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: Fix errors in lfs documentation
18 Details : Fixes man pages
22 Description: don't use GFP_* in generic Lustre code.
23 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
27 Description: obdclass is limited by single OBD_ALLOC(idarray)
28 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
32 Description: warnings with build patchless client with vanila 2.6.19 and up
33 Details : change old ctl_table style and replace ctl_table/ctl_table_header
34 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
38 Frequency : Only for SLES
39 Description: sanity_quota fail test_1
40 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
41 file, which makes TSTID[2] inunique.
45 Frequency : Only for SLES9
46 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
47 Details : Update target/ChangeLog/which_patch .
51 Frequency : Only for test_10 in sanity-quota.sh
52 Description: a bug in quota test code
53 Details : checking a test flag in a wrong place
57 Frequency : Only for test_13 in sanity-quota.sh
58 Description: a bug in quota test script
59 Details : shouldn't check overfull allocated quota
63 Description: jbd statistics
64 Details : Port older jbd statistics patch for sles10
66 Severity : enhancement
68 Description: Add printing inode info into message about error in writepage.
73 Description: reply_lock_interpret crash due to race with it and lock cancel.
74 Details : Do not replay locks that are being cancelled. Do not reference
75 locks by their address during replay, just by their handle.
79 Frequency : When flocks are used.
80 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
81 Details : Flock locks might destroy just granted lock if it could be merged
82 with another existing flock, this is done in completion handler,
83 so teach ldlm_cli_enquque_fini that this is a valid case for
89 Description: Oops in read and write path when failing to allocate lock.
90 Details : Check if lock allocation failed and return error back.
94 Description: lstripe command fails for valid OST index
95 Details : The stripe offset is compared to lov->desc.ld_tgt_count
96 instead of lov->desc.ld_active_tgt_count.
100 Description: sanity.sh failed test 103
101 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
106 Description: df doesn't work properly if diskfs blocksize != 4K
107 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
111 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
112 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
116 Description: obd_health_check_timeout too short
117 Details : set obd_health_check_timeout as 1.5x of obd_timeout
121 Description: llapi_file_create() does not allow some changes
122 Details : add llapi_file_open() that allows specifying the mode and
123 open flags, and also returns an open file handle.
127 Description: merge and cleanup kernel patches.
128 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
132 Description: Add journal checksum support.(Kernel part)
133 Details : The journal checksum feature adds two new flags i.e
134 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
135 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
136 indicates that the commit block contains the checksum for
137 the blocks described by the descriptor blocks. Now commit
138 record can be sent to disk without waiting for descriptor
139 blocks to be written to disk. This behavior is controlled
140 using JBD2_FEATURE_ASYNC_COMMIT flag.
144 Description: OSS needs mutliple precreate threads
145 Details : Add ability to start more than one create thread per OSS.
149 Description: Sanity.sh test_65k failed: setstripe should have succeeded
150 Details : execute lfs setstripe on client
154 Description: mds_obd_create error creating tmp object
155 Details : When the user sets quota on root, llog will be affected and can't
156 create files and write files.
159 Frequency : Always on ia64 patchless client, and possibly others.
161 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
162 Details : This allows the patchless client to be loaded on architectures
167 Description: RedHat Update kernel for RHEL5
168 Details : Add the kernel config file for RHEL5.
172 Description: RedHat Update kernel for RHEL5
173 Details : Modify the kernel config file more closer RHEL5.
177 Description: Build failure against Centos5 (RHEL5)
178 Details : Define PAGE_SIZE when it isn't present.
182 Description: test_7 on sanity-quota.sh can't be used on separate machines
183 Details : add facet to handle it
187 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
188 Details : with NFS, the anon dentry's parent was set to itself in
189 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
190 corresponding dentry other than use rec->ur_name.
192 Severity : enhancement
194 Description: lfs setstripe enhancement
195 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
199 Description: testing performance impact of enabling checksumming
200 Details : enable checksum by default, allow --disable-checksum
201 configure option and "-o nochecksum" mount option
205 Description: lustre support for RHEL5
206 Details : Add support for RHEL5.
210 Description: Client eviction due to failover config
211 Details : after a connection loss, the lustre client should attempt to
212 reconnect to the last active server first before trying the
213 other potential connections.
217 Description: when mds and osts use different quota unit(32bit and 64bit),
218 quota will be released repeatly.
219 Details : void sending multiple quota reqs to mds, which will keep the
220 status between the reqs.
224 Description: osts not allocated evenly to files
225 Details : change the condition to increase offset_idx
229 Description: Tune the kernel for good SCSI performance.
230 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
231 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
236 Description: rewrite ext2-derived code in obdclass/uuid.c
237 Details : rewrite inherited code (uuid parsing code from ext2 utils)
238 from scratch preserving functionality.
239 --------------------------------------------------------------------------------
241 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
243 * Support for kernels:
244 2.6.5-7.283 (SLES 9),
245 2.6.9-55.EL (RHEL 4),
246 2.6.16.46-0.14 (SLES 10),
247 2.6.18.8 vanilla (kernel.org)
248 * Client support for unpatched kernels:
249 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
250 2.6.16 - 2.6.22 vanilla (kernel.org)
251 * Due to recently discovered recovery problems, we do not recommend
252 using patchless RHEL 4 clients with this or any earlier release.
253 * Recommended e2fsprogs version: 1.39.cfs8
254 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
255 * Starting with this release, the ldiskfs backing filesystem required
256 by Lustre is now in its own package, lustre-ldiskfs. This package
257 should be installed. It is versioned separately from Lustre and
258 may be released separately in future.
262 Description: block reactivating mgc import until all deactivates complete
263 Details : Fix race when failing back MDT/MGS to itself (testing)
265 Severity : enhancement
267 Description: add optional extra BUILD_VERSION info
268 Details : add a new environment variable (namely LUSTRE_VERS) which allows
269 to override the lustre version.
272 Frequency : 2.6.18 servers only
274 Description: ll_kern_mount() doesn't release the module reference
275 Details : The ldiskfs module reference count never drops down to 0
276 because ll_kern_mount() doesn't release the module reference.
281 Description: server LBUG when using old ost_num_threads parameter
282 Details : Accept the old ost_num_threads parameter but warn that it
283 is deprecated, and fix an off-by-one error that caused an LBUG.
288 Description: Transient SCSI error results in persistent IO issue
289 Details : iobuf->dr_error is not reinitialized to 0 between two
293 Frequency : sometimes when underlying device returns I/O errors
295 Description: OSTs not going read-only during write failures
296 Details : OSTs are not remounted read-only when the journal commit threads
297 get I/O errors because fsfilt_ext3 calls journal_start/stop()
298 instead of the ext3 wrappers.
302 Description: poor connect scaling with increasing client count
303 Details : Don't run filter_grant_sanity_check for more than 100 exports
304 to improve scaling for large numbers of clients.
307 Frequency : SLES10 only
309 Description: sanity-quota.sh quotacheck failed: rc = -22
310 Details : Quotas cannot be enabled on SLES10.
313 Frequency : liblustre clients only
315 Description: getdirentries does not give error when run on compute nodes
316 Details : getdirentries does not fail when the size specified as an argument
317 is too small to contain at least one entry
319 Severity : enhancement
321 Description: Add LNET router traceability for debug purposes
322 Details : If a checksum failure occurs with a router as part of the
323 IO path, the NID of the last router that forwarded the bulk data
324 is printed so it can be identified.
329 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
330 Details : Due to a race condition, liblustre clients were occasionally
333 Severity : enhancement
335 Description: lfs setstripe use optional parameters instead of postional
338 Severity : enhancement
340 Description: Nanosecond timestamp support for ldiskfs
341 Details : The on-disk ldiskfs filesystem has added support for nanosecond
342 resolution timestamps. There is not yet support for this at
343 the Lustre filesystem level.
346 Frequency : during server recovery
348 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
349 Details : request with rq_no_resend flag not awake l_wait_event if they get a
353 Frequency : nfs export on patchless client
355 Description: connectathon hang when test nfs export over patchless client
356 Details : Disconnected dentry cannot be found with lookup, so we do not need
357 to unhash it or make it invalid
360 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
361 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
362 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
367 Description: open req refcounting wrong on reconnect
368 Details : If reconnect happened between getting open reply from server and
369 call to mdc_set_replay_data in ll_file_open, we will schedule
370 replay for unreferenced request that we are about to free.
371 Subsequent close will crash in variety of ways.
372 Check that request is still eligible for replay in
373 mdc_set_replay_data().
378 Description: disable writes to filesystem when reading health_check file
379 Details : the default for reading the health_check proc file has changed
380 to NOT do a journal transaction and write to disk, because this
381 can cause reads of the /proc file to hang and block HA state
382 checking on a healthy but otherwise heavily loaded system. It
383 is possible to return to the previous behaviour during configure
384 with --enable-health-write.
386 Severity : enhancement
388 Description: 64-bit inode version
389 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
390 to the inode. This will be required for version-based recovery.
395 Description: MDS fails to start if a duplicate client export is detected
396 Details : in some rare cases it was possible for a client to connect to
397 an MDS multiple times. Upon recovery the MDS would detect this
398 and fail during startup. Handle this more gracefully.
400 Severity : enhancement
402 Description: Add -o localflock option to simulate old noflock
404 Details : This will achieve local-only flock/fcntl locks
410 Description: log_commit_thread vs filter_destroy race leads to crash
411 Details : Take import reference before releasing llog record semaphore
416 Description: Wrong request locking in request set processing
417 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
418 imp_delayed_list, in this place should be used imp_lock.
421 Frequency : when reconnection
423 Description: Grant Leak when osc reconnect to OST
424 Details : When osc reconnect ost, OST(filter) should check whether it
425 should grant more space to client by comparing fed_grant and
426 cl_avail_grant, and return the granted space to client instead
427 of "new granted" space, because client will call osc_init_grant
428 to update the client grant space info.
431 Frequency : when client reconnect to OST
433 Description: Grant Leak when osc do resend and replay bulk write
434 Details : When osc reconnect to OST, OST(filter)should clear grant info of
435 bulk write request, because the grant info while be sync between
436 OSC and OST when reconnect, and we should ignore the grant info
437 these of resend/replay write req.
442 Description: Grant space more than avaiable left space sometimes.
443 Details : When then OST is about to be full, if two bulk writing from
444 different clients came to OST. Accord the avaliable space of the
445 OST, the first req should be permitted, and the second one
446 should be denied by ENOSPC. But if the seconde arrived before
447 the first one is commited. The OST might wrongly permit second
448 writing, which will cause grant space > avaiable space.
451 Frequency : when client is evicted
453 Description: Grant might be wrongly erased when osc is evicted by OST
454 Details : when the import is evicted by server, it will fork another
455 thread ptlrpc_invalidate_import_thread to invalidate the
456 import, where the grant will be set to 0. While the original
457 thread will update the grant it got when connecting. So if
458 the former happened latter, the grant will be wrongly errased
459 because of this race.
464 Description: Checking Stale with correct fid
465 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
466 to check whether it is stale, because sometimes, we want the
467 enqueue happened anyway, and op_data.fid2 will not be initialized.
469 Severity : enhancement
471 Description: update patchless client
472 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
475 Frequency : only with 2.4 kernel
477 Description: random memory corruption
478 Details : size of struct ll_inode_info is to big for union inode.u and this
479 can be cause of random memory corruption.
484 Description: Memory leak in recovery
485 Details : Lov_mds_md was not free in an error handler in mds_create_object.
486 It should also check obd_fail before fsfilt_start, otherwise if
487 fsfilt_start return -EROFS,(failover mds during mds recovery).
488 then the req will return with repmsg->transno = 0 and rc = EROFS.
489 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
490 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
491 be freed no matter whether fsfilt_commit success or not.
494 Frequency : only with huge count clients
496 Description: Prevents from taking the superblock lock in llap_from_page for
498 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
499 for llap_from_page call in ll_removepage prevents from taking the
500 superblock lock for a soon died page.
505 Description: Not check open intent error before release open handle
506 Details : in some rare cases, the open intent error is not checked before
507 release open handle, which may cause
508 ASSERTION(open_req->rq_transno != 0), because it tries to release
509 the failed open handle.
514 Description: Set cat log bitmap only after create log success.
515 Details : in some rare cases, the cat log bitmap is set too early. and it
516 should be set only after create log success.
520 Description: Accessing a block bevice can re-enable I/O when Lustre is
521 tearing down a device.
522 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
527 Description: service threads may hog cpus when there are a lot of requests
529 Details : Insert cond_resched to give other threads a chance to use some of
535 Description: the cat log was not initialized in recovery
536 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
537 unlink log on mds will not be initialized until mds post
538 recovery. And also in mds post recovery, the unlink log will
539 initialization will be done asynchronausly, so there will be race
540 between add unlink log and unlink log initialization.
544 Description: brw_stats were being printed incorrectly
545 Details : brw_stats were being printed as log2 but all of them were not
546 recorded as log2. Also remove some code duplication arising from
547 filter_tally_{read,write}.
551 Frequency : rare, only in recovery.
552 Description: ASSERTION(req->rq_type != LI_POISON) failed
553 Details : imp_lock should be held while iterating over imp_sending_list for
554 prevent destroy request after get timeout in ptlrpc_queue_wait.
558 Description: replay-single.sh test 52 fails
559 Details : A lock's skiplist need to be cleanup when it being unlinked
560 from its resource list.
564 Description: Short directio read returns full requested size rather than
566 Details : Direct I/O operations should return actual amount of bytes
567 transferred rather than requested size.
569 Severity : enhancement
571 Description: metadata RPC reduction (e.g. for rm performance)
572 Details : decrease the amount of synchronous RPC between clients and servers
573 by canceling conflicing lock before the operation on the client side
574 and packing thier handles into the main operation RPC to server.
576 Severity : enhancement
578 Description: Async OSC create to avoid the blocking unnecessarily.
579 Details : If a OST has no remain object, system will block on the creating
580 when need to create a new object on this OST. Now, ways use
581 pre-created objects when available, instead of blocking on an
582 empty osc while others are not empty. If we must block, we block
583 for the shortest possible period of time.
585 --------------------------------------------------------------------------------
587 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
592 Frequency : on some architectures
594 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
595 Details : Uninitialized flags sometimes cause configuration commands to
599 Frequency : patchless clients only
601 Description: missing __iget() symbol export
602 Details : The __iget() symbol export is missing. To avoid the need for
603 this on patchless clients the deathrow inode reaper is turned
604 off, and we depend on the VM to clean up old inodes. This
605 dependency was during via the fix for bug 12181.
607 --------------------------------------------------------------------------------
609 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
611 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
612 INTEROPERATE with older versions automatically. In many cases a
613 special upgrade step is needed. Please read the
614 user documentation before upgrading any part of a live system.
615 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
616 version will not interoperate with 1.6 betas before beta5 (1.5.95).
617 * WARNING: Lustre configuration and startup changes are required with
618 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
624 Severity : enhancement
626 Description: MountConf
627 Details : Lustre configuration is now managed via mkfs and mount
628 commands instead of lmc and lconf. New obd types (MGS, MGC)
629 are added for dynamic configuration management. See
630 https://mail.clusterfs.com/wikis/lustre/MountConf for
633 Severity : enhancement
635 Description: dynamic OST addition
636 Details : OSTs can now be added to a live filesystem
638 Severity : enhancement
640 Description: startup order invariance
641 Details : MDTs and OSTs can be started in any order. Clients only
642 require the MDT to complete startup.
644 Severity : enhancement
646 Description: parallel, asynchronous orphan cleanup
647 Details : orphan cleanup is now performed in separate threads for each
648 OST, allowing parallel non-blocking operation.
650 Severity : enhancement
652 Description: optimized stripe assignment
653 Details : stripe assignments are now made based on ost space available,
654 ost previous usage, and OSS previous usage, in order to try
655 to optimize storage space and networking resources.
657 Severity : enhancement
659 Description: Permanently set tunables
660 Details : All writable /proc/fs/lustre tunables can now be permanently
661 set on a per-server basis, at mkfs time or on a live
664 Severity : enhancement
666 Description: Lustre message v2
667 Details : Add lustre message format v2.
669 Severity : enhancement
671 Description: client OST exclusion list
672 Details : Clients can be started with a list of OSTs that should be
673 declared "inactive" for known non-responsive OSTs.
677 Description: SPEC SFS validation failure on NFS v2 over lustre.
678 Details : Changes the blocksize for regular files to be 2x RPC size,
679 and not depend on stripe size.
681 Severity : enhancement
683 Description: Multiple MD RPCs in flight.
684 Details : Further unserialise some read-only MDS RPCs - learn about intents.
685 To avoid overly-overloading MDS, introduce a limit on number of
686 MDS RPCs in flight for a single client and add /proc controls
687 to adjust this limit.
689 Severity : enhancement
691 Description: client read/write statistics
692 Details : Add client read/write call usage stats for performance
693 analysis of user processes.
694 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
695 file access. extents_stats shows chunk size distribution.
696 extents_stats_per_process show chunk size distribution per
699 Severity : enhancement
701 Description: mds statistics
702 Details : Add detailed mds operations statistics in
703 /proc/fs/lustre/mds/*/stats.
707 Description: Failure of copying files with lustre special EAs.
708 Details : Client side always return success for setxattr call for lustre
709 special xattr (currently only "trusted.lov").
714 Description: Refcount LNET uuids
715 Details : The global LNET uuid list grew linearly with every startup;
716 refcount repeated list entries instead of always adding to
719 Severity : enhancement
721 Description: Dynamic service threads
722 Details : Within a small range, start extra service threads
723 automatically when the request queue builds up.
726 Frequency : mixed-endian client/server environments
728 Description: mixed-endian crashes
729 Details : The new msg_v2 system had some failures in mixed-endian
732 Severity : enhancement
734 Description: Easy OST removal
735 Details : OSTs can be permanently deactivated with e.g. 'lctl
736 conf_param lustre-OST0001.osc.active=0'
738 Severity : enhancement
740 Description: MGS proc entries
741 Details : Added basic proc entries for the MGS showing what filesystems
744 Severity : enhancement
746 Description: provide MGS failover
747 Details : Added config lock reacquisition after MGS server failover.
749 Severity : enhancement
751 Description: add Linux 2.4 support
752 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
756 Description: a large application tries to do I/O to the same resource and dies
758 Details : Check the req->rq_arrival time after the call to
759 ost_brw_lock_get(), but before we do anything about
760 processing it & sending the BULK transfer request. This
761 should help move old stale pending locks off the queue as
762 quickly as obd_timeout.
765 Frequency : when an incorrect nid is specified during startup
767 Description: ptlrpc connect to non-existant node causes kernel crash
768 Details : LNET can't be re-entered from an event callback, which
769 happened when we expire a message after the export has been
770 cleaned up. Instead, hand the zombie cleanup off to another
773 Severity : enhancement
775 Description: plain/inodebits lock performance improvement
776 Details : Grouping plain/inodebits in granted list by their request modes
777 and bits policy, thus improving the performance of search through
781 Frequency : only if OST filesystem is corrupted
783 Description: client incorrectly hits assertion in ptlrpc_replay_req()
784 Details : for a short time RPCs with bulk IO are in the replay list,
785 but replay of bulk IOs is unimplemented. If the OST filesystem
786 is corrupted due to disk cache incoherency and then replay is
787 started it is possible to trip an assertion. Avoid putting
788 committed RPCs into the replay list at all to avoid this issue.
791 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
793 Description: System hang on startup
794 Details : This bug allowed the liblustre (e.g. catamount) client to
795 return to the app before handling all startup RPCs. This
796 could leave the node unresponsive to lustre network traffic
797 and manifested as a server ptllnd timeout.
799 Severity : enhancement
801 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
802 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
803 boolean to control whether to print peer debug info when a
804 client's RPC times out.
807 Frequency : only for kernels with patches from Lustre below 1.4.3
809 Description: Remove old rdonly API
810 Details : Remove old rdonly API which unused from at least lustre 1.4.3
813 Frequency : only for devices with external journals
815 Description: Set external device read-only also
816 Details : During a commanded failover stop, we set the disk device
817 read-only while the server shuts down. We now also set any
818 external journal device read-only at the same time.
821 Frequency : when upgrading from 1.4 while trying to change parameters
823 Description: The wrong (new) MDC name was used when setting parameters for
824 upgraded MDT's. Also allows changing of OSC (and MDC)
825 parameters if --writeconf is specified at tunefs upgrade time.
828 Frequency : when setting specific ost indicies
830 Description: QOS code breaks on skipped indicies
831 Details : Add checks for missing OST indicies in the QOS code, so OSTs
832 created with --index need not be sequential.
834 Severity : enhancement
836 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
837 Details : The uninit_groups feature works in conjunction with the kernel
838 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
839 up the pass1 processing of e2fsck. This is a read-only feature
840 in ldiskfs2 only, so older kernels and current ldiskfs cannot
841 mount filesystems that have had this feature enabled.
843 Severity : enhancement
845 Description: Improve multi-block allocation algorithm to avoid fragmentation
846 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
847 allocation locality and avoid filesystem fragmentation.
849 ------------------------------------------------------------------------------
851 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
853 * Support for kernels:
854 2.4.21-47.0.1.EL (RHEL 3)
856 2.6.9-42.0.10.EL (RHEL 4)
857 2.6.12.6 vanilla (kernel.org)
858 2.6.16.27-0.9 (SLES 10)
859 * Recommended e2fsprogs version: 1.39.cfs5
861 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
865 Frequency : occasional, depends on client load and configuration
866 Bugzilla : 12181, 12203
867 Description: data loss for recently-modified files
869 Details : In some cases it is possible that recently written or created
870 files may not be written to disk in a timely manner (this should
871 normally be within 30s unless client IO load is very high).
872 The problem appears as zero-length files or files that are a
873 multiple of 1MB in size after a client crash or client eviction
874 that are missing data at the end of the file.
876 This problem is more likely to be hit on clients where files are
877 repeatedly created and unlinked in the same directory, clients
878 have a large amount of RAM, have many CPUs, the filesystem has
879 many OSTs, the clients are rebooted frequently, and/or the files
880 are not accessed by other nodes after being written.
882 The presence of the problem can be detected by looking at
883 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
884 smaller than the second (nr_unused) then dirty files will not
885 be flushed automatically to disk. "sync; sleep 10" should be
886 run several times on the node before unmounting it to update
887 Lustre (this is also safe to run on nodes without this problem).
889 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
890 that can cause this same problem, so customers using that kernel
891 also need to update the kernel in addition to Lustre. In order
892 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
894 It is normal that files written just before a client crash (less
895 than 30s) may not yet have been flushed to disk, even for local
899 Frequency : frequent on thin XT3 nodes
901 Description: UUID collision on thin XT3 Linux nodes
902 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
903 randomly, since we relied on an insufficiently-seeded PRNG.
908 Description: OSS hangs after "All ost request buffers busy"
909 Details : A deadlock between quota and journal operations caused OSS
910 hangs after printing "All ost request buffers busy."
913 Frequency : always on liblustre builds
915 Description: Cleanup compiler warnings on liblustre
918 Frequency : always on liblustre builds on XT3
920 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
924 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
925 > 32000 subdirectories
926 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
927 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
928 subdirectory count crosses 32000. This will aid e2fsck to
929 correctly handle more than 32000 subdirectories.
932 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
934 Description: System hang on startup
935 Details : This bug allowed the liblustre (e.g. catamount) client to
936 return to the app before handling all startup RPCs. This
937 could leave the node unresponsive to lustre network traffic
938 and manifested as a server ptllnd timeout.
940 Severity : enhancement
942 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
943 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
944 boolean to control whether to print peer debug info when a
945 client's RPC times out.
950 Description: make O_SYNC working on 2.6 kernels
951 Details : 2.6 kernels use different method for mark pages for write,
952 so need add a code to lustre for O_SYNC work.
957 Description: Failure to close file and release space on NFS
958 Details : Put inode details into lock acquired in ll_intent_file_open.
959 Use mdc_intent_lock in ll_intent_open to properly
960 detect all kind of errors unhandled by mdc_enqueue.
965 Description: proc file read during shutdown sometimes raced obd removal,
967 Details : Add lock to prevent obd access after proc file removal.
970 Frequency : Only for files larger than 4GB on 32-bit clients.
972 Description: improperly doing page alignment of locks
973 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
974 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
979 Description: RPCs being resent when they shouldn't be
980 Details : Some RPCs that should not be resent are being resent. This
981 can cause inconsistencies in the RPC state machine. Do not
982 resend such requests.
985 Frequency : rare, only with NFS export
987 Description: Crash on NFS re-export node
988 Details : under very unusual load conditions an assertion is hit in
989 ll_intent_file_open()
992 Frequency : only if OST filesystem is corrupted
994 Description: client incorrectly hits assertion in ptlrpc_replay_req()
995 Details : for a short time RPCs with bulk IO are in the replay list,
996 but replay of bulk IOs is unimplemented. If the OST filesystem
997 is corrupted due to disk cache incoherency and then replay is
998 started it is possible to trip an assertion. Avoid putting
999 committed RPCs into the replay list at all to avoid this issue.
1004 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1005 Details : Large single O_DIRECT read and write calls can fail to allocate
1006 a sufficiently large buffer to process the request. In case of
1007 allocation failure the allocation is retried with a smaller
1008 buffer and broken into smaller requests.
1010 Severity : enhancement
1012 Description: Add -o localflock option to simulate old noflock behaviour.
1013 Details : This will achieve local-only flock/fcntl locks coherentness.
1018 Description: versioning check is incomplete
1019 Details : Checking the version difference of client vs. server, report
1020 error if the gap is too big.
1025 Description: add support PG_writeback bit
1026 Details : add support for PG_writeback bit for Lustre, for more carefull
1027 work with page cache in 2.6 kernel. This also fix some deadlocks
1028 and remove hack for work O_SYNC with 2.6 kernel.
1030 Severity : enhancement
1032 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1033 Details : The uninit_groups feature works in conjunction with the kernel
1034 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1035 up the pass1 processing of e2fsck. This is a read-only feature
1036 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1037 mount filesystems that have had this feature enabled.
1039 Severity : enhancement
1041 Description: Improve multi-block allocation algorithm to avoid fragmentation
1042 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1043 allocation locality and avoid filesystem fragmentation.
1045 ------------------------------------------------------------------------------
1047 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1049 * Support for kernels:
1050 2.6.9-42.0.3.EL (RHEL 4)
1051 2.6.5-7.276 (SLES 9)
1052 2.4.21-47.0.1.EL (RHEL 3)
1053 2.6.12.6 vanilla (kernel.org)
1054 2.6.16.21-0.8 (SLES10)
1055 * Recommended e2fsprogs version: 1.39.cfs2-0
1057 * The backwards-compatible /proc/sys/portals symlink has been removed
1058 in this release. Before upgrading, please ensure that you change
1059 any configuration scripts or /etc/sysctl.conf files that access
1060 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1061 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1062 in advance of the upgrade on any system running Lustre 1.4.6 or
1063 newer, since /proc/sys/lnet was added in that version.
1064 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1068 Frequency : only when quota is used
1070 Description: avoid scanning export list for quota master
1071 Details : Change the algorithms to avoid scanning export list in order
1072 to improve the efficiency.
1075 Frequency : MDS failover only, very rarely
1077 Description: "went back in time" messages on mds failover
1078 Details : The greatest transno may be lost when the current operation
1079 finishes with an error (transno==0) and the client's last_rcvd
1080 record is over-written. Save the greatest transno in the
1081 mds_last_transno for this case.
1084 Frequency : always for specific kernels and striping counts
1086 Description: client may get "Matching packet too big" without ACL support
1087 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1088 when trying to access files in certain configurations. The
1089 clients should in fact be denied when mounting because they do
1090 not understand ACLs.
1093 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1095 Description: many clients connecting with IO in progress causes connect timeouts
1096 Details : Avoid synchronous journal commits to avoid delays caused by many
1097 clients connecting/disconnecting when bulk IO is in progress.
1098 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1099 OST_IO_PORTAL to avoid delays behind potentially many pending
1103 Frequency : occasionally with multiple writers to a single file
1105 Description: shared writes to file may result in wrong size reported by stat()
1106 Details : Allow growing of kms when extent lock is cancelled
1109 Frequency : always with random mmap IO to multi-striped file
1111 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1112 Details : Only if the hole is at the end of OST object so that kms is too
1113 small. Fix is to increase kms accordingly in ll_nopage.
1116 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1118 Description: writes to a missing object would leak memory on the OST
1119 Details : If there is an inconsistency between the MDS and OST filesystems,
1120 such that the MDS references an object that doesn't exist, writes
1121 to that object will leak memory due to incorrect cleanup in the
1122 error handling path, eventually running out of memory on the OST.
1127 Description: Creating too long symlink causes lustre errors
1128 Details : Check symlink and name lengths before sending requests to MDS.
1131 Frequency : only if flock is enabled (not on by default)
1133 Description: posix locks not released on fd closure on 2.6.9+
1134 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1135 kernels, this caused such locks not to be released on fd close and
1136 then assertions on fs unmount about still used locks.
1139 Frequency : MDS failover only, very rarely
1141 Description: clients may get ASSERTION(granted_lock != NULL)
1142 Details : When request was taking a long time, and a client was resending
1143 a getattr by name lock request. The were multiple lock requests
1144 with the same client lock handle and
1145 mds_getattr_name->fixup_handle_for_resent_request found one of the
1146 lock handles but later failed with ASSERTION(granted_lock != NULL).
1151 Description: handle->h_buffer_credits > 0, assertion failure
1152 Details : h_buffer_credits is zero after truncate, causing assertion
1153 failure. This patch extends the transaction or creates a new
1157 Frequency : NFS re-export or patchless client
1158 Bugzilla : 11179, 10796
1159 Description: Crash on NFS re-export node (__d_move)
1160 Details : We do not want to hash the dentry if we don't have a lock.
1161 But if this dentry is later used in d_move, we'd hit uninitialised
1162 list head d_hash, so we just do this to init d_hash field but
1163 leave dentry unhashed.
1166 Frequency : NFS re-export or patchless client
1168 Description: NFS exports has problem with symbolic link
1169 Details : lustre client didn't properly install dentry when re-exported
1170 to NFS or running patchless client.
1173 Frequency : NFS re-export or patchless client
1175 Description: Various nfs/patchless fixes.
1176 Details : fixes reuse disconected alias for lookup process - this fixes
1177 warning "find_exported_dentry: npd != pd",
1178 fix permission error with open files at nfs.
1179 fix apply umask when do revalidate.
1182 Frequency : occasional
1184 Description: Crash on NFS re-export node
1185 Details : calling clear_page() on the wrong pointer triggered oops in
1186 generic_mapping_read().
1189 Frequency : rarely, using O_DIRECT IO
1191 Description: unaligned directio crashes client with LASSERT
1192 Details : check for unaligned buffers before trying any requests.
1195 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1197 Description: stale data returned from RAID cache
1198 Details : If only a small amount of IO is done to the RAID device before
1199 reading it again it is possible to get stale data from the RAID
1200 cache instead of reading it from disk.
1203 Frequency : always for sles10 kernel
1205 Description: sles10 support
1206 Details : ll_follow_link: compile fixes and using of nd_set_link
1207 under newer kernels.
1210 Frequency : depends on arch, kernel and compiler version, always on sles10
1213 Description: recursive or deep enough symlinks cause stack overflow
1214 Details : getting rid of large stack-allocated variable in
1218 Frequency : depends on hardware
1220 Description: lustre write performance loss in the SLES10 kernel
1221 Details : the performance loss is caused by using of write barriers in the
1222 ext3 code. The SLES10 kernel turns barrier support on by
1223 default. The fix is to undo that change for ldiskfs.
1225 ------------------------------------------------------------------------------
1227 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1229 * Support for kernels:
1230 2.6.9-42.0.3EL (RHEL 4)
1231 2.6.5-7.276 (SLES 9)
1232 2.4.21-47.0.1.EL (RHEL 3)
1233 2.6.12.6 vanilla (kernel.org)
1237 Frequency : quota enabled and large files being deleted
1239 Description: releasing more than 4GB of quota at once hangs OST
1240 Details : If a user deletes more than 4GB of files on a single OST it
1241 will cause the OST to spin in an infinite loop. Release
1242 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1247 Description: statfs data retrieved from /proc may be stale or zero
1248 Details : When reading per-device statfs data from /proc, in the
1249 {kbytes,files}_{total,free,avail} files, it may appear
1250 as zero or be out of date.
1253 Frequency : systems with MD RAID1 external journal devices
1255 Description: lconf's call to blkid is confused by RAID1 journal devices
1256 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1257 of returning all block devices that match the journal UUID.
1260 Frequency : always, for aggregate stripe size over 4GB
1262 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1263 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1264 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1267 Frequency : always if "lfs find" used on a local file/directory
1269 Description: "lfs find" segfaults if used on a local file/directory
1270 Details : The case where a directory component was not specified wasn't
1271 handled correctly. Handle this properly.
1274 Frequency : always on ppc64
1276 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1277 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1278 perform endianness conversion.
1281 Frequency : rarely (truncate to non-zero file size after write under load)
1282 Bugzilla : 10730, 10687
1283 Description: Files padded with zeros to next 4K multiple
1284 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1285 it is possible that files that are truncated to a non-zero size
1286 immediately after being written are filled with zero bytes beyond
1287 the truncated size. No file data is lost.
1289 Severity : enhancement
1291 Description: Allow recovery/failover for liblustre clients.
1292 Details : liblustre clients were unaware of failover configurations until
1295 Severity : enhancement
1297 Description: user file locks should fail when not mounting with flock option
1298 Details : Set up an error-returning stub in ll_file_operations.lock field
1299 to prevent incorrect behaviour when client is mounted without
1300 flock option. Also, set up properly f_op->flock field for
1304 Frequency : always on ia64
1306 Description: "lfs df" loops on printing out MDS statfs information
1307 Details : The obd_ioctl_data was not initialized and in some systems
1308 this caused a failure during the ioctl that did not return
1309 an error. Initialize the struct and return an error on failure.
1312 Frequency : SLES 9 only
1314 Description: Error of copying files with lustre special EAs as root
1315 Details : Client side always return success for setxattr call for lustre
1316 special xattr (currently only "trusted.lov").
1319 Frequency : rarely on clusters with both ia64+i386 clients
1321 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1322 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1323 clients are doing concurrent writes to the same file it is
1324 possible that the ia64 clients may LASSERT because the OST
1325 extent locks are not PAGE_SIZE aligned. Ensure that grown
1326 locks are always aligned on the request boundary.
1329 Frequency : specific use, occasional
1331 Description: Overwriting in use executable truncates on-disk binary image
1332 Details : If one node attempts to overwrite an executable in use by
1333 another node, we now correctly return ETXTBSY instead of
1334 truncating the file.
1336 Severity : enhancement
1338 Description: Async OSC create to avoid the blocking unnecessarily.
1339 Details : If a OST has no remain object, system will block on the creating
1340 when need to create a new object on this OST. Now, ways use
1341 pre-created objects when available, instead of blocking on an
1342 empty osc while others are not empty. If we must block, we block
1343 for the shortest possible period of time.
1348 Description: chmod on Lustre root is propagated to other clients
1349 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1350 invalid by the follow_mount time.
1353 Frequency : liblustre clients only
1355 Description: Race in 'instant cancel' lock handling could lead to such locks
1356 never to be granted in case of SMP MDS
1357 Details : Do not destroy not yet granted but cbpending locks in
1361 Frequency : replay/resend of open
1363 Description: non null lock assetion failure in mds_intent_policy
1364 Details : Trying to replay/resend lockless open requests resulted in
1365 mds_open() returning 0 with no lock. Now it sets a flag if
1366 a lock is going to be returned.
1368 Severity : enhancement
1370 Description: Checksum enhancements
1371 Details : New checksum enhancements allow for resending RPCs that failed
1374 Severity : enhancement
1376 Description: Tunables on number of dirty pages in cacche
1377 Details : Allow to set limit on number of dirty pages cached.
1382 Description: client crash on unmount - lock still has references
1383 Details : In some error handling cases it was possible to leak a lock
1384 reference on a client while accessing a file. This was not
1385 harmful to the client during operation, but would cause the
1386 client to crash when the filesystem is unmounted.
1389 Frequency : specific case, rare
1391 Description: ETXTBSY on mds though file not in use by client
1392 Details : ETXTBSY is no longer incorrectly returned when attempting to
1393 chmod or chown a directory that the user previously tried to
1394 execute or a currently-executing binary.
1397 Frequency : extremely rare except on liblustre-based clients
1399 Description: Lustre space not freed when files are deleted
1400 Details : Clean up open-unlinked files after client eviction. Previously
1401 the unlink was skipped and the files remained as orphans.
1406 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1407 Details : In some cases it was possible to send a blocking callback to
1408 a client doing a glimpse, even though that client didn't get
1409 a lock granted. When the glimpse lock is cancelled on the OST
1410 the freed lock is left on the waiting list and corrupted the list.
1413 Frequency : all core dumps
1415 Description: Broke core dumps to lustre
1416 Details : Negative dentry may be unhashed if parent does not have UPDATE
1417 lock, but some callers, e.g. do_coredump, expect dentry to be
1418 hashed after successful create, hash it in ll_create_it.
1420 ------------------------------------------------------------------------------
1422 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1424 * Support for kernels:
1425 2.6.9-42.0.2.EL (RHEL 4)
1426 2.6.5-7.276 (SLES 9)
1427 2.4.21-40.EL (RHEL 3)
1428 2.6.12.6 vanilla (kernel.org)
1432 Frequency : always on RHEL 3
1434 Description: Number of open files grows over time
1435 Details : The number of open files grows over time, whether or not
1436 Lustre is started. This was due to a filp leak introduced
1437 by one of our kernel patches.
1439 ------------------------------------------------------------------------------
1441 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1443 * Support for kernels:
1444 2.6.9-42.EL (RHEL 4)
1445 2.6.5-7.276 (SLES 9)
1446 2.4.21-40.EL (RHEL 3)
1447 2.6.12.6 vanilla (kernel.org)
1452 Bugzilla : 5719, 9635, 9792, 9684
1453 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1454 Details : If a server is under heavy load and cannot reply to new
1455 connection requests before the client resends the (re)connect,
1456 the connection handling code can behave badly if two service
1457 threads are concurrently handing separate (re)connections from
1458 the same client. Add better locking to the connection handling
1459 code, and ensure that only a single connection will be processed
1460 for a given client UUID, even if the lock is dropped.
1462 Severity : enhancement
1464 Description: add TCP zero-copy support to kernel
1465 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1466 sends if the hardware supports scatter-gather and checksumming.
1467 This allows socklnd to do client-write and server-read more
1468 efficiently and reduce CPU utilization from skbuf copying.
1471 Frequency : only if NFS exporting from client
1473 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1474 Details : If setting attributes on a file created under NFS that had
1475 never been opened it would be possible to oops the client
1476 if the file had no objects.
1479 Frequency : always for liblustre
1481 Description: liblustre client does MDS+OSTs setattr RPC for each write
1482 Details : When doing a write from a liblustre client, the client
1483 incorrectly issued an RPC to the MDS and each OST the file was
1484 striped over in order to update the timestamps. When writing
1485 with small chunks and many clients this could overwhelm the MDS
1486 with RPCs. In all cases it would slow down the write because
1487 these RPCs are unnecessary.
1489 Severity : enhancement
1491 Description: allow number of MDS service threads to be changed at module load
1492 Details : It is now possible to change the number of MDS service threads
1493 running. Adding "options mds mds_num_threads={N}" to the MDS's
1494 /etc/modprobe.conf will set the number of threads for the next
1495 time Lustre is restarted (assuming the "mds" module is also
1496 reloaded at that time). The default number of threads will
1497 stay the same, 32 for most systems.
1502 Description: OST crash if filesystem is unformatted or corrupt
1503 Details : If an OST is started on a device that has never been formatted
1504 or if the filesystem is corrupt and cannot even mount then the
1505 error handling cleanup routines would dereference a NULL pointer.
1510 Description: NULL pointer deref in llap_from_page.
1511 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1512 incorrect) mapping if the page was truncated/reclaimed while it was
1513 searched for. Check for this condition and skip such pages when
1514 doing readahead. Introduce extra check to llap_from_page() to
1515 verify page->mapping->host is non-NULL (so page is not anonymous).
1518 Frequency : Sometimes when using sys_sendfile
1520 Description: "page not covered by a lock" warnings from ll_readpage
1521 Details : sendfile called ll_readpage without right page locks present.
1522 Now we introduced ll_file_sendfile that does necessary locking
1523 around call to generic_file_sendfile() much like we do in
1527 Frequency : with certain MDS communication failures at client mount time
1529 Description: NULL pointer deref after failed client mount
1530 Details : a client connection request may delayed by the network layer
1531 and not be sent until after the PTLRPC layer has timed out the
1532 request. If the client fails the mount immediately it will try
1533 to clean up before the network times out the request. Add a
1534 reference from the request import to the obd device and delay
1535 the cleanup until the network drops the request.
1538 Frequency : occasionally during client (re)connect
1540 Description: assertion failure during client (re)connect
1541 Details : processing a client connection request may be delayed by the
1542 client or server longer than the client connect timeout. This
1543 causes the client to resend the connection request. If the
1544 original connection request is replied in this interval, the
1545 client may trip an assertion failure in ptlrpc_connect_interpret()
1546 which thought it would be the only running connect process.
1549 Frequency : only with obd_echo servers and clients that are rebooted
1551 Description: kernel BUG accessing uninitialized data structure
1552 Details : When running an obd_echo server it did not start the ping_evictor
1553 thread, and when a client was evicted an uninitialized data
1554 structure was accessed. Start the ping_evictor in the RPC
1555 service startup instead of the OBD startup.
1557 Severity : enhancement
1558 Bugzilla : 10193 (patchless)
1559 Description: Remove dependency on various unexported kernel interfaces.
1560 Details : No longer need reparent_to_init, exit_mm, exit_files,
1561 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1564 Frequency : rare (only users of deprecated and unsupported LDAP config)
1566 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1567 Details : LDAP apparently contains 'lustreName' attributes instead of
1568 'name'. A simple remapping of the name is sufficient.
1571 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1573 Description: waiting_locks_callback trips kernel BUG if client is evicted
1574 Details : Running with the dump_on_timeout debug flag turned on makes
1575 it possible that the waiting_locks_callback() can try to dump
1576 the Lustre kernel debug logs from an interrupt handler. Defer
1577 this log dumping to the expired_lock_main() thread.
1579 Severity : enhancement
1581 Description: Support NFS exporting on 2.6 kernels.
1582 Details : Implement non-rawops metadata methods for NFS server to use without
1583 changing NFS server code.
1586 Frequency : very rare (synthetic metadata workload only)
1588 Description: two racing renames might cause an MDS thread to deadlock
1589 Details : Running the "racer" program may cause one MDS thread to rename
1590 a file from being the source of a rename to being the target of
1591 a rename at exactly the same time that another thread is doing
1592 so, and the second thread has already enqueued these locks after
1593 doing a lookup of the target and is trying to relock them in
1594 order. Ensure that we don't try to re-lock the same resource.
1597 Frequency : only very large systems with liblustre clients
1599 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1600 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1601 all OSTs in parallel. This allows the eviction of stale liblustre
1602 clients to proceed much faster than if they were done in series,
1603 and also offers similar improvements for other set_info RPCs.
1608 Description: excessive CPU usage during initial read phase on client
1609 Details : During the initial read phase on a client, it would agressively
1610 retry readahead on the file, consuming too much CPU and impacting
1611 performance (since 1.4.5.8). Improve the readahead algorithm
1612 to avoid this, and also improve some other common cases (read
1613 of small files in particular, where "small" is files smaller than
1614 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1619 Description: MDS crash when receiving packet with unknown intent.
1620 Details : Do not LBUG in unknown intent case, just return -EFAULT
1622 Severity : enhancement
1623 Bugzilla : 9293, 9385
1624 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1625 Details : Do not serialize getattr (non-intent version) and statfs.
1628 Frequency : occasional, when OST network is overloaded/intermittent
1630 Description: client evicted by OST after bulk IO timeout
1631 Details : If a client sends a bulk IO request (read or write) the OST
1632 may evict the client if it is unresposive to its data GET/PUT
1633 request. This is incorrect if the network is overloaded (takes
1634 too long to transfer the RPC data) or dropped the OST GET/PUT
1635 request. There is no need to evict the client at all, since
1636 the pinger and/or lock callbacks will handle this, and the
1637 client can restart the bulk request.
1640 Frequency : Always when mmapping file with no objects
1642 Description: client crashes when mmapping file with no objects
1643 Details : Check that we actually have objects in a file before doing any
1644 operations on objects in ll_vm_open, ll_vm_close and
1650 Description: Request leak when working with deleted CWD
1651 Details : Introduce advanced request refcount tracking for requests
1652 referenced from lustre intent.
1654 Severity : Enhancement
1656 Description: Cache open file handles on client.
1657 Details : MDS now will return special lock along with openhandle, if
1658 requested and client is allowed to hold openhandle, even if unused,
1659 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1660 closing files for every read/write openration.
1662 Severity : Enhancement
1664 Description: Cache open negative dentries on client when possible.
1665 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1666 negative dentries on lock revocation.
1671 Description: Remounting a client read-only wasn't possible with a zconf mount
1672 Details : It wasn't possible to remount a client read-only with llmount.
1674 Severity : enhancement
1675 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1676 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1677 MPICH is included to add Lustre-specific ADIO interfaces.
1678 This is based closely on the UFS ADIO layer and only differs
1679 in file creation, in order to allow the OST striping to be set.
1680 This is user-contributed code and not supported by CFS.
1685 Description: extended inode attributes (immutable, append-only) work improperly
1686 when 2.4 and 2.6 kernels are used on client/server or vice versa
1687 Details : Introduce kernel-independent values for these flags.
1689 Severity : enhancement
1692 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1693 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1694 granularity. Now, Fractional MB granularity is be supported,
1695 this is very useful for low memory system.
1697 Severity : enhancement
1699 Description: Getattr by fid
1700 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1701 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1702 have locks and avoids getattr rpc requests for it on every stat.
1705 Frequency : Always, for filesystems larger than 2TB
1707 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1708 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1709 filesystems larger than 2TB to be mounted with mballoc enabled.
1712 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1714 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1715 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1716 filesystems larger than 2TB is mounted then the kernel will
1717 silently corrupt the start of the filesystem. CONFIG_LBD is
1718 enabled for all CFS-supported kernels, but the possibility of
1719 this happening with a modified kernel config exists.
1721 Severity : enhancement
1723 Description: add client O_DIRECT support for 2.6 kernels
1724 Details : It is now possible to do O_DIRECT reads and writes to files
1725 in the Lustre client mountpoint on 2.6 kernel clients.
1727 Severity : enhancement
1729 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1730 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1731 parallel, not waiting for response from every OST before sending
1732 a rpc to the next OST.
1737 Description: setattr vs write race when updating file timestamps
1738 Details : Client processes that update a file timestamp into the past
1739 right after writing to the file (e.g. tar) it is possible that
1740 the updated file modification time can be reset to the current
1741 time due to a race between processing the setattr and write RPC.
1743 Severity : enhancement
1745 Description: Bring 'lfs find' closer in line with regular Linux find.
1746 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1747 -print0 options and obtains all the needed info through the lustre
1750 Severity : enhancement
1752 Description: support up to 1024 configured devices on one node
1753 Details : change obd_dev array from statically allocated to dynamically
1754 allocated structs as they are first used to reduce memory usage
1759 Description: Flush dirty partially truncated pages during truncate
1760 Details : Immediatelly flush partially truncated pages in filter_setattr,
1761 this way we completely avoid having any pages in page cache on OST
1762 and can retire ugly workarounds during writes to flush such pages.
1767 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1768 Details : protect inode from truncation within vfs_unlink() context
1769 just take a reference before calling vfs_unlink() and release it
1770 when parent's i_sem is free.
1775 Description: last_id value checked outside lock on OST caused LASSERT failure
1776 Details : If there were multiple MDS->OST object precreate requests in
1777 flight, it was possible that the OST's last object id was checked
1778 outside a lock and incorrectly tripped an assertion. Move checks
1779 inside locks, and discard old precreate requests.
1782 Frequency : always, if extents are used on OSTs
1784 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1785 Details : OSTs using the extents format would not zero the high 16 bits of
1786 the index physical block number. This is not a problem for any
1787 OST filesystems smaller than 16TB, and no kernels support ext3
1788 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1789 new/modified files) and can be fixed for existing filesystems
1790 with e2fsprogs-1.39-cfs1.
1795 Description: import connection selection may be incorrect if timer wraps
1796 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1797 import connections to be ignored if the 32-bit jiffies counter
1798 wraps. Use a 64-bit jiffies counter.
1801 Frequency : very large clusters immediately after boot
1803 Description: LNET request buffers exhausted under heavy short-term load
1804 Details : If a large number of client requests are generated on a service
1805 that has previously never seen so many requests it is possible
1806 that the request buffer growth cannot keep up with the spike in
1807 demand. Instead of dropping incoming requests, they are held in
1808 the LND until the RPC service can accept more requests.
1811 Frequency : Sometimes during replay
1813 Description: Assertion failure in ll_local_open after replay.
1814 Details : If replay happened on an open request reply before we were able
1815 to set replay handler, reply will become not swabbed tripping the
1816 assertion in ll_local_open. Now we set the handler right after
1817 recognising of open request
1820 Frequency : very rare
1822 Description: kernel reports "badness in vsnprintf"
1823 Details : Reading from the "recovery_status" /proc file in small chunks
1824 may cause a negative length in lprocfs_obd_rd_recovery_status()
1825 call to vsnprintf() (which is otherwise harmless). Exit early
1826 if there is no more space in the output buffer.
1828 Severity : enhancement
1830 Description: clear OBD RPC statistics by writing to them
1831 Details : It is now possible to clear the OBD RPC statistics by writing
1832 to the "stats" file.
1837 Description: Client mtime is not the same on different clients after utimes
1838 Details : In some cases, the client was using the utimes() syscall on
1839 a file cached on another node. The clients now validate the
1840 ctime from the MDS + OSTs to determine which one is right.
1845 Description: Inability to activate failout mode
1846 Details : lconf script incorrectly assumed that in pythong string's numeric
1847 value is used in comparisons.
1850 Frequency : always with multiple stripes per file
1852 Description: Inefficient object allocation for mutli-stripe files
1853 Details : When selecting which OSTs to stripe files over, for files with
1854 a stripe count that divides evenly into the number of OSTs,
1855 the MDS is always picking the same starting OST for each file.
1856 Return the OST selection heuristic to the original design.
1861 Description: mount failures may take full timeout to return an error
1862 Details : Under some heavy load conditions it is possible that a
1863 failed mount can wait for the full obd_timeout interval,
1864 possibly several minutes, before reporting an error.
1865 Instead return an error as soon as the status is known.
1867 Frequency : quota enabled and large files being deleted
1869 Description: releasing more than 4GB of quota at once hangs OST
1870 Details : If a user deletes more than 4GB of files on a single OST it
1871 will cause the OST to spin in an infinite loop. Release
1872 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1877 Description: statfs data retrieved from /proc may be stale or zero
1878 Details : When reading per-device statfs data from /proc, in the
1879 {kbytes,files}_{total,free,avail} files, it may appear
1880 as zero or be out of date.
1883 Frequency : systems with MD RAID1 external journal devices
1885 Description: lconf's call to blkid is confused by RAID1 journal devices
1886 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1887 of returning all block devices that match the journal UUID.
1890 Frequency : always, for aggregate stripe size over 4GB
1892 Description: assertion fails when trying to use 4GB stripe size
1893 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1894 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1897 Frequency : always on ppc64
1899 Description: the first write on an ext3 filesystem with mballoc got stuck
1900 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1901 perform endianness conversion.
1903 ------------------------------------------------------------------------------
1905 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1907 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1908 INTEROPERATE with older versions automatically. Please read the
1909 user documentation before upgrading any part of a live system.
1910 * WARNING: Lustre networking configuration changes are required with
1911 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1914 * Support for kernels:
1915 2.6.9-22.0.2.EL (RHEL 4)
1916 2.6.5-7.244 (SLES 9)
1917 2.6.12.6 vanilla (kernel.org)
1920 Severity : enhancement
1921 Bugzilla : 7981/8208
1922 Description: Introduced Lustre Networking (LNET)
1923 Details : LNET is new networking infrastructure for Lustre, it includes
1924 a reorganized network configuration mode (see the user
1925 documentation for full details) as well as support for routing
1926 between different network fabrics. Lustre Networking Devices
1927 (LNDs) for the supported network fabrics have also been
1928 created for this new infrastructure.
1930 Severity : enhancement
1931 Description: Introduced Access control lists
1932 Details : clients can set ACLs on files and directories in order to have
1933 more fine-grained permissions than the standard Unix UGO+RWX.
1934 The MDS must be started with the "-o acl" mount option.
1936 Severity : enhancement
1937 Description: Introduced filesystem quotas
1938 Details : Administrators may now establish per-user quotas on the
1941 Severity : enhancement
1943 Description: Configuration change for the XT3
1944 The PTLLND is now used to run Lustre over Portals on the XT3
1945 The configure option(s) --with-cray-portals are no longer used.
1946 Rather --with-portals=<path-to-portals-includes> is used to
1947 enable building on the XT3. In addition to enable XT3 specific
1948 features the option --enable-cray-xt3 must be used.
1953 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1954 Details : A race between allocating a new llcd and re-getting the llcd_lock
1955 allowed another thread to grab newly-allocated llcd.
1957 Severity : enhancement
1959 Description: 2.6 OST async journal commit and locking fix to improve performance
1960 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1961 2.6 kernels are now async as they already were in 2.4 kernels so
1962 that they can commit concurrently with the network bulk transfer.
1963 For block-allocated files the filter allocation semaphore is held
1964 to avoid filesystem fragmentation during allocation. BKL lock
1965 removed for 2.6 xattr operations where it is no longer needed.
1970 Description: lconf incorrectly determined whether two IP networks could talk
1971 Details : In some more complicated routing and multiple-network
1972 configurations, lconf will avoid trying to make a network
1973 connection to a disjoint part of the IP space. It was doing the
1974 math incorrectly for one set of cases.
1979 Description: Fix for potential infinite loop processing records in an llog.
1980 Details : If an llog record is corrupted/zeroed, it is possible to loop
1981 forever in llog_process(). Validate the llog record length
1982 and skip the remainder of the block on error.
1985 Frequency : occasional (liblustre only)
1987 Description: liblustre could not open files whose last component is a symlink
1988 Details : sysio_path_walk() would incorrectly pass the open intent to
1989 intermediate path components.
1992 Frequency : rare (liblustre only with non-standard tuning)
1993 Bugzilla : 7201 (7350)
1994 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1995 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1996 after it has been released, only do this for referenced locks
1998 Severity : enhancement
2000 Description: specifying an (invalid) directory default stripe_size of -1
2001 would reset the directory default striping
2002 Details : stripe_size -1 was used internally to signal directory stripe
2003 removal, now use "all default" to signal dir stripe removal
2004 as a directory striping of "all default" is not useful
2007 Frequency : common for large clusters running liblustre clients
2009 Description: doing an ls when liblustre clients are running is slow
2010 Details : sending a glimpse AST to a liblustre client waits for every AST
2011 to time out, as liblustre clients will not respond. Since they
2012 cannot cache data we refresh the OST lock LVB from disk instead.
2014 Severity : enhancement
2016 Description: doing an ls at the same time as file IO can be slow
2017 Details : enqueue and other "small" requests can be blocked behind many
2018 large IO requests. Create a new OST IO portal for non-IO
2019 requests so they can be processed faster.
2022 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2024 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2025 created file with mode 000
2026 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2027 allow an ftruncate() as a non-root user to a file with mode 000.
2028 Lustre now respects this flag to disable mode checks when
2029 truncating a file owned by the user
2032 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2035 Description: Revoking locks from clients that went dead or catatonic might take
2037 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2038 cancellation of such locks instant on servers without waiting for
2039 any reply from clients. Clients drops these locks when cancel
2040 notification from server is received without replying.
2043 Frequency : liblustre-only, when liblustre client dies or becomes busy
2045 Description: Doing ls on Linux clients can take a long time with active
2047 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2048 granting such locks to it in the first place if possible. Locks
2049 are taken by proxy on the OST during the read or write and
2050 dropped immediately afterward. Add connect flags handling, do
2051 not grant locks to liblustre clients for glimpse ASTs.
2053 Severity : enhancement
2055 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2056 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2057 sequential reads and behaves badly with applications doing random
2058 reads. Improve it by reading ahead at least read region, and
2059 avoiding excessive large RPC for small reads.
2061 Severity : enhancement
2063 Description: Creating more than 1000 files for a single job may cause a load
2064 imbalance on the OSTs if there are also a large number of OSTs.
2065 Details : qos_prep_create() uses an OST index reseed value that is an
2066 even multiple of the number of available OSTs so that if the
2067 reseed happens in the middle of the object allocation it will
2068 still utilize the OSTs as uniformly as possible.
2073 Description: OST or MDS may oops in ping_evictor_main()
2074 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2075 but doesn't restart at beginning of obd_exports_timed list
2078 Severity : enhancement
2080 Description: improve by-nid export eviction on the MDS and OST
2081 Details : allow multiple exports with the same NID to be evicted at one
2082 time without re-searching the exports list.
2085 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2087 Description: MDS may oops in groups_free()
2088 Details : in rare race conditions a newly allocated group_info struct is
2089 freed again, and this can be NULL. The 2.4 compatibility code
2090 for groups_free() checked for a NULL pointer, but 2.6 did not.
2093 Frequency : common for liblustre clients doing little filesystem IO
2094 Bugzilla : 9352, 7313
2095 Description: server may evict liblustre clients accessing contended locks
2096 Details : if a client is granted a lock or receives a completion AST
2097 with a blocking AST already set it would not reply to the AST
2098 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2101 Frequency : lfs setstripe, only systems with more than 160 OSTs
2103 Description: unable to set striping with a starting offset beyond OST 160
2104 Details : llapi_create_file() incorrectly limited the starting stripe
2105 index to the maximum single-file stripe count.
2108 Frequency : LDAP users only
2110 Description: lconf did not handle in-kernel recovery with LDAP properly
2111 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2113 Severity : enhancement
2115 Description: bind OST threads to NUMA nodes to improve performance
2116 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2117 node and do their allocations there to localize memory access
2119 Severity : enhancement
2121 Description: llmount can determine client NID directly from Myrinet (GM)
2122 Details : the client NID code from gmnalnid was moved directly into
2123 llmount, removing the need to use this or specifying the
2124 client NID explicitly when mounting GM clients with zeroconf
2127 Frequency : if client is started with down MDS
2129 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2130 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2131 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2132 ptlrpc_import_delayed_req() because we didn't check if the
2133 request was interrupted, and we also didn't break out of the
2134 event loop if there was no timeout
2139 Description: data loss during non-page-aligned writes to a single file from
2140 both multiple nodes and multiple threads on one node at same time
2141 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2142 inconsistency led to false short-reads, that were cached and later
2143 used by ->prepare_write() to fill in partially written page,
2144 leading to data loss.
2147 Frequency : always, if lconf --abort_recovery used
2149 Description: lconf --abort_recovery fails with 'Operation not supported'
2150 Details : lconf was attempting to abort recovery on the MDT device and not
2153 Severity : enhancement
2155 Description: remove cleanup logs
2156 Details : replace lconf-generated cleanup logs with lustre internal
2157 cleanup routines. Eliminates the need for client-cleanup and
2160 Severity : enhancement
2162 Description: add support for EAs (user and system) on lustre filesystems
2163 Details : it is now possible to store extended attributes in the Lustre
2164 client filesystem, and with the user_xattr mount option it
2165 is possible to allow users to store EAs on their files also
2167 Severity : enhancement
2169 Description: Add possibility (config option) to show minimal available OST free
2171 Details : When compiled with --enable-mindf configure option, statfs(2)
2172 (and so, df) will return least minimal free space available from
2173 all OSTs as amount of free space on FS, instead of summary of
2174 free spaces of all OSTs.
2176 Severity : enhancement
2178 Description: do not expand extent locks acquired on OST-side
2179 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2180 by server: they are not cached anyway.
2183 Frequency : when mmap is used/binaries executed from Lustre
2185 Description: Unmmap pages before throwing them away from read cache.
2186 Details : llap_shrink cache now attempts to unmap pages before discarding
2187 them (if unmapping failed - do not discard). SLES9 kernel has
2188 extra checks that trigger if this unmapping is not done first.
2193 Description: lconf didn't resolve symlinks before checking to see whether a
2194 given mountpoint was already in use
2197 Frequency : when migrating failover services
2198 Bugzilla : 6395, 9514
2199 Description: When migrating a subset of services from a node (e.g. failback
2200 from a failover service node) the remaining services would
2201 time out and evict clients.
2202 Details : lconf --force (implied by --failover) sets the global obd_timeout
2203 to 5 seconds in order to quickly disconnect, but this caused
2204 other RPCs to time out too quickly. Do not change the global
2205 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2207 Severity : enhancement
2208 Frequency : if MDS is started with down OST
2209 Bugzilla : 9439,5706
2210 Description: Allow startup/shutdown of an MDS without depending on the
2211 availability of the OSTs.
2212 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2213 Add appropriate locking and lov-osc refcounts for safe
2214 cleaning. Add osc abort_inflight calls in case the
2215 synchronize never started.
2218 Frequency : occasional (Cray XT3 only)
2220 Description: root not authorized to access files in CRAY_PORTALS environment
2221 Details : The client process capabilities were not honoured on the MDS in
2222 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2223 been accessed by an authorized user then root was able to access
2224 the file on the local client also. The root user capabilities
2225 are now allowed on the MDS, as this environment has secure UID.
2228 Frequency : occasional
2230 Description: ldiskfs "too long searching" message happens too often
2231 Details : A debugging message (otherwise harmless) prints too often on
2232 the OST console. This has been reduced to only happen when
2233 there are fragmentation problems on the filesystem.
2238 Description: Division by zero in statfs when all OSCs are inactive
2239 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2240 lov_statfs divides by value returned by lov_get_stripecnt().
2244 Bugzilla : 9489, 3273
2245 Description: First write from each client to each OST was only 4kB in size,
2246 to initialize client writeback cache, which caused sub-optimal
2247 RPCs and poor layout on disk for the first writen file.
2248 Details : Clients now request an initial cache grant at (re)connect time
2249 and so that they can start streaming writes to the cache right
2250 away and always do full-sized RPCs if there is enough data.
2251 If the OST is rebooted the client also re-establishes its grant
2252 so that client cached writes will be honoured under the grant.
2257 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2258 Details : Now I/O RPCs go to different portal number and (presumably) fast
2259 lock requests (and glimses) and other RPCs get their own service
2260 threads pool that should be able to service those RPCs
2263 Severity : enhancement
2265 Description: Ability to exchange lustre version between client and servers and
2266 issue warnings at client side if client is too old. Also for
2267 liblustre clients there is ability to refuse connection of too old
2269 Details : New 'version' field is added to connect data structure that is
2270 filled with version info. That info is later checked by server and
2274 Frequency : rare, liblustre only.
2275 Bugzilla : 9296, 9581
2276 Description: Two simultaneous writes from liblustre at offset within same page
2277 might proceed at the same time overwriting eachother with stale
2279 Details : I/O lock withing llu_file_prwv was released too early, before data
2280 actually was hitting the wire. Extended lock-holding time until
2281 server acknowledges receiving data.
2284 Frequency : extremely rare. Never observed in practice.
2286 Description: avoid generating lustre_handle cookie of 0.
2287 Details : class_handle_hash() generates handle cookies by incrementing
2288 global counter, and can hit 0 occasionaly (this is unlikely, but
2289 not impossible, because initial value of cookie counter is
2290 selected randonly). Value of 0 is used as a sentinel meaning
2291 "unassigned handle" --- avoid it. Also coalesce two critical
2292 sections in this function into one.
2294 Severity : enhancement
2296 Description: allow liblustre clients to delegate truncate locking to OST
2297 Details : To avoid overhead of locking, liblustre client instructs OST to
2298 take extent lock in ost_punch() on client's behalf. New connection
2299 flag is added to handle backward compatibility.
2301 Severity : enhancement
2302 Bugzilla : 4928, 7341, 9758
2303 Description: allow number of OST service threads to be specified
2304 Details : a module parameter allows the number of OST service threads
2305 to be specified via "options ost ost_num_threads={N}" in the
2306 OSS's /etc/modules.conf or /etc/modprobe.conf.
2310 Bugzilla : 6146, 9635, 9895
2311 Description: servers crash with bad pointer in target_handle_connect()
2312 Details : In rare cases when a client is reconnecting it was possible that
2313 the connection request was the last reference for that export.
2314 We would temporarily drop the export reference and get a new
2315 one, but this may have been the last reference and the export
2316 was just destroyed. Get new reference before dropping old one.
2318 Severity : enhancement
2319 Frequency : if client is started with failover MDS
2321 Description: Allow multiple MDS hostnames in the mount command
2322 Details : Try to read the configuration from all specified MDS
2323 hostnames during a client mount in case the "primary"
2326 Severity : enhancement
2328 Description: Stop sending data to evicted clients as soon as possible.
2329 Details : Check if the client we are about to send or are sending data to
2330 was evicted already. (Check is done every second of waiting,
2331 for which l_wait_event interface was extended to allow checking
2332 of exit condition at specified intervals).
2335 Frequency : rare, normally only when NFS exporting is done from client
2337 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2339 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2342 Frequency : always, if srand() is called before liblustre initialization
2344 Description: Liblustre uses system PRNG disturbing its usage by user application
2345 Details : Introduce internal to lustre fast and high-quality PRNG for
2346 lustre usage and make liblustre and some other places in generic
2347 lustre code to use it.
2349 Severity : enhancement
2350 Bugzilla : 9477, 9557, 9870
2351 Description: Verify that the MDS configuration logs are updated when xml is
2352 Details : Check if the .xml configuration logs are newer than the config
2353 logs stored on the MDS and report an error if this is the case.
2354 Request --write-conf, or allow starting with --old_conf.
2356 Severity : enhancement
2358 Description: Handle symlinks in the path when checking if Lustre is mounted.
2359 Details : Resolve intermediate symlinks when checking if a client has
2360 mounted a filesystem to avoid duplicate client mounts.
2365 Description: lconf can hit an error exception but still return success.
2366 Details : The lconf command catches the Command error exception at the top
2367 level script context and will exit with the associated exit
2368 status, but doesn't ensure that this exit status is non-zero.
2373 Description: failure of ptlrpc thread startup can cause oops
2374 Details : Starting a ptlrpc service thread can fail if there are a large
2375 number of threads or the server memory is very fragmented.
2376 Handle this without oopsing.
2379 Frequency : always, only if liblustre and non-default acceptor port was used
2381 Description: liblustre cannot connect to servers with non-default acceptor port
2382 Details : tcpnal_set_default_params() was not called and was therefore
2383 ignoring the environment varaible TCPNAL_PORT, as well as other
2384 TCPNAL_ environment variables
2389 Description: two objects could be created on the same OST for a single file
2390 Details : If an OST is down, in some cases it was possible to create two
2391 objects on a single OST for a single file. No problems other
2392 than potential performance impact and spurious error messages.
2396 Bugzilla : 5681, 9562
2397 Description: Client may oops in ll_unhash_aliases
2398 Details : Client dcache may become inconsistent in race condition.
2399 In some cases "getcwd" can fail if the current directory is
2405 Description: Inode refcounting problems in NFS export code
2406 Details : link_raw functions used to call d_instantiate without obtaining
2407 extra inode reference first.
2411 Bugzilla : 9942, 9903
2412 Description: Referencing freed requests leading to crash, memleaks with NFS.
2413 Details : We used to require that call to ll_revalidate_it was always
2414 followed by ll_lookup_it. Also with revalidate_special() it is
2415 possible to call ll_revalidate_it() twice for the same dentry
2416 even if first occurence returned success. This fix changes semantic
2417 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2418 reference on a request referred from the intent.
2419 ll_intent_release() then releases such a request.
2422 Frequency : rare, normally benchmark loads only
2424 Description: unlinked inodes were kept in memory on the client
2425 Details : If a client is repeatedly creating and unlinking files it
2426 can accumulate a lot of stale inodes in the inode slab cache.
2427 If there is no other client load running this can cause the
2428 client node to run out of memory. Instead flush old inodes
2429 from client cache that have the same inode number as a new inode.
2432 Frequency : SLES9 2.6.5 kernel and long filenames only
2433 Bugzilla : 9969, 10379
2434 Description: utime reports stale NFS file handle
2435 Details : SLES9 uses out-of-dentry names in some cases, which confused
2436 the lustre dentry revalidation. Change it to always use the
2440 Frequency : rare, unless heavy write-truncate concurrency is continuous
2441 Bugzilla : 4180, 6984, 7171, 9963, 9331
2442 Description: OST becomes very slow and/or deadlocked during object unlink
2443 Details : filter_destroy() was holding onto the parent directory lock
2444 while truncating+unlinking objects. For very large objects this
2445 may block other threads for a long time and slow overall OST
2446 responsiveness. It may also be possible to get a lock ordering
2447 deadlock in this case, or run out of journal credits because of
2448 the combined truncate+unlink. Solution is to do object truncate
2449 first in one transaction without parent lock, and then do the
2450 final unlink in a new transaction with the parent lock. This
2451 reduces the lock hold time dramatically.
2454 Frequency : rare, 2.4 kernels only
2456 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2457 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2458 before unmounting their filesystem in order to avoid deadlock.
2459 The kernel_locked() function in 2.4 kernels only checks whether
2460 the kernel lock is held, not whether it is this process that is
2461 holding it as 2.6 kernels do.
2466 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2467 Details : The client ptlrpc code may be trying to reconnect to a down
2468 server before a previous connection attempt has timed out.
2469 Increase the reconnect interval to be longer than the connection
2470 timeout interval to avoid sending duplicate connections to
2474 Frequency : echo_client brw_test command
2476 Description: fix echo_client to work with OST preallocated code
2477 Details : OST preallocation code (5137) didn't take echo_client IO path
2478 into account: echo_client calls filter methods outside of any
2479 OST thread and, hence, there is no per-thread preallocated
2480 pages and buffers to use. Solution: hijack pga pages for IO. As
2481 a byproduct, this avoids unnecessary data copying.
2485 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2486 Description: Client can oops in mdc_commit_close() after open replay
2487 Details : It was possible for the MDS to return an open request with no
2488 transaction number in mds_finish_transno() if the client was
2489 evicted, but without actually returning an error. Clients
2490 would later try to replay that open and may trip an assertion
2491 Simplify the client close codepath, and always return an error
2492 from the MDS in case the open is not successful.
2495 Frequency : rare, 2.6 OSTs only
2497 Description: OST may deadlock under high load on fragmented files
2498 Details : If there was a heavy load and highly-fragmented OST filesystems
2499 it was possible to have all the OST threads deadlock waiting on
2500 allocation of biovecs, because the biovecs were not released
2501 until the entire RPC IO was completed. Instead, release biovecs
2502 as soon as they are complete to ensure forward IO progress.
2504 Severity : enhancement
2506 Description: Support for specifying external journal device at mount
2507 Details : If an OST or MDS device is formatted with an external journal
2508 device, this device major/minor is stored in the ext3 superblock
2509 and may not be valid for failover. Allow detecting and
2510 specifying the external journal at mount time.
2515 Description: Mounting an MDS with pending unlinked files may cause oops
2516 Details : target_finish_recovery() calls mds_postrecov() which returned
2517 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2518 considers this an error and immediately begins cleaning up the
2519 lov, just after starting the mds_lov process
2521 Severity : enhancement
2523 Description: Implement 'lfs df' to report actual free space on per-OST basis
2524 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2525 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2526 usage of objects; '-h' to report in human readable format.
2528 ------------------------------------------------------------------------------
2530 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2537 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2538 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2539 at startup during orphan unlinking. Instead, skip buddy update
2540 before setup, as it will be regenerated after recovery is complete.
2545 Description: If an OST is inactive, its locks might reference stale inodes.
2546 Details : lov_change_cbdata() must iterate over all namespaces, even if
2547 they are inactive to clear inode references from the lock.
2549 Severity : enhancement
2550 Frequency : occasional, if non-standard max_dirty_mb used
2552 Description: Client will block write RPCs if not enough grant
2553 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2554 then the client will block writes while waiting for another RPC
2555 to complete instead of consuming its dirty limit. With change
2556 we get improved performance when max_dirty_mb is small.
2558 Severity : enhancement
2559 Bugzilla : 3389, 6253
2560 Description: Add support for supplementary groups on the MDS.
2561 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2562 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2563 lookups for user supplementary groups into a cache.
2567 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2568 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2569 for liblustre O_EXCL is enforced by MDS.
2573 Description: readdir never returns NULL in liblustre.
2574 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2575 dentry in struct dirent.
2579 Frequency : liblustre only, depends on application IO pattern
2580 Description: liblustre clients evicted if not contacting servers
2581 Details : Don't put liblustre clients into the ping_evictor list, so
2582 they will not be evicted by the pinger ever.
2584 Severity : enhancement
2586 Description: Add ability to evict clients by NID from MDS.
2587 Details : By echoing "nid:$NID" string into
2588 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2589 $NID would be instantly evicted from this MDS and from all active
2590 OSTs connected to it.
2594 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2595 Details : lookup_it_finish() used to query file size from OSTs that was not
2600 Description: service threads change working directory to that of init
2601 Details : Starting lustre service threads may pin the working directory
2602 of the parent thread, making that filesystem busy. Threads
2603 now change to the working directory of init to avoid this.
2607 Frequency : during shutdown only
2608 Description: shutdown with a failed MDS or OST can cause unmount to hang
2609 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2614 Frequency : 2.6 only
2615 Description: chmod/chown may include an extra supplementary group
2616 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2617 supplementary group and if none is specified this is used.
2620 Bugzilla : 5479 (6816)
2622 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2623 Details : If lookup is in progress on a file that is unlinked we might try
2624 to revalidate the inode and fail in revalidate after lookup is
2625 complete and ll_file_open() enqueues the open again but
2626 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2629 Frequency : always, if lconf --abort_recovery used
2631 Description: lconf --abort_recovery fails with 'Operation not supported'
2632 Details : lconf was attempting to abort recovery on the MDT device and not
2635 ------------------------------------------------------------------------------
2637 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2642 Frequency : rare (only unsupported configurations with a node running as an
2644 Bugzilla : 6514, 5137
2645 Description: Mounting a Lustre file system on a node running as an OST could
2647 Details : OSTs now preallocates memory needed to write out data at
2648 startup, instead of when needed, to avoid having to
2649 allocate memory in possibly low memory situations.
2650 Specifically, if the file system is mounted on on OST,
2651 memory pressure could force it to try to write out data,
2652 which it needed to allocate memory to do. Due to the low
2653 memory, it would be unable to do so and the node would
2654 become unresponsive.
2656 Severity : enhancement
2658 Description: Addition of lconf --service command line option
2659 Details : lconf now accepts a '--service <arg>' option, which is
2660 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2662 Severity : enhancement
2664 Description: Failover mode is now the default for OSTs.
2665 Details : By default, OSTs will now run in failover mode. To return to
2666 the old behaviour, add '--failout' to the lmc line for OSTs.
2668 Severity : enhancement
2670 Description: Health checks are now provided for MDS and OSTs
2671 Details : Additional detailed health check information on MSD and OSTs
2672 is now provided through the procfs health_check value.
2675 Frequency : occasional, depends on IO load
2677 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2678 after numerous create/delete cycles
2679 Details : The ext3 inode allocation policy would not allocate new inodes
2680 very well on the OSTs because there are no new directories
2681 being created. Instead we look for groups with free space if
2682 the parent directories are nearly full.
2687 Description: Network or server problems during mount may cause partially
2688 mounted clients instead of returning an error.
2689 Details : The config llog parsing code may overwrite the error return
2690 code during mount error handling, returning success instead
2696 Description: MDS can fail to allocate large reply buffers
2697 Details : After long uptimes the MDS can fail to allocate large reply
2698 buffers (e.g. zconf client mount config records) due to memory
2699 fragmentation or consumption by the buffer cache. Preallocate
2700 some large reply buffers so that these replies can be sent even
2701 under memory pressure.
2705 Frequency : rare (liblustre)
2706 Description: fsx running with liblustre complained that using truncate() to
2707 extend the file doesn't work. This patch corrects that issue.
2708 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2709 ATTR_SIZE and lsm use in llu_setattr_raw.
2713 Frequency : rare, only 2.6 kernels
2714 Description: Unusual file access patterns on the MDS may result in inode
2715 data being lost in very rare circumstances.
2716 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2717 code in the 2.6 kernel caused the inode and/or EA data not to
2718 be read from disk, causing single-file corruption.
2722 Frequency : rare, only 2.6 filesystems using extents
2723 Description: Heavy concurrent write and delete load may cause data corruption.
2724 Details : It was possible under high-load situations to have an extent
2725 metadata block in the block device cache from a just-unlinked
2726 file overwrite a newly-allocated data block. We now unmap any
2727 metadata buffers that alias just-allocated data blocks.
2731 Frequency : filesystems with default stripe_count larger than 77
2732 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2733 Details : lconf specifies an inode size of 4096 bytes when the default
2734 stripe_count is larger than 77. This conflicts with the default
2735 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2736 this case to avoid pinning too much memory for large EAs.
2738 ------------------------------------------------------------------------------
2740 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2745 Frequency : rare (extremely heavy IO load with hundreds of clients)
2747 Description: Client is evicted, gets IO error writing to file
2748 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2749 caused clients to be evicted for AST timeouts. The fixes in
2750 bug 5192 mean we no longer need to have such short AST timeouts
2751 so ldlm_timeout has been increased.
2754 Frequency : occasional during --force or --failover shutdown under load
2755 Bugzilla : 5949, 4834
2756 Description: Server oops/LBUG if stopped with --force or --failover under load
2757 Details : a collection of import/export refcount and cleanup ordering
2758 issues fixed for safer force cleanup
2761 Frequency : only filesystems larger than 120 OSTs
2762 Bugzilla : 5990, 6223
2763 Description: lfs getstripe would oops on a very large filesystem
2764 Details : lov_getconfig used kfree on vmalloc'd memory
2767 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2768 Bugzilla : 6242, 6243
2769 Description: reading from files that had been truncated to a non-zero size
2770 but never opened returned no data
2771 Details : ll_file_read() reads zeros from no-object files to EOF
2776 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2777 Details : There's little harm in aborting MDS/OSS recovery and letting it
2778 try again, so I removed the LASSERT and return an error instead.
2780 Severity : enhancement
2782 Description: New debugging infrastructure for tracking down data corruption
2783 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2784 (b) cover more of the client-side code path, and (c) try to narrow
2785 down where problems occurred
2789 Bugzilla : 3819, 4364, 4397, 6313
2790 Description: Racing close and eviction MDS could cause assertion in mds_close
2791 Details : It was possible to get multiple mfd references during close and
2792 client eviction, leading to one thread referencing a freed mfd.
2794 Severity: : enhancement
2795 Bugzilla : 3262, 6359
2796 Description: Attempts to reconnect to servers are now more aggressive.
2797 Details : This builds on the enhanced upcall-less recovery that was added
2798 in 1.4.2. When trying to reconnect to servers, clients will
2799 now try each server in the failover group every 10 seconds. By
2800 default, clients would previously try one server every 25 seconds.
2805 Description: After recovery, certain operations trigger a failed
2806 assertion on a client.
2807 Details : Failing over an mds, using lconf -d --failover, while a
2808 client was doing a readdir() call would cause the client to
2809 LBUG after recovery completed and the readdir() was resent.
2811 Severity : enhancement
2813 Description: Default groups are now added by lconf
2814 Details : You can now run lconf --group <servicename> without having to
2815 manually add groups with lmc.
2818 Frequency : occasional
2820 Description: Nodes with an elan id of 0 trigger a failed assertion
2823 Frequency : always when accessing e.g. tty/console device nodes
2825 Description: tty and some other devices nodes cannot be used on lustre
2826 Details : file's private_data field is used by device data and lustre
2827 values in there got lost. New field was added to struct file to
2828 store fs-specific private data.
2831 Frequency : when exporting Lustre via NFS
2833 Description: NFSD failed occasionally when looking up a path component
2834 Details : NFSD is looking up ".." which was broken in ext3 directories
2835 that had grown large enough to become hashed.
2838 Frequency : Clusters with multiple interfaces not on the same subnet
2840 Description: Nodes will repeatedly try to reconnect to an interface which it
2841 cannot reach and report an error to the log.
2842 Details : Extra peer list entries will be created by lconf with some peers
2843 unreachable. lconf now validates the peer before adding it.
2846 Frequency : Only if a default stripe is set on the filesystem root.
2848 Description: Setting a default stripe on the filesystem root prevented the
2849 filesystem from being remounted.
2850 Details : The client was sending extra request flags in the root getattr
2851 request and did not allocate a reply buffer for the dir EA.
2854 Frequency : occasional, higher if lots of files are accessed by one client
2855 Bugzilla : 6159, 6097
2856 Description: Client trips assertion regarding lsm mismatch/magic
2857 Details : While revalidating inodes the VFS looks up inodes with ifind()
2858 and in rare cases can find an inode that is being freed.
2859 The ll_test_inode() code will free the lsm during ifind()
2860 when it finds an existing inode and then the VFS later attaches
2861 this free lsm to a new inode.
2865 Bugzilla : 6422, 7030
2866 Description: MDS deadlock between mkdir and client eviction
2867 Details : Creating a new file via mkdir or mknod (starting a transaction
2868 and getting the ns lock) can deadlock with client eviction
2869 (gets ns lock and trying to finish a synchronous transaction).
2872 Frequency : occasional
2873 Description: While starting a server, the fsfilt_ext3 module could not be
2875 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2876 kernels. Previously, lconf would still use the ext3 name
2877 when trying to load modules. Now, it will correctly use
2878 ext3 on 2.4 and ldiskfs on 2.6.
2880 Severity : enhancement
2881 Description: The default stripe count has been changed to 1
2882 Details : The interpretation of the default stripe count (0, to lfs
2883 or lmc) has been changed to mean striping across a single
2884 OST, rather than all available. For general usage we have
2885 found a stripe count of 1 or 2 works best.
2887 Severity : enhancement
2888 Description: Add support for compiling against Cray portals.
2889 Details : Conditional compiling for some areas that are different
2893 Frequency : occasional
2894 Bugzilla : 6409, 6834
2895 Description: Creating files with an explicit stripe count may lead to
2896 a failed assertion on the MDS
2897 Details : If some OSTs are full or unavailable, creating files may
2898 trigger a failed assertion on the MDS. Now, Lustre will
2899 try to use other servers or return an error to the
2903 Frequency : occasional
2905 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2906 caused lock timeout during readahead (since 1.4.2).
2907 Details : Processes doing readahead might match a lock that hasn't been
2908 granted yet if there are overlapping and conflicting lock
2909 requests. The readahead process waits on ungranted lock
2910 (original lock is CBPENDING), while OST waits for that process
2911 to cancel CBPENDING read lock and eventually evicts client.
2913 Severity : enhancement
2915 Description: Initial enabling of flock support for clients
2916 Details : Implements fcntl advisory locking and file status functions.
2917 This feature is provided as an optional mount flag (default
2918 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2919 locking are implemented yet, and those that are are not guaranteed
2920 to be completely correct in production environments.
2921 mount -t lustre -o [flock|noflock] ...
2924 Frequency : occasional
2926 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2927 assertion in the ext3 JBD (journaling) layer.
2928 Details : The b_committed_data struct is protected by the big kernel lock
2929 in 2.4 kernels, serializing journal_commit_transaction() and
2930 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2931 access to this struct. In 2.6 kernels there is finer grained
2932 locking to improve SMP performance of the JBD layer.
2936 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2941 Description: merge and cleanup kernel patches.
2944 -----------------------------------------------------------------------------
2946 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2948 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2949 versions of Lustre. Please update all servers and clients to
2950 version 1.4.2 or later at the same time. You must also run
2951 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2952 to update the configuration logs.
2954 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2955 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2956 - swab llog records of type '0' so we get proper header size/idx (5861)
2957 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2958 - fix rename of one directory over another leaking an inode (5953)
2959 - avoid SetPageDirty on 2.6 (5981)
2960 - don't re-add just-being-destroyed locks to the waiting list (5653)
2961 - when creating new directories, inherit the parent's custom
2962 striping settings if present parent (3048)
2963 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2964 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2965 - don't hold client locks on temporary worklist from l_lru (5666)
2966 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2967 - automatically evict dead clients (5921)
2968 - Update file size properly in create+truncate+fstat case (6196)
2969 - Do not unhash mountpoint dentries, do not allow removal of
2971 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2972 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2973 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2974 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2975 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2976 - Correctly handle reads of files with no objects (6243)
2977 - lctl recover will also mark a device active if deactivate used (5933)
2979 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2980 - allow --write-conf on an MDS with different nettype than client (5619)
2981 - don't write config llogs to MDS for mounts not from that MDS (5617)
2982 - lconf should create multiple TCP connections from a client (5201)
2983 - init scripts are now turned off by default; run chkconfig --on
2984 lustre and chkconfig --on lustrefs to use them
2985 - upcalls are no longer needed for clients to recover to failover
2987 - add --abort-recovery option to lconf to abort recovery on device
2989 - add support for an arbitrary number of OSTs (3026)
2990 - Quota support protocol changes.
2991 - forward compatibility changes to wire structs (6007)
2992 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2993 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2994 - improved "lustre status" script
2995 - initialize blocksize for non-regular files (6062)
2996 - added --disable-server and --disable-client configure options (5782)
2997 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2998 - Vanilla 2.4.29 support
2999 - increase maximum number of obd devices to 520 (6242)
3000 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3001 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3002 SLES 9 2.6.5 kernels
3003 - limit stripes per file to 160 (the maximum EA size) (6093)
3005 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3008 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3009 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3010 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3011 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3012 - small ext3 extents cleanups and fixes (5733)
3013 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3014 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3015 - avoid needless client->OST connect, fix handle mismatch (5317)
3016 - fix DLM error path that led to out-of-sync client, long delays (5779)
3017 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3018 - fix several locking issues related to i_size (5492,5624,5654,5672)
3019 - don't move pending lock onto export if it is already evicted (5683)
3020 - fix kernel oops when creating .foo in unlinked directory (5548)
3021 - fix deadlock in obdfilter statistics vs. object create (5811)
3022 - use time_{before,after} to avoid timer jiffies wrap (5882)
3023 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3024 - Do not leak request if server was not able to process it (5154)
3025 - If mds_open unable to find parent dir, make that negative lookup(5154)
3026 - don't create new directories with extent-mapping (5909, 5936)
3028 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3029 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3030 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3031 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3032 - support for RHEL4 kernel on i686 (5773)
3033 - provide error messages when incompatible logs are encountered (5898)
3035 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3036 * version 1.4.0.10 (1.4.1 release candidate 1)
3038 - don't keep a lock reference when lock is not granted (4238)
3039 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3040 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3041 - drop import inflight refcount on signal_completed_replay error (5255)
3042 - unlock page after async write error during send (3677)
3043 - handle missing objects in filter_preprw_read properly (5265)
3044 - no transno return for symlink open, don't save no-trasno open (3440)
3045 - don't try to complete elan receive that already failed (4012)
3046 - free RPC server reply state on error (5406)
3047 - clean up thread from ptlrpc_start_thread() on error (5160)
3048 - readahead could read extra page into cache that wasn't ejected (5388)
3049 - prevent races in class_attach/setup/cleanup/detach (5260)
3050 - don't dereference de->d_inode after l_dput of de (5458)
3051 - use "int" for stripe value returned from lock_to_stripe (5544)
3052 - mballoc allocation and error-checking fixes in 2.6 (5504)
3053 - block device patches to fix I/O request sizes in 2.6 (5482)
3054 - look up hostnames for IB nals (5602)
3055 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3056 - don't start multiple acceptors for the same port (5277)
3057 - fix incorrect LASSERT in mds_getattr_name (5635)
3058 - export a proc file for general "ping" checking (5628)
3059 - fix "lfs check" to not block when the MDS is down (5628)
3061 - service request history (4965)
3062 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3063 - create an "evict_client" /proc entry on OSTs, like the MDS has
3064 - fix mount usage message, return errors per mount(8) (5168)
3065 - change grep [] to grep "[]" in tests so they work in more UMLs
3066 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3067 - remove ancient LOV_MAGIC_V0 EA support (5047)
3068 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3069 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3070 - allow readahead allocations to fail when low on memory (5383)
3071 - mmap locking landed again, after considerable improvement (2828)
3072 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3074 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3077 - send OST transaction number in read/write reply to free req (4966)
3078 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3079 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3080 - the watchdog thread now runs as interruptible (5246)
3081 - flock/lockf fixes (but it's still disabled, pending 5135)
3082 - don't use EXT3 constants in llite code (5094)
3083 - memory shortage at startup could cause assertion (5176)
3085 - reorganization of lov code
3086 - single portals codebase
3088 - add extents/mballoc support (5025)
3089 - direct I/O reads in the obdfilter (4048)
3090 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3092 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3095 - send OST transaction number in read/write reply to free req (4966)
3096 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3097 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3098 - the watchdog thread now runs as interruptible (5246)
3099 - handle missing objects in filter_preprw_read properly (5265)
3100 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3101 - drop import inflight refcount on signal_completed_replay error (5255)
3102 - unlock page after async write error during send (3677)
3103 - return original error code on reconstructed replies (3761)
3104 - no transno return for symlink open, don't save no-trasno open (3440)
3106 - add pid to ldlm debugging output (4922)
3107 - bump the watchdog timeouts -- we can't handle 30sec yet
3108 - extra debugging for orphan dentry/inode bug (5259)
3110 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3113 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3114 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3115 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3116 - fix NULL dereference in /proc/sys/portals/routes (4827)
3117 - allow failed mdc_close() operations to be interrupted (4561)
3118 - stop precreate on OST before MDS would time out on it (4778)
3119 - don't send partial-page writes before EOF from client (4410)
3120 - discard client grant for sub-page writes on large-page clients (4520)
3121 - don't free dentries not owned by NFS code, check generation (4806)
3122 - fix lsm leak if mds_create_objects() fails (4801)
3123 - limit debug_daemon file size, always print CERROR messages (4789)
3124 - use transno after validating reply (3892)
3125 - process timed out requests if import state changes (3754)
3126 - update mtime on OST during writes, return in glimpse (4829)
3127 - add mkfsoptions to LDAP (4679)
3128 - use ->max_readahead method instead of zapping global ra (5039)
3129 - don't interrupt __l_wait_event() during strace
3131 - add software watchdogs to catch hung threads quickly (4941)
3132 - make lustrefs init script start after nfs is mounted
3133 - fix CWARN/ERROR duplication (4930)
3134 - return async write errors to application if possible (2248)
3135 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3136 - print NAL number in %x format (4645)
3137 - update barely-supported suse-2.4.21-171 series (4842)
3138 - support for sles 9 %post scripts
3139 - support for building 2.6 kernel-source packages
3140 - support for sles km_* packages
3142 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3145 - ignore -ENOENT errors in osc_destroy (3639)
3146 - notify osc create thread that OSC is being cleaned up (4600)
3147 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3148 - reconstruct ost_handle() like mds_handle() (4657)
3149 - create a new thread to do import eviction to avoid deadlock (3969)
3150 - let lconf resolve symlinked-to devices (4629)
3151 - don't unlink "objects" from directory with default EA (4554)
3152 - hold socknal file ref over connect in case target is down (4394)
3153 - allow more than 32000 subdirectories in a single directory (3244)
3154 - fix blocks count for O_DIRECT writes (3751)
3155 - OST returns ENOSPC from object create when no space left (4539)
3156 - don't send truncate RPC if file size isn't changing (4410)
3157 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3158 - bind to privileged port in socknal and tcpnal (3689)
3160 - rate limit CERROR/CWARN console message to avoid overload (4519)
3161 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3162 - basic mmap support (3918)
3163 - kernel patch series update from b1_4 (4711)
3165 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3168 - avoid crash during MDS cleanup with OST shut down (2775)
3169 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3170 - don't use bad inodes on the MDS (3744)
3171 - dynamic object preallocation to improve recovery speed (4236)
3172 - don't hold spinlock over lock dumping or change debug flags (4401)
3173 - don't zero obd_dev when it is force cleaned (3651)
3174 - print grants to console if they go negative (4431)
3175 - "lctl deactivate" will stop automatic recovery attempts (3406)
3176 - look for existing locks in ldlm_handle_enqueue() (3764)
3177 - don't resolve lock handle twice in recovery avoiding race (4401)
3178 - revalidate should check working dir is a directory (4134)
3180 - don't always mark "slow" obdfilter messages as errors (4418)
3182 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3185 - don't close LustreDB during write_conf until it is done (3860)
3186 - fix typo in lconf for_each_profile (3821)
3187 - allow dumping logs from multiple threads at one time (3820)
3188 - don't allow multiple threads in OSC recovery (3812)
3189 - fix debug_size parameters (3864)
3190 - fix mds_postrecov to initialize import for llog ctxt (3121)
3191 - replace config semaphore with spinlock (3306)
3192 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3193 - don't allow enqueue to complete on a destroyed export (3822)
3194 - down write_lock before checking llog header bitmap (3825)
3195 - recover from lock replay timeout (3764)
3196 - up llog sem before sending rpc (3652)
3197 - reduce ns lock hold times when setting kms (3267)
3198 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3199 - fix NULL deref and obd_dev leak on setup error (3312)
3200 - replace some LBUG about llog ops with error handling (3841)
3201 - don't match INVALID dentries from d_lookup and spin (3784)
3202 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3203 - fix invalid assertion in ptlrpc_set_wait (3880)
3205 - add libwrap support for the TCP acceptor (3996)
3206 - add /proc/sys/portals/routes for non-root route listing (3994)
3207 - allow setting MDS UUID in .xml (2580)
3208 - print the stack of a process that LBUGs (4228)
3210 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3213 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3214 - make sure to unset replay flag from failed open requests (3440)
3215 - if default stripe count is 0, use OST count for inode size (3636)
3216 - update parent mtime/ctime on client for create/unlink (2611)
3217 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3218 - free recovery state on server during a forced cleanup (3571)
3219 - unregister_reply for resent reqs (3063)
3220 - loop back devices mounting and status check on 2.6 (3563)
3221 - fix resource-creation race that can provoke i_size == 0 (3513)
3222 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3223 - more debugging for page-accounting assertion (3746)
3224 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3225 - avoid deadlock after precreation failure (3758)
3226 - fix race and lock order deadlock in orphan handling (3450, 3750)
3227 - add validity checks when grabbing inodes from l_ast_data (3599)
3229 - add /proc/.../recovery_status to obdfilter (3428)
3230 - lightweight CDEBUG infrastructure, debug daemon (3668)
3231 - change default OSC RPC parameters to be better on small clusters
3232 - turn off OST read cache for files smaller than 32MB
3233 - install man pages and include them in rpms (3100)
3234 - add new init script for (un)mounting lustre filesystems (2593)
3235 - run chkconfig in %post for init scripts (3701)
3236 - drop scimac NAL (unmaintained)
3238 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3241 - clean kiobufs before and after use (3485)
3242 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3243 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3244 - change init script to fail more clearly if not run as root (1528)
3245 - allow clients to reconnect during replay (1742)
3246 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3247 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3248 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3249 - fix cond_resched (3554)
3250 - only evict unfinished clients after recovery (3515)
3251 - allow bulk resend, prevent data loss (3570)
3252 - dynamic ptlrpc request buffer allocation (2102)
3253 - don't allow unlinking open directory if it isn't empty (2904)
3254 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3255 - remove extraneous obd dereference causing LASSERT failure (3334)
3256 - don't use get_cycles() when creating temp. files on the mds (3156)
3257 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3258 - handle EEXIST for set-stripe, set proper directory name (3336)
3260 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3261 - fix ksocknal_fmb_callback() error messages (2918)
3263 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3266 - don't copy lvb into (possibly NULL) reply on error (2983)
3267 - don't deref dentry after dput, don't free lvb on error (2922)
3268 - use the kms to determine writeback rpc length (2947)
3269 - increment oti_logcookies when osc is inactive (2948)
3270 - update client's i_blocks count via lvb messages (2543)
3271 - handle intent open/close of special files properly (1557)
3272 - mount MDS with errors=remount-ro, like obdfilter (2009)
3273 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3274 - don't use cancelling-locks' kms values (2947)
3275 - use highest lock extent for kms, not last one (2925)
3276 - don't dereference ERR_PTR() dentry in error handling path (3107)
3277 - fix thread race in portals_debug_dumplog() (3122)
3278 - create lprocfs device entries at setup instead of at attach (1519)
3279 - common AST error handler, don't evict client on completion race (3145)
3280 - zero nameidata in detach_mnt in 2.6 (3118)
3281 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3282 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3283 - initialize RPC timeout timer earlier for 2.6 (3219)
3284 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3285 - print nal/nid for unknown nid (3258)
3286 - additional checks for oscc recovery before doing precreate (3284)
3287 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3288 - don't crash in mdc_close for bad permissions on open (3285)
3289 - zero i_rdev for non-device files (3147)
3290 - clear page->private before handing to FS, better assertion (3119)
3291 - tune the read pipeline (3236)
3292 - fix incorrect decref of invalidated dentry (2350)
3293 - provide read-ahead stats and refine rpc in flight stats (3328)
3294 - don't hold journal transaction open across create RPC (3313)
3295 - update atime on MDS at close time (3265)
3296 - close LDAP connection when recovering to avoid server load (3315)
3297 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3298 - don't leak open file on MDS after open resend (3325)
3299 - serialize filter_precreate and filter_destroy_precreated (3329)
3300 - loop device shouldn't call sync_dev() for nul device (3092)
3301 - clear page cache after eviction (2766)
3302 - resynchronize MDS->OST in background (2824)
3303 - refuse to mount the same filesystem twice on same mountpoint (3394)
3304 - allow llmount to create routes for mounting behind routers (3320)
3305 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3306 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3307 - fix rare problem with rename on htree directories (3417)
3309 - allow default OST striping configuration per directory (1414)
3310 - fix compilation for qswnal for 2.6 kernels (3125)
3311 - increase maximum number of MDS request buffers for large systems
3312 - change liblustreapi to be useful for external progs like lfsck (3098)
3313 - increase local configuration timeout for slow disks (3353)
3314 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3316 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3319 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3320 - don't overwrite extent policy data in reply if lock was blocked (2901)
3321 - drop filter export grants atomically with removal from device (2663)
3322 - del obd_self_export from work_list in class_disconnect_exports (2908)
3323 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3324 - swab reply message in mdc_close, other PPC fixes (2464)
3325 - fix destroying of named logs (2325)
3326 - overwrite old logs when running lconf --write_conf (2264)
3327 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3328 - fix race in target_handle_connect (2898)
3329 - mds_reint_create() should take same inode create lock (2926)
3330 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3331 - don't close files for self_export to avoid uninitialized obd (2936)
3332 - allow MDS with the same name as client node (2939)
3333 - hold dentry reference for closed log files for unlink (2325)
3334 - reserve space for all logs during transactions (2059)
3335 - don't evict page beyond end of stripe extent (2925)
3336 - don't oops on a deleted current working directory (2399)
3337 - handle hard links to targets without a parent properly (2517)
3338 - don't dereference NULL lock when racing during eviction (2867)
3339 - don't grow lock extents when lots of conflicting locks (2919)
3341 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3344 - account for cache space usage on clients to avoid data loss (974)
3345 - lfsck support in lustre kernel code (2349)
3346 - reduce journal credits needed for BRW writes (2370)
3347 - orphan handling to avoid losing space on client/server crashes
3348 - ptlrpcd can be blocked, stopping ALL progress (2477)
3349 - use lock value blocks to assist in proper KMS, faster stat (1021)
3350 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3351 - recovery for initial connections (2355)
3352 - fixes for mds_cleanup_orphans (1934)
3353 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3354 - block all file creations until orphan recovery completes (1901)
3355 - client remove rq_connection from request struct (2423)
3356 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3357 - recovery timer race (2670)
3358 - mdc_close recovey bug (2532)
3359 - ptlrpc cleanup bug (2710)
3360 - mds timeout on local locks (2588)
3361 - namespace lock held during RPCs (2431)
3362 - handle interrupted sync write properly (2503)
3363 - don't try to handle a message that hasn't been replied to (2699)
3364 - client assert failure during cleanup after abort recovery (2701)
3365 - leak mdc device after failed mount (2712)
3366 - ptlrpc_check_set allows timedout requests to complete (2714)
3367 - wait for inflight reqs when ptlrpcd finishes (2710)
3368 - make sure unregistered services are removed from the srv_list
3369 - reset bulk XID's when resending them (caught by 1138 test)
3370 - unregister_bulk after timeout
3371 - fix lconf error (2694)
3372 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3373 - readahead locks pages, leaves pending causing memory pressure (2673)
3374 - increase OST request buffers to 4096 on large machines (2729)
3375 - fix up permission of existing directories in simple_mkdir (2661)
3376 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3377 - don't assign transno to errored transactions (2742)
3378 - don't delete objects on OST if given a bogus objid from MDS (2751)
3379 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3380 - if rq_no_resend, then timeout request after recovery (2432)
3381 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3382 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3383 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3384 - MDS recovery shouldn't delete objects that it has given out (2730)
3385 - if enqueue arrives after completion, don't clobber LVB (2819)
3386 - don't unlock pages twice when trigger_group_io returns error (2814)
3387 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3388 - don't write pages to disk if there was an error (1450)
3389 - don't ping imports that have recovery disabled (2676)
3390 - take buffered bytes into account when balancing socknal conn (2817)
3391 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3392 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3393 - remove little-endian swabbing of llog records (1987)
3394 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3395 - retry reposting request buffers if they fail (1191)
3396 - grow extent at grant time to avoid granting a revoked lock (2809)
3397 - lock revoke doesn't evict page if covered by a second lock (2765)
3398 - disable VM readahead to avoid reading outside lock extents (2805)
3400 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3401 - updated kernel patches for hp-2.4.20 kernel (2681)
3403 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3406 - fix truncated write corruption (2366)
3407 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3409 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3410 - ensure len(uuid) < 37 in lmc (1171)
3411 - fix ia64 OOPS in llog_test (2255)
3412 - zero end of page at obdfilter for partial page writes (2648)
3413 - don't leave stale dentries around after renames (bug 2428)
3414 - fix timeouts when evicting a client with a single lock held (2642)
3415 - set deadline for the initial HELLO message to drain (2634)
3416 - print out dotted-quad IP addresses in the socknal (2302)
3418 - additional debugging for MDS client eviction problem (2443)
3419 - fix mkfsoptions support for osts (2603, 2604)
3421 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3424 - add series for the vanilla 2.6.0 kernel
3425 - add series for the vanilla 2.4.24 kernel
3426 - add series for a cray x86/64 UL kernel drop
3427 - fix xattr patches for the vanilla 2.4.19 series
3429 - generate true UUIDs in lmc (1171)
3430 - have portals stack dumping break in UML (2466)
3431 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3432 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3433 - wake callback waiting threads on client eviction (2460)
3434 - Add --ptldebug and --subsystem to lmc (1719)
3435 - update assertion to allow safe interrupt allocation
3436 - set rq_no_resend for cancel requests (2432)
3437 - recalculate ptlrpcd timeout after resend (2494)
3438 - call vfs_rmdir when removing pending directories (2368)
3439 - fix renaming a file to itself (2429)
3440 - lmc creates a default one-stripe lov (2454)
3441 - expand procfs space to handle large clusters (2326)
3442 - increase UML stack to avoid overflow
3443 - update lconf's list of debug and subsystem masks
3444 - fix lfs find --obd (2510)
3445 - /proc tunable for disabling filter read caching (2591)
3446 - stop rpm packages from altering slapd.conf (2301)
3447 - disable nagle in the socknal under 0conf (2578)
3448 - choose mds inode size based on stripe count (2572)
3449 - fix kernel-source rpm problems (2516)
3451 - add --disable-doc to avoid pdf generation (2421)
3452 - update documentation, tests, type-os, comments
3453 - avoid format warnings on ia64
3454 - remove the TOE NAL
3455 - tiny code cleanups by removing unused fields
3457 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3460 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3461 - remove the most blatant lies from BUILDING file (2371)
3462 - change default debug level to reasonable production setting
3463 - reduce client side cache size to reduce cache flush time
3464 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3465 - make TCP zerocopy and pinger support enabled by default (2476)
3466 - sync writes completed after process exits caused crashes (2319)
3467 - maintain correct mount count on the MDS (2356)
3468 - backout 1557, because 2316 wasn't really fixed
3469 - better file I/O statistics gathering in /proc
3470 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3471 - another kernel patch to fix zero-copy TCP function export
3472 - don't take duplicate lock when processing re-sent getattr (2420)
3473 - lctl uses obd_self_export instead of creating new conn (2353)
3474 - MDS/OST recovery case which requires object creation asserted (2425)
3475 - move lfs from /usr/sbin to /usr/bin in packages
3476 - fix race between mds_client_add and mds_client_free (2417)
3477 - use kmalloc instead of slabs in portals (2430)
3478 - don't create duplicate records when a failover MDS is present (2442)
3479 - remove unnecessary mount age check (2332)
3480 - don't remove directory inodes from locks prematurely (2451)
3481 - don't break if MDS service name is the same as hostname (2103)
3482 - fix races in client write RPC generation when cache full (2482)
3484 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3487 - remove now-unused request->rq_obd (278)
3488 - if an allocation fails, print out how much memory we've used (1933)
3489 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3490 - add the "configurable stack size" patch to most series files (1256)
3491 - ability to write large log records, for 100+ OST configs (2306)
3492 - fix NULL deref when filter_prep fails (2314)
3493 - fix operator precedence error in filter_sync
3494 - dynamic allocation of socknal TX descriptors (2315)
3495 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3496 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3497 - ensure that CWARN messages in Portals always get to the syslog
3498 - __init/__exit are not for prototype decls (ldlm_init/exit)
3499 - x86-64 compile warning fixes
3500 - fix gateway LMC keyword conflict (2318)
3501 - fix MDS lock inversions in getattr/reint paths (1844)
3502 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3503 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3504 - DLM race condition prevented some lock evictions (2328)
3505 - ENOMEM detection and retry on socknal sends (2230)
3506 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3507 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3508 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3509 - better error messages when a client is rejected during recovery (1505)
3510 - avoid cancelling locks which were never granted, after failure (2330)
3511 - fix i_sem/journal inversion in mds_client_add (2333)
3512 - fix truncate/getattr lock cycle deadlock (2334)
3513 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3514 - fix two rare exit paths which could leak an l_lock() ref (2321)
3515 - fencepost error in MDS/OST orphan recovery (2226)
3516 - make log record alignment 8 bytes (1988)
3517 - lstripe now fails when requested offset > ost_count (2237)
3518 - ensure that all kernel series have a complete list.h (1607)
3519 - fix crashes in special-file operations (2316)
3520 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3522 - allow configurable automake binary, for testing new versions
3523 - small update to the lfs documentation
3525 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3527 * fix negative export reference count in fsfilt_sync (2312)
3529 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3530 * release candidate 0.9.1
3532 - orphans are moved into the PENDING directory for possible recovery
3533 - replayed opens now open by fid for orphan/rename safety (1042)
3534 - last close of an orphan inode generates a transno (683)
3535 - chdir() and mount() now pin the directory entry (1020)
3536 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3537 - discard very old requests without processing them (1502)
3538 - remove some common, well-understood CERRORs (1505)
3539 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3540 - clear "grant" flags in OST replies until OST grant code lands (1644)
3541 - fix read performance by not clobbering i_blksize on client (1598)
3542 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3543 - make LRU size a /proc tunable, clears locks when reduced (707)
3544 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3545 - server locks take references on exports now (1558)
3546 - build fixes for 2.4.20-rh trees (1663)
3547 - return an error from lov_create if all OSCs are inactive (1751)
3548 - fix import levels when a reconnect happens without a timeout (1597)
3549 - exit early from mds_open if we get a lookup error (1749)
3550 - partial page read at EOF wouldn't wait for disk before sending (1642)
3551 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3552 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3553 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3554 - write/truncate lock inversion (1639)
3555 - Don't auto-load obdclass, portals modules during cleanup (1495)
3556 - fix timestamps from jumping to "now" (1763)
3557 - extra journal assertions (1648)
3558 - add an extra multiunlink test (1771)
3559 - fix read_record/write_record API (1776)
3560 - fix leak of offset_extent, possible incorrect i_size later (1772)
3561 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3562 - Debugging for the kqswnal_get_idle_tx problems (1820)
3563 - Allow recovery to be attempted multiple times (1536)
3564 - Write out MDS last_rcvd file after it is first created (1600)
3565 - Fix tx_descriptor leak in failed transmit situations (1827)
3566 - ext3 journaling fixes for assertion failure after IO error (1871)
3567 - class_export_put() on freed export after completion AST error (1896)
3568 - Fix revalidate looping in VFS (1322)
3569 - Don't access a freed export during MDS_REINT timeout (1521)
3570 - Add open-unlink recovery support on the MDS (1673,1764)
3571 - Return an error if no MDS data was read from last_rcvd (1946)
3572 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3573 - Don't setup a disk device that doesn't match exported UUID (317)
3574 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3575 - avoid committing NULL handle in force close
3576 - local.sh is now a one-stripe LOV configuration
3577 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3578 - fix ext3 htree duplicate directory entry corruption (1516)
3579 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3580 - update to new LOV EA format (2097)
3581 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3583 2003-06-15 Phil Schwan <phil@clusterfs.com>
3586 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3587 - per-import/export recovery handling (958, 931, 959)
3588 - multiple last-rcvd slots, for serving multiple FSes (949)
3589 - connections are again shared between multiple imp/exports (963, 964)
3590 - "umount -f" would hang if any requests needed to be sent (393, 978)
3591 - avoid pinning large req buffer by copying for queued messages (989)
3592 - add "uuid" to "lctl device" command to help upcalls (991)
3593 - "open" RPCs with transnos would confuse recovery counters (1037)
3594 - do proper endian conversion of all wire messages (288, 340, 891)
3595 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3596 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3597 - fix ptlrpc_connection leak in target_handle_connect (1174)
3598 - fix import refcounting bug in OST and MDS cleanup (1134)
3599 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3600 - fix very unlikely obd_types race condition (501)
3601 - remove osc_open hack for echo_client (1187)
3602 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3603 - a failure in read_inode2 leads to deadlock (1139)
3604 - cancel ack-locks as soon as transaction is committed (1072)
3605 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3606 - make sure to commitrw after any preprw to avoid deadlock (1162)
3607 - failing to execute a file in a lustre FS would lock inode (1203)
3608 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3609 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3610 - obd_uuid2tgt would walk off the end of the list (1255)
3611 - on IA64 the osc would give portals incorrect bulk size (1258)
3612 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3613 - fix lock inversion caused by new llite matching code (1282)
3614 - limit the number of dirty pages on a client to 10MB (1286)
3615 - timed out locks were not being corrected cancelled (1289)
3616 - fix O_DIRECT above 4GB on IA-32 (1292)
3617 * major user-visible changes
3618 - fail out/fail over policy now controlled by the upcall (993)
3620 - add OBD_PING to check server availability and failure (954)
3621 - lustre messages are now sent in sending host order (288, 340, 891)
3622 - add eadatalen to MDS getattr reply (340)
3623 - OST read replies may contain second buffer, with per-page status (593)
3625 2003-03-11 Phil Schwan <phil@clusterfs.com>
3628 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3629 - failed open()s could cause deadlock; fixed (867, 869)
3630 - stop cancelling OST locks when files are closed (481)
3631 - overlapping XID spaces caused network corruption (851, 853)
3632 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3633 - setattr_raw added, to do single-RPC, server-side setattrs
3634 - lmc/lconf syntax change for OST UUIDs
3635 - fix crashy race condition between ptlrpc_free_req and osc_close
3636 - don't use request in mdc_enqueue if we hit a timeout (889)
3637 - don't set the inode i_size for regular files from the MDS (896)
3638 - handle out of order completion AST (842)
3639 - don't LBUG if a lock request times out after receiving AST (913)
3640 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3641 - if a bad lock AST arrives, send an error instead of dropping entirely
3642 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3643 - fix leak in bulk IO when only partially completed (899, 900, 926)
3644 - fix O_DIRECT for ia64 (55)
3645 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3646 - C-z after timeout could hang a process forever; fixed (977)
3648 - client-side I/O cache (678, 924, 929, 941, 970)
3650 - READPAGE and SETATTRs which don't take server-side locks get
3653 2003-02-11 Phil Schwan <phil@clusterfs.com>
3656 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3657 - Fix fsfilt_extN_readpage() to read a full page of directory
3658 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3659 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3660 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3661 - Most intent operations were removed, in favour of a new RPC mode
3662 that does a single RPC to the server and bypasses most of the VFS
3663 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3664 - Aggressively cancel local locks on DLM servers
3665 - mds_reint_unlink sends EA to the client if it's the last nlink.
3666 client uses that EA to unlink OST objects.
3667 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3668 - recursive symlinks were fixed (439)
3669 - fixed NULL deref in DEBUG_REQ
3670 - filter_update_lastobjid no longer calls sync, which annoyed extN
3671 - fixed multi-client small-writes to a single file problem (445)
3672 - fixed mtime updates during file writes (607)
3673 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3674 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3675 - fixed Linux OST size reporting problem (444, 656)
3676 - OST now updates object mtime with writes or setattr (607, 619)
3677 - client verifies file size before zeroing page past EOF (445)
3678 - OST now writes last allocated objid to disk with allocation (108)
3679 - LOV on echo now works (409)
3681 - mds_reint_unlink sends a new buffer, with the EA included. this
3682 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3683 set if a regular file was being unlinked, and it was the last link
3684 - use PtlGet from the target for bulk writes (315)
3685 - OST now updates object mtime with writes or setattr (607, 619)
3686 - LDLM now has a grant-time callback to revalidate locked items, if
3688 - Many MDS operations were reorganized to combat race conditions
3690 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3691 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3693 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3696 - Fully reactivate OST imports after reconnection (512, others)
3697 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3698 - More graceful error handling for truncating on dead OST (515)
3699 - Don't error out unless we're actually accessing dead stripes (474)
3700 - Fix garbage sizes when stripes are missing (410)
3701 - LRU counters were broken, causing constant lock purge (433, 432)
3702 - garbage on read from stripes with failed OSTs (441)
3703 - mark OSCs as active before reconnecting during recovery (438)
3704 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3705 - lfind did not preserve OST order in output (443)
3706 - symlinks cause hung clients, incorrect data (439)
3707 - stop dereferencing request after dropping refcount (457)
3708 - don't LASSERT(spin_is_locked) on non-SMP (455)
3709 - fixes for many rename() bugs
3710 - fstat didn't correctly synchronize attributes (399)
3711 - server must handle lock cancellation during blocking AST prep (487)
3712 - bulk descriptors were free()d too soon (511)
3713 - fix paths in lconf, which would load incorrect modules (451, 507)
3714 - fix confusing lconf 'host not found' error message (386)
3715 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3716 - fix race condition in mdc_blocking_ast() for inode access (526)
3717 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3718 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3719 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3720 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3721 - fix lstripe in several ways (kernel side) (527)
3722 - fix request leak in ldlm_cli_enqueue (262)
3723 - incorrect OSC was marked inactive after OST failure
3724 - call mds_fs_cleanup before unmounting filesystem (524)
3725 - fix races between taking ns_lock and ldlm_lock_change_resource
3726 - fix races updating LOV export open file list
3727 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3728 - fix recovery NULL deref in ldlm_cli_cancel_unused
3729 - fix some DLM races by using new hash table for lock handles (419)
3730 - permit the client to specify desired inodes, at replay
3731 - duplicate requests when we queue them for replay reintegration
3732 - fix last_rcvd offset calculation
3733 - sync after each recovered transaction, so we always make progress
3734 - never, not always, ERESTART requests without transnos
3735 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3737 - skip replay if the MDS says that the client is already connected
3738 - don't check for a recovery-enabled export to match lctl's UUID
3739 - don't INC_USE_COUNT for phantom exports
3740 - don't crash when cleaning up phantom exports (567)
3741 - don't double-finish or set replay data for errored mdc_open requests
3742 - abort requests when they time out, so we don't get old replies
3743 - send/receive replies for AST messages again
3744 - if the client says that it doesn't have the lock, cancel it on the
3746 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3747 mark it as destroyed, it will all work out when decref is called
3748 - fix module use counts (22, 581)
3750 - ASTs now expect a reply (server cancels lock on error reply)
3752 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3755 - fix many simultaneous client startup (392)
3756 - fix dentry->d_it clobbering
3757 - credentials weren't being shipped for readdir/getattr operations
3758 - remove invalid assertions triggered during some concurrent MD
3760 - proper Lustre versions added (336, 389)
3761 - fix memory leak for create error case (398)
3762 - fix LOV locking bug that would get cli/srv out of sync
3763 - fix echo client over LOV (409)
3764 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3765 - fix double-O_EXCL intent crash (424)
3766 - avoid sending multiple lock CANCELs (352)
3768 - MDS can do multi-client recovery (modulo bugs in new code)
3770 - many updates, edits, cleanups
3772 2002-11-18 Phil Schwan <phil@clusterfs.com>
3775 - fix null d_it dereference (346)
3776 - fix full OST/dbench hang (333)
3777 - fix permission problem with file removal (286)
3778 - fix removal of OSCs from LOV when they fail
3779 - fix NULL deref during bulk timeout (214)
3780 - fix problems related to multiple filesystems on one MDS (241)
3781 - fixed serious subtle metadata locking bugs
3782 - free locks on clients when inodes are removed due to memory
3784 - fix inode pointer in lock data (285)
3785 - partial support for multiple MDS on a single host (241)
3786 - data locks weren't cancelled at clear_inode time (290, 311)
3787 - intent locks could lead to unbounded lock growth (205)
3788 - added a maximum lock count, an LRU list, and a flusher
3789 - fix multiple rename (365)
3790 - properly abstracted the echo client
3791 - OSC locked 1 byte too many; fixed
3792 - rewrote brw callback code:
3793 - fixed recovery bugs related to LOVs (306)
3794 - fixed too-many-pages-in-one-write crash (191)
3795 - fixed (again) crash in sync_io_timeout (214)
3796 - probably fixed callback-related race (385)
3798 - Add capability to MDS protocol
3799 - LDLM cancellations and callbacks on different portals
3801 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3804 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3805 - defer open object destruction to close time (601981, 138)
3806 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3807 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3808 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3809 - don't call obd_setattr in ll_file_release for destroyed objects
3810 * protocol change to lustre_msg: move |version| and add |flags|
3811 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3812 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3813 * added replay of create, unlink, link and rename operations during
3814 MDS failover; recovery should be much more robust now
3815 * remove failed OSCs from LOVs (only lov_create uses this so far)
3816 * the lustre-HOWTO was brought (more) up to date (582544)
3818 2002-10-23 Phil Schwan <phil@clusterfs.com>
3821 - in-use dentries weren't being reused properly (617851)
3822 - prevent multiple LDLM setup (599178)
3823 - fix LOV size calculations for truncate (617853)
3824 - fix client handling of MDS intent errors (POSIX)
3825 - fix permission bug in lovstripe.c test (624321)
3826 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3827 - truncate past end of file could corrupt data
3828 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3829 - a race in recovery could return ETIMEDOUT to apps (623947)
3830 - building outside the source directory was fixed
3831 * the lustre-HOWTO was brought (more) up to date (582544)
3832 * major progress was made on recovery functionality
3834 2002-10-10 Phil Schwan <phil@clusterfs.com>
3837 - recovery deadlock fix
3838 - rm -rf causes LBUG fix (617817)
3839 - file open by multiple tasks fix (618962)
3840 - directory permissions bugs (602707 and 620007)
3841 - journal_stop fixed with locking (611313)
3842 - O_APPEND failures resolved (618273, perhaps 614459)
3843 - lconf PATH fix (619770)
3844 - IA64 build fix (621450)
3845 - RPC buffer sizes scale with amount of memory
3847 2002-10-01 Phil Schwan <phil@clusterfs.com>
3850 - locks would be cancelled without throwing away data pages,
3851 resulting in inconsistent data (605627)
3852 - inode attributes were not always being refreshed (605627, 612449)
3853 - lconf now continues to cleanup after lctl reports an error
3854 - MDS now enforces user permissions (602707)
3855 - lprocfs cleanup fixed, but not yet enabled (614157)
3856 - fixed infinite server hang, should a client not respond to an AST
3857 - avoid going into recovery if user calls readlink() with a buffer
3858 that's too small (613941)
3859 - AST RPCs no longer require replies (614867) -- this may be changed
3860 - don't crash server if client sends an IOV that's too big (611336)
3861 - fixed lock conversion deadlock (611892)
3862 - fixed the following of symlinks (614622)
3863 * recovery: the server can remove locks from a client that dies, other
3864 clients can make progress
3865 * more extN patch fixes
3866 * compile-time configurable ptlrpc buffer allocations
3868 - collaborative read cache document
3869 - Lustre Lite Performance CDR document-in-progress
3871 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3874 - fix typo in patch-2.4.18
3876 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3879 - clear ptlrpc request each time in handle_incoming_request()
3880 - unlink of files now destroys the object on the OST
3882 2002-09-19 Peter Braam <braam@clusterfs.com>
3884 * add hard link support
3885 * change obdfile creation method
3886 * kernel patch changed
3888 2002-09-19 Peter Braam <braam@clusterfs.com>
3891 - stack overflow bug in extN fixed
3893 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3895 * documentation updates
3896 - add man pages for config tools
3897 - update tests/README to describe testing with new config tools
3898 - finish metadata API descriptions
3899 * bug fixes and cleanups
3900 - statfs workaround for 16TB limit
3901 - LOV stripe allocation improved, can stripe on subset of OSTs
3902 - LOV file size/IO offset was wrong for files > 4GB in size
3903 - object EA data was being dropped, caused files to be unreadable
3904 - memory overflow with non-LOV OST caused memory corruption
3905 - fixed regression tests to work with new config tools, obdfilter
3906 - fixed bug when directory size became larger than 1 block
3907 - fixed bug (for single client case) when PWD was deleted
3908 - invalidate local directory pages when doing intent-based ops
3909 - avoid LDLM oops when lock callback contained bad data
3911 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3913 * documentation updates
3914 * bug fixes and cleanups
3915 - configuration tools
3918 - 64-bit compile warnings
3919 - 64-bit internal statfs data
3921 * test_brw on persistent OST devices
3923 * lprocfs (disabled)
3925 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3927 * documentation updates
3928 * bug fixes and cleanups
3929 * configuration tools
3931 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3934 * many small fixes to 0_5_4
3935 * io/network handling
3936 * thinkos in MDS operations
3938 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3941 * crucial basic fixes to 0.5.3
3942 * IOR, Iozone work over Elan
3945 2002-08-07 Phil Schwan <phil@clusterfs.com>
3946 * version 0_5_3, our first alpha
3947 * we use the new Portals iovs
3948 * documentation updates
3949 * bug fixes and cleanups
3950 * small changes in the DLM wire protocol
3952 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3953 * version 0_5_1 with some initial stability,
3954 * locking on MD and file I/O.
3955 * documentation updates
3956 * several bug fixes since 0.5.0
3957 * small changes in wire protocol
3959 2002-07-18 Phil Schwan <phil@clusterfs.com>
3961 * delivered as Lustre Light Alpha
3962 * fixed a crash after handling invalid MDS requests
3963 * fixed directory pages for architectures with non-4k pages sizes
3965 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3966 * release version v0_4_4
3967 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3969 * Fixes a number of interruption problems with OST operations.
3970 * Update documentation for portals header changes
3971 * Move all wire protocol structs/defines to lustre_idl.h
3972 * Fixes symlink length bug.
3973 * Add tcpdump to repository.
3975 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3976 * release version v0_4_3
3977 * Fixes statfs for inodes on extN.
3978 * Fixes bug in runtests which would delete /etc/hosts.
3979 * Use 64-bit object IDs wherever possible (not into VFS though)
3980 Remove ost_get_info, which is unused by lustre, and out of date.
3982 2002-07-03 Peter Braam <braam@clusterfs.com>
3983 * release version v0_4_2 Fixes a lookup error (type not passed)
3984 * move forward to head of Portals
3985 * move forward to latest Lustre kernel
3987 2002-06-25 Peter Braam <braam@clusterfs.com>
3988 * release version v0_4_1. Hopefully stable on single node use.