurely (2451)
3346 - don't break if MDS service name is the same as hostname (2103)
3347 - fix races in client write RPC generation when cache full (2482)
3349 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3352 - remove now-unused request->rq_obd (278)
3353 - if an allocation fails, print out how much memory we've used (1933)
3354 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3355 - add the "configurable stack size" patch to most series files (1256)
3356 - ability to write large log records, for 100+ OST configs (2306)
3357 - fix NULL deref when filter_prep fails (2314)
3358 - fix operator precedence error in filter_sync
3359 - dynamic allocation of socknal TX descriptors (2315)
3360 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3361 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3362 - ensure that CWARN messages in Portals always get to the syslog
3363 - __init/__exit are not for prototype decls (ldlm_init/exit)
3364 - x86-64 compile warning fixes
3365 - fix gateway LMC keyword conflict (2318)
3366 - fix MDS lock inversions in getattr/reint paths (1844)
3367 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3368 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3369 - DLM race condition prevented some lock evictions (2328)
3370 - ENOMEM detection and retry on socknal sends (2230)
3371 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3372 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3373 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3374 - better error messages when a client is rejected during recovery (1505)
3375 - avoid cancelling locks which were never granted, after failure (2330)
3376 - fix i_sem/journal inversion in mds_client_add (2333)
3377 - fix truncate/getattr lock cycle deadlock (2334)
3378 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3379 - fix two rare exit paths which could leak an l_lock() ref (2321)
3380 - fencepost error in MDS/OST orphan recovery (2226)
3381 - make log record alignment 8 bytes (1988)
3382 - lstripe now fails when requested offset > ost_count (2237)
3383 - ensure that all kernel series have a complete list.h (1607)
3384 - fix crashes in special-file operations (2316)
3385 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3387 - allow configurable automake binary, for testing new versions
3388 - small update to the lfs documentation
3390 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3392 * fix negative export reference count in fsfilt_sync (2312)
3394 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3395 * release candidate 0.9.1
3397 - orphans are moved into the PENDING directory for possible recovery
3398 - replayed opens now open by fid for orphan/rename safety (1042)
3399 - last close of an orphan inode generates a transno (683)
3400 - chdir() and mount() now pin the directory entry (1020)
3401 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3402 - discard very old requests without processing them (1502)
3403 - remove some common, well-understood CERRORs (1505)
3404 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3405 - clear "grant" flags in OST replies until OST grant code lands (1644)
3406 - fix read performance by not clobbering i_blksize on client (1598)
3407 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3408 - make LRU size a /proc tunable, clears locks when reduced (707)
3409 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3410 - server locks take references on exports now (1558)
3411 - build fixes for 2.4.20-rh trees (1663)
3412 - return an error from lov_create if all OSCs are inactive (1751)
3413 - fix import levels when a reconnect happens without a timeout (1597)
3414 - exit early from mds_open if we get a lookup error (1749)
3415 - partial page read at EOF wouldn't wait for disk before sending (1642)
3416 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3417 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3418 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3419 - write/truncate lock inversion (1639)
3420 - Don't auto-load obdclass, portals modules during cleanup (1495)
3421 - fix timestamps from jumping to "now" (1763)
3422 - extra journal assertions (1648)
3423 - add an extra multiunlink test (1771)
3424 - fix read_record/write_record API (1776)
3425 - fix leak of offset_extent, possible incorrect i_size later (1772)
3426 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3427 - Debugging for the kqswnal_get_idle_tx problems (1820)
3428 - Allow recovery to be attempted multiple times (1536)
3429 - Write out MDS last_rcvd file after it is first created (1600)
3430 - Fix tx_descriptor leak in failed transmit situations (1827)
3431 - ext3 journaling fixes for assertion failure after IO error (1871)
3432 - class_export_put() on freed export after completion AST error (1896)
3433 - Fix revalidate looping in VFS (1322)
3434 - Don't access a freed export during MDS_REINT timeout (1521)
3435 - Add open-unlink recovery support on the MDS (1673,1764)
3436 - Return an error if no MDS data was read from last_rcvd (1946)
3437 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3438 - Don't setup a disk device that doesn't match exported UUID (317)
3439 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3440 - avoid committing NULL handle in force close
3441 - local.sh is now a one-stripe LOV configuration
3442 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3443 - fix ext3 htree duplicate directory entry corruption (1516)
3444 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3445 - update to new LOV EA format (2097)
3446 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3448 2003-06-15 Phil Schwan <phil@clusterfs.com>
3451 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3452 - per-import/export recovery handling (958, 931, 959)
3453 - multiple last-rcvd slots, for serving multiple FSes (949)
3454 - connections are again shared between multiple imp/exports (963, 964)
3455 - "umount -f" would hang if any requests needed to be sent (393, 978)
3456 - avoid pinning large req buffer by copying for queued messages (989)
3457 - add "uuid" to "lctl device" command to help upcalls (991)
3458 - "open" RPCs with transnos would confuse recovery counters (1037)
3459 - do proper endian conversion of all wire messages (288, 340, 891)
3460 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3461 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3462 - fix ptlrpc_connection leak in target_handle_connect (1174)
3463 - fix import refcounting bug in OST and MDS cleanup (1134)
3464 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3465 - fix very unlikely obd_types race condition (501)
3466 - remove osc_open hack for echo_client (1187)
3467 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3468 - a failure in read_inode2 leads to deadlock (1139)
3469 - cancel ack-locks as soon as transaction is committed (1072)
3470 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3471 - make sure to commitrw after any preprw to avoid deadlock (1162)
3472 - failing to execute a file in a lustre FS would lock inode (1203)
3473 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3474 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3475 - obd_uuid2tgt would walk off the end of the list (1255)
3476 - on IA64 the osc would give portals incorrect bulk size (1258)
3477 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3478 - fix lock inversion caused by new llite matching code (1282)
3479 - limit the number of dirty pages on a client to 10MB (1286)
3480 - timed out locks were not being corrected cancelled (1289)
3481 - fix O_DIRECT above 4GB on IA-32 (1292)
3482 * major user-visible changes
3483 - fail out/fail over policy now controlled by the upcall (993)
3485 - add OBD_PING to check server availability and failure (954)
3486 - lustre messages are now sent in sending host order (288, 340, 891)
3487 - add eadatalen to MDS getattr reply (340)
3488 - OST read replies may contain second buffer, with per-page status (593)
3490 2003-03-11 Phil Schwan <phil@clusterfs.com>
3493 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3494 - failed open()s could cause deadlock; fixed (867, 869)
3495 - stop cancelling OST locks when files are closed (481)
3496 - overlapping XID spaces caused network corruption (851, 853)
3497 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3498 - setattr_raw added, to do single-RPC, server-side setattrs
3499 - lmc/lconf syntax change for OST UUIDs
3500 - fix crashy race condition between ptlrpc_free_req and osc_close
3501 - don't use request in mdc_enqueue if we hit a timeout (889)
3502 - don't set the inode i_size for regular files from the MDS (896)
3503 - handle out of order completion AST (842)
3504 - don't LBUG if a lock request times out after receiving AST (913)
3505 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3506 - if a bad lock AST arrives, send an error instead of dropping entirely
3507 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3508 - fix leak in bulk IO when only partially completed (899, 900, 926)
3509 - fix O_DIRECT for ia64 (55)
3510 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3511 - C-z after timeout could hang a process forever; fixed (977)
3513 - client-side I/O cache (678, 924, 929, 941, 970)
3515 - READPAGE and SETATTRs which don't take server-side locks get
3518 2003-02-11 Phil Schwan <phil@clusterfs.com>
3521 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3522 - Fix fsfilt_extN_readpage() to read a full page of directory
3523 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3524 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3525 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3526 - Most intent operations were removed, in favour of a new RPC mode
3527 that does a single RPC to the server and bypasses most of the VFS
3528 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3529 - Aggressively cancel local locks on DLM servers
3530 - mds_reint_unlink sends EA to the client if it's the last nlink.
3531 client uses that EA to unlink OST objects.
3532 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3533 - recursive symlinks were fixed (439)
3534 - fixed NULL deref in DEBUG_REQ
3535 - filter_update_lastobjid no longer calls sync, which annoyed extN
3536 - fixed multi-client small-writes to a single file problem (445)
3537 - fixed mtime updates during file writes (607)
3538 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3539 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3540 - fixed Linux OST size reporting problem (444, 656)
3541 - OST now updates object mtime with writes or setattr (607, 619)
3542 - client verifies file size before zeroing page past EOF (445)
3543 - OST now writes last allocated objid to disk with allocation (108)
3544 - LOV on echo now works (409)
3546 - mds_reint_unlink sends a new buffer, with the EA included. this
3547 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3548 set if a regular file was being unlinked, and it was the last link
3549 - use PtlGet from the target for bulk writes (315)
3550 - OST now updates object mtime with writes or setattr (607, 619)
3551 - LDLM now has a grant-time callback to revalidate locked items, if
3553 - Many MDS operations were reorganized to combat race conditions
3555 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3556 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3558 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3561 - Fully reactivate OST imports after reconnection (512, others)
3562 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3563 - More graceful error handling for truncating on dead OST (515)
3564 - Don't error out unless we're actually accessing dead stripes (474)
3565 - Fix garbage sizes when stripes are missing (410)
3566 - LRU counters were broken, causing constant lock purge (433, 432)
3567 - garbage on read from stripes with failed OSTs (441)
3568 - mark OSCs as active before reconnecting during recovery (438)
3569 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3570 - lfind did not preserve OST order in output (443)
3571 - symlinks cause hung clients, incorrect data (439)
3572 - stop dereferencing request after dropping refcount (457)
3573 - don't LASSERT(spin_is_locked) on non-SMP (455)
3574 - fixes for many rename() bugs
3575 - fstat didn't correctly synchronize attributes (399)
3576 - server must handle lock cancellation during blocking AST prep (487)
3577 - bulk descriptors were free()d too soon (511)
3578 - fix paths in lconf, which would load incorrect modules (451, 507)
3579 - fix confusing lconf 'host not found' error message (386)
3580 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3581 - fix race condition in mdc_blocking_ast() for inode access (526)
3582 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3583 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3584 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3585 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3586 - fix lstripe in several ways (kernel side) (527)
3587 - fix request leak in ldlm_cli_enqueue (262)
3588 - incorrect OSC was marked inactive after OST failure
3589 - call mds_fs_cleanup before unmounting filesystem (524)
3590 - fix races between taking ns_lock and ldlm_lock_change_resource
3591 - fix races updating LOV export open file list
3592 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3593 - fix recovery NULL deref in ldlm_cli_cancel_unused
3594 - fix some DLM races by using new hash table for lock handles (419)
3595 - permit the client to specify desired inodes, at replay
3596 - duplicate requests when we queue them for replay reintegration
3597 - fix last_rcvd offset calculation
3598 - sync after each recovered transaction, so we always make progress
3599 - never, not always, ERESTART requests without transnos
3600 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3602 - skip replay if the MDS says that the client is already connected
3603 - don't check for a recovery-enabled export to match lctl's UUID
3604 - don't INC_USE_COUNT for phantom exports
3605 - don't crash when cleaning up phantom exports (567)
3606 - don't double-finish or set replay data for errored mdc_open requests
3607 - abort requests when they time out, so we don't get old replies
3608 - send/receive replies for AST messages again
3609 - if the client says that it doesn't have the lock, cancel it on the
3611 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3612 mark it as destroyed, it will all work out when decref is called
3613 - fix module use counts (22, 581)
3615 - ASTs now expect a reply (server cancels lock on error reply)
3617 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3620 - fix many simultaneous client startup (392)
3621 - fix dentry->d_it clobbering
3622 - credentials weren't being shipped for readdir/getattr operations
3623 - remove invalid assertions triggered during some concurrent MD
3625 - proper Lustre versions added (336, 389)
3626 - fix memory leak for create error case (398)
3627 - fix LOV locking bug that would get cli/srv out of sync
3628 - fix echo client over LOV (409)
3629 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3630 - fix double-O_EXCL intent crash (424)
3631 - avoid sending multiple lock CANCELs (352)
3633 - MDS can do multi-client recovery (modulo bugs in new code)
3635 - many updates, edits, cleanups
3637 2002-11-18 Phil Schwan <phil@clusterfs.com>
3640 - fix null d_it dereference (346)
3641 - fix full OST/dbench hang (333)
3642 - fix permission problem with file removal (286)
3643 - fix removal of OSCs from LOV when they fail
3644 - fix NULL deref during bulk timeout (214)
3645 - fix problems related to multiple filesystems on one MDS (241)
3646 - fixed serious subtle metadata locking bugs
3647 - free locks on clients when inodes are removed due to memory
3649 - fix inode pointer in lock data (285)
3650 - partial support for multiple MDS on a single host (241)
3651 - data locks weren't cancelled at clear_inode time (290, 311)
3652 - intent locks could lead to unbounded lock growth (205)
3653 - added a maximum lock count, an LRU list, and a flusher
3654 - fix multiple rename (365)
3655 - properly abstracted the echo client
3656 - OSC locked 1 byte too many; fixed
3657 - rewrote brw callback code:
3658 - fixed recovery bugs related to LOVs (306)
3659 - fixed too-many-pages-in-one-write crash (191)
3660 - fixed (again) crash in sync_io_timeout (214)
3661 - probably fixed callback-related race (385)
3663 - Add capability to MDS protocol
3664 - LDLM cancellations and callbacks on different portals
3666 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3669 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3670 - defer open object destruction to close time (601981, 138)
3671 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3672 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3673 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3674 - don't call obd_setattr in ll_file_release for destroyed objects
3675 * protocol change to lustre_msg: move |version| and add |flags|
3676 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3677 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3678 * added replay of create, unlink, link and rename operations during
3679 MDS failover; recovery should be much more robust now
3680 * remove failed OSCs from LOVs (only lov_create uses this so far)
3681 * the lustre-HOWTO was brought (more) up to date (582544)
3683 2002-10-23 Phil Schwan <phil@clusterfs.com>
3686 - in-use dentries weren't being reused properly (617851)
3687 - prevent multiple LDLM setup (599178)
3688 - fix LOV size calculations for truncate (617853)
3689 - fix client handling of MDS intent errors (POSIX)
3690 - fix permission bug in lovstripe.c test (624321)
3691 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3692 - truncate past end of file could corrupt data
3693 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3694 - a race in recovery could return ETIMEDOUT to apps (623947)
3695 - building outside the source directory was fixed
3696 * the lustre-HOWTO was brought (more) up to date (582544)
3697 * major progress was made on recovery functionality
3699 2002-10-10 Phil Schwan <phil@clusterfs.com>
3702 - recovery deadlock fix
3703 - rm -rf causes LBUG fix (617817)
3704 - file open by multiple tasks fix (618962)
3705 - directory permissions bugs (602707 and 620007)
3706 - journal_stop fixed with locking (611313)
3707 - O_APPEND failures resolved (618273, perhaps 614459)
3708 - lconf PATH fix (619770)
3709 - IA64 build fix (621450)
3710 - RPC buffer sizes scale with amount of memory
3712 2002-10-01 Phil Schwan <phil@clusterfs.com>
3715 - locks would be cancelled without throwing away data pages,
3716 resulting in inconsistent data (605627)
3717 - inode attributes were not always being refreshed (605627, 612449)
3718 - lconf now continues to cleanup after lctl reports an error
3719 - MDS now enforces user permissions (602707)
3720 - lprocfs cleanup fixed, but not yet enabled (614157)
3721 - fixed infinite server hang, should a client not respond to an AST
3722 - avoid going into recovery if user calls readlink() with a buffer
3723 that's too small (613941)
3724 - AST RPCs no longer require replies (614867) -- this may be changed
3725 - don't crash server if client sends an IOV that's too big (611336)
3726 - fixed lock conversion deadlock (611892)
3727 - fixed the following of symlinks (614622)
3728 * recovery: the server can remove locks from a client that dies, other
3729 clients can make progress
3730 * more extN patch fixes
3731 * compile-time configurable ptlrpc buffer allocations
3733 - collaborative read cache document
3734 - Lustre Lite Performance CDR document-in-progress
3736 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3739 - fix typo in patch-2.4.18
3741 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3744 - clear ptlrpc request each time in handle_incoming_request()
3745 - unlink of files now destroys the object on the OST
3747 2002-09-19 Peter Braam <braam@clusterfs.com>
3749 * add hard link support
3750 * change obdfile creation method
3751 * kernel patch changed
3753 2002-09-19 Peter Braam <braam@clusterfs.com>
3756 - stack overflow bug in extN fixed
3758 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3760 * documentation updates
3761 - add man pages for config tools
3762 - update tests/README to describe testing with new config tools
3763 - finish metadata API descriptions
3764 * bug fixes and cleanups
3765 - statfs workaround for 16TB limit
3766 - LOV stripe allocation improved, can stripe on subset of OSTs
3767 - LOV file size/IO offset was wrong for files > 4GB in size
3768 - object EA data was being dropped, caused files to be unreadable
3769 - memory overflow with non-LOV OST caused memory corruption
3770 - fixed regression tests to work with new config tools, obdfilter
3771 - fixed bug when directory size became larger than 1 block
3772 - fixed bug (for single client case) when PWD was deleted
3773 - invalidate local directory pages when doing intent-based ops
3774 - avoid LDLM oops when lock callback contained bad data
3776 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3778 * documentation updates
3779 * bug fixes and cleanups
3780 - configuration tools
3783 - 64-bit compile warnings
3784 - 64-bit internal statfs data
3786 * test_brw on persistent OST devices
3788 * lprocfs (disabled)
3790 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3792 * documentation updates
3793 * bug fixes and cleanups
3794 * configuration tools
3796 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3799 * many small fixes to 0_5_4
3800 * io/network handling
3801 * thinkos in MDS operations
3803 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3806 * crucial basic fixes to 0.5.3
3807 * IOR, Iozone work over Elan
3810 2002-08-07 Phil Schwan <phil@clusterfs.com>
3811 * version 0_5_3, our first alpha
3812 * we use the new Portals iovs
3813 * documentation updates
3814 * bug fixes and cleanups
3815 * small changes in the DLM wire protocol
3817 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3818 * version 0_5_1 with some initial stability,
3819 * locking on MD and file I/O.
3820 * documentation updates
3821 * several bug fixes since 0.5.0
3822 * small changes in wire protocol
3824 2002-07-18 Phil Schwan <phil@clusterfs.com>
3826 * delivered as Lustre Light Alpha
3827 * fixed a crash after handling invalid MDS requests
3828 * fixed directory pages for architectures with non-4k pages sizes
3830 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3831 * release version v0_4_4
3832 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3834 * Fixes a number of interruption problems with OST operations.
3835 * Update documentation for portals header changes
3836 * Move all wire protocol structs/defines to lustre_idl.h
3837 * Fixes symlink length bug.
3838 * Add tcpdump to repository.
3840 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3841 * release version v0_4_3
3842 * Fixes statfs for inodes on extN.
3843 * Fixes bug in runtests which would delete /etc/hosts.
3844 * Use 64-bit object IDs wherever possible (not into VFS though)
3845 Remove ost_get_info, which is unused by lustre, and out of date.
3847 2002-07-03 Peter Braam <braam@clusterfs.com>
3848 * release version v0_4_2 Fixes a lookup error (type not passed)
3849 * move forward to head of Portals
3850 * move forward to latest Lustre kernel
3852 2002-06-25 Peter Braam <braam@clusterfs.com>
3853 * release version v0_4_1. Hopefully stable on single node use.