5 2.6.16.46-0.14 (SLES 10),
6 2.6.18-8.1.8.EL5 (RHEL 5).
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: don't use GFP_* in generic Lustre code.
18 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
22 Description: obdclass is limited by single OBD_ALLOC(idarray)
23 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
27 Description: warnings with build patchless client with vanila 2.6.19 and up
28 Details : change old ctl_table style and replace ctl_table/ctl_table_header
29 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
33 Frequency : Only for SLES
34 Description: sanity_quota fail test_1
35 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
36 file, which makes TSTID[2] inunique.
40 Frequency : Only for SLES9
41 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
42 Details : Update target/ChangeLog/which_patch .
46 Frequency : Only for test_10 in sanity-quota.sh
47 Description: a bug in quota test code
48 Details : checking a test flag in a wrong place
52 Frequency : Only for test_13 in sanity-quota.sh
53 Description: a bug in quota test script
54 Details : shouldn't check overfull allocated quota
58 Description: jbd statistics
59 Details : Port older jbd statistics patch for sles10
61 Severity : enhancement
63 Description: Add printing inode info into message about error in writepage.
68 Description: reply_lock_interpret crash due to race with it and lock cancel.
69 Details : Do not replay locks that are being cancelled. Do not reference
70 locks by their address during replay, just by their handle.
74 Frequency : When flocks are used.
75 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
76 Details : Flock locks might destroy just granted lock if it could be merged
77 with another existing flock, this is done in completion handler,
78 so teach ldlm_cli_enquque_fini that this is a valid case for
84 Description: Oops in read and write path when failing to allocate lock.
85 Details : Check if lock allocation failed and return error back.
89 Description: lstripe command fails for valid OST index
90 Details : The stripe offset is compared to lov->desc.ld_tgt_count
91 instead of lov->desc.ld_active_tgt_count.
95 Description: sanity.sh failed test 103
96 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
101 Description: df doesn't work properly if diskfs blocksize != 4K
102 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
106 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
107 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
111 Description: obd_health_check_timeout too short
112 Details : set obd_health_check_timeout as 1.5x of obd_timeout
116 Description: llapi_file_create() does not allow some changes
117 Details : add llapi_file_open() that allows specifying the mode and
118 open flags, and also returns an open file handle.
122 Description: merge and cleanup kernel patches.
123 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
127 Description: Add journal checksum support.(Kernel part)
128 Details : The journal checksum feature adds two new flags i.e
129 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
130 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
131 indicates that the commit block contains the checksum for
132 the blocks described by the descriptor blocks. Now commit
133 record can be sent to disk without waiting for descriptor
134 blocks to be written to disk. This behavior is controlled
135 using JBD2_FEATURE_ASYNC_COMMIT flag.
139 Description: OSS needs mutliple precreate threads
140 Details : Add ability to start more than one create thread per OSS.
144 Description: Sanity.sh test_65k failed: setstripe should have succeeded
145 Details : execute lfs setstripe on client
149 Description: mds_obd_create error creating tmp object
150 Details : When the user sets quota on root, llog will be affected and can't
151 create files and write files.
154 Frequency : Always on ia64 patchless client, and possibly others.
156 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
157 Details : This allows the patchless client to be loaded on architectures
162 Description: RedHat Update kernel for RHEL5
163 Details : Add the kernel config file for RHEL5.
167 Description: RedHat Update kernel for RHEL5
168 Details : Modify the kernel config file more closer RHEL5.
172 Description: Build failure against Centos5 (RHEL5)
173 Details : Define PAGE_SIZE when it isn't present.
177 Description: test_7 on sanity-quota.sh can't be used on separate machines
178 Details : add facet to handle it
182 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
183 Details : with NFS, the anon dentry's parent was set to itself in
184 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
185 corresponding dentry other than use rec->ur_name.
187 Severity : enhancement
189 Description: lfs setstripe enhancement
190 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
194 Description: testing performance impact of enabling checksumming
195 Details : enable checksum by default, allow --disable-checksum
196 configure option and "-o nochecksum" mount option
200 Description: lustre support for RHEL5
201 Details : Add support for RHEL5.
205 Description: Client eviction due to failover config
206 Details : after a connection loss, the lustre client should attempt to
207 reconnect to the last active server first before trying the
208 other potential connections.
212 Description: when mds and osts use different quota unit(32bit and 64bit),
213 quota will be released repeatly.
214 Details : void sending multiple quota reqs to mds, which will keep the
215 status between the reqs.
219 Description: osts not allocated evenly to files
220 Details : change the condition to increase offset_idx
224 Description: Tune the kernel for good SCSI performance.
225 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
226 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
231 Description: rewrite ext2-derived code in obdclass/uuid.c
232 Details : rewrite inherited code (uuid parsing code from ext2 utils)
233 from scratch preserving functionality.
234 --------------------------------------------------------------------------------
236 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
238 * Support for kernels:
239 2.6.5-7.283 (SLES 9),
240 2.6.9-55.EL (RHEL 4),
241 2.6.16.46-0.14 (SLES 10),
242 2.6.18.8 vanilla (kernel.org)
243 * Client support for unpatched kernels:
244 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
245 2.6.16 - 2.6.22 vanilla (kernel.org)
246 * Due to recently discovered recovery problems, we do not recommend
247 using patchless RHEL 4 clients with this or any earlier release.
248 * Recommended e2fsprogs version: 1.39.cfs8
249 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
250 * Starting with this release, the ldiskfs backing filesystem required
251 by Lustre is now in its own package, lustre-ldiskfs. This package
252 should be installed. It is versioned separately from Lustre and
253 may be released separately in future.
257 Description: block reactivating mgc import until all deactivates complete
258 Details : Fix race when failing back MDT/MGS to itself (testing)
260 Severity : enhancement
262 Description: add optional extra BUILD_VERSION info
263 Details : add a new environment variable (namely LUSTRE_VERS) which allows
264 to override the lustre version.
267 Frequency : 2.6.18 servers only
269 Description: ll_kern_mount() doesn't release the module reference
270 Details : The ldiskfs module reference count never drops down to 0
271 because ll_kern_mount() doesn't release the module reference.
276 Description: server LBUG when using old ost_num_threads parameter
277 Details : Accept the old ost_num_threads parameter but warn that it
278 is deprecated, and fix an off-by-one error that caused an LBUG.
283 Description: Transient SCSI error results in persistent IO issue
284 Details : iobuf->dr_error is not reinitialized to 0 between two
288 Frequency : sometimes when underlying device returns I/O errors
290 Description: OSTs not going read-only during write failures
291 Details : OSTs are not remounted read-only when the journal commit threads
292 get I/O errors because fsfilt_ext3 calls journal_start/stop()
293 instead of the ext3 wrappers.
297 Description: poor connect scaling with increasing client count
298 Details : Don't run filter_grant_sanity_check for more than 100 exports
299 to improve scaling for large numbers of clients.
302 Frequency : SLES10 only
304 Description: sanity-quota.sh quotacheck failed: rc = -22
305 Details : Quotas cannot be enabled on SLES10.
308 Frequency : liblustre clients only
310 Description: getdirentries does not give error when run on compute nodes
311 Details : getdirentries does not fail when the size specified as an argument
312 is too small to contain at least one entry
314 Severity : enhancement
316 Description: Add LNET router traceability for debug purposes
317 Details : If a checksum failure occurs with a router as part of the
318 IO path, the NID of the last router that forwarded the bulk data
319 is printed so it can be identified.
324 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
325 Details : Due to a race condition, liblustre clients were occasionally
328 Severity : enhancement
330 Description: lfs setstripe use optional parameters instead of postional
333 Severity : enhancement
335 Description: Nanosecond timestamp support for ldiskfs
336 Details : The on-disk ldiskfs filesystem has added support for nanosecond
337 resolution timestamps. There is not yet support for this at
338 the Lustre filesystem level.
341 Frequency : during server recovery
343 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
344 Details : request with rq_no_resend flag not awake l_wait_event if they get a
348 Frequency : nfs export on patchless client
350 Description: connectathon hang when test nfs export over patchless client
351 Details : Disconnected dentry cannot be found with lookup, so we do not need
352 to unhash it or make it invalid
355 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
356 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
357 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
362 Description: open req refcounting wrong on reconnect
363 Details : If reconnect happened between getting open reply from server and
364 call to mdc_set_replay_data in ll_file_open, we will schedule
365 replay for unreferenced request that we are about to free.
366 Subsequent close will crash in variety of ways.
367 Check that request is still eligible for replay in
368 mdc_set_replay_data().
373 Description: disable writes to filesystem when reading health_check file
374 Details : the default for reading the health_check proc file has changed
375 to NOT do a journal transaction and write to disk, because this
376 can cause reads of the /proc file to hang and block HA state
377 checking on a healthy but otherwise heavily loaded system. It
378 is possible to return to the previous behaviour during configure
379 with --enable-health-write.
381 Severity : enhancement
383 Description: 64-bit inode version
384 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
385 to the inode. This will be required for version-based recovery.
390 Description: MDS fails to start if a duplicate client export is detected
391 Details : in some rare cases it was possible for a client to connect to
392 an MDS multiple times. Upon recovery the MDS would detect this
393 and fail during startup. Handle this more gracefully.
395 Severity : enhancement
397 Description: Add -o localflock option to simulate old noflock
399 Details : This will achieve local-only flock/fcntl locks
405 Description: log_commit_thread vs filter_destroy race leads to crash
406 Details : Take import reference before releasing llog record semaphore
411 Description: Wrong request locking in request set processing
412 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
413 imp_delayed_list, in this place should be used imp_lock.
416 Frequency : when reconnection
418 Description: Grant Leak when osc reconnect to OST
419 Details : When osc reconnect ost, OST(filter) should check whether it
420 should grant more space to client by comparing fed_grant and
421 cl_avail_grant, and return the granted space to client instead
422 of "new granted" space, because client will call osc_init_grant
423 to update the client grant space info.
426 Frequency : when client reconnect to OST
428 Description: Grant Leak when osc do resend and replay bulk write
429 Details : When osc reconnect to OST, OST(filter)should clear grant info of
430 bulk write request, because the grant info while be sync between
431 OSC and OST when reconnect, and we should ignore the grant info
432 these of resend/replay write req.
437 Description: Grant space more than avaiable left space sometimes.
438 Details : When then OST is about to be full, if two bulk writing from
439 different clients came to OST. Accord the avaliable space of the
440 OST, the first req should be permitted, and the second one
441 should be denied by ENOSPC. But if the seconde arrived before
442 the first one is commited. The OST might wrongly permit second
443 writing, which will cause grant space > avaiable space.
446 Frequency : when client is evicted
448 Description: Grant might be wrongly erased when osc is evicted by OST
449 Details : when the import is evicted by server, it will fork another
450 thread ptlrpc_invalidate_import_thread to invalidate the
451 import, where the grant will be set to 0. While the original
452 thread will update the grant it got when connecting. So if
453 the former happened latter, the grant will be wrongly errased
454 because of this race.
459 Description: Checking Stale with correct fid
460 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
461 to check whether it is stale, because sometimes, we want the
462 enqueue happened anyway, and op_data.fid2 will not be initialized.
464 Severity : enhancement
466 Description: update patchless client
467 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
470 Frequency : only with 2.4 kernel
472 Description: random memory corruption
473 Details : size of struct ll_inode_info is to big for union inode.u and this
474 can be cause of random memory corruption.
479 Description: Memory leak in recovery
480 Details : Lov_mds_md was not free in an error handler in mds_create_object.
481 It should also check obd_fail before fsfilt_start, otherwise if
482 fsfilt_start return -EROFS,(failover mds during mds recovery).
483 then the req will return with repmsg->transno = 0 and rc = EROFS.
484 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
485 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
486 be freed no matter whether fsfilt_commit success or not.
489 Frequency : only with huge count clients
491 Description: Prevents from taking the superblock lock in llap_from_page for
493 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
494 for llap_from_page call in ll_removepage prevents from taking the
495 superblock lock for a soon died page.
500 Description: Not check open intent error before release open handle
501 Details : in some rare cases, the open intent error is not checked before
502 release open handle, which may cause
503 ASSERTION(open_req->rq_transno != 0), because it tries to release
504 the failed open handle.
509 Description: Set cat log bitmap only after create log success.
510 Details : in some rare cases, the cat log bitmap is set too early. and it
511 should be set only after create log success.
515 Description: Accessing a block bevice can re-enable I/O when Lustre is
516 tearing down a device.
517 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
522 Description: service threads may hog cpus when there are a lot of requests
524 Details : Insert cond_resched to give other threads a chance to use some of
530 Description: the cat log was not initialized in recovery
531 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
532 unlink log on mds will not be initialized until mds post
533 recovery. And also in mds post recovery, the unlink log will
534 initialization will be done asynchronausly, so there will be race
535 between add unlink log and unlink log initialization.
539 Description: brw_stats were being printed incorrectly
540 Details : brw_stats were being printed as log2 but all of them were not
541 recorded as log2. Also remove some code duplication arising from
542 filter_tally_{read,write}.
546 Frequency : rare, only in recovery.
547 Description: ASSERTION(req->rq_type != LI_POISON) failed
548 Details : imp_lock should be held while iterating over imp_sending_list for
549 prevent destroy request after get timeout in ptlrpc_queue_wait.
553 Description: replay-single.sh test 52 fails
554 Details : A lock's skiplist need to be cleanup when it being unlinked
555 from its resource list.
559 Description: Short directio read returns full requested size rather than
561 Details : Direct I/O operations should return actual amount of bytes
562 transferred rather than requested size.
564 Severity : enhancement
566 Description: metadata RPC reduction (e.g. for rm performance)
567 Details : decrease the amount of synchronous RPC between clients and servers
568 by canceling conflicing lock before the operation on the client side
569 and packing thier handles into the main operation RPC to server.
571 Severity : enhancement
573 Description: Async OSC create to avoid the blocking unnecessarily.
574 Details : If a OST has no remain object, system will block on the creating
575 when need to create a new object on this OST. Now, ways use
576 pre-created objects when available, instead of blocking on an
577 empty osc while others are not empty. If we must block, we block
578 for the shortest possible period of time.
580 --------------------------------------------------------------------------------
582 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
587 Frequency : on some architectures
589 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
590 Details : Uninitialized flags sometimes cause configuration commands to
594 Frequency : patchless clients only
596 Description: missing __iget() symbol export
597 Details : The __iget() symbol export is missing. To avoid the need for
598 this on patchless clients the deathrow inode reaper is turned
599 off, and we depend on the VM to clean up old inodes. This
600 dependency was during via the fix for bug 12181.
602 --------------------------------------------------------------------------------
604 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
606 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
607 INTEROPERATE with older versions automatically. In many cases a
608 special upgrade step is needed. Please read the
609 user documentation before upgrading any part of a live system.
610 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
611 version will not interoperate with 1.6 betas before beta5 (1.5.95).
612 * WARNING: Lustre configuration and startup changes are required with
613 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
619 Severity : enhancement
621 Description: MountConf
622 Details : Lustre configuration is now managed via mkfs and mount
623 commands instead of lmc and lconf. New obd types (MGS, MGC)
624 are added for dynamic configuration management. See
625 https://mail.clusterfs.com/wikis/lustre/MountConf for
628 Severity : enhancement
630 Description: dynamic OST addition
631 Details : OSTs can now be added to a live filesystem
633 Severity : enhancement
635 Description: startup order invariance
636 Details : MDTs and OSTs can be started in any order. Clients only
637 require the MDT to complete startup.
639 Severity : enhancement
641 Description: parallel, asynchronous orphan cleanup
642 Details : orphan cleanup is now performed in separate threads for each
643 OST, allowing parallel non-blocking operation.
645 Severity : enhancement
647 Description: optimized stripe assignment
648 Details : stripe assignments are now made based on ost space available,
649 ost previous usage, and OSS previous usage, in order to try
650 to optimize storage space and networking resources.
652 Severity : enhancement
654 Description: Permanently set tunables
655 Details : All writable /proc/fs/lustre tunables can now be permanently
656 set on a per-server basis, at mkfs time or on a live
659 Severity : enhancement
661 Description: Lustre message v2
662 Details : Add lustre message format v2.
664 Severity : enhancement
666 Description: client OST exclusion list
667 Details : Clients can be started with a list of OSTs that should be
668 declared "inactive" for known non-responsive OSTs.
672 Description: SPEC SFS validation failure on NFS v2 over lustre.
673 Details : Changes the blocksize for regular files to be 2x RPC size,
674 and not depend on stripe size.
676 Severity : enhancement
678 Description: Multiple MD RPCs in flight.
679 Details : Further unserialise some read-only MDS RPCs - learn about intents.
680 To avoid overly-overloading MDS, introduce a limit on number of
681 MDS RPCs in flight for a single client and add /proc controls
682 to adjust this limit.
684 Severity : enhancement
686 Description: client read/write statistics
687 Details : Add client read/write call usage stats for performance
688 analysis of user processes.
689 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
690 file access. extents_stats shows chunk size distribution.
691 extents_stats_per_process show chunk size distribution per
694 Severity : enhancement
696 Description: mds statistics
697 Details : Add detailed mds operations statistics in
698 /proc/fs/lustre/mds/*/stats.
702 Description: Failure of copying files with lustre special EAs.
703 Details : Client side always return success for setxattr call for lustre
704 special xattr (currently only "trusted.lov").
709 Description: Refcount LNET uuids
710 Details : The global LNET uuid list grew linearly with every startup;
711 refcount repeated list entries instead of always adding to
714 Severity : enhancement
716 Description: Dynamic service threads
717 Details : Within a small range, start extra service threads
718 automatically when the request queue builds up.
721 Frequency : mixed-endian client/server environments
723 Description: mixed-endian crashes
724 Details : The new msg_v2 system had some failures in mixed-endian
727 Severity : enhancement
729 Description: Easy OST removal
730 Details : OSTs can be permanently deactivated with e.g. 'lctl
731 conf_param lustre-OST0001.osc.active=0'
733 Severity : enhancement
735 Description: MGS proc entries
736 Details : Added basic proc entries for the MGS showing what filesystems
739 Severity : enhancement
741 Description: provide MGS failover
742 Details : Added config lock reacquisition after MGS server failover.
744 Severity : enhancement
746 Description: add Linux 2.4 support
747 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
751 Description: a large application tries to do I/O to the same resource and dies
753 Details : Check the req->rq_arrival time after the call to
754 ost_brw_lock_get(), but before we do anything about
755 processing it & sending the BULK transfer request. This
756 should help move old stale pending locks off the queue as
757 quickly as obd_timeout.
760 Frequency : when an incorrect nid is specified during startup
762 Description: ptlrpc connect to non-existant node causes kernel crash
763 Details : LNET can't be re-entered from an event callback, which
764 happened when we expire a message after the export has been
765 cleaned up. Instead, hand the zombie cleanup off to another
768 Severity : enhancement
770 Description: plain/inodebits lock performance improvement
771 Details : Grouping plain/inodebits in granted list by their request modes
772 and bits policy, thus improving the performance of search through
776 Frequency : only if OST filesystem is corrupted
778 Description: client incorrectly hits assertion in ptlrpc_replay_req()
779 Details : for a short time RPCs with bulk IO are in the replay list,
780 but replay of bulk IOs is unimplemented. If the OST filesystem
781 is corrupted due to disk cache incoherency and then replay is
782 started it is possible to trip an assertion. Avoid putting
783 committed RPCs into the replay list at all to avoid this issue.
786 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
788 Description: System hang on startup
789 Details : This bug allowed the liblustre (e.g. catamount) client to
790 return to the app before handling all startup RPCs. This
791 could leave the node unresponsive to lustre network traffic
792 and manifested as a server ptllnd timeout.
794 Severity : enhancement
796 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
797 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
798 boolean to control whether to print peer debug info when a
799 client's RPC times out.
802 Frequency : only for kernels with patches from Lustre below 1.4.3
804 Description: Remove old rdonly API
805 Details : Remove old rdonly API which unused from at least lustre 1.4.3
808 Frequency : only for devices with external journals
810 Description: Set external device read-only also
811 Details : During a commanded failover stop, we set the disk device
812 read-only while the server shuts down. We now also set any
813 external journal device read-only at the same time.
816 Frequency : when upgrading from 1.4 while trying to change parameters
818 Description: The wrong (new) MDC name was used when setting parameters for
819 upgraded MDT's. Also allows changing of OSC (and MDC)
820 parameters if --writeconf is specified at tunefs upgrade time.
823 Frequency : when setting specific ost indicies
825 Description: QOS code breaks on skipped indicies
826 Details : Add checks for missing OST indicies in the QOS code, so OSTs
827 created with --index need not be sequential.
829 Severity : enhancement
831 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
832 Details : The uninit_groups feature works in conjunction with the kernel
833 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
834 up the pass1 processing of e2fsck. This is a read-only feature
835 in ldiskfs2 only, so older kernels and current ldiskfs cannot
836 mount filesystems that have had this feature enabled.
838 Severity : enhancement
840 Description: Improve multi-block allocation algorithm to avoid fragmentation
841 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
842 allocation locality and avoid filesystem fragmentation.
844 ------------------------------------------------------------------------------
846 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
848 * Support for kernels:
849 2.4.21-47.0.1.EL (RHEL 3)
851 2.6.9-42.0.10.EL (RHEL 4)
852 2.6.12.6 vanilla (kernel.org)
853 2.6.16.27-0.9 (SLES 10)
854 * Recommended e2fsprogs version: 1.39.cfs5
856 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
860 Frequency : occasional, depends on client load and configuration
861 Bugzilla : 12181, 12203
862 Description: data loss for recently-modified files
864 Details : In some cases it is possible that recently written or created
865 files may not be written to disk in a timely manner (this should
866 normally be within 30s unless client IO load is very high).
867 The problem appears as zero-length files or files that are a
868 multiple of 1MB in size after a client crash or client eviction
869 that are missing data at the end of the file.
871 This problem is more likely to be hit on clients where files are
872 repeatedly created and unlinked in the same directory, clients
873 have a large amount of RAM, have many CPUs, the filesystem has
874 many OSTs, the clients are rebooted frequently, and/or the files
875 are not accessed by other nodes after being written.
877 The presence of the problem can be detected by looking at
878 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
879 smaller than the second (nr_unused) then dirty files will not
880 be flushed automatically to disk. "sync; sleep 10" should be
881 run several times on the node before unmounting it to update
882 Lustre (this is also safe to run on nodes without this problem).
884 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
885 that can cause this same problem, so customers using that kernel
886 also need to update the kernel in addition to Lustre. In order
887 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
889 It is normal that files written just before a client crash (less
890 than 30s) may not yet have been flushed to disk, even for local
894 Frequency : frequent on thin XT3 nodes
896 Description: UUID collision on thin XT3 Linux nodes
897 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
898 randomly, since we relied on an insufficiently-seeded PRNG.
903 Description: OSS hangs after "All ost request buffers busy"
904 Details : A deadlock between quota and journal operations caused OSS
905 hangs after printing "All ost request buffers busy."
908 Frequency : always on liblustre builds
910 Description: Cleanup compiler warnings on liblustre
913 Frequency : always on liblustre builds on XT3
915 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
919 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
920 > 32000 subdirectories
921 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
922 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
923 subdirectory count crosses 32000. This will aid e2fsck to
924 correctly handle more than 32000 subdirectories.
927 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
929 Description: System hang on startup
930 Details : This bug allowed the liblustre (e.g. catamount) client to
931 return to the app before handling all startup RPCs. This
932 could leave the node unresponsive to lustre network traffic
933 and manifested as a server ptllnd timeout.
935 Severity : enhancement
937 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
938 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
939 boolean to control whether to print peer debug info when a
940 client's RPC times out.
945 Description: make O_SYNC working on 2.6 kernels
946 Details : 2.6 kernels use different method for mark pages for write,
947 so need add a code to lustre for O_SYNC work.
952 Description: Failure to close file and release space on NFS
953 Details : Put inode details into lock acquired in ll_intent_file_open.
954 Use mdc_intent_lock in ll_intent_open to properly
955 detect all kind of errors unhandled by mdc_enqueue.
960 Description: proc file read during shutdown sometimes raced obd removal,
962 Details : Add lock to prevent obd access after proc file removal.
965 Frequency : Only for files larger than 4GB on 32-bit clients.
967 Description: improperly doing page alignment of locks
968 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
969 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
974 Description: RPCs being resent when they shouldn't be
975 Details : Some RPCs that should not be resent are being resent. This
976 can cause inconsistencies in the RPC state machine. Do not
977 resend such requests.
980 Frequency : rare, only with NFS export
982 Description: Crash on NFS re-export node
983 Details : under very unusual load conditions an assertion is hit in
984 ll_intent_file_open()
987 Frequency : only if OST filesystem is corrupted
989 Description: client incorrectly hits assertion in ptlrpc_replay_req()
990 Details : for a short time RPCs with bulk IO are in the replay list,
991 but replay of bulk IOs is unimplemented. If the OST filesystem
992 is corrupted due to disk cache incoherency and then replay is
993 started it is possible to trip an assertion. Avoid putting
994 committed RPCs into the replay list at all to avoid this issue.
999 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1000 Details : Large single O_DIRECT read and write calls can fail to allocate
1001 a sufficiently large buffer to process the request. In case of
1002 allocation failure the allocation is retried with a smaller
1003 buffer and broken into smaller requests.
1005 Severity : enhancement
1007 Description: Add -o localflock option to simulate old noflock behaviour.
1008 Details : This will achieve local-only flock/fcntl locks coherentness.
1013 Description: versioning check is incomplete
1014 Details : Checking the version difference of client vs. server, report
1015 error if the gap is too big.
1020 Description: add support PG_writeback bit
1021 Details : add support for PG_writeback bit for Lustre, for more carefull
1022 work with page cache in 2.6 kernel. This also fix some deadlocks
1023 and remove hack for work O_SYNC with 2.6 kernel.
1025 Severity : enhancement
1027 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1028 Details : The uninit_groups feature works in conjunction with the kernel
1029 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1030 up the pass1 processing of e2fsck. This is a read-only feature
1031 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1032 mount filesystems that have had this feature enabled.
1034 Severity : enhancement
1036 Description: Improve multi-block allocation algorithm to avoid fragmentation
1037 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1038 allocation locality and avoid filesystem fragmentation.
1040 ------------------------------------------------------------------------------
1042 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1044 * Support for kernels:
1045 2.6.9-42.0.3.EL (RHEL 4)
1046 2.6.5-7.276 (SLES 9)
1047 2.4.21-47.0.1.EL (RHEL 3)
1048 2.6.12.6 vanilla (kernel.org)
1049 2.6.16.21-0.8 (SLES10)
1050 * Recommended e2fsprogs version: 1.39.cfs2-0
1052 * The backwards-compatible /proc/sys/portals symlink has been removed
1053 in this release. Before upgrading, please ensure that you change
1054 any configuration scripts or /etc/sysctl.conf files that access
1055 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1056 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1057 in advance of the upgrade on any system running Lustre 1.4.6 or
1058 newer, since /proc/sys/lnet was added in that version.
1059 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1063 Frequency : only when quota is used
1065 Description: avoid scanning export list for quota master
1066 Details : Change the algorithms to avoid scanning export list in order
1067 to improve the efficiency.
1070 Frequency : MDS failover only, very rarely
1072 Description: "went back in time" messages on mds failover
1073 Details : The greatest transno may be lost when the current operation
1074 finishes with an error (transno==0) and the client's last_rcvd
1075 record is over-written. Save the greatest transno in the
1076 mds_last_transno for this case.
1079 Frequency : always for specific kernels and striping counts
1081 Description: client may get "Matching packet too big" without ACL support
1082 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1083 when trying to access files in certain configurations. The
1084 clients should in fact be denied when mounting because they do
1085 not understand ACLs.
1088 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1090 Description: many clients connecting with IO in progress causes connect timeouts
1091 Details : Avoid synchronous journal commits to avoid delays caused by many
1092 clients connecting/disconnecting when bulk IO is in progress.
1093 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1094 OST_IO_PORTAL to avoid delays behind potentially many pending
1098 Frequency : occasionally with multiple writers to a single file
1100 Description: shared writes to file may result in wrong size reported by stat()
1101 Details : Allow growing of kms when extent lock is cancelled
1104 Frequency : always with random mmap IO to multi-striped file
1106 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1107 Details : Only if the hole is at the end of OST object so that kms is too
1108 small. Fix is to increase kms accordingly in ll_nopage.
1111 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1113 Description: writes to a missing object would leak memory on the OST
1114 Details : If there is an inconsistency between the MDS and OST filesystems,
1115 such that the MDS references an object that doesn't exist, writes
1116 to that object will leak memory due to incorrect cleanup in the
1117 error handling path, eventually running out of memory on the OST.
1122 Description: Creating too long symlink causes lustre errors
1123 Details : Check symlink and name lengths before sending requests to MDS.
1126 Frequency : only if flock is enabled (not on by default)
1128 Description: posix locks not released on fd closure on 2.6.9+
1129 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1130 kernels, this caused such locks not to be released on fd close and
1131 then assertions on fs unmount about still used locks.
1134 Frequency : MDS failover only, very rarely
1136 Description: clients may get ASSERTION(granted_lock != NULL)
1137 Details : When request was taking a long time, and a client was resending
1138 a getattr by name lock request. The were multiple lock requests
1139 with the same client lock handle and
1140 mds_getattr_name->fixup_handle_for_resent_request found one of the
1141 lock handles but later failed with ASSERTION(granted_lock != NULL).
1146 Description: handle->h_buffer_credits > 0, assertion failure
1147 Details : h_buffer_credits is zero after truncate, causing assertion
1148 failure. This patch extends the transaction or creates a new
1152 Frequency : NFS re-export or patchless client
1153 Bugzilla : 11179, 10796
1154 Description: Crash on NFS re-export node (__d_move)
1155 Details : We do not want to hash the dentry if we don't have a lock.
1156 But if this dentry is later used in d_move, we'd hit uninitialised
1157 list head d_hash, so we just do this to init d_hash field but
1158 leave dentry unhashed.
1161 Frequency : NFS re-export or patchless client
1163 Description: NFS exports has problem with symbolic link
1164 Details : lustre client didn't properly install dentry when re-exported
1165 to NFS or running patchless client.
1168 Frequency : NFS re-export or patchless client
1170 Description: Various nfs/patchless fixes.
1171 Details : fixes reuse disconected alias for lookup process - this fixes
1172 warning "find_exported_dentry: npd != pd",
1173 fix permission error with open files at nfs.
1174 fix apply umask when do revalidate.
1177 Frequency : occasional
1179 Description: Crash on NFS re-export node
1180 Details : calling clear_page() on the wrong pointer triggered oops in
1181 generic_mapping_read().
1184 Frequency : rarely, using O_DIRECT IO
1186 Description: unaligned directio crashes client with LASSERT
1187 Details : check for unaligned buffers before trying any requests.
1190 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1192 Description: stale data returned from RAID cache
1193 Details : If only a small amount of IO is done to the RAID device before
1194 reading it again it is possible to get stale data from the RAID
1195 cache instead of reading it from disk.
1198 Frequency : always for sles10 kernel
1200 Description: sles10 support
1201 Details : ll_follow_link: compile fixes and using of nd_set_link
1202 under newer kernels.
1205 Frequency : depends on arch, kernel and compiler version, always on sles10
1208 Description: recursive or deep enough symlinks cause stack overflow
1209 Details : getting rid of large stack-allocated variable in
1213 Frequency : depends on hardware
1215 Description: lustre write performance loss in the SLES10 kernel
1216 Details : the performance loss is caused by using of write barriers in the
1217 ext3 code. The SLES10 kernel turns barrier support on by
1218 default. The fix is to undo that change for ldiskfs.
1220 ------------------------------------------------------------------------------
1222 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1224 * Support for kernels:
1225 2.6.9-42.0.3EL (RHEL 4)
1226 2.6.5-7.276 (SLES 9)
1227 2.4.21-47.0.1.EL (RHEL 3)
1228 2.6.12.6 vanilla (kernel.org)
1232 Frequency : quota enabled and large files being deleted
1234 Description: releasing more than 4GB of quota at once hangs OST
1235 Details : If a user deletes more than 4GB of files on a single OST it
1236 will cause the OST to spin in an infinite loop. Release
1237 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1242 Description: statfs data retrieved from /proc may be stale or zero
1243 Details : When reading per-device statfs data from /proc, in the
1244 {kbytes,files}_{total,free,avail} files, it may appear
1245 as zero or be out of date.
1248 Frequency : systems with MD RAID1 external journal devices
1250 Description: lconf's call to blkid is confused by RAID1 journal devices
1251 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1252 of returning all block devices that match the journal UUID.
1255 Frequency : always, for aggregate stripe size over 4GB
1257 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1258 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1259 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1262 Frequency : always if "lfs find" used on a local file/directory
1264 Description: "lfs find" segfaults if used on a local file/directory
1265 Details : The case where a directory component was not specified wasn't
1266 handled correctly. Handle this properly.
1269 Frequency : always on ppc64
1271 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1272 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1273 perform endianness conversion.
1276 Frequency : rarely (truncate to non-zero file size after write under load)
1277 Bugzilla : 10730, 10687
1278 Description: Files padded with zeros to next 4K multiple
1279 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1280 it is possible that files that are truncated to a non-zero size
1281 immediately after being written are filled with zero bytes beyond
1282 the truncated size. No file data is lost.
1284 Severity : enhancement
1286 Description: Allow recovery/failover for liblustre clients.
1287 Details : liblustre clients were unaware of failover configurations until
1290 Severity : enhancement
1292 Description: user file locks should fail when not mounting with flock option
1293 Details : Set up an error-returning stub in ll_file_operations.lock field
1294 to prevent incorrect behaviour when client is mounted without
1295 flock option. Also, set up properly f_op->flock field for
1299 Frequency : always on ia64
1301 Description: "lfs df" loops on printing out MDS statfs information
1302 Details : The obd_ioctl_data was not initialized and in some systems
1303 this caused a failure during the ioctl that did not return
1304 an error. Initialize the struct and return an error on failure.
1307 Frequency : SLES 9 only
1309 Description: Error of copying files with lustre special EAs as root
1310 Details : Client side always return success for setxattr call for lustre
1311 special xattr (currently only "trusted.lov").
1314 Frequency : rarely on clusters with both ia64+i386 clients
1316 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1317 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1318 clients are doing concurrent writes to the same file it is
1319 possible that the ia64 clients may LASSERT because the OST
1320 extent locks are not PAGE_SIZE aligned. Ensure that grown
1321 locks are always aligned on the request boundary.
1324 Frequency : specific use, occasional
1326 Description: Overwriting in use executable truncates on-disk binary image
1327 Details : If one node attempts to overwrite an executable in use by
1328 another node, we now correctly return ETXTBSY instead of
1329 truncating the file.
1331 Severity : enhancement
1333 Description: Async OSC create to avoid the blocking unnecessarily.
1334 Details : If a OST has no remain object, system will block on the creating
1335 when need to create a new object on this OST. Now, ways use
1336 pre-created objects when available, instead of blocking on an
1337 empty osc while others are not empty. If we must block, we block
1338 for the shortest possible period of time.
1343 Description: chmod on Lustre root is propagated to other clients
1344 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1345 invalid by the follow_mount time.
1348 Frequency : liblustre clients only
1350 Description: Race in 'instant cancel' lock handling could lead to such locks
1351 never to be granted in case of SMP MDS
1352 Details : Do not destroy not yet granted but cbpending locks in
1356 Frequency : replay/resend of open
1358 Description: non null lock assetion failure in mds_intent_policy
1359 Details : Trying to replay/resend lockless open requests resulted in
1360 mds_open() returning 0 with no lock. Now it sets a flag if
1361 a lock is going to be returned.
1363 Severity : enhancement
1365 Description: Checksum enhancements
1366 Details : New checksum enhancements allow for resending RPCs that failed
1369 Severity : enhancement
1371 Description: Tunables on number of dirty pages in cacche
1372 Details : Allow to set limit on number of dirty pages cached.
1377 Description: client crash on unmount - lock still has references
1378 Details : In some error handling cases it was possible to leak a lock
1379 reference on a client while accessing a file. This was not
1380 harmful to the client during operation, but would cause the
1381 client to crash when the filesystem is unmounted.
1384 Frequency : specific case, rare
1386 Description: ETXTBSY on mds though file not in use by client
1387 Details : ETXTBSY is no longer incorrectly returned when attempting to
1388 chmod or chown a directory that the user previously tried to
1389 execute or a currently-executing binary.
1392 Frequency : extremely rare except on liblustre-based clients
1394 Description: Lustre space not freed when files are deleted
1395 Details : Clean up open-unlinked files after client eviction. Previously
1396 the unlink was skipped and the files remained as orphans.
1401 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1402 Details : In some cases it was possible to send a blocking callback to
1403 a client doing a glimpse, even though that client didn't get
1404 a lock granted. When the glimpse lock is cancelled on the OST
1405 the freed lock is left on the waiting list and corrupted the list.
1408 Frequency : all core dumps
1410 Description: Broke core dumps to lustre
1411 Details : Negative dentry may be unhashed if parent does not have UPDATE
1412 lock, but some callers, e.g. do_coredump, expect dentry to be
1413 hashed after successful create, hash it in ll_create_it.
1415 ------------------------------------------------------------------------------
1417 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1419 * Support for kernels:
1420 2.6.9-42.0.2.EL (RHEL 4)
1421 2.6.5-7.276 (SLES 9)
1422 2.4.21-40.EL (RHEL 3)
1423 2.6.12.6 vanilla (kernel.org)
1427 Frequency : always on RHEL 3
1429 Description: Number of open files grows over time
1430 Details : The number of open files grows over time, whether or not
1431 Lustre is started. This was due to a filp leak introduced
1432 by one of our kernel patches.
1434 ------------------------------------------------------------------------------
1436 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1438 * Support for kernels:
1439 2.6.9-42.EL (RHEL 4)
1440 2.6.5-7.276 (SLES 9)
1441 2.4.21-40.EL (RHEL 3)
1442 2.6.12.6 vanilla (kernel.org)
1447 Bugzilla : 5719, 9635, 9792, 9684
1448 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1449 Details : If a server is under heavy load and cannot reply to new
1450 connection requests before the client resends the (re)connect,
1451 the connection handling code can behave badly if two service
1452 threads are concurrently handing separate (re)connections from
1453 the same client. Add better locking to the connection handling
1454 code, and ensure that only a single connection will be processed
1455 for a given client UUID, even if the lock is dropped.
1457 Severity : enhancement
1459 Description: add TCP zero-copy support to kernel
1460 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1461 sends if the hardware supports scatter-gather and checksumming.
1462 This allows socklnd to do client-write and server-read more
1463 efficiently and reduce CPU utilization from skbuf copying.
1466 Frequency : only if NFS exporting from client
1468 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1469 Details : If setting attributes on a file created under NFS that had
1470 never been opened it would be possible to oops the client
1471 if the file had no objects.
1474 Frequency : always for liblustre
1476 Description: liblustre client does MDS+OSTs setattr RPC for each write
1477 Details : When doing a write from a liblustre client, the client
1478 incorrectly issued an RPC to the MDS and each OST the file was
1479 striped over in order to update the timestamps. When writing
1480 with small chunks and many clients this could overwhelm the MDS
1481 with RPCs. In all cases it would slow down the write because
1482 these RPCs are unnecessary.
1484 Severity : enhancement
1486 Description: allow number of MDS service threads to be changed at module load
1487 Details : It is now possible to change the number of MDS service threads
1488 running. Adding "options mds mds_num_threads={N}" to the MDS's
1489 /etc/modprobe.conf will set the number of threads for the next
1490 time Lustre is restarted (assuming the "mds" module is also
1491 reloaded at that time). The default number of threads will
1492 stay the same, 32 for most systems.
1497 Description: OST crash if filesystem is unformatted or corrupt
1498 Details : If an OST is started on a device that has never been formatted
1499 or if the filesystem is corrupt and cannot even mount then the
1500 error handling cleanup routines would dereference a NULL pointer.
1505 Description: NULL pointer deref in llap_from_page.
1506 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1507 incorrect) mapping if the page was truncated/reclaimed while it was
1508 searched for. Check for this condition and skip such pages when
1509 doing readahead. Introduce extra check to llap_from_page() to
1510 verify page->mapping->host is non-NULL (so page is not anonymous).
1513 Frequency : Sometimes when using sys_sendfile
1515 Description: "page not covered by a lock" warnings from ll_readpage
1516 Details : sendfile called ll_readpage without right page locks present.
1517 Now we introduced ll_file_sendfile that does necessary locking
1518 around call to generic_file_sendfile() much like we do in
1522 Frequency : with certain MDS communication failures at client mount time
1524 Description: NULL pointer deref after failed client mount
1525 Details : a client connection request may delayed by the network layer
1526 and not be sent until after the PTLRPC layer has timed out the
1527 request. If the client fails the mount immediately it will try
1528 to clean up before the network times out the request. Add a
1529 reference from the request import to the obd device and delay
1530 the cleanup until the network drops the request.
1533 Frequency : occasionally during client (re)connect
1535 Description: assertion failure during client (re)connect
1536 Details : processing a client connection request may be delayed by the
1537 client or server longer than the client connect timeout. This
1538 causes the client to resend the connection request. If the
1539 original connection request is replied in this interval, the
1540 client may trip an assertion failure in ptlrpc_connect_interpret()
1541 which thought it would be the only running connect process.
1544 Frequency : only with obd_echo servers and clients that are rebooted
1546 Description: kernel BUG accessing uninitialized data structure
1547 Details : When running an obd_echo server it did not start the ping_evictor
1548 thread, and when a client was evicted an uninitialized data
1549 structure was accessed. Start the ping_evictor in the RPC
1550 service startup instead of the OBD startup.
1552 Severity : enhancement
1553 Bugzilla : 10193 (patchless)
1554 Description: Remove dependency on various unexported kernel interfaces.
1555 Details : No longer need reparent_to_init, exit_mm, exit_files,
1556 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1559 Frequency : rare (only users of deprecated and unsupported LDAP config)
1561 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1562 Details : LDAP apparently contains 'lustreName' attributes instead of
1563 'name'. A simple remapping of the name is sufficient.
1566 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1568 Description: waiting_locks_callback trips kernel BUG if client is evicted
1569 Details : Running with the dump_on_timeout debug flag turned on makes
1570 it possible that the waiting_locks_callback() can try to dump
1571 the Lustre kernel debug logs from an interrupt handler. Defer
1572 this log dumping to the expired_lock_main() thread.
1574 Severity : enhancement
1576 Description: Support NFS exporting on 2.6 kernels.
1577 Details : Implement non-rawops metadata methods for NFS server to use without
1578 changing NFS server code.
1581 Frequency : very rare (synthetic metadata workload only)
1583 Description: two racing renames might cause an MDS thread to deadlock
1584 Details : Running the "racer" program may cause one MDS thread to rename
1585 a file from being the source of a rename to being the target of
1586 a rename at exactly the same time that another thread is doing
1587 so, and the second thread has already enqueued these locks after
1588 doing a lookup of the target and is trying to relock them in
1589 order. Ensure that we don't try to re-lock the same resource.
1592 Frequency : only very large systems with liblustre clients
1594 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1595 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1596 all OSTs in parallel. This allows the eviction of stale liblustre
1597 clients to proceed much faster than if they were done in series,
1598 and also offers similar improvements for other set_info RPCs.
1603 Description: excessive CPU usage during initial read phase on client
1604 Details : During the initial read phase on a client, it would agressively
1605 retry readahead on the file, consuming too much CPU and impacting
1606 performance (since 1.4.5.8). Improve the readahead algorithm
1607 to avoid this, and also improve some other common cases (read
1608 of small files in particular, where "small" is files smaller than
1609 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1614 Description: MDS crash when receiving packet with unknown intent.
1615 Details : Do not LBUG in unknown intent case, just return -EFAULT
1617 Severity : enhancement
1618 Bugzilla : 9293, 9385
1619 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1620 Details : Do not serialize getattr (non-intent version) and statfs.
1623 Frequency : occasional, when OST network is overloaded/intermittent
1625 Description: client evicted by OST after bulk IO timeout
1626 Details : If a client sends a bulk IO request (read or write) the OST
1627 may evict the client if it is unresposive to its data GET/PUT
1628 request. This is incorrect if the network is overloaded (takes
1629 too long to transfer the RPC data) or dropped the OST GET/PUT
1630 request. There is no need to evict the client at all, since
1631 the pinger and/or lock callbacks will handle this, and the
1632 client can restart the bulk request.
1635 Frequency : Always when mmapping file with no objects
1637 Description: client crashes when mmapping file with no objects
1638 Details : Check that we actually have objects in a file before doing any
1639 operations on objects in ll_vm_open, ll_vm_close and
1645 Description: Request leak when working with deleted CWD
1646 Details : Introduce advanced request refcount tracking for requests
1647 referenced from lustre intent.
1649 Severity : Enhancement
1651 Description: Cache open file handles on client.
1652 Details : MDS now will return special lock along with openhandle, if
1653 requested and client is allowed to hold openhandle, even if unused,
1654 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1655 closing files for every read/write openration.
1657 Severity : Enhancement
1659 Description: Cache open negative dentries on client when possible.
1660 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1661 negative dentries on lock revocation.
1666 Description: Remounting a client read-only wasn't possible with a zconf mount
1667 Details : It wasn't possible to remount a client read-only with llmount.
1669 Severity : enhancement
1670 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1671 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1672 MPICH is included to add Lustre-specific ADIO interfaces.
1673 This is based closely on the UFS ADIO layer and only differs
1674 in file creation, in order to allow the OST striping to be set.
1675 This is user-contributed code and not supported by CFS.
1680 Description: extended inode attributes (immutable, append-only) work improperly
1681 when 2.4 and 2.6 kernels are used on client/server or vice versa
1682 Details : Introduce kernel-independent values for these flags.
1684 Severity : enhancement
1687 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1688 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1689 granularity. Now, Fractional MB granularity is be supported,
1690 this is very useful for low memory system.
1692 Severity : enhancement
1694 Description: Getattr by fid
1695 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1696 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1697 have locks and avoids getattr rpc requests for it on every stat.
1700 Frequency : Always, for filesystems larger than 2TB
1702 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1703 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1704 filesystems larger than 2TB to be mounted with mballoc enabled.
1707 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1709 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1710 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1711 filesystems larger than 2TB is mounted then the kernel will
1712 silently corrupt the start of the filesystem. CONFIG_LBD is
1713 enabled for all CFS-supported kernels, but the possibility of
1714 this happening with a modified kernel config exists.
1716 Severity : enhancement
1718 Description: add client O_DIRECT support for 2.6 kernels
1719 Details : It is now possible to do O_DIRECT reads and writes to files
1720 in the Lustre client mountpoint on 2.6 kernel clients.
1722 Severity : enhancement
1724 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1725 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1726 parallel, not waiting for response from every OST before sending
1727 a rpc to the next OST.
1732 Description: setattr vs write race when updating file timestamps
1733 Details : Client processes that update a file timestamp into the past
1734 right after writing to the file (e.g. tar) it is possible that
1735 the updated file modification time can be reset to the current
1736 time due to a race between processing the setattr and write RPC.
1738 Severity : enhancement
1740 Description: Bring 'lfs find' closer in line with regular Linux find.
1741 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1742 -print0 options and obtains all the needed info through the lustre
1745 Severity : enhancement
1747 Description: support up to 1024 configured devices on one node
1748 Details : change obd_dev array from statically allocated to dynamically
1749 allocated structs as they are first used to reduce memory usage
1754 Description: Flush dirty partially truncated pages during truncate
1755 Details : Immediatelly flush partially truncated pages in filter_setattr,
1756 this way we completely avoid having any pages in page cache on OST
1757 and can retire ugly workarounds during writes to flush such pages.
1762 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1763 Details : protect inode from truncation within vfs_unlink() context
1764 just take a reference before calling vfs_unlink() and release it
1765 when parent's i_sem is free.
1770 Description: last_id value checked outside lock on OST caused LASSERT failure
1771 Details : If there were multiple MDS->OST object precreate requests in
1772 flight, it was possible that the OST's last object id was checked
1773 outside a lock and incorrectly tripped an assertion. Move checks
1774 inside locks, and discard old precreate requests.
1777 Frequency : always, if extents are used on OSTs
1779 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1780 Details : OSTs using the extents format would not zero the high 16 bits of
1781 the index physical block number. This is not a problem for any
1782 OST filesystems smaller than 16TB, and no kernels support ext3
1783 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1784 new/modified files) and can be fixed for existing filesystems
1785 with e2fsprogs-1.39-cfs1.
1790 Description: import connection selection may be incorrect if timer wraps
1791 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1792 import connections to be ignored if the 32-bit jiffies counter
1793 wraps. Use a 64-bit jiffies counter.
1796 Frequency : very large clusters immediately after boot
1798 Description: LNET request buffers exhausted under heavy short-term load
1799 Details : If a large number of client requests are generated on a service
1800 that has previously never seen so many requests it is possible
1801 that the request buffer growth cannot keep up with the spike in
1802 demand. Instead of dropping incoming requests, they are held in
1803 the LND until the RPC service can accept more requests.
1806 Frequency : Sometimes during replay
1808 Description: Assertion failure in ll_local_open after replay.
1809 Details : If replay happened on an open request reply before we were able
1810 to set replay handler, reply will become not swabbed tripping the
1811 assertion in ll_local_open. Now we set the handler right after
1812 recognising of open request
1815 Frequency : very rare
1817 Description: kernel reports "badness in vsnprintf"
1818 Details : Reading from the "recovery_status" /proc file in small chunks
1819 may cause a negative length in lprocfs_obd_rd_recovery_status()
1820 call to vsnprintf() (which is otherwise harmless). Exit early
1821 if there is no more space in the output buffer.
1823 Severity : enhancement
1825 Description: clear OBD RPC statistics by writing to them
1826 Details : It is now possible to clear the OBD RPC statistics by writing
1827 to the "stats" file.
1832 Description: Client mtime is not the same on different clients after utimes
1833 Details : In some cases, the client was using the utimes() syscall on
1834 a file cached on another node. The clients now validate the
1835 ctime from the MDS + OSTs to determine which one is right.
1840 Description: Inability to activate failout mode
1841 Details : lconf script incorrectly assumed that in pythong string's numeric
1842 value is used in comparisons.
1845 Frequency : always with multiple stripes per file
1847 Description: Inefficient object allocation for mutli-stripe files
1848 Details : When selecting which OSTs to stripe files over, for files with
1849 a stripe count that divides evenly into the number of OSTs,
1850 the MDS is always picking the same starting OST for each file.
1851 Return the OST selection heuristic to the original design.
1856 Description: mount failures may take full timeout to return an error
1857 Details : Under some heavy load conditions it is possible that a
1858 failed mount can wait for the full obd_timeout interval,
1859 possibly several minutes, before reporting an error.
1860 Instead return an error as soon as the status is known.
1862 Frequency : quota enabled and large files being deleted
1864 Description: releasing more than 4GB of quota at once hangs OST
1865 Details : If a user deletes more than 4GB of files on a single OST it
1866 will cause the OST to spin in an infinite loop. Release
1867 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1872 Description: statfs data retrieved from /proc may be stale or zero
1873 Details : When reading per-device statfs data from /proc, in the
1874 {kbytes,files}_{total,free,avail} files, it may appear
1875 as zero or be out of date.
1878 Frequency : systems with MD RAID1 external journal devices
1880 Description: lconf's call to blkid is confused by RAID1 journal devices
1881 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1882 of returning all block devices that match the journal UUID.
1885 Frequency : always, for aggregate stripe size over 4GB
1887 Description: assertion fails when trying to use 4GB stripe size
1888 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1889 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1892 Frequency : always on ppc64
1894 Description: the first write on an ext3 filesystem with mballoc got stuck
1895 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1896 perform endianness conversion.
1898 ------------------------------------------------------------------------------
1900 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1902 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1903 INTEROPERATE with older versions automatically. Please read the
1904 user documentation before upgrading any part of a live system.
1905 * WARNING: Lustre networking configuration changes are required with
1906 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1909 * Support for kernels:
1910 2.6.9-22.0.2.EL (RHEL 4)
1911 2.6.5-7.244 (SLES 9)
1912 2.6.12.6 vanilla (kernel.org)
1915 Severity : enhancement
1916 Bugzilla : 7981/8208
1917 Description: Introduced Lustre Networking (LNET)
1918 Details : LNET is new networking infrastructure for Lustre, it includes
1919 a reorganized network configuration mode (see the user
1920 documentation for full details) as well as support for routing
1921 between different network fabrics. Lustre Networking Devices
1922 (LNDs) for the supported network fabrics have also been
1923 created for this new infrastructure.
1925 Severity : enhancement
1926 Description: Introduced Access control lists
1927 Details : clients can set ACLs on files and directories in order to have
1928 more fine-grained permissions than the standard Unix UGO+RWX.
1929 The MDS must be started with the "-o acl" mount option.
1931 Severity : enhancement
1932 Description: Introduced filesystem quotas
1933 Details : Administrators may now establish per-user quotas on the
1936 Severity : enhancement
1938 Description: Configuration change for the XT3
1939 The PTLLND is now used to run Lustre over Portals on the XT3
1940 The configure option(s) --with-cray-portals are no longer used.
1941 Rather --with-portals=<path-to-portals-includes> is used to
1942 enable building on the XT3. In addition to enable XT3 specific
1943 features the option --enable-cray-xt3 must be used.
1948 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1949 Details : A race between allocating a new llcd and re-getting the llcd_lock
1950 allowed another thread to grab newly-allocated llcd.
1952 Severity : enhancement
1954 Description: 2.6 OST async journal commit and locking fix to improve performance
1955 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1956 2.6 kernels are now async as they already were in 2.4 kernels so
1957 that they can commit concurrently with the network bulk transfer.
1958 For block-allocated files the filter allocation semaphore is held
1959 to avoid filesystem fragmentation during allocation. BKL lock
1960 removed for 2.6 xattr operations where it is no longer needed.
1965 Description: lconf incorrectly determined whether two IP networks could talk
1966 Details : In some more complicated routing and multiple-network
1967 configurations, lconf will avoid trying to make a network
1968 connection to a disjoint part of the IP space. It was doing the
1969 math incorrectly for one set of cases.
1974 Description: Fix for potential infinite loop processing records in an llog.
1975 Details : If an llog record is corrupted/zeroed, it is possible to loop
1976 forever in llog_process(). Validate the llog record length
1977 and skip the remainder of the block on error.
1980 Frequency : occasional (liblustre only)
1982 Description: liblustre could not open files whose last component is a symlink
1983 Details : sysio_path_walk() would incorrectly pass the open intent to
1984 intermediate path components.
1987 Frequency : rare (liblustre only with non-standard tuning)
1988 Bugzilla : 7201 (7350)
1989 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1990 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1991 after it has been released, only do this for referenced locks
1993 Severity : enhancement
1995 Description: specifying an (invalid) directory default stripe_size of -1
1996 would reset the directory default striping
1997 Details : stripe_size -1 was used internally to signal directory stripe
1998 removal, now use "all default" to signal dir stripe removal
1999 as a directory striping of "all default" is not useful
2002 Frequency : common for large clusters running liblustre clients
2004 Description: doing an ls when liblustre clients are running is slow
2005 Details : sending a glimpse AST to a liblustre client waits for every AST
2006 to time out, as liblustre clients will not respond. Since they
2007 cannot cache data we refresh the OST lock LVB from disk instead.
2009 Severity : enhancement
2011 Description: doing an ls at the same time as file IO can be slow
2012 Details : enqueue and other "small" requests can be blocked behind many
2013 large IO requests. Create a new OST IO portal for non-IO
2014 requests so they can be processed faster.
2017 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2019 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2020 created file with mode 000
2021 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2022 allow an ftruncate() as a non-root user to a file with mode 000.
2023 Lustre now respects this flag to disable mode checks when
2024 truncating a file owned by the user
2027 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2030 Description: Revoking locks from clients that went dead or catatonic might take
2032 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2033 cancellation of such locks instant on servers without waiting for
2034 any reply from clients. Clients drops these locks when cancel
2035 notification from server is received without replying.
2038 Frequency : liblustre-only, when liblustre client dies or becomes busy
2040 Description: Doing ls on Linux clients can take a long time with active
2042 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2043 granting such locks to it in the first place if possible. Locks
2044 are taken by proxy on the OST during the read or write and
2045 dropped immediately afterward. Add connect flags handling, do
2046 not grant locks to liblustre clients for glimpse ASTs.
2048 Severity : enhancement
2050 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2051 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2052 sequential reads and behaves badly with applications doing random
2053 reads. Improve it by reading ahead at least read region, and
2054 avoiding excessive large RPC for small reads.
2056 Severity : enhancement
2058 Description: Creating more than 1000 files for a single job may cause a load
2059 imbalance on the OSTs if there are also a large number of OSTs.
2060 Details : qos_prep_create() uses an OST index reseed value that is an
2061 even multiple of the number of available OSTs so that if the
2062 reseed happens in the middle of the object allocation it will
2063 still utilize the OSTs as uniformly as possible.
2068 Description: OST or MDS may oops in ping_evictor_main()
2069 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2070 but doesn't restart at beginning of obd_exports_timed list
2073 Severity : enhancement
2075 Description: improve by-nid export eviction on the MDS and OST
2076 Details : allow multiple exports with the same NID to be evicted at one
2077 time without re-searching the exports list.
2080 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2082 Description: MDS may oops in groups_free()
2083 Details : in rare race conditions a newly allocated group_info struct is
2084 freed again, and this can be NULL. The 2.4 compatibility code
2085 for groups_free() checked for a NULL pointer, but 2.6 did not.
2088 Frequency : common for liblustre clients doing little filesystem IO
2089 Bugzilla : 9352, 7313
2090 Description: server may evict liblustre clients accessing contended locks
2091 Details : if a client is granted a lock or receives a completion AST
2092 with a blocking AST already set it would not reply to the AST
2093 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2096 Frequency : lfs setstripe, only systems with more than 160 OSTs
2098 Description: unable to set striping with a starting offset beyond OST 160
2099 Details : llapi_create_file() incorrectly limited the starting stripe
2100 index to the maximum single-file stripe count.
2103 Frequency : LDAP users only
2105 Description: lconf did not handle in-kernel recovery with LDAP properly
2106 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2108 Severity : enhancement
2110 Description: bind OST threads to NUMA nodes to improve performance
2111 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2112 node and do their allocations there to localize memory access
2114 Severity : enhancement
2116 Description: llmount can determine client NID directly from Myrinet (GM)
2117 Details : the client NID code from gmnalnid was moved directly into
2118 llmount, removing the need to use this or specifying the
2119 client NID explicitly when mounting GM clients with zeroconf
2122 Frequency : if client is started with down MDS
2124 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2125 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2126 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2127 ptlrpc_import_delayed_req() because we didn't check if the
2128 request was interrupted, and we also didn't break out of the
2129 event loop if there was no timeout
2134 Description: data loss during non-page-aligned writes to a single file from
2135 both multiple nodes and multiple threads on one node at same time
2136 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2137 inconsistency led to false short-reads, that were cached and later
2138 used by ->prepare_write() to fill in partially written page,
2139 leading to data loss.
2142 Frequency : always, if lconf --abort_recovery used
2144 Description: lconf --abort_recovery fails with 'Operation not supported'
2145 Details : lconf was attempting to abort recovery on the MDT device and not
2148 Severity : enhancement
2150 Description: remove cleanup logs
2151 Details : replace lconf-generated cleanup logs with lustre internal
2152 cleanup routines. Eliminates the need for client-cleanup and
2155 Severity : enhancement
2157 Description: add support for EAs (user and system) on lustre filesystems
2158 Details : it is now possible to store extended attributes in the Lustre
2159 client filesystem, and with the user_xattr mount option it
2160 is possible to allow users to store EAs on their files also
2162 Severity : enhancement
2164 Description: Add possibility (config option) to show minimal available OST free
2166 Details : When compiled with --enable-mindf configure option, statfs(2)
2167 (and so, df) will return least minimal free space available from
2168 all OSTs as amount of free space on FS, instead of summary of
2169 free spaces of all OSTs.
2171 Severity : enhancement
2173 Description: do not expand extent locks acquired on OST-side
2174 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2175 by server: they are not cached anyway.
2178 Frequency : when mmap is used/binaries executed from Lustre
2180 Description: Unmmap pages before throwing them away from read cache.
2181 Details : llap_shrink cache now attempts to unmap pages before discarding
2182 them (if unmapping failed - do not discard). SLES9 kernel has
2183 extra checks that trigger if this unmapping is not done first.
2188 Description: lconf didn't resolve symlinks before checking to see whether a
2189 given mountpoint was already in use
2192 Frequency : when migrating failover services
2193 Bugzilla : 6395, 9514
2194 Description: When migrating a subset of services from a node (e.g. failback
2195 from a failover service node) the remaining services would
2196 time out and evict clients.
2197 Details : lconf --force (implied by --failover) sets the global obd_timeout
2198 to 5 seconds in order to quickly disconnect, but this caused
2199 other RPCs to time out too quickly. Do not change the global
2200 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2202 Severity : enhancement
2203 Frequency : if MDS is started with down OST
2204 Bugzilla : 9439,5706
2205 Description: Allow startup/shutdown of an MDS without depending on the
2206 availability of the OSTs.
2207 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2208 Add appropriate locking and lov-osc refcounts for safe
2209 cleaning. Add osc abort_inflight calls in case the
2210 synchronize never started.
2213 Frequency : occasional (Cray XT3 only)
2215 Description: root not authorized to access files in CRAY_PORTALS environment
2216 Details : The client process capabilities were not honoured on the MDS in
2217 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2218 been accessed by an authorized user then root was able to access
2219 the file on the local client also. The root user capabilities
2220 are now allowed on the MDS, as this environment has secure UID.
2223 Frequency : occasional
2225 Description: ldiskfs "too long searching" message happens too often
2226 Details : A debugging message (otherwise harmless) prints too often on
2227 the OST console. This has been reduced to only happen when
2228 there are fragmentation problems on the filesystem.
2233 Description: Division by zero in statfs when all OSCs are inactive
2234 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2235 lov_statfs divides by value returned by lov_get_stripecnt().
2239 Bugzilla : 9489, 3273
2240 Description: First write from each client to each OST was only 4kB in size,
2241 to initialize client writeback cache, which caused sub-optimal
2242 RPCs and poor layout on disk for the first writen file.
2243 Details : Clients now request an initial cache grant at (re)connect time
2244 and so that they can start streaming writes to the cache right
2245 away and always do full-sized RPCs if there is enough data.
2246 If the OST is rebooted the client also re-establishes its grant
2247 so that client cached writes will be honoured under the grant.
2252 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2253 Details : Now I/O RPCs go to different portal number and (presumably) fast
2254 lock requests (and glimses) and other RPCs get their own service
2255 threads pool that should be able to service those RPCs
2258 Severity : enhancement
2260 Description: Ability to exchange lustre version between client and servers and
2261 issue warnings at client side if client is too old. Also for
2262 liblustre clients there is ability to refuse connection of too old
2264 Details : New 'version' field is added to connect data structure that is
2265 filled with version info. That info is later checked by server and
2269 Frequency : rare, liblustre only.
2270 Bugzilla : 9296, 9581
2271 Description: Two simultaneous writes from liblustre at offset within same page
2272 might proceed at the same time overwriting eachother with stale
2274 Details : I/O lock withing llu_file_prwv was released too early, before data
2275 actually was hitting the wire. Extended lock-holding time until
2276 server acknowledges receiving data.
2279 Frequency : extremely rare. Never observed in practice.
2281 Description: avoid generating lustre_handle cookie of 0.
2282 Details : class_handle_hash() generates handle cookies by incrementing
2283 global counter, and can hit 0 occasionaly (this is unlikely, but
2284 not impossible, because initial value of cookie counter is
2285 selected randonly). Value of 0 is used as a sentinel meaning
2286 "unassigned handle" --- avoid it. Also coalesce two critical
2287 sections in this function into one.
2289 Severity : enhancement
2291 Description: allow liblustre clients to delegate truncate locking to OST
2292 Details : To avoid overhead of locking, liblustre client instructs OST to
2293 take extent lock in ost_punch() on client's behalf. New connection
2294 flag is added to handle backward compatibility.
2296 Severity : enhancement
2297 Bugzilla : 4928, 7341, 9758
2298 Description: allow number of OST service threads to be specified
2299 Details : a module parameter allows the number of OST service threads
2300 to be specified via "options ost ost_num_threads={N}" in the
2301 OSS's /etc/modules.conf or /etc/modprobe.conf.
2305 Bugzilla : 6146, 9635, 9895
2306 Description: servers crash with bad pointer in target_handle_connect()
2307 Details : In rare cases when a client is reconnecting it was possible that
2308 the connection request was the last reference for that export.
2309 We would temporarily drop the export reference and get a new
2310 one, but this may have been the last reference and the export
2311 was just destroyed. Get new reference before dropping old one.
2313 Severity : enhancement
2314 Frequency : if client is started with failover MDS
2316 Description: Allow multiple MDS hostnames in the mount command
2317 Details : Try to read the configuration from all specified MDS
2318 hostnames during a client mount in case the "primary"
2321 Severity : enhancement
2323 Description: Stop sending data to evicted clients as soon as possible.
2324 Details : Check if the client we are about to send or are sending data to
2325 was evicted already. (Check is done every second of waiting,
2326 for which l_wait_event interface was extended to allow checking
2327 of exit condition at specified intervals).
2330 Frequency : rare, normally only when NFS exporting is done from client
2332 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2334 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2337 Frequency : always, if srand() is called before liblustre initialization
2339 Description: Liblustre uses system PRNG disturbing its usage by user application
2340 Details : Introduce internal to lustre fast and high-quality PRNG for
2341 lustre usage and make liblustre and some other places in generic
2342 lustre code to use it.
2344 Severity : enhancement
2345 Bugzilla : 9477, 9557, 9870
2346 Description: Verify that the MDS configuration logs are updated when xml is
2347 Details : Check if the .xml configuration logs are newer than the config
2348 logs stored on the MDS and report an error if this is the case.
2349 Request --write-conf, or allow starting with --old_conf.
2351 Severity : enhancement
2353 Description: Handle symlinks in the path when checking if Lustre is mounted.
2354 Details : Resolve intermediate symlinks when checking if a client has
2355 mounted a filesystem to avoid duplicate client mounts.
2360 Description: lconf can hit an error exception but still return success.
2361 Details : The lconf command catches the Command error exception at the top
2362 level script context and will exit with the associated exit
2363 status, but doesn't ensure that this exit status is non-zero.
2368 Description: failure of ptlrpc thread startup can cause oops
2369 Details : Starting a ptlrpc service thread can fail if there are a large
2370 number of threads or the server memory is very fragmented.
2371 Handle this without oopsing.
2374 Frequency : always, only if liblustre and non-default acceptor port was used
2376 Description: liblustre cannot connect to servers with non-default acceptor port
2377 Details : tcpnal_set_default_params() was not called and was therefore
2378 ignoring the environment varaible TCPNAL_PORT, as well as other
2379 TCPNAL_ environment variables
2384 Description: two objects could be created on the same OST for a single file
2385 Details : If an OST is down, in some cases it was possible to create two
2386 objects on a single OST for a single file. No problems other
2387 than potential performance impact and spurious error messages.
2391 Bugzilla : 5681, 9562
2392 Description: Client may oops in ll_unhash_aliases
2393 Details : Client dcache may become inconsistent in race condition.
2394 In some cases "getcwd" can fail if the current directory is
2400 Description: Inode refcounting problems in NFS export code
2401 Details : link_raw functions used to call d_instantiate without obtaining
2402 extra inode reference first.
2406 Bugzilla : 9942, 9903
2407 Description: Referencing freed requests leading to crash, memleaks with NFS.
2408 Details : We used to require that call to ll_revalidate_it was always
2409 followed by ll_lookup_it. Also with revalidate_special() it is
2410 possible to call ll_revalidate_it() twice for the same dentry
2411 even if first occurence returned success. This fix changes semantic
2412 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2413 reference on a request referred from the intent.
2414 ll_intent_release() then releases such a request.
2417 Frequency : rare, normally benchmark loads only
2419 Description: unlinked inodes were kept in memory on the client
2420 Details : If a client is repeatedly creating and unlinking files it
2421 can accumulate a lot of stale inodes in the inode slab cache.
2422 If there is no other client load running this can cause the
2423 client node to run out of memory. Instead flush old inodes
2424 from client cache that have the same inode number as a new inode.
2427 Frequency : SLES9 2.6.5 kernel and long filenames only
2428 Bugzilla : 9969, 10379
2429 Description: utime reports stale NFS file handle
2430 Details : SLES9 uses out-of-dentry names in some cases, which confused
2431 the lustre dentry revalidation. Change it to always use the
2435 Frequency : rare, unless heavy write-truncate concurrency is continuous
2436 Bugzilla : 4180, 6984, 7171, 9963, 9331
2437 Description: OST becomes very slow and/or deadlocked during object unlink
2438 Details : filter_destroy() was holding onto the parent directory lock
2439 while truncating+unlinking objects. For very large objects this
2440 may block other threads for a long time and slow overall OST
2441 responsiveness. It may also be possible to get a lock ordering
2442 deadlock in this case, or run out of journal credits because of
2443 the combined truncate+unlink. Solution is to do object truncate
2444 first in one transaction without parent lock, and then do the
2445 final unlink in a new transaction with the parent lock. This
2446 reduces the lock hold time dramatically.
2449 Frequency : rare, 2.4 kernels only
2451 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2452 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2453 before unmounting their filesystem in order to avoid deadlock.
2454 The kernel_locked() function in 2.4 kernels only checks whether
2455 the kernel lock is held, not whether it is this process that is
2456 holding it as 2.6 kernels do.
2461 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2462 Details : The client ptlrpc code may be trying to reconnect to a down
2463 server before a previous connection attempt has timed out.
2464 Increase the reconnect interval to be longer than the connection
2465 timeout interval to avoid sending duplicate connections to
2469 Frequency : echo_client brw_test command
2471 Description: fix echo_client to work with OST preallocated code
2472 Details : OST preallocation code (5137) didn't take echo_client IO path
2473 into account: echo_client calls filter methods outside of any
2474 OST thread and, hence, there is no per-thread preallocated
2475 pages and buffers to use. Solution: hijack pga pages for IO. As
2476 a byproduct, this avoids unnecessary data copying.
2480 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2481 Description: Client can oops in mdc_commit_close() after open replay
2482 Details : It was possible for the MDS to return an open request with no
2483 transaction number in mds_finish_transno() if the client was
2484 evicted, but without actually returning an error. Clients
2485 would later try to replay that open and may trip an assertion
2486 Simplify the client close codepath, and always return an error
2487 from the MDS in case the open is not successful.
2490 Frequency : rare, 2.6 OSTs only
2492 Description: OST may deadlock under high load on fragmented files
2493 Details : If there was a heavy load and highly-fragmented OST filesystems
2494 it was possible to have all the OST threads deadlock waiting on
2495 allocation of biovecs, because the biovecs were not released
2496 until the entire RPC IO was completed. Instead, release biovecs
2497 as soon as they are complete to ensure forward IO progress.
2499 Severity : enhancement
2501 Description: Support for specifying external journal device at mount
2502 Details : If an OST or MDS device is formatted with an external journal
2503 device, this device major/minor is stored in the ext3 superblock
2504 and may not be valid for failover. Allow detecting and
2505 specifying the external journal at mount time.
2510 Description: Mounting an MDS with pending unlinked files may cause oops
2511 Details : target_finish_recovery() calls mds_postrecov() which returned
2512 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2513 considers this an error and immediately begins cleaning up the
2514 lov, just after starting the mds_lov process
2516 Severity : enhancement
2518 Description: Implement 'lfs df' to report actual free space on per-OST basis
2519 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2520 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2521 usage of objects; '-h' to report in human readable format.
2523 ------------------------------------------------------------------------------
2525 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2532 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2533 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2534 at startup during orphan unlinking. Instead, skip buddy update
2535 before setup, as it will be regenerated after recovery is complete.
2540 Description: If an OST is inactive, its locks might reference stale inodes.
2541 Details : lov_change_cbdata() must iterate over all namespaces, even if
2542 they are inactive to clear inode references from the lock.
2544 Severity : enhancement
2545 Frequency : occasional, if non-standard max_dirty_mb used
2547 Description: Client will block write RPCs if not enough grant
2548 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2549 then the client will block writes while waiting for another RPC
2550 to complete instead of consuming its dirty limit. With change
2551 we get improved performance when max_dirty_mb is small.
2553 Severity : enhancement
2554 Bugzilla : 3389, 6253
2555 Description: Add support for supplementary groups on the MDS.
2556 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2557 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2558 lookups for user supplementary groups into a cache.
2562 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2563 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2564 for liblustre O_EXCL is enforced by MDS.
2568 Description: readdir never returns NULL in liblustre.
2569 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2570 dentry in struct dirent.
2574 Frequency : liblustre only, depends on application IO pattern
2575 Description: liblustre clients evicted if not contacting servers
2576 Details : Don't put liblustre clients into the ping_evictor list, so
2577 they will not be evicted by the pinger ever.
2579 Severity : enhancement
2581 Description: Add ability to evict clients by NID from MDS.
2582 Details : By echoing "nid:$NID" string into
2583 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2584 $NID would be instantly evicted from this MDS and from all active
2585 OSTs connected to it.
2589 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2590 Details : lookup_it_finish() used to query file size from OSTs that was not
2595 Description: service threads change working directory to that of init
2596 Details : Starting lustre service threads may pin the working directory
2597 of the parent thread, making that filesystem busy. Threads
2598 now change to the working directory of init to avoid this.
2602 Frequency : during shutdown only
2603 Description: shutdown with a failed MDS or OST can cause unmount to hang
2604 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2609 Frequency : 2.6 only
2610 Description: chmod/chown may include an extra supplementary group
2611 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2612 supplementary group and if none is specified this is used.
2615 Bugzilla : 5479 (6816)
2617 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2618 Details : If lookup is in progress on a file that is unlinked we might try
2619 to revalidate the inode and fail in revalidate after lookup is
2620 complete and ll_file_open() enqueues the open again but
2621 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2624 Frequency : always, if lconf --abort_recovery used
2626 Description: lconf --abort_recovery fails with 'Operation not supported'
2627 Details : lconf was attempting to abort recovery on the MDT device and not
2630 ------------------------------------------------------------------------------
2632 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2637 Frequency : rare (only unsupported configurations with a node running as an
2639 Bugzilla : 6514, 5137
2640 Description: Mounting a Lustre file system on a node running as an OST could
2642 Details : OSTs now preallocates memory needed to write out data at
2643 startup, instead of when needed, to avoid having to
2644 allocate memory in possibly low memory situations.
2645 Specifically, if the file system is mounted on on OST,
2646 memory pressure could force it to try to write out data,
2647 which it needed to allocate memory to do. Due to the low
2648 memory, it would be unable to do so and the node would
2649 become unresponsive.
2651 Severity : enhancement
2653 Description: Addition of lconf --service command line option
2654 Details : lconf now accepts a '--service <arg>' option, which is
2655 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2657 Severity : enhancement
2659 Description: Failover mode is now the default for OSTs.
2660 Details : By default, OSTs will now run in failover mode. To return to
2661 the old behaviour, add '--failout' to the lmc line for OSTs.
2663 Severity : enhancement
2665 Description: Health checks are now provided for MDS and OSTs
2666 Details : Additional detailed health check information on MSD and OSTs
2667 is now provided through the procfs health_check value.
2670 Frequency : occasional, depends on IO load
2672 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2673 after numerous create/delete cycles
2674 Details : The ext3 inode allocation policy would not allocate new inodes
2675 very well on the OSTs because there are no new directories
2676 being created. Instead we look for groups with free space if
2677 the parent directories are nearly full.
2682 Description: Network or server problems during mount may cause partially
2683 mounted clients instead of returning an error.
2684 Details : The config llog parsing code may overwrite the error return
2685 code during mount error handling, returning success instead
2691 Description: MDS can fail to allocate large reply buffers
2692 Details : After long uptimes the MDS can fail to allocate large reply
2693 buffers (e.g. zconf client mount config records) due to memory
2694 fragmentation or consumption by the buffer cache. Preallocate
2695 some large reply buffers so that these replies can be sent even
2696 under memory pressure.
2700 Frequency : rare (liblustre)
2701 Description: fsx running with liblustre complained that using truncate() to
2702 extend the file doesn't work. This patch corrects that issue.
2703 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2704 ATTR_SIZE and lsm use in llu_setattr_raw.
2708 Frequency : rare, only 2.6 kernels
2709 Description: Unusual file access patterns on the MDS may result in inode
2710 data being lost in very rare circumstances.
2711 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2712 code in the 2.6 kernel caused the inode and/or EA data not to
2713 be read from disk, causing single-file corruption.
2717 Frequency : rare, only 2.6 filesystems using extents
2718 Description: Heavy concurrent write and delete load may cause data corruption.
2719 Details : It was possible under high-load situations to have an extent
2720 metadata block in the block device cache from a just-unlinked
2721 file overwrite a newly-allocated data block. We now unmap any
2722 metadata buffers that alias just-allocated data blocks.
2726 Frequency : filesystems with default stripe_count larger than 77
2727 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2728 Details : lconf specifies an inode size of 4096 bytes when the default
2729 stripe_count is larger than 77. This conflicts with the default
2730 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2731 this case to avoid pinning too much memory for large EAs.
2733 ------------------------------------------------------------------------------
2735 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2740 Frequency : rare (extremely heavy IO load with hundreds of clients)
2742 Description: Client is evicted, gets IO error writing to file
2743 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2744 caused clients to be evicted for AST timeouts. The fixes in
2745 bug 5192 mean we no longer need to have such short AST timeouts
2746 so ldlm_timeout has been increased.
2749 Frequency : occasional during --force or --failover shutdown under load
2750 Bugzilla : 5949, 4834
2751 Description: Server oops/LBUG if stopped with --force or --failover under load
2752 Details : a collection of import/export refcount and cleanup ordering
2753 issues fixed for safer force cleanup
2756 Frequency : only filesystems larger than 120 OSTs
2757 Bugzilla : 5990, 6223
2758 Description: lfs getstripe would oops on a very large filesystem
2759 Details : lov_getconfig used kfree on vmalloc'd memory
2762 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2763 Bugzilla : 6242, 6243
2764 Description: reading from files that had been truncated to a non-zero size
2765 but never opened returned no data
2766 Details : ll_file_read() reads zeros from no-object files to EOF
2771 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2772 Details : There's little harm in aborting MDS/OSS recovery and letting it
2773 try again, so I removed the LASSERT and return an error instead.
2775 Severity : enhancement
2777 Description: New debugging infrastructure for tracking down data corruption
2778 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2779 (b) cover more of the client-side code path, and (c) try to narrow
2780 down where problems occurred
2784 Bugzilla : 3819, 4364, 4397, 6313
2785 Description: Racing close and eviction MDS could cause assertion in mds_close
2786 Details : It was possible to get multiple mfd references during close and
2787 client eviction, leading to one thread referencing a freed mfd.
2789 Severity: : enhancement
2790 Bugzilla : 3262, 6359
2791 Description: Attempts to reconnect to servers are now more aggressive.
2792 Details : This builds on the enhanced upcall-less recovery that was added
2793 in 1.4.2. When trying to reconnect to servers, clients will
2794 now try each server in the failover group every 10 seconds. By
2795 default, clients would previously try one server every 25 seconds.
2800 Description: After recovery, certain operations trigger a failed
2801 assertion on a client.
2802 Details : Failing over an mds, using lconf -d --failover, while a
2803 client was doing a readdir() call would cause the client to
2804 LBUG after recovery completed and the readdir() was resent.
2806 Severity : enhancement
2808 Description: Default groups are now added by lconf
2809 Details : You can now run lconf --group <servicename> without having to
2810 manually add groups with lmc.
2813 Frequency : occasional
2815 Description: Nodes with an elan id of 0 trigger a failed assertion
2818 Frequency : always when accessing e.g. tty/console device nodes
2820 Description: tty and some other devices nodes cannot be used on lustre
2821 Details : file's private_data field is used by device data and lustre
2822 values in there got lost. New field was added to struct file to
2823 store fs-specific private data.
2826 Frequency : when exporting Lustre via NFS
2828 Description: NFSD failed occasionally when looking up a path component
2829 Details : NFSD is looking up ".." which was broken in ext3 directories
2830 that had grown large enough to become hashed.
2833 Frequency : Clusters with multiple interfaces not on the same subnet
2835 Description: Nodes will repeatedly try to reconnect to an interface which it
2836 cannot reach and report an error to the log.
2837 Details : Extra peer list entries will be created by lconf with some peers
2838 unreachable. lconf now validates the peer before adding it.
2841 Frequency : Only if a default stripe is set on the filesystem root.
2843 Description: Setting a default stripe on the filesystem root prevented the
2844 filesystem from being remounted.
2845 Details : The client was sending extra request flags in the root getattr
2846 request and did not allocate a reply buffer for the dir EA.
2849 Frequency : occasional, higher if lots of files are accessed by one client
2850 Bugzilla : 6159, 6097
2851 Description: Client trips assertion regarding lsm mismatch/magic
2852 Details : While revalidating inodes the VFS looks up inodes with ifind()
2853 and in rare cases can find an inode that is being freed.
2854 The ll_test_inode() code will free the lsm during ifind()
2855 when it finds an existing inode and then the VFS later attaches
2856 this free lsm to a new inode.
2860 Bugzilla : 6422, 7030
2861 Description: MDS deadlock between mkdir and client eviction
2862 Details : Creating a new file via mkdir or mknod (starting a transaction
2863 and getting the ns lock) can deadlock with client eviction
2864 (gets ns lock and trying to finish a synchronous transaction).
2867 Frequency : occasional
2868 Description: While starting a server, the fsfilt_ext3 module could not be
2870 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2871 kernels. Previously, lconf would still use the ext3 name
2872 when trying to load modules. Now, it will correctly use
2873 ext3 on 2.4 and ldiskfs on 2.6.
2875 Severity : enhancement
2876 Description: The default stripe count has been changed to 1
2877 Details : The interpretation of the default stripe count (0, to lfs
2878 or lmc) has been changed to mean striping across a single
2879 OST, rather than all available. For general usage we have
2880 found a stripe count of 1 or 2 works best.
2882 Severity : enhancement
2883 Description: Add support for compiling against Cray portals.
2884 Details : Conditional compiling for some areas that are different
2888 Frequency : occasional
2889 Bugzilla : 6409, 6834
2890 Description: Creating files with an explicit stripe count may lead to
2891 a failed assertion on the MDS
2892 Details : If some OSTs are full or unavailable, creating files may
2893 trigger a failed assertion on the MDS. Now, Lustre will
2894 try to use other servers or return an error to the
2898 Frequency : occasional
2900 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2901 caused lock timeout during readahead (since 1.4.2).
2902 Details : Processes doing readahead might match a lock that hasn't been
2903 granted yet if there are overlapping and conflicting lock
2904 requests. The readahead process waits on ungranted lock
2905 (original lock is CBPENDING), while OST waits for that process
2906 to cancel CBPENDING read lock and eventually evicts client.
2908 Severity : enhancement
2910 Description: Initial enabling of flock support for clients
2911 Details : Implements fcntl advisory locking and file status functions.
2912 This feature is provided as an optional mount flag (default
2913 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2914 locking are implemented yet, and those that are are not guaranteed
2915 to be completely correct in production environments.
2916 mount -t lustre -o [flock|noflock] ...
2919 Frequency : occasional
2921 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2922 assertion in the ext3 JBD (journaling) layer.
2923 Details : The b_committed_data struct is protected by the big kernel lock
2924 in 2.4 kernels, serializing journal_commit_transaction() and
2925 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2926 access to this struct. In 2.6 kernels there is finer grained
2927 locking to improve SMP performance of the JBD layer.
2931 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2936 Description: merge and cleanup kernel patches.
2939 -----------------------------------------------------------------------------
2941 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2943 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2944 versions of Lustre. Please update all servers and clients to
2945 version 1.4.2 or later at the same time. You must also run
2946 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2947 to update the configuration logs.
2949 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2950 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2951 - swab llog records of type '0' so we get proper header size/idx (5861)
2952 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2953 - fix rename of one directory over another leaking an inode (5953)
2954 - avoid SetPageDirty on 2.6 (5981)
2955 - don't re-add just-being-destroyed locks to the waiting list (5653)
2956 - when creating new directories, inherit the parent's custom
2957 striping settings if present parent (3048)
2958 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2959 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2960 - don't hold client locks on temporary worklist from l_lru (5666)
2961 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2962 - automatically evict dead clients (5921)
2963 - Update file size properly in create+truncate+fstat case (6196)
2964 - Do not unhash mountpoint dentries, do not allow removal of
2966 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2967 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2968 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2969 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2970 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2971 - Correctly handle reads of files with no objects (6243)
2972 - lctl recover will also mark a device active if deactivate used (5933)
2974 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2975 - allow --write-conf on an MDS with different nettype than client (5619)
2976 - don't write config llogs to MDS for mounts not from that MDS (5617)
2977 - lconf should create multiple TCP connections from a client (5201)
2978 - init scripts are now turned off by default; run chkconfig --on
2979 lustre and chkconfig --on lustrefs to use them
2980 - upcalls are no longer needed for clients to recover to failover
2982 - add --abort-recovery option to lconf to abort recovery on device
2984 - add support for an arbitrary number of OSTs (3026)
2985 - Quota support protocol changes.
2986 - forward compatibility changes to wire structs (6007)
2987 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2988 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2989 - improved "lustre status" script
2990 - initialize blocksize for non-regular files (6062)
2991 - added --disable-server and --disable-client configure options (5782)
2992 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2993 - Vanilla 2.4.29 support
2994 - increase maximum number of obd devices to 520 (6242)
2995 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2996 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2997 SLES 9 2.6.5 kernels
2998 - limit stripes per file to 160 (the maximum EA size) (6093)
3000 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3003 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3004 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3005 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3006 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3007 - small ext3 extents cleanups and fixes (5733)
3008 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3009 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3010 - avoid needless client->OST connect, fix handle mismatch (5317)
3011 - fix DLM error path that led to out-of-sync client, long delays (5779)
3012 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3013 - fix several locking issues related to i_size (5492,5624,5654,5672)
3014 - don't move pending lock onto export if it is already evicted (5683)
3015 - fix kernel oops when creating .foo in unlinked directory (5548)
3016 - fix deadlock in obdfilter statistics vs. object create (5811)
3017 - use time_{before,after} to avoid timer jiffies wrap (5882)
3018 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3019 - Do not leak request if server was not able to process it (5154)
3020 - If mds_open unable to find parent dir, make that negative lookup(5154)
3021 - don't create new directories with extent-mapping (5909, 5936)
3023 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3024 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3025 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3026 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3027 - support for RHEL4 kernel on i686 (5773)
3028 - provide error messages when incompatible logs are encountered (5898)
3030 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3031 * version 1.4.0.10 (1.4.1 release candidate 1)
3033 - don't keep a lock reference when lock is not granted (4238)
3034 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3035 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3036 - drop import inflight refcount on signal_completed_replay error (5255)
3037 - unlock page after async write error during send (3677)
3038 - handle missing objects in filter_preprw_read properly (5265)
3039 - no transno return for symlink open, don't save no-trasno open (3440)
3040 - don't try to complete elan receive that already failed (4012)
3041 - free RPC server reply state on error (5406)
3042 - clean up thread from ptlrpc_start_thread() on error (5160)
3043 - readahead could read extra page into cache that wasn't ejected (5388)
3044 - prevent races in class_attach/setup/cleanup/detach (5260)
3045 - don't dereference de->d_inode after l_dput of de (5458)
3046 - use "int" for stripe value returned from lock_to_stripe (5544)
3047 - mballoc allocation and error-checking fixes in 2.6 (5504)
3048 - block device patches to fix I/O request sizes in 2.6 (5482)
3049 - look up hostnames for IB nals (5602)
3050 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3051 - don't start multiple acceptors for the same port (5277)
3052 - fix incorrect LASSERT in mds_getattr_name (5635)
3053 - export a proc file for general "ping" checking (5628)
3054 - fix "lfs check" to not block when the MDS is down (5628)
3056 - service request history (4965)
3057 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3058 - create an "evict_client" /proc entry on OSTs, like the MDS has
3059 - fix mount usage message, return errors per mount(8) (5168)
3060 - change grep [] to grep "[]" in tests so they work in more UMLs
3061 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3062 - remove ancient LOV_MAGIC_V0 EA support (5047)
3063 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3064 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3065 - allow readahead allocations to fail when low on memory (5383)
3066 - mmap locking landed again, after considerable improvement (2828)
3067 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3069 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3072 - send OST transaction number in read/write reply to free req (4966)
3073 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3074 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3075 - the watchdog thread now runs as interruptible (5246)
3076 - flock/lockf fixes (but it's still disabled, pending 5135)
3077 - don't use EXT3 constants in llite code (5094)
3078 - memory shortage at startup could cause assertion (5176)
3080 - reorganization of lov code
3081 - single portals codebase
3083 - add extents/mballoc support (5025)
3084 - direct I/O reads in the obdfilter (4048)
3085 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3087 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3090 - send OST transaction number in read/write reply to free req (4966)
3091 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3092 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3093 - the watchdog thread now runs as interruptible (5246)
3094 - handle missing objects in filter_preprw_read properly (5265)
3095 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3096 - drop import inflight refcount on signal_completed_replay error (5255)
3097 - unlock page after async write error during send (3677)
3098 - return original error code on reconstructed replies (3761)
3099 - no transno return for symlink open, don't save no-trasno open (3440)
3101 - add pid to ldlm debugging output (4922)
3102 - bump the watchdog timeouts -- we can't handle 30sec yet
3103 - extra debugging for orphan dentry/inode bug (5259)
3105 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3108 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3109 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3110 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3111 - fix NULL dereference in /proc/sys/portals/routes (4827)
3112 - allow failed mdc_close() operations to be interrupted (4561)
3113 - stop precreate on OST before MDS would time out on it (4778)
3114 - don't send partial-page writes before EOF from client (4410)
3115 - discard client grant for sub-page writes on large-page clients (4520)
3116 - don't free dentries not owned by NFS code, check generation (4806)
3117 - fix lsm leak if mds_create_objects() fails (4801)
3118 - limit debug_daemon file size, always print CERROR messages (4789)
3119 - use transno after validating reply (3892)
3120 - process timed out requests if import state changes (3754)
3121 - update mtime on OST during writes, return in glimpse (4829)
3122 - add mkfsoptions to LDAP (4679)
3123 - use ->max_readahead method instead of zapping global ra (5039)
3124 - don't interrupt __l_wait_event() during strace
3126 - add software watchdogs to catch hung threads quickly (4941)
3127 - make lustrefs init script start after nfs is mounted
3128 - fix CWARN/ERROR duplication (4930)
3129 - return async write errors to application if possible (2248)
3130 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3131 - print NAL number in %x format (4645)
3132 - update barely-supported suse-2.4.21-171 series (4842)
3133 - support for sles 9 %post scripts
3134 - support for building 2.6 kernel-source packages
3135 - support for sles km_* packages
3137 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3140 - ignore -ENOENT errors in osc_destroy (3639)
3141 - notify osc create thread that OSC is being cleaned up (4600)
3142 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3143 - reconstruct ost_handle() like mds_handle() (4657)
3144 - create a new thread to do import eviction to avoid deadlock (3969)
3145 - let lconf resolve symlinked-to devices (4629)
3146 - don't unlink "objects" from directory with default EA (4554)
3147 - hold socknal file ref over connect in case target is down (4394)
3148 - allow more than 32000 subdirectories in a single directory (3244)
3149 - fix blocks count for O_DIRECT writes (3751)
3150 - OST returns ENOSPC from object create when no space left (4539)
3151 - don't send truncate RPC if file size isn't changing (4410)
3152 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3153 - bind to privileged port in socknal and tcpnal (3689)
3155 - rate limit CERROR/CWARN console message to avoid overload (4519)
3156 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3157 - basic mmap support (3918)
3158 - kernel patch series update from b1_4 (4711)
3160 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3163 - avoid crash during MDS cleanup with OST shut down (2775)
3164 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3165 - don't use bad inodes on the MDS (3744)
3166 - dynamic object preallocation to improve recovery speed (4236)
3167 - don't hold spinlock over lock dumping or change debug flags (4401)
3168 - don't zero obd_dev when it is force cleaned (3651)
3169 - print grants to console if they go negative (4431)
3170 - "lctl deactivate" will stop automatic recovery attempts (3406)
3171 - look for existing locks in ldlm_handle_enqueue() (3764)
3172 - don't resolve lock handle twice in recovery avoiding race (4401)
3173 - revalidate should check working dir is a directory (4134)
3175 - don't always mark "slow" obdfilter messages as errors (4418)
3177 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3180 - don't close LustreDB during write_conf until it is done (3860)
3181 - fix typo in lconf for_each_profile (3821)
3182 - allow dumping logs from multiple threads at one time (3820)
3183 - don't allow multiple threads in OSC recovery (3812)
3184 - fix debug_size parameters (3864)
3185 - fix mds_postrecov to initialize import for llog ctxt (3121)
3186 - replace config semaphore with spinlock (3306)
3187 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3188 - don't allow enqueue to complete on a destroyed export (3822)
3189 - down write_lock before checking llog header bitmap (3825)
3190 - recover from lock replay timeout (3764)
3191 - up llog sem before sending rpc (3652)
3192 - reduce ns lock hold times when setting kms (3267)
3193 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3194 - fix NULL deref and obd_dev leak on setup error (3312)
3195 - replace some LBUG about llog ops with error handling (3841)
3196 - don't match INVALID dentries from d_lookup and spin (3784)
3197 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3198 - fix invalid assertion in ptlrpc_set_wait (3880)
3200 - add libwrap support for the TCP acceptor (3996)
3201 - add /proc/sys/portals/routes for non-root route listing (3994)
3202 - allow setting MDS UUID in .xml (2580)
3203 - print the stack of a process that LBUGs (4228)
3205 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3208 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3209 - make sure to unset replay flag from failed open requests (3440)
3210 - if default stripe count is 0, use OST count for inode size (3636)
3211 - update parent mtime/ctime on client for create/unlink (2611)
3212 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3213 - free recovery state on server during a forced cleanup (3571)
3214 - unregister_reply for resent reqs (3063)
3215 - loop back devices mounting and status check on 2.6 (3563)
3216 - fix resource-creation race that can provoke i_size == 0 (3513)
3217 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3218 - more debugging for page-accounting assertion (3746)
3219 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3220 - avoid deadlock after precreation failure (3758)
3221 - fix race and lock order deadlock in orphan handling (3450, 3750)
3222 - add validity checks when grabbing inodes from l_ast_data (3599)
3224 - add /proc/.../recovery_status to obdfilter (3428)
3225 - lightweight CDEBUG infrastructure, debug daemon (3668)
3226 - change default OSC RPC parameters to be better on small clusters
3227 - turn off OST read cache for files smaller than 32MB
3228 - install man pages and include them in rpms (3100)
3229 - add new init script for (un)mounting lustre filesystems (2593)
3230 - run chkconfig in %post for init scripts (3701)
3231 - drop scimac NAL (unmaintained)
3233 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3236 - clean kiobufs before and after use (3485)
3237 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3238 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3239 - change init script to fail more clearly if not run as root (1528)
3240 - allow clients to reconnect during replay (1742)
3241 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3242 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3243 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3244 - fix cond_resched (3554)
3245 - only evict unfinished clients after recovery (3515)
3246 - allow bulk resend, prevent data loss (3570)
3247 - dynamic ptlrpc request buffer allocation (2102)
3248 - don't allow unlinking open directory if it isn't empty (2904)
3249 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3250 - remove extraneous obd dereference causing LASSERT failure (3334)
3251 - don't use get_cycles() when creating temp. files on the mds (3156)
3252 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3253 - handle EEXIST for set-stripe, set proper directory name (3336)
3255 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3256 - fix ksocknal_fmb_callback() error messages (2918)
3258 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3261 - don't copy lvb into (possibly NULL) reply on error (2983)
3262 - don't deref dentry after dput, don't free lvb on error (2922)
3263 - use the kms to determine writeback rpc length (2947)
3264 - increment oti_logcookies when osc is inactive (2948)
3265 - update client's i_blocks count via lvb messages (2543)
3266 - handle intent open/close of special files properly (1557)
3267 - mount MDS with errors=remount-ro, like obdfilter (2009)
3268 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3269 - don't use cancelling-locks' kms values (2947)
3270 - use highest lock extent for kms, not last one (2925)
3271 - don't dereference ERR_PTR() dentry in error handling path (3107)
3272 - fix thread race in portals_debug_dumplog() (3122)
3273 - create lprocfs device entries at setup instead of at attach (1519)
3274 - common AST error handler, don't evict client on completion race (3145)
3275 - zero nameidata in detach_mnt in 2.6 (3118)
3276 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3277 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3278 - initialize RPC timeout timer earlier for 2.6 (3219)
3279 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3280 - print nal/nid for unknown nid (3258)
3281 - additional checks for oscc recovery before doing precreate (3284)
3282 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3283 - don't crash in mdc_close for bad permissions on open (3285)
3284 - zero i_rdev for non-device files (3147)
3285 - clear page->private before handing to FS, better assertion (3119)
3286 - tune the read pipeline (3236)
3287 - fix incorrect decref of invalidated dentry (2350)
3288 - provide read-ahead stats and refine rpc in flight stats (3328)
3289 - don't hold journal transaction open across create RPC (3313)
3290 - update atime on MDS at close time (3265)
3291 - close LDAP connection when recovering to avoid server load (3315)
3292 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3293 - don't leak open file on MDS after open resend (3325)
3294 - serialize filter_precreate and filter_destroy_precreated (3329)
3295 - loop device shouldn't call sync_dev() for nul device (3092)
3296 - clear page cache after eviction (2766)
3297 - resynchronize MDS->OST in background (2824)
3298 - refuse to mount the same filesystem twice on same mountpoint (3394)
3299 - allow llmount to create routes for mounting behind routers (3320)
3300 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3301 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3302 - fix rare problem with rename on htree directories (3417)
3304 - allow default OST striping configuration per directory (1414)
3305 - fix compilation for qswnal for 2.6 kernels (3125)
3306 - increase maximum number of MDS request buffers for large systems
3307 - change liblustreapi to be useful for external progs like lfsck (3098)
3308 - increase local configuration timeout for slow disks (3353)
3309 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3311 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3314 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3315 - don't overwrite extent policy data in reply if lock was blocked (2901)
3316 - drop filter export grants atomically with removal from device (2663)
3317 - del obd_self_export from work_list in class_disconnect_exports (2908)
3318 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3319 - swab reply message in mdc_close, other PPC fixes (2464)
3320 - fix destroying of named logs (2325)
3321 - overwrite old logs when running lconf --write_conf (2264)
3322 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3323 - fix race in target_handle_connect (2898)
3324 - mds_reint_create() should take same inode create lock (2926)
3325 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3326 - don't close files for self_export to avoid uninitialized obd (2936)
3327 - allow MDS with the same name as client node (2939)
3328 - hold dentry reference for closed log files for unlink (2325)
3329 - reserve space for all logs during transactions (2059)
3330 - don't evict page beyond end of stripe extent (2925)
3331 - don't oops on a deleted current working directory (2399)
3332 - handle hard links to targets without a parent properly (2517)
3333 - don't dereference NULL lock when racing during eviction (2867)
3334 - don't grow lock extents when lots of conflicting locks (2919)
3336 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3339 - account for cache space usage on clients to avoid data loss (974)
3340 - lfsck support in lustre kernel code (2349)
3341 - reduce journal credits needed for BRW writes (2370)
3342 - orphan handling to avoid losing space on client/server crashes
3343 - ptlrpcd can be blocked, stopping ALL progress (2477)
3344 - use lock value blocks to assist in proper KMS, faster stat (1021)
3345 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3346 - recovery for initial connections (2355)
3347 - fixes for mds_cleanup_orphans (1934)
3348 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3349 - block all file creations until orphan recovery completes (1901)
3350 - client remove rq_connection from request struct (2423)
3351 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3352 - recovery timer race (2670)
3353 - mdc_close recovey bug (2532)
3354 - ptlrpc cleanup bug (2710)
3355 - mds timeout on local locks (2588)
3356 - namespace lock held during RPCs (2431)
3357 - handle interrupted sync write properly (2503)
3358 - don't try to handle a message that hasn't been replied to (2699)
3359 - client assert failure during cleanup after abort recovery (2701)
3360 - leak mdc device after failed mount (2712)
3361 - ptlrpc_check_set allows timedout requests to complete (2714)
3362 - wait for inflight reqs when ptlrpcd finishes (2710)
3363 - make sure unregistered services are removed from the srv_list
3364 - reset bulk XID's when resending them (caught by 1138 test)
3365 - unregister_bulk after timeout
3366 - fix lconf error (2694)
3367 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3368 - readahead locks pages, leaves pending causing memory pressure (2673)
3369 - increase OST request buffers to 4096 on large machines (2729)
3370 - fix up permission of existing directories in simple_mkdir (2661)
3371 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3372 - don't assign transno to errored transactions (2742)
3373 - don't delete objects on OST if given a bogus objid from MDS (2751)
3374 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3375 - if rq_no_resend, then timeout request after recovery (2432)
3376 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3377 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3378 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3379 - MDS recovery shouldn't delete objects that it has given out (2730)
3380 - if enqueue arrives after completion, don't clobber LVB (2819)
3381 - don't unlock pages twice when trigger_group_io returns error (2814)
3382 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3383 - don't write pages to disk if there was an error (1450)
3384 - don't ping imports that have recovery disabled (2676)
3385 - take buffered bytes into account when balancing socknal conn (2817)
3386 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3387 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3388 - remove little-endian swabbing of llog records (1987)
3389 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3390 - retry reposting request buffers if they fail (1191)
3391 - grow extent at grant time to avoid granting a revoked lock (2809)
3392 - lock revoke doesn't evict page if covered by a second lock (2765)
3393 - disable VM readahead to avoid reading outside lock extents (2805)
3395 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3396 - updated kernel patches for hp-2.4.20 kernel (2681)
3398 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3401 - fix truncated write corruption (2366)
3402 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3404 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3405 - ensure len(uuid) < 37 in lmc (1171)
3406 - fix ia64 OOPS in llog_test (2255)
3407 - zero end of page at obdfilter for partial page writes (2648)
3408 - don't leave stale dentries around after renames (bug 2428)
3409 - fix timeouts when evicting a client with a single lock held (2642)
3410 - set deadline for the initial HELLO message to drain (2634)
3411 - print out dotted-quad IP addresses in the socknal (2302)
3413 - additional debugging for MDS client eviction problem (2443)
3414 - fix mkfsoptions support for osts (2603, 2604)
3416 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3419 - add series for the vanilla 2.6.0 kernel
3420 - add series for the vanilla 2.4.24 kernel
3421 - add series for a cray x86/64 UL kernel drop
3422 - fix xattr patches for the vanilla 2.4.19 series
3424 - generate true UUIDs in lmc (1171)
3425 - have portals stack dumping break in UML (2466)
3426 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3427 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3428 - wake callback waiting threads on client eviction (2460)
3429 - Add --ptldebug and --subsystem to lmc (1719)
3430 - update assertion to allow safe interrupt allocation
3431 - set rq_no_resend for cancel requests (2432)
3432 - recalculate ptlrpcd timeout after resend (2494)
3433 - call vfs_rmdir when removing pending directories (2368)
3434 - fix renaming a file to itself (2429)
3435 - lmc creates a default one-stripe lov (2454)
3436 - expand procfs space to handle large clusters (2326)
3437 - increase UML stack to avoid overflow
3438 - update lconf's list of debug and subsystem masks
3439 - fix lfs find --obd (2510)
3440 - /proc tunable for disabling filter read caching (2591)
3441 - stop rpm packages from altering slapd.conf (2301)
3442 - disable nagle in the socknal under 0conf (2578)
3443 - choose mds inode size based on stripe count (2572)
3444 - fix kernel-source rpm problems (2516)
3446 - add --disable-doc to avoid pdf generation (2421)
3447 - update documentation, tests, type-os, comments
3448 - avoid format warnings on ia64
3449 - remove the TOE NAL
3450 - tiny code cleanups by removing unused fields
3452 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3455 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3456 - remove the most blatant lies from BUILDING file (2371)
3457 - change default debug level to reasonable production setting
3458 - reduce client side cache size to reduce cache flush time
3459 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3460 - make TCP zerocopy and pinger support enabled by default (2476)
3461 - sync writes completed after process exits caused crashes (2319)
3462 - maintain correct mount count on the MDS (2356)
3463 - backout 1557, because 2316 wasn't really fixed
3464 - better file I/O statistics gathering in /proc
3465 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3466 - another kernel patch to fix zero-copy TCP function export
3467 - don't take duplicate lock when processing re-sent getattr (2420)
3468 - lctl uses obd_self_export instead of creating new conn (2353)
3469 - MDS/OST recovery case which requires object creation asserted (2425)
3470 - move lfs from /usr/sbin to /usr/bin in packages
3471 - fix race between mds_client_add and mds_client_free (2417)
3472 - use kmalloc instead of slabs in portals (2430)
3473 - don't create duplicate records when a failover MDS is present (2442)
3474 - remove unnecessary mount age check (2332)
3475 - don't remove directory inodes from locks prematurely (2451)
3476 - don't break if MDS service name is the same as hostname (2103)
3477 - fix races in client write RPC generation when cache full (2482)
3479 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3482 - remove now-unused request->rq_obd (278)
3483 - if an allocation fails, print out how much memory we've used (1933)
3484 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3485 - add the "configurable stack size" patch to most series files (1256)
3486 - ability to write large log records, for 100+ OST configs (2306)
3487 - fix NULL deref when filter_prep fails (2314)
3488 - fix operator precedence error in filter_sync
3489 - dynamic allocation of socknal TX descriptors (2315)
3490 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3491 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3492 - ensure that CWARN messages in Portals always get to the syslog
3493 - __init/__exit are not for prototype decls (ldlm_init/exit)
3494 - x86-64 compile warning fixes
3495 - fix gateway LMC keyword conflict (2318)
3496 - fix MDS lock inversions in getattr/reint paths (1844)
3497 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3498 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3499 - DLM race condition prevented some lock evictions (2328)
3500 - ENOMEM detection and retry on socknal sends (2230)
3501 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3502 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3503 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3504 - better error messages when a client is rejected during recovery (1505)
3505 - avoid cancelling locks which were never granted, after failure (2330)
3506 - fix i_sem/journal inversion in mds_client_add (2333)
3507 - fix truncate/getattr lock cycle deadlock (2334)
3508 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3509 - fix two rare exit paths which could leak an l_lock() ref (2321)
3510 - fencepost error in MDS/OST orphan recovery (2226)
3511 - make log record alignment 8 bytes (1988)
3512 - lstripe now fails when requested offset > ost_count (2237)
3513 - ensure that all kernel series have a complete list.h (1607)
3514 - fix crashes in special-file operations (2316)
3515 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3517 - allow configurable automake binary, for testing new versions
3518 - small update to the lfs documentation
3520 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3522 * fix negative export reference count in fsfilt_sync (2312)
3524 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3525 * release candidate 0.9.1
3527 - orphans are moved into the PENDING directory for possible recovery
3528 - replayed opens now open by fid for orphan/rename safety (1042)
3529 - last close of an orphan inode generates a transno (683)
3530 - chdir() and mount() now pin the directory entry (1020)
3531 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3532 - discard very old requests without processing them (1502)
3533 - remove some common, well-understood CERRORs (1505)
3534 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3535 - clear "grant" flags in OST replies until OST grant code lands (1644)
3536 - fix read performance by not clobbering i_blksize on client (1598)
3537 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3538 - make LRU size a /proc tunable, clears locks when reduced (707)
3539 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3540 - server locks take references on exports now (1558)
3541 - build fixes for 2.4.20-rh trees (1663)
3542 - return an error from lov_create if all OSCs are inactive (1751)
3543 - fix import levels when a reconnect happens without a timeout (1597)
3544 - exit early from mds_open if we get a lookup error (1749)
3545 - partial page read at EOF wouldn't wait for disk before sending (1642)
3546 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3547 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3548 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3549 - write/truncate lock inversion (1639)
3550 - Don't auto-load obdclass, portals modules during cleanup (1495)
3551 - fix timestamps from jumping to "now" (1763)
3552 - extra journal assertions (1648)
3553 - add an extra multiunlink test (1771)
3554 - fix read_record/write_record API (1776)
3555 - fix leak of offset_extent, possible incorrect i_size later (1772)
3556 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3557 - Debugging for the kqswnal_get_idle_tx problems (1820)
3558 - Allow recovery to be attempted multiple times (1536)
3559 - Write out MDS last_rcvd file after it is first created (1600)
3560 - Fix tx_descriptor leak in failed transmit situations (1827)
3561 - ext3 journaling fixes for assertion failure after IO error (1871)
3562 - class_export_put() on freed export after completion AST error (1896)
3563 - Fix revalidate looping in VFS (1322)
3564 - Don't access a freed export during MDS_REINT timeout (1521)
3565 - Add open-unlink recovery support on the MDS (1673,1764)
3566 - Return an error if no MDS data was read from last_rcvd (1946)
3567 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3568 - Don't setup a disk device that doesn't match exported UUID (317)
3569 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3570 - avoid committing NULL handle in force close
3571 - local.sh is now a one-stripe LOV configuration
3572 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3573 - fix ext3 htree duplicate directory entry corruption (1516)
3574 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3575 - update to new LOV EA format (2097)
3576 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3578 2003-06-15 Phil Schwan <phil@clusterfs.com>
3581 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3582 - per-import/export recovery handling (958, 931, 959)
3583 - multiple last-rcvd slots, for serving multiple FSes (949)
3584 - connections are again shared between multiple imp/exports (963, 964)
3585 - "umount -f" would hang if any requests needed to be sent (393, 978)
3586 - avoid pinning large req buffer by copying for queued messages (989)
3587 - add "uuid" to "lctl device" command to help upcalls (991)
3588 - "open" RPCs with transnos would confuse recovery counters (1037)
3589 - do proper endian conversion of all wire messages (288, 340, 891)
3590 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3591 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3592 - fix ptlrpc_connection leak in target_handle_connect (1174)
3593 - fix import refcounting bug in OST and MDS cleanup (1134)
3594 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3595 - fix very unlikely obd_types race condition (501)
3596 - remove osc_open hack for echo_client (1187)
3597 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3598 - a failure in read_inode2 leads to deadlock (1139)
3599 - cancel ack-locks as soon as transaction is committed (1072)
3600 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3601 - make sure to commitrw after any preprw to avoid deadlock (1162)
3602 - failing to execute a file in a lustre FS would lock inode (1203)
3603 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3604 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3605 - obd_uuid2tgt would walk off the end of the list (1255)
3606 - on IA64 the osc would give portals incorrect bulk size (1258)
3607 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3608 - fix lock inversion caused by new llite matching code (1282)
3609 - limit the number of dirty pages on a client to 10MB (1286)
3610 - timed out locks were not being corrected cancelled (1289)
3611 - fix O_DIRECT above 4GB on IA-32 (1292)
3612 * major user-visible changes
3613 - fail out/fail over policy now controlled by the upcall (993)
3615 - add OBD_PING to check server availability and failure (954)
3616 - lustre messages are now sent in sending host order (288, 340, 891)
3617 - add eadatalen to MDS getattr reply (340)
3618 - OST read replies may contain second buffer, with per-page status (593)
3620 2003-03-11 Phil Schwan <phil@clusterfs.com>
3623 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3624 - failed open()s could cause deadlock; fixed (867, 869)
3625 - stop cancelling OST locks when files are closed (481)
3626 - overlapping XID spaces caused network corruption (851, 853)
3627 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3628 - setattr_raw added, to do single-RPC, server-side setattrs
3629 - lmc/lconf syntax change for OST UUIDs
3630 - fix crashy race condition between ptlrpc_free_req and osc_close
3631 - don't use request in mdc_enqueue if we hit a timeout (889)
3632 - don't set the inode i_size for regular files from the MDS (896)
3633 - handle out of order completion AST (842)
3634 - don't LBUG if a lock request times out after receiving AST (913)
3635 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3636 - if a bad lock AST arrives, send an error instead of dropping entirely
3637 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3638 - fix leak in bulk IO when only partially completed (899, 900, 926)
3639 - fix O_DIRECT for ia64 (55)
3640 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3641 - C-z after timeout could hang a process forever; fixed (977)
3643 - client-side I/O cache (678, 924, 929, 941, 970)
3645 - READPAGE and SETATTRs which don't take server-side locks get
3648 2003-02-11 Phil Schwan <phil@clusterfs.com>
3651 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3652 - Fix fsfilt_extN_readpage() to read a full page of directory
3653 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3654 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3655 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3656 - Most intent operations were removed, in favour of a new RPC mode
3657 that does a single RPC to the server and bypasses most of the VFS
3658 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3659 - Aggressively cancel local locks on DLM servers
3660 - mds_reint_unlink sends EA to the client if it's the last nlink.
3661 client uses that EA to unlink OST objects.
3662 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3663 - recursive symlinks were fixed (439)
3664 - fixed NULL deref in DEBUG_REQ
3665 - filter_update_lastobjid no longer calls sync, which annoyed extN
3666 - fixed multi-client small-writes to a single file problem (445)
3667 - fixed mtime updates during file writes (607)
3668 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3669 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3670 - fixed Linux OST size reporting problem (444, 656)
3671 - OST now updates object mtime with writes or setattr (607, 619)
3672 - client verifies file size before zeroing page past EOF (445)
3673 - OST now writes last allocated objid to disk with allocation (108)
3674 - LOV on echo now works (409)
3676 - mds_reint_unlink sends a new buffer, with the EA included. this
3677 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3678 set if a regular file was being unlinked, and it was the last link
3679 - use PtlGet from the target for bulk writes (315)
3680 - OST now updates object mtime with writes or setattr (607, 619)
3681 - LDLM now has a grant-time callback to revalidate locked items, if
3683 - Many MDS operations were reorganized to combat race conditions
3685 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3686 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3688 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3691 - Fully reactivate OST imports after reconnection (512, others)
3692 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3693 - More graceful error handling for truncating on dead OST (515)
3694 - Don't error out unless we're actually accessing dead stripes (474)
3695 - Fix garbage sizes when stripes are missing (410)
3696 - LRU counters were broken, causing constant lock purge (433, 432)
3697 - garbage on read from stripes with failed OSTs (441)
3698 - mark OSCs as active before reconnecting during recovery (438)
3699 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3700 - lfind did not preserve OST order in output (443)
3701 - symlinks cause hung clients, incorrect data (439)
3702 - stop dereferencing request after dropping refcount (457)
3703 - don't LASSERT(spin_is_locked) on non-SMP (455)
3704 - fixes for many rename() bugs
3705 - fstat didn't correctly synchronize attributes (399)
3706 - server must handle lock cancellation during blocking AST prep (487)
3707 - bulk descriptors were free()d too soon (511)
3708 - fix paths in lconf, which would load incorrect modules (451, 507)
3709 - fix confusing lconf 'host not found' error message (386)
3710 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3711 - fix race condition in mdc_blocking_ast() for inode access (526)
3712 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3713 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3714 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3715 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3716 - fix lstripe in several ways (kernel side) (527)
3717 - fix request leak in ldlm_cli_enqueue (262)
3718 - incorrect OSC was marked inactive after OST failure
3719 - call mds_fs_cleanup before unmounting filesystem (524)
3720 - fix races between taking ns_lock and ldlm_lock_change_resource
3721 - fix races updating LOV export open file list
3722 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3723 - fix recovery NULL deref in ldlm_cli_cancel_unused
3724 - fix some DLM races by using new hash table for lock handles (419)
3725 - permit the client to specify desired inodes, at replay
3726 - duplicate requests when we queue them for replay reintegration
3727 - fix last_rcvd offset calculation
3728 - sync after each recovered transaction, so we always make progress
3729 - never, not always, ERESTART requests without transnos
3730 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3732 - skip replay if the MDS says that the client is already connected
3733 - don't check for a recovery-enabled export to match lctl's UUID
3734 - don't INC_USE_COUNT for phantom exports
3735 - don't crash when cleaning up phantom exports (567)
3736 - don't double-finish or set replay data for errored mdc_open requests
3737 - abort requests when they time out, so we don't get old replies
3738 - send/receive replies for AST messages again
3739 - if the client says that it doesn't have the lock, cancel it on the
3741 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3742 mark it as destroyed, it will all work out when decref is called
3743 - fix module use counts (22, 581)
3745 - ASTs now expect a reply (server cancels lock on error reply)
3747 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3750 - fix many simultaneous client startup (392)
3751 - fix dentry->d_it clobbering
3752 - credentials weren't being shipped for readdir/getattr operations
3753 - remove invalid assertions triggered during some concurrent MD
3755 - proper Lustre versions added (336, 389)
3756 - fix memory leak for create error case (398)
3757 - fix LOV locking bug that would get cli/srv out of sync
3758 - fix echo client over LOV (409)
3759 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3760 - fix double-O_EXCL intent crash (424)
3761 - avoid sending multiple lock CANCELs (352)
3763 - MDS can do multi-client recovery (modulo bugs in new code)
3765 - many updates, edits, cleanups
3767 2002-11-18 Phil Schwan <phil@clusterfs.com>
3770 - fix null d_it dereference (346)
3771 - fix full OST/dbench hang (333)
3772 - fix permission problem with file removal (286)
3773 - fix removal of OSCs from LOV when they fail
3774 - fix NULL deref during bulk timeout (214)
3775 - fix problems related to multiple filesystems on one MDS (241)
3776 - fixed serious subtle metadata locking bugs
3777 - free locks on clients when inodes are removed due to memory
3779 - fix inode pointer in lock data (285)
3780 - partial support for multiple MDS on a single host (241)
3781 - data locks weren't cancelled at clear_inode time (290, 311)
3782 - intent locks could lead to unbounded lock growth (205)
3783 - added a maximum lock count, an LRU list, and a flusher
3784 - fix multiple rename (365)
3785 - properly abstracted the echo client
3786 - OSC locked 1 byte too many; fixed
3787 - rewrote brw callback code:
3788 - fixed recovery bugs related to LOVs (306)
3789 - fixed too-many-pages-in-one-write crash (191)
3790 - fixed (again) crash in sync_io_timeout (214)
3791 - probably fixed callback-related race (385)
3793 - Add capability to MDS protocol
3794 - LDLM cancellations and callbacks on different portals
3796 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3799 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3800 - defer open object destruction to close time (601981, 138)
3801 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3802 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3803 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3804 - don't call obd_setattr in ll_file_release for destroyed objects
3805 * protocol change to lustre_msg: move |version| and add |flags|
3806 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3807 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3808 * added replay of create, unlink, link and rename operations during
3809 MDS failover; recovery should be much more robust now
3810 * remove failed OSCs from LOVs (only lov_create uses this so far)
3811 * the lustre-HOWTO was brought (more) up to date (582544)
3813 2002-10-23 Phil Schwan <phil@clusterfs.com>
3816 - in-use dentries weren't being reused properly (617851)
3817 - prevent multiple LDLM setup (599178)
3818 - fix LOV size calculations for truncate (617853)
3819 - fix client handling of MDS intent errors (POSIX)
3820 - fix permission bug in lovstripe.c test (624321)
3821 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3822 - truncate past end of file could corrupt data
3823 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3824 - a race in recovery could return ETIMEDOUT to apps (623947)
3825 - building outside the source directory was fixed
3826 * the lustre-HOWTO was brought (more) up to date (582544)
3827 * major progress was made on recovery functionality
3829 2002-10-10 Phil Schwan <phil@clusterfs.com>
3832 - recovery deadlock fix
3833 - rm -rf causes LBUG fix (617817)
3834 - file open by multiple tasks fix (618962)
3835 - directory permissions bugs (602707 and 620007)
3836 - journal_stop fixed with locking (611313)
3837 - O_APPEND failures resolved (618273, perhaps 614459)
3838 - lconf PATH fix (619770)
3839 - IA64 build fix (621450)
3840 - RPC buffer sizes scale with amount of memory
3842 2002-10-01 Phil Schwan <phil@clusterfs.com>
3845 - locks would be cancelled without throwing away data pages,
3846 resulting in inconsistent data (605627)
3847 - inode attributes were not always being refreshed (605627, 612449)
3848 - lconf now continues to cleanup after lctl reports an error
3849 - MDS now enforces user permissions (602707)
3850 - lprocfs cleanup fixed, but not yet enabled (614157)
3851 - fixed infinite server hang, should a client not respond to an AST
3852 - avoid going into recovery if user calls readlink() with a buffer
3853 that's too small (613941)
3854 - AST RPCs no longer require replies (614867) -- this may be changed
3855 - don't crash server if client sends an IOV that's too big (611336)
3856 - fixed lock conversion deadlock (611892)
3857 - fixed the following of symlinks (614622)
3858 * recovery: the server can remove locks from a client that dies, other
3859 clients can make progress
3860 * more extN patch fixes
3861 * compile-time configurable ptlrpc buffer allocations
3863 - collaborative read cache document
3864 - Lustre Lite Performance CDR document-in-progress
3866 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3869 - fix typo in patch-2.4.18
3871 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3874 - clear ptlrpc request each time in handle_incoming_request()
3875 - unlink of files now destroys the object on the OST
3877 2002-09-19 Peter Braam <braam@clusterfs.com>
3879 * add hard link support
3880 * change obdfile creation method
3881 * kernel patch changed
3883 2002-09-19 Peter Braam <braam@clusterfs.com>
3886 - stack overflow bug in extN fixed
3888 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3890 * documentation updates
3891 - add man pages for config tools
3892 - update tests/README to describe testing with new config tools
3893 - finish metadata API descriptions
3894 * bug fixes and cleanups
3895 - statfs workaround for 16TB limit
3896 - LOV stripe allocation improved, can stripe on subset of OSTs
3897 - LOV file size/IO offset was wrong for files > 4GB in size
3898 - object EA data was being dropped, caused files to be unreadable
3899 - memory overflow with non-LOV OST caused memory corruption
3900 - fixed regression tests to work with new config tools, obdfilter
3901 - fixed bug when directory size became larger than 1 block
3902 - fixed bug (for single client case) when PWD was deleted
3903 - invalidate local directory pages when doing intent-based ops
3904 - avoid LDLM oops when lock callback contained bad data
3906 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3908 * documentation updates
3909 * bug fixes and cleanups
3910 - configuration tools
3913 - 64-bit compile warnings
3914 - 64-bit internal statfs data
3916 * test_brw on persistent OST devices
3918 * lprocfs (disabled)
3920 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3922 * documentation updates
3923 * bug fixes and cleanups
3924 * configuration tools
3926 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3929 * many small fixes to 0_5_4
3930 * io/network handling
3931 * thinkos in MDS operations
3933 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3936 * crucial basic fixes to 0.5.3
3937 * IOR, Iozone work over Elan
3940 2002-08-07 Phil Schwan <phil@clusterfs.com>
3941 * version 0_5_3, our first alpha
3942 * we use the new Portals iovs
3943 * documentation updates
3944 * bug fixes and cleanups
3945 * small changes in the DLM wire protocol
3947 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3948 * version 0_5_1 with some initial stability,
3949 * locking on MD and file I/O.
3950 * documentation updates
3951 * several bug fixes since 0.5.0
3952 * small changes in wire protocol
3954 2002-07-18 Phil Schwan <phil@clusterfs.com>
3956 * delivered as Lustre Light Alpha
3957 * fixed a crash after handling invalid MDS requests
3958 * fixed directory pages for architectures with non-4k pages sizes
3960 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3961 * release version v0_4_4
3962 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3964 * Fixes a number of interruption problems with OST operations.
3965 * Update documentation for portals header changes
3966 * Move all wire protocol structs/defines to lustre_idl.h
3967 * Fixes symlink length bug.
3968 * Add tcpdump to repository.
3970 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3971 * release version v0_4_3
3972 * Fixes statfs for inodes on extN.
3973 * Fixes bug in runtests which would delete /etc/hosts.
3974 * Use 64-bit object IDs wherever possible (not into VFS though)
3975 Remove ost_get_info, which is unused by lustre, and out of date.
3977 2002-07-03 Peter Braam <braam@clusterfs.com>
3978 * release version v0_4_2 Fixes a lookup error (type not passed)
3979 * move forward to head of Portals
3980 * move forward to latest Lustre kernel
3982 2002-06-25 Peter Braam <braam@clusterfs.com>
3983 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob