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.
242 Description: "lfs find -obd UUID" prints directories
243 Details : "lfs find -obd UUID" will return all directory names instead
244 of just file names. It is incorrect because the directories
245 do not reside on the OSTs.
249 Description: use i_size_read and i_size_write in 2.6 port
250 Details : replace inode->i_size access with i_size_read/write()
252 --------------------------------------------------------------------------------
254 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
256 * Support for kernels:
257 2.6.5-7.283 (SLES 9),
258 2.6.9-55.EL (RHEL 4),
259 2.6.16.46-0.14 (SLES 10),
260 2.6.18.8 vanilla (kernel.org)
261 * Client support for unpatched kernels:
262 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
263 2.6.16 - 2.6.22 vanilla (kernel.org)
264 * Due to recently discovered recovery problems, we do not recommend
265 using patchless RHEL 4 clients with this or any earlier release.
266 * Recommended e2fsprogs version: 1.39.cfs8
267 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
268 * Starting with this release, the ldiskfs backing filesystem required
269 by Lustre is now in its own package, lustre-ldiskfs. This package
270 should be installed. It is versioned separately from Lustre and
271 may be released separately in future.
275 Description: block reactivating mgc import until all deactivates complete
276 Details : Fix race when failing back MDT/MGS to itself (testing)
278 Severity : enhancement
280 Description: add optional extra BUILD_VERSION info
281 Details : add a new environment variable (namely LUSTRE_VERS) which allows
282 to override the lustre version.
285 Frequency : 2.6.18 servers only
287 Description: ll_kern_mount() doesn't release the module reference
288 Details : The ldiskfs module reference count never drops down to 0
289 because ll_kern_mount() doesn't release the module reference.
294 Description: server LBUG when using old ost_num_threads parameter
295 Details : Accept the old ost_num_threads parameter but warn that it
296 is deprecated, and fix an off-by-one error that caused an LBUG.
301 Description: Transient SCSI error results in persistent IO issue
302 Details : iobuf->dr_error is not reinitialized to 0 between two
306 Frequency : sometimes when underlying device returns I/O errors
308 Description: OSTs not going read-only during write failures
309 Details : OSTs are not remounted read-only when the journal commit threads
310 get I/O errors because fsfilt_ext3 calls journal_start/stop()
311 instead of the ext3 wrappers.
315 Description: poor connect scaling with increasing client count
316 Details : Don't run filter_grant_sanity_check for more than 100 exports
317 to improve scaling for large numbers of clients.
320 Frequency : SLES10 only
322 Description: sanity-quota.sh quotacheck failed: rc = -22
323 Details : Quotas cannot be enabled on SLES10.
326 Frequency : liblustre clients only
328 Description: getdirentries does not give error when run on compute nodes
329 Details : getdirentries does not fail when the size specified as an argument
330 is too small to contain at least one entry
332 Severity : enhancement
334 Description: Add LNET router traceability for debug purposes
335 Details : If a checksum failure occurs with a router as part of the
336 IO path, the NID of the last router that forwarded the bulk data
337 is printed so it can be identified.
342 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
343 Details : Due to a race condition, liblustre clients were occasionally
346 Severity : enhancement
348 Description: lfs setstripe use optional parameters instead of postional
351 Severity : enhancement
353 Description: Nanosecond timestamp support for ldiskfs
354 Details : The on-disk ldiskfs filesystem has added support for nanosecond
355 resolution timestamps. There is not yet support for this at
356 the Lustre filesystem level.
359 Frequency : during server recovery
361 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
362 Details : request with rq_no_resend flag not awake l_wait_event if they get a
366 Frequency : nfs export on patchless client
368 Description: connectathon hang when test nfs export over patchless client
369 Details : Disconnected dentry cannot be found with lookup, so we do not need
370 to unhash it or make it invalid
373 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
374 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
375 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
380 Description: open req refcounting wrong on reconnect
381 Details : If reconnect happened between getting open reply from server and
382 call to mdc_set_replay_data in ll_file_open, we will schedule
383 replay for unreferenced request that we are about to free.
384 Subsequent close will crash in variety of ways.
385 Check that request is still eligible for replay in
386 mdc_set_replay_data().
391 Description: disable writes to filesystem when reading health_check file
392 Details : the default for reading the health_check proc file has changed
393 to NOT do a journal transaction and write to disk, because this
394 can cause reads of the /proc file to hang and block HA state
395 checking on a healthy but otherwise heavily loaded system. It
396 is possible to return to the previous behaviour during configure
397 with --enable-health-write.
399 Severity : enhancement
401 Description: 64-bit inode version
402 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
403 to the inode. This will be required for version-based recovery.
408 Description: MDS fails to start if a duplicate client export is detected
409 Details : in some rare cases it was possible for a client to connect to
410 an MDS multiple times. Upon recovery the MDS would detect this
411 and fail during startup. Handle this more gracefully.
413 Severity : enhancement
415 Description: Add -o localflock option to simulate old noflock
417 Details : This will achieve local-only flock/fcntl locks
423 Description: log_commit_thread vs filter_destroy race leads to crash
424 Details : Take import reference before releasing llog record semaphore
429 Description: Wrong request locking in request set processing
430 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
431 imp_delayed_list, in this place should be used imp_lock.
434 Frequency : when reconnection
436 Description: Grant Leak when osc reconnect to OST
437 Details : When osc reconnect ost, OST(filter) should check whether it
438 should grant more space to client by comparing fed_grant and
439 cl_avail_grant, and return the granted space to client instead
440 of "new granted" space, because client will call osc_init_grant
441 to update the client grant space info.
444 Frequency : when client reconnect to OST
446 Description: Grant Leak when osc do resend and replay bulk write
447 Details : When osc reconnect to OST, OST(filter)should clear grant info of
448 bulk write request, because the grant info while be sync between
449 OSC and OST when reconnect, and we should ignore the grant info
450 these of resend/replay write req.
455 Description: Grant space more than avaiable left space sometimes.
456 Details : When then OST is about to be full, if two bulk writing from
457 different clients came to OST. Accord the avaliable space of the
458 OST, the first req should be permitted, and the second one
459 should be denied by ENOSPC. But if the seconde arrived before
460 the first one is commited. The OST might wrongly permit second
461 writing, which will cause grant space > avaiable space.
464 Frequency : when client is evicted
466 Description: Grant might be wrongly erased when osc is evicted by OST
467 Details : when the import is evicted by server, it will fork another
468 thread ptlrpc_invalidate_import_thread to invalidate the
469 import, where the grant will be set to 0. While the original
470 thread will update the grant it got when connecting. So if
471 the former happened latter, the grant will be wrongly errased
472 because of this race.
477 Description: Checking Stale with correct fid
478 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
479 to check whether it is stale, because sometimes, we want the
480 enqueue happened anyway, and op_data.fid2 will not be initialized.
482 Severity : enhancement
484 Description: update patchless client
485 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
488 Frequency : only with 2.4 kernel
490 Description: random memory corruption
491 Details : size of struct ll_inode_info is to big for union inode.u and this
492 can be cause of random memory corruption.
497 Description: Memory leak in recovery
498 Details : Lov_mds_md was not free in an error handler in mds_create_object.
499 It should also check obd_fail before fsfilt_start, otherwise if
500 fsfilt_start return -EROFS,(failover mds during mds recovery).
501 then the req will return with repmsg->transno = 0 and rc = EROFS.
502 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
503 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
504 be freed no matter whether fsfilt_commit success or not.
507 Frequency : only with huge count clients
509 Description: Prevents from taking the superblock lock in llap_from_page for
511 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
512 for llap_from_page call in ll_removepage prevents from taking the
513 superblock lock for a soon died page.
518 Description: Not check open intent error before release open handle
519 Details : in some rare cases, the open intent error is not checked before
520 release open handle, which may cause
521 ASSERTION(open_req->rq_transno != 0), because it tries to release
522 the failed open handle.
527 Description: Set cat log bitmap only after create log success.
528 Details : in some rare cases, the cat log bitmap is set too early. and it
529 should be set only after create log success.
533 Description: Accessing a block bevice can re-enable I/O when Lustre is
534 tearing down a device.
535 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
540 Description: service threads may hog cpus when there are a lot of requests
542 Details : Insert cond_resched to give other threads a chance to use some of
548 Description: the cat log was not initialized in recovery
549 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
550 unlink log on mds will not be initialized until mds post
551 recovery. And also in mds post recovery, the unlink log will
552 initialization will be done asynchronausly, so there will be race
553 between add unlink log and unlink log initialization.
557 Description: brw_stats were being printed incorrectly
558 Details : brw_stats were being printed as log2 but all of them were not
559 recorded as log2. Also remove some code duplication arising from
560 filter_tally_{read,write}.
564 Frequency : rare, only in recovery.
565 Description: ASSERTION(req->rq_type != LI_POISON) failed
566 Details : imp_lock should be held while iterating over imp_sending_list for
567 prevent destroy request after get timeout in ptlrpc_queue_wait.
571 Description: replay-single.sh test 52 fails
572 Details : A lock's skiplist need to be cleanup when it being unlinked
573 from its resource list.
577 Description: Short directio read returns full requested size rather than
579 Details : Direct I/O operations should return actual amount of bytes
580 transferred rather than requested size.
582 Severity : enhancement
584 Description: metadata RPC reduction (e.g. for rm performance)
585 Details : decrease the amount of synchronous RPC between clients and servers
586 by canceling conflicing lock before the operation on the client side
587 and packing thier handles into the main operation RPC to server.
589 Severity : enhancement
591 Description: Async OSC create to avoid the blocking unnecessarily.
592 Details : If a OST has no remain object, system will block on the creating
593 when need to create a new object on this OST. Now, ways use
594 pre-created objects when available, instead of blocking on an
595 empty osc while others are not empty. If we must block, we block
596 for the shortest possible period of time.
598 --------------------------------------------------------------------------------
600 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
605 Frequency : on some architectures
607 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
608 Details : Uninitialized flags sometimes cause configuration commands to
612 Frequency : patchless clients only
614 Description: missing __iget() symbol export
615 Details : The __iget() symbol export is missing. To avoid the need for
616 this on patchless clients the deathrow inode reaper is turned
617 off, and we depend on the VM to clean up old inodes. This
618 dependency was during via the fix for bug 12181.
620 --------------------------------------------------------------------------------
622 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
624 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
625 INTEROPERATE with older versions automatically. In many cases a
626 special upgrade step is needed. Please read the
627 user documentation before upgrading any part of a live system.
628 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
629 version will not interoperate with 1.6 betas before beta5 (1.5.95).
630 * WARNING: Lustre configuration and startup changes are required with
631 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
637 Severity : enhancement
639 Description: MountConf
640 Details : Lustre configuration is now managed via mkfs and mount
641 commands instead of lmc and lconf. New obd types (MGS, MGC)
642 are added for dynamic configuration management. See
643 https://mail.clusterfs.com/wikis/lustre/MountConf for
646 Severity : enhancement
648 Description: dynamic OST addition
649 Details : OSTs can now be added to a live filesystem
651 Severity : enhancement
653 Description: startup order invariance
654 Details : MDTs and OSTs can be started in any order. Clients only
655 require the MDT to complete startup.
657 Severity : enhancement
659 Description: parallel, asynchronous orphan cleanup
660 Details : orphan cleanup is now performed in separate threads for each
661 OST, allowing parallel non-blocking operation.
663 Severity : enhancement
665 Description: optimized stripe assignment
666 Details : stripe assignments are now made based on ost space available,
667 ost previous usage, and OSS previous usage, in order to try
668 to optimize storage space and networking resources.
670 Severity : enhancement
672 Description: Permanently set tunables
673 Details : All writable /proc/fs/lustre tunables can now be permanently
674 set on a per-server basis, at mkfs time or on a live
677 Severity : enhancement
679 Description: Lustre message v2
680 Details : Add lustre message format v2.
682 Severity : enhancement
684 Description: client OST exclusion list
685 Details : Clients can be started with a list of OSTs that should be
686 declared "inactive" for known non-responsive OSTs.
690 Description: SPEC SFS validation failure on NFS v2 over lustre.
691 Details : Changes the blocksize for regular files to be 2x RPC size,
692 and not depend on stripe size.
694 Severity : enhancement
696 Description: Multiple MD RPCs in flight.
697 Details : Further unserialise some read-only MDS RPCs - learn about intents.
698 To avoid overly-overloading MDS, introduce a limit on number of
699 MDS RPCs in flight for a single client and add /proc controls
700 to adjust this limit.
702 Severity : enhancement
704 Description: client read/write statistics
705 Details : Add client read/write call usage stats for performance
706 analysis of user processes.
707 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
708 file access. extents_stats shows chunk size distribution.
709 extents_stats_per_process show chunk size distribution per
712 Severity : enhancement
714 Description: mds statistics
715 Details : Add detailed mds operations statistics in
716 /proc/fs/lustre/mds/*/stats.
720 Description: Failure of copying files with lustre special EAs.
721 Details : Client side always return success for setxattr call for lustre
722 special xattr (currently only "trusted.lov").
727 Description: Refcount LNET uuids
728 Details : The global LNET uuid list grew linearly with every startup;
729 refcount repeated list entries instead of always adding to
732 Severity : enhancement
734 Description: Dynamic service threads
735 Details : Within a small range, start extra service threads
736 automatically when the request queue builds up.
739 Frequency : mixed-endian client/server environments
741 Description: mixed-endian crashes
742 Details : The new msg_v2 system had some failures in mixed-endian
745 Severity : enhancement
747 Description: Easy OST removal
748 Details : OSTs can be permanently deactivated with e.g. 'lctl
749 conf_param lustre-OST0001.osc.active=0'
751 Severity : enhancement
753 Description: MGS proc entries
754 Details : Added basic proc entries for the MGS showing what filesystems
757 Severity : enhancement
759 Description: provide MGS failover
760 Details : Added config lock reacquisition after MGS server failover.
762 Severity : enhancement
764 Description: add Linux 2.4 support
765 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
769 Description: a large application tries to do I/O to the same resource and dies
771 Details : Check the req->rq_arrival time after the call to
772 ost_brw_lock_get(), but before we do anything about
773 processing it & sending the BULK transfer request. This
774 should help move old stale pending locks off the queue as
775 quickly as obd_timeout.
778 Frequency : when an incorrect nid is specified during startup
780 Description: ptlrpc connect to non-existant node causes kernel crash
781 Details : LNET can't be re-entered from an event callback, which
782 happened when we expire a message after the export has been
783 cleaned up. Instead, hand the zombie cleanup off to another
786 Severity : enhancement
788 Description: plain/inodebits lock performance improvement
789 Details : Grouping plain/inodebits in granted list by their request modes
790 and bits policy, thus improving the performance of search through
794 Frequency : only if OST filesystem is corrupted
796 Description: client incorrectly hits assertion in ptlrpc_replay_req()
797 Details : for a short time RPCs with bulk IO are in the replay list,
798 but replay of bulk IOs is unimplemented. If the OST filesystem
799 is corrupted due to disk cache incoherency and then replay is
800 started it is possible to trip an assertion. Avoid putting
801 committed RPCs into the replay list at all to avoid this issue.
804 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
806 Description: System hang on startup
807 Details : This bug allowed the liblustre (e.g. catamount) client to
808 return to the app before handling all startup RPCs. This
809 could leave the node unresponsive to lustre network traffic
810 and manifested as a server ptllnd timeout.
812 Severity : enhancement
814 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
815 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
816 boolean to control whether to print peer debug info when a
817 client's RPC times out.
820 Frequency : only for kernels with patches from Lustre below 1.4.3
822 Description: Remove old rdonly API
823 Details : Remove old rdonly API which unused from at least lustre 1.4.3
826 Frequency : only for devices with external journals
828 Description: Set external device read-only also
829 Details : During a commanded failover stop, we set the disk device
830 read-only while the server shuts down. We now also set any
831 external journal device read-only at the same time.
834 Frequency : when upgrading from 1.4 while trying to change parameters
836 Description: The wrong (new) MDC name was used when setting parameters for
837 upgraded MDT's. Also allows changing of OSC (and MDC)
838 parameters if --writeconf is specified at tunefs upgrade time.
841 Frequency : when setting specific ost indicies
843 Description: QOS code breaks on skipped indicies
844 Details : Add checks for missing OST indicies in the QOS code, so OSTs
845 created with --index need not be sequential.
847 Severity : enhancement
849 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
850 Details : The uninit_groups feature works in conjunction with the kernel
851 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
852 up the pass1 processing of e2fsck. This is a read-only feature
853 in ldiskfs2 only, so older kernels and current ldiskfs cannot
854 mount filesystems that have had this feature enabled.
856 Severity : enhancement
858 Description: Improve multi-block allocation algorithm to avoid fragmentation
859 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
860 allocation locality and avoid filesystem fragmentation.
862 ------------------------------------------------------------------------------
864 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
866 * Support for kernels:
867 2.4.21-47.0.1.EL (RHEL 3)
869 2.6.9-42.0.10.EL (RHEL 4)
870 2.6.12.6 vanilla (kernel.org)
871 2.6.16.27-0.9 (SLES 10)
872 * Recommended e2fsprogs version: 1.39.cfs5
874 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
878 Frequency : occasional, depends on client load and configuration
879 Bugzilla : 12181, 12203
880 Description: data loss for recently-modified files
882 Details : In some cases it is possible that recently written or created
883 files may not be written to disk in a timely manner (this should
884 normally be within 30s unless client IO load is very high).
885 The problem appears as zero-length files or files that are a
886 multiple of 1MB in size after a client crash or client eviction
887 that are missing data at the end of the file.
889 This problem is more likely to be hit on clients where files are
890 repeatedly created and unlinked in the same directory, clients
891 have a large amount of RAM, have many CPUs, the filesystem has
892 many OSTs, the clients are rebooted frequently, and/or the files
893 are not accessed by other nodes after being written.
895 The presence of the problem can be detected by looking at
896 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
897 smaller than the second (nr_unused) then dirty files will not
898 be flushed automatically to disk. "sync; sleep 10" should be
899 run several times on the node before unmounting it to update
900 Lustre (this is also safe to run on nodes without this problem).
902 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
903 that can cause this same problem, so customers using that kernel
904 also need to update the kernel in addition to Lustre. In order
905 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
907 It is normal that files written just before a client crash (less
908 than 30s) may not yet have been flushed to disk, even for local
912 Frequency : frequent on thin XT3 nodes
914 Description: UUID collision on thin XT3 Linux nodes
915 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
916 randomly, since we relied on an insufficiently-seeded PRNG.
921 Description: OSS hangs after "All ost request buffers busy"
922 Details : A deadlock between quota and journal operations caused OSS
923 hangs after printing "All ost request buffers busy."
926 Frequency : always on liblustre builds
928 Description: Cleanup compiler warnings on liblustre
931 Frequency : always on liblustre builds on XT3
933 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
937 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
938 > 32000 subdirectories
939 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
940 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
941 subdirectory count crosses 32000. This will aid e2fsck to
942 correctly handle more than 32000 subdirectories.
945 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
947 Description: System hang on startup
948 Details : This bug allowed the liblustre (e.g. catamount) client to
949 return to the app before handling all startup RPCs. This
950 could leave the node unresponsive to lustre network traffic
951 and manifested as a server ptllnd timeout.
953 Severity : enhancement
955 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
956 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
957 boolean to control whether to print peer debug info when a
958 client's RPC times out.
963 Description: make O_SYNC working on 2.6 kernels
964 Details : 2.6 kernels use different method for mark pages for write,
965 so need add a code to lustre for O_SYNC work.
970 Description: Failure to close file and release space on NFS
971 Details : Put inode details into lock acquired in ll_intent_file_open.
972 Use mdc_intent_lock in ll_intent_open to properly
973 detect all kind of errors unhandled by mdc_enqueue.
978 Description: proc file read during shutdown sometimes raced obd removal,
980 Details : Add lock to prevent obd access after proc file removal.
983 Frequency : Only for files larger than 4GB on 32-bit clients.
985 Description: improperly doing page alignment of locks
986 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
987 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
992 Description: RPCs being resent when they shouldn't be
993 Details : Some RPCs that should not be resent are being resent. This
994 can cause inconsistencies in the RPC state machine. Do not
995 resend such requests.
998 Frequency : rare, only with NFS export
1000 Description: Crash on NFS re-export node
1001 Details : under very unusual load conditions an assertion is hit in
1002 ll_intent_file_open()
1005 Frequency : only if OST filesystem is corrupted
1007 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1008 Details : for a short time RPCs with bulk IO are in the replay list,
1009 but replay of bulk IOs is unimplemented. If the OST filesystem
1010 is corrupted due to disk cache incoherency and then replay is
1011 started it is possible to trip an assertion. Avoid putting
1012 committed RPCs into the replay list at all to avoid this issue.
1017 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1018 Details : Large single O_DIRECT read and write calls can fail to allocate
1019 a sufficiently large buffer to process the request. In case of
1020 allocation failure the allocation is retried with a smaller
1021 buffer and broken into smaller requests.
1023 Severity : enhancement
1025 Description: Add -o localflock option to simulate old noflock behaviour.
1026 Details : This will achieve local-only flock/fcntl locks coherentness.
1031 Description: versioning check is incomplete
1032 Details : Checking the version difference of client vs. server, report
1033 error if the gap is too big.
1038 Description: add support PG_writeback bit
1039 Details : add support for PG_writeback bit for Lustre, for more carefull
1040 work with page cache in 2.6 kernel. This also fix some deadlocks
1041 and remove hack for work O_SYNC with 2.6 kernel.
1043 Severity : enhancement
1045 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1046 Details : The uninit_groups feature works in conjunction with the kernel
1047 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1048 up the pass1 processing of e2fsck. This is a read-only feature
1049 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1050 mount filesystems that have had this feature enabled.
1052 Severity : enhancement
1054 Description: Improve multi-block allocation algorithm to avoid fragmentation
1055 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1056 allocation locality and avoid filesystem fragmentation.
1058 ------------------------------------------------------------------------------
1060 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1062 * Support for kernels:
1063 2.6.9-42.0.3.EL (RHEL 4)
1064 2.6.5-7.276 (SLES 9)
1065 2.4.21-47.0.1.EL (RHEL 3)
1066 2.6.12.6 vanilla (kernel.org)
1067 2.6.16.21-0.8 (SLES10)
1068 * Recommended e2fsprogs version: 1.39.cfs2-0
1070 * The backwards-compatible /proc/sys/portals symlink has been removed
1071 in this release. Before upgrading, please ensure that you change
1072 any configuration scripts or /etc/sysctl.conf files that access
1073 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1074 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1075 in advance of the upgrade on any system running Lustre 1.4.6 or
1076 newer, since /proc/sys/lnet was added in that version.
1077 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1081 Frequency : only when quota is used
1083 Description: avoid scanning export list for quota master
1084 Details : Change the algorithms to avoid scanning export list in order
1085 to improve the efficiency.
1088 Frequency : MDS failover only, very rarely
1090 Description: "went back in time" messages on mds failover
1091 Details : The greatest transno may be lost when the current operation
1092 finishes with an error (transno==0) and the client's last_rcvd
1093 record is over-written. Save the greatest transno in the
1094 mds_last_transno for this case.
1097 Frequency : always for specific kernels and striping counts
1099 Description: client may get "Matching packet too big" without ACL support
1100 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1101 when trying to access files in certain configurations. The
1102 clients should in fact be denied when mounting because they do
1103 not understand ACLs.
1106 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1108 Description: many clients connecting with IO in progress causes connect timeouts
1109 Details : Avoid synchronous journal commits to avoid delays caused by many
1110 clients connecting/disconnecting when bulk IO is in progress.
1111 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1112 OST_IO_PORTAL to avoid delays behind potentially many pending
1116 Frequency : occasionally with multiple writers to a single file
1118 Description: shared writes to file may result in wrong size reported by stat()
1119 Details : Allow growing of kms when extent lock is cancelled
1122 Frequency : always with random mmap IO to multi-striped file
1124 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1125 Details : Only if the hole is at the end of OST object so that kms is too
1126 small. Fix is to increase kms accordingly in ll_nopage.
1129 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1131 Description: writes to a missing object would leak memory on the OST
1132 Details : If there is an inconsistency between the MDS and OST filesystems,
1133 such that the MDS references an object that doesn't exist, writes
1134 to that object will leak memory due to incorrect cleanup in the
1135 error handling path, eventually running out of memory on the OST.
1140 Description: Creating too long symlink causes lustre errors
1141 Details : Check symlink and name lengths before sending requests to MDS.
1144 Frequency : only if flock is enabled (not on by default)
1146 Description: posix locks not released on fd closure on 2.6.9+
1147 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1148 kernels, this caused such locks not to be released on fd close and
1149 then assertions on fs unmount about still used locks.
1152 Frequency : MDS failover only, very rarely
1154 Description: clients may get ASSERTION(granted_lock != NULL)
1155 Details : When request was taking a long time, and a client was resending
1156 a getattr by name lock request. The were multiple lock requests
1157 with the same client lock handle and
1158 mds_getattr_name->fixup_handle_for_resent_request found one of the
1159 lock handles but later failed with ASSERTION(granted_lock != NULL).
1164 Description: handle->h_buffer_credits > 0, assertion failure
1165 Details : h_buffer_credits is zero after truncate, causing assertion
1166 failure. This patch extends the transaction or creates a new
1170 Frequency : NFS re-export or patchless client
1171 Bugzilla : 11179, 10796
1172 Description: Crash on NFS re-export node (__d_move)
1173 Details : We do not want to hash the dentry if we don't have a lock.
1174 But if this dentry is later used in d_move, we'd hit uninitialised
1175 list head d_hash, so we just do this to init d_hash field but
1176 leave dentry unhashed.
1179 Frequency : NFS re-export or patchless client
1181 Description: NFS exports has problem with symbolic link
1182 Details : lustre client didn't properly install dentry when re-exported
1183 to NFS or running patchless client.
1186 Frequency : NFS re-export or patchless client
1188 Description: Various nfs/patchless fixes.
1189 Details : fixes reuse disconected alias for lookup process - this fixes
1190 warning "find_exported_dentry: npd != pd",
1191 fix permission error with open files at nfs.
1192 fix apply umask when do revalidate.
1195 Frequency : occasional
1197 Description: Crash on NFS re-export node
1198 Details : calling clear_page() on the wrong pointer triggered oops in
1199 generic_mapping_read().
1202 Frequency : rarely, using O_DIRECT IO
1204 Description: unaligned directio crashes client with LASSERT
1205 Details : check for unaligned buffers before trying any requests.
1208 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1210 Description: stale data returned from RAID cache
1211 Details : If only a small amount of IO is done to the RAID device before
1212 reading it again it is possible to get stale data from the RAID
1213 cache instead of reading it from disk.
1216 Frequency : always for sles10 kernel
1218 Description: sles10 support
1219 Details : ll_follow_link: compile fixes and using of nd_set_link
1220 under newer kernels.
1223 Frequency : depends on arch, kernel and compiler version, always on sles10
1226 Description: recursive or deep enough symlinks cause stack overflow
1227 Details : getting rid of large stack-allocated variable in
1231 Frequency : depends on hardware
1233 Description: lustre write performance loss in the SLES10 kernel
1234 Details : the performance loss is caused by using of write barriers in the
1235 ext3 code. The SLES10 kernel turns barrier support on by
1236 default. The fix is to undo that change for ldiskfs.
1238 ------------------------------------------------------------------------------
1240 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1242 * Support for kernels:
1243 2.6.9-42.0.3EL (RHEL 4)
1244 2.6.5-7.276 (SLES 9)
1245 2.4.21-47.0.1.EL (RHEL 3)
1246 2.6.12.6 vanilla (kernel.org)
1250 Frequency : quota enabled and large files being deleted
1252 Description: releasing more than 4GB of quota at once hangs OST
1253 Details : If a user deletes more than 4GB of files on a single OST it
1254 will cause the OST to spin in an infinite loop. Release
1255 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1260 Description: statfs data retrieved from /proc may be stale or zero
1261 Details : When reading per-device statfs data from /proc, in the
1262 {kbytes,files}_{total,free,avail} files, it may appear
1263 as zero or be out of date.
1266 Frequency : systems with MD RAID1 external journal devices
1268 Description: lconf's call to blkid is confused by RAID1 journal devices
1269 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1270 of returning all block devices that match the journal UUID.
1273 Frequency : always, for aggregate stripe size over 4GB
1275 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1276 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1277 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1280 Frequency : always if "lfs find" used on a local file/directory
1282 Description: "lfs find" segfaults if used on a local file/directory
1283 Details : The case where a directory component was not specified wasn't
1284 handled correctly. Handle this properly.
1287 Frequency : always on ppc64
1289 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1290 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1291 perform endianness conversion.
1294 Frequency : rarely (truncate to non-zero file size after write under load)
1295 Bugzilla : 10730, 10687
1296 Description: Files padded with zeros to next 4K multiple
1297 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1298 it is possible that files that are truncated to a non-zero size
1299 immediately after being written are filled with zero bytes beyond
1300 the truncated size. No file data is lost.
1302 Severity : enhancement
1304 Description: Allow recovery/failover for liblustre clients.
1305 Details : liblustre clients were unaware of failover configurations until
1308 Severity : enhancement
1310 Description: user file locks should fail when not mounting with flock option
1311 Details : Set up an error-returning stub in ll_file_operations.lock field
1312 to prevent incorrect behaviour when client is mounted without
1313 flock option. Also, set up properly f_op->flock field for
1317 Frequency : always on ia64
1319 Description: "lfs df" loops on printing out MDS statfs information
1320 Details : The obd_ioctl_data was not initialized and in some systems
1321 this caused a failure during the ioctl that did not return
1322 an error. Initialize the struct and return an error on failure.
1325 Frequency : SLES 9 only
1327 Description: Error of copying files with lustre special EAs as root
1328 Details : Client side always return success for setxattr call for lustre
1329 special xattr (currently only "trusted.lov").
1332 Frequency : rarely on clusters with both ia64+i386 clients
1334 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1335 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1336 clients are doing concurrent writes to the same file it is
1337 possible that the ia64 clients may LASSERT because the OST
1338 extent locks are not PAGE_SIZE aligned. Ensure that grown
1339 locks are always aligned on the request boundary.
1342 Frequency : specific use, occasional
1344 Description: Overwriting in use executable truncates on-disk binary image
1345 Details : If one node attempts to overwrite an executable in use by
1346 another node, we now correctly return ETXTBSY instead of
1347 truncating the file.
1349 Severity : enhancement
1351 Description: Async OSC create to avoid the blocking unnecessarily.
1352 Details : If a OST has no remain object, system will block on the creating
1353 when need to create a new object on this OST. Now, ways use
1354 pre-created objects when available, instead of blocking on an
1355 empty osc while others are not empty. If we must block, we block
1356 for the shortest possible period of time.
1361 Description: chmod on Lustre root is propagated to other clients
1362 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1363 invalid by the follow_mount time.
1366 Frequency : liblustre clients only
1368 Description: Race in 'instant cancel' lock handling could lead to such locks
1369 never to be granted in case of SMP MDS
1370 Details : Do not destroy not yet granted but cbpending locks in
1374 Frequency : replay/resend of open
1376 Description: non null lock assetion failure in mds_intent_policy
1377 Details : Trying to replay/resend lockless open requests resulted in
1378 mds_open() returning 0 with no lock. Now it sets a flag if
1379 a lock is going to be returned.
1381 Severity : enhancement
1383 Description: Checksum enhancements
1384 Details : New checksum enhancements allow for resending RPCs that failed
1387 Severity : enhancement
1389 Description: Tunables on number of dirty pages in cacche
1390 Details : Allow to set limit on number of dirty pages cached.
1395 Description: client crash on unmount - lock still has references
1396 Details : In some error handling cases it was possible to leak a lock
1397 reference on a client while accessing a file. This was not
1398 harmful to the client during operation, but would cause the
1399 client to crash when the filesystem is unmounted.
1402 Frequency : specific case, rare
1404 Description: ETXTBSY on mds though file not in use by client
1405 Details : ETXTBSY is no longer incorrectly returned when attempting to
1406 chmod or chown a directory that the user previously tried to
1407 execute or a currently-executing binary.
1410 Frequency : extremely rare except on liblustre-based clients
1412 Description: Lustre space not freed when files are deleted
1413 Details : Clean up open-unlinked files after client eviction. Previously
1414 the unlink was skipped and the files remained as orphans.
1419 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1420 Details : In some cases it was possible to send a blocking callback to
1421 a client doing a glimpse, even though that client didn't get
1422 a lock granted. When the glimpse lock is cancelled on the OST
1423 the freed lock is left on the waiting list and corrupted the list.
1426 Frequency : all core dumps
1428 Description: Broke core dumps to lustre
1429 Details : Negative dentry may be unhashed if parent does not have UPDATE
1430 lock, but some callers, e.g. do_coredump, expect dentry to be
1431 hashed after successful create, hash it in ll_create_it.
1433 ------------------------------------------------------------------------------
1435 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1437 * Support for kernels:
1438 2.6.9-42.0.2.EL (RHEL 4)
1439 2.6.5-7.276 (SLES 9)
1440 2.4.21-40.EL (RHEL 3)
1441 2.6.12.6 vanilla (kernel.org)
1445 Frequency : always on RHEL 3
1447 Description: Number of open files grows over time
1448 Details : The number of open files grows over time, whether or not
1449 Lustre is started. This was due to a filp leak introduced
1450 by one of our kernel patches.
1452 ------------------------------------------------------------------------------
1454 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1456 * Support for kernels:
1457 2.6.9-42.EL (RHEL 4)
1458 2.6.5-7.276 (SLES 9)
1459 2.4.21-40.EL (RHEL 3)
1460 2.6.12.6 vanilla (kernel.org)
1465 Bugzilla : 5719, 9635, 9792, 9684
1466 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1467 Details : If a server is under heavy load and cannot reply to new
1468 connection requests before the client resends the (re)connect,
1469 the connection handling code can behave badly if two service
1470 threads are concurrently handing separate (re)connections from
1471 the same client. Add better locking to the connection handling
1472 code, and ensure that only a single connection will be processed
1473 for a given client UUID, even if the lock is dropped.
1475 Severity : enhancement
1477 Description: add TCP zero-copy support to kernel
1478 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1479 sends if the hardware supports scatter-gather and checksumming.
1480 This allows socklnd to do client-write and server-read more
1481 efficiently and reduce CPU utilization from skbuf copying.
1484 Frequency : only if NFS exporting from client
1486 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1487 Details : If setting attributes on a file created under NFS that had
1488 never been opened it would be possible to oops the client
1489 if the file had no objects.
1492 Frequency : always for liblustre
1494 Description: liblustre client does MDS+OSTs setattr RPC for each write
1495 Details : When doing a write from a liblustre client, the client
1496 incorrectly issued an RPC to the MDS and each OST the file was
1497 striped over in order to update the timestamps. When writing
1498 with small chunks and many clients this could overwhelm the MDS
1499 with RPCs. In all cases it would slow down the write because
1500 these RPCs are unnecessary.
1502 Severity : enhancement
1504 Description: allow number of MDS service threads to be changed at module load
1505 Details : It is now possible to change the number of MDS service threads
1506 running. Adding "options mds mds_num_threads={N}" to the MDS's
1507 /etc/modprobe.conf will set the number of threads for the next
1508 time Lustre is restarted (assuming the "mds" module is also
1509 reloaded at that time). The default number of threads will
1510 stay the same, 32 for most systems.
1515 Description: OST crash if filesystem is unformatted or corrupt
1516 Details : If an OST is started on a device that has never been formatted
1517 or if the filesystem is corrupt and cannot even mount then the
1518 error handling cleanup routines would dereference a NULL pointer.
1523 Description: NULL pointer deref in llap_from_page.
1524 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1525 incorrect) mapping if the page was truncated/reclaimed while it was
1526 searched for. Check for this condition and skip such pages when
1527 doing readahead. Introduce extra check to llap_from_page() to
1528 verify page->mapping->host is non-NULL (so page is not anonymous).
1531 Frequency : Sometimes when using sys_sendfile
1533 Description: "page not covered by a lock" warnings from ll_readpage
1534 Details : sendfile called ll_readpage without right page locks present.
1535 Now we introduced ll_file_sendfile that does necessary locking
1536 around call to generic_file_sendfile() much like we do in
1540 Frequency : with certain MDS communication failures at client mount time
1542 Description: NULL pointer deref after failed client mount
1543 Details : a client connection request may delayed by the network layer
1544 and not be sent until after the PTLRPC layer has timed out the
1545 request. If the client fails the mount immediately it will try
1546 to clean up before the network times out the request. Add a
1547 reference from the request import to the obd device and delay
1548 the cleanup until the network drops the request.
1551 Frequency : occasionally during client (re)connect
1553 Description: assertion failure during client (re)connect
1554 Details : processing a client connection request may be delayed by the
1555 client or server longer than the client connect timeout. This
1556 causes the client to resend the connection request. If the
1557 original connection request is replied in this interval, the
1558 client may trip an assertion failure in ptlrpc_connect_interpret()
1559 which thought it would be the only running connect process.
1562 Frequency : only with obd_echo servers and clients that are rebooted
1564 Description: kernel BUG accessing uninitialized data structure
1565 Details : When running an obd_echo server it did not start the ping_evictor
1566 thread, and when a client was evicted an uninitialized data
1567 structure was accessed. Start the ping_evictor in the RPC
1568 service startup instead of the OBD startup.
1570 Severity : enhancement
1571 Bugzilla : 10193 (patchless)
1572 Description: Remove dependency on various unexported kernel interfaces.
1573 Details : No longer need reparent_to_init, exit_mm, exit_files,
1574 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1577 Frequency : rare (only users of deprecated and unsupported LDAP config)
1579 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1580 Details : LDAP apparently contains 'lustreName' attributes instead of
1581 'name'. A simple remapping of the name is sufficient.
1584 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1586 Description: waiting_locks_callback trips kernel BUG if client is evicted
1587 Details : Running with the dump_on_timeout debug flag turned on makes
1588 it possible that the waiting_locks_callback() can try to dump
1589 the Lustre kernel debug logs from an interrupt handler. Defer
1590 this log dumping to the expired_lock_main() thread.
1592 Severity : enhancement
1594 Description: Support NFS exporting on 2.6 kernels.
1595 Details : Implement non-rawops metadata methods for NFS server to use without
1596 changing NFS server code.
1599 Frequency : very rare (synthetic metadata workload only)
1601 Description: two racing renames might cause an MDS thread to deadlock
1602 Details : Running the "racer" program may cause one MDS thread to rename
1603 a file from being the source of a rename to being the target of
1604 a rename at exactly the same time that another thread is doing
1605 so, and the second thread has already enqueued these locks after
1606 doing a lookup of the target and is trying to relock them in
1607 order. Ensure that we don't try to re-lock the same resource.
1610 Frequency : only very large systems with liblustre clients
1612 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1613 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1614 all OSTs in parallel. This allows the eviction of stale liblustre
1615 clients to proceed much faster than if they were done in series,
1616 and also offers similar improvements for other set_info RPCs.
1621 Description: excessive CPU usage during initial read phase on client
1622 Details : During the initial read phase on a client, it would agressively
1623 retry readahead on the file, consuming too much CPU and impacting
1624 performance (since 1.4.5.8). Improve the readahead algorithm
1625 to avoid this, and also improve some other common cases (read
1626 of small files in particular, where "small" is files smaller than
1627 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1632 Description: MDS crash when receiving packet with unknown intent.
1633 Details : Do not LBUG in unknown intent case, just return -EFAULT
1635 Severity : enhancement
1636 Bugzilla : 9293, 9385
1637 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1638 Details : Do not serialize getattr (non-intent version) and statfs.
1641 Frequency : occasional, when OST network is overloaded/intermittent
1643 Description: client evicted by OST after bulk IO timeout
1644 Details : If a client sends a bulk IO request (read or write) the OST
1645 may evict the client if it is unresposive to its data GET/PUT
1646 request. This is incorrect if the network is overloaded (takes
1647 too long to transfer the RPC data) or dropped the OST GET/PUT
1648 request. There is no need to evict the client at all, since
1649 the pinger and/or lock callbacks will handle this, and the
1650 client can restart the bulk request.
1653 Frequency : Always when mmapping file with no objects
1655 Description: client crashes when mmapping file with no objects
1656 Details : Check that we actually have objects in a file before doing any
1657 operations on objects in ll_vm_open, ll_vm_close and
1663 Description: Request leak when working with deleted CWD
1664 Details : Introduce advanced request refcount tracking for requests
1665 referenced from lustre intent.
1667 Severity : Enhancement
1669 Description: Cache open file handles on client.
1670 Details : MDS now will return special lock along with openhandle, if
1671 requested and client is allowed to hold openhandle, even if unused,
1672 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1673 closing files for every read/write openration.
1675 Severity : Enhancement
1677 Description: Cache open negative dentries on client when possible.
1678 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1679 negative dentries on lock revocation.
1684 Description: Remounting a client read-only wasn't possible with a zconf mount
1685 Details : It wasn't possible to remount a client read-only with llmount.
1687 Severity : enhancement
1688 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1689 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1690 MPICH is included to add Lustre-specific ADIO interfaces.
1691 This is based closely on the UFS ADIO layer and only differs
1692 in file creation, in order to allow the OST striping to be set.
1693 This is user-contributed code and not supported by CFS.
1698 Description: extended inode attributes (immutable, append-only) work improperly
1699 when 2.4 and 2.6 kernels are used on client/server or vice versa
1700 Details : Introduce kernel-independent values for these flags.
1702 Severity : enhancement
1705 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1706 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1707 granularity. Now, Fractional MB granularity is be supported,
1708 this is very useful for low memory system.
1710 Severity : enhancement
1712 Description: Getattr by fid
1713 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1714 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1715 have locks and avoids getattr rpc requests for it on every stat.
1718 Frequency : Always, for filesystems larger than 2TB
1720 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1721 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1722 filesystems larger than 2TB to be mounted with mballoc enabled.
1725 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1727 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1728 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1729 filesystems larger than 2TB is mounted then the kernel will
1730 silently corrupt the start of the filesystem. CONFIG_LBD is
1731 enabled for all CFS-supported kernels, but the possibility of
1732 this happening with a modified kernel config exists.
1734 Severity : enhancement
1736 Description: add client O_DIRECT support for 2.6 kernels
1737 Details : It is now possible to do O_DIRECT reads and writes to files
1738 in the Lustre client mountpoint on 2.6 kernel clients.
1740 Severity : enhancement
1742 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1743 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1744 parallel, not waiting for response from every OST before sending
1745 a rpc to the next OST.
1750 Description: setattr vs write race when updating file timestamps
1751 Details : Client processes that update a file timestamp into the past
1752 right after writing to the file (e.g. tar) it is possible that
1753 the updated file modification time can be reset to the current
1754 time due to a race between processing the setattr and write RPC.
1756 Severity : enhancement
1758 Description: Bring 'lfs find' closer in line with regular Linux find.
1759 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1760 -print0 options and obtains all the needed info through the lustre
1763 Severity : enhancement
1765 Description: support up to 1024 configured devices on one node
1766 Details : change obd_dev array from statically allocated to dynamically
1767 allocated structs as they are first used to reduce memory usage
1772 Description: Flush dirty partially truncated pages during truncate
1773 Details : Immediatelly flush partially truncated pages in filter_setattr,
1774 this way we completely avoid having any pages in page cache on OST
1775 and can retire ugly workarounds during writes to flush such pages.
1780 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1781 Details : protect inode from truncation within vfs_unlink() context
1782 just take a reference before calling vfs_unlink() and release it
1783 when parent's i_sem is free.
1788 Description: last_id value checked outside lock on OST caused LASSERT failure
1789 Details : If there were multiple MDS->OST object precreate requests in
1790 flight, it was possible that the OST's last object id was checked
1791 outside a lock and incorrectly tripped an assertion. Move checks
1792 inside locks, and discard old precreate requests.
1795 Frequency : always, if extents are used on OSTs
1797 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1798 Details : OSTs using the extents format would not zero the high 16 bits of
1799 the index physical block number. This is not a problem for any
1800 OST filesystems smaller than 16TB, and no kernels support ext3
1801 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1802 new/modified files) and can be fixed for existing filesystems
1803 with e2fsprogs-1.39-cfs1.
1808 Description: import connection selection may be incorrect if timer wraps
1809 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1810 import connections to be ignored if the 32-bit jiffies counter
1811 wraps. Use a 64-bit jiffies counter.
1814 Frequency : very large clusters immediately after boot
1816 Description: LNET request buffers exhausted under heavy short-term load
1817 Details : If a large number of client requests are generated on a service
1818 that has previously never seen so many requests it is possible
1819 that the request buffer growth cannot keep up with the spike in
1820 demand. Instead of dropping incoming requests, they are held in
1821 the LND until the RPC service can accept more requests.
1824 Frequency : Sometimes during replay
1826 Description: Assertion failure in ll_local_open after replay.
1827 Details : If replay happened on an open request reply before we were able
1828 to set replay handler, reply will become not swabbed tripping the
1829 assertion in ll_local_open. Now we set the handler right after
1830 recognising of open request
1833 Frequency : very rare
1835 Description: kernel reports "badness in vsnprintf"
1836 Details : Reading from the "recovery_status" /proc file in small chunks
1837 may cause a negative length in lprocfs_obd_rd_recovery_status()
1838 call to vsnprintf() (which is otherwise harmless). Exit early
1839 if there is no more space in the output buffer.
1841 Severity : enhancement
1843 Description: clear OBD RPC statistics by writing to them
1844 Details : It is now possible to clear the OBD RPC statistics by writing
1845 to the "stats" file.
1850 Description: Client mtime is not the same on different clients after utimes
1851 Details : In some cases, the client was using the utimes() syscall on
1852 a file cached on another node. The clients now validate the
1853 ctime from the MDS + OSTs to determine which one is right.
1858 Description: Inability to activate failout mode
1859 Details : lconf script incorrectly assumed that in pythong string's numeric
1860 value is used in comparisons.
1863 Frequency : always with multiple stripes per file
1865 Description: Inefficient object allocation for mutli-stripe files
1866 Details : When selecting which OSTs to stripe files over, for files with
1867 a stripe count that divides evenly into the number of OSTs,
1868 the MDS is always picking the same starting OST for each file.
1869 Return the OST selection heuristic to the original design.
1874 Description: mount failures may take full timeout to return an error
1875 Details : Under some heavy load conditions it is possible that a
1876 failed mount can wait for the full obd_timeout interval,
1877 possibly several minutes, before reporting an error.
1878 Instead return an error as soon as the status is known.
1880 Frequency : quota enabled and large files being deleted
1882 Description: releasing more than 4GB of quota at once hangs OST
1883 Details : If a user deletes more than 4GB of files on a single OST it
1884 will cause the OST to spin in an infinite loop. Release
1885 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1890 Description: statfs data retrieved from /proc may be stale or zero
1891 Details : When reading per-device statfs data from /proc, in the
1892 {kbytes,files}_{total,free,avail} files, it may appear
1893 as zero or be out of date.
1896 Frequency : systems with MD RAID1 external journal devices
1898 Description: lconf's call to blkid is confused by RAID1 journal devices
1899 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1900 of returning all block devices that match the journal UUID.
1903 Frequency : always, for aggregate stripe size over 4GB
1905 Description: assertion fails when trying to use 4GB stripe size
1906 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1907 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1910 Frequency : always on ppc64
1912 Description: the first write on an ext3 filesystem with mballoc got stuck
1913 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1914 perform endianness conversion.
1916 ------------------------------------------------------------------------------
1918 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1920 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1921 INTEROPERATE with older versions automatically. Please read the
1922 user documentation before upgrading any part of a live system.
1923 * WARNING: Lustre networking configuration changes are required with
1924 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1927 * Support for kernels:
1928 2.6.9-22.0.2.EL (RHEL 4)
1929 2.6.5-7.244 (SLES 9)
1930 2.6.12.6 vanilla (kernel.org)
1933 Severity : enhancement
1934 Bugzilla : 7981/8208
1935 Description: Introduced Lustre Networking (LNET)
1936 Details : LNET is new networking infrastructure for Lustre, it includes
1937 a reorganized network configuration mode (see the user
1938 documentation for full details) as well as support for routing
1939 between different network fabrics. Lustre Networking Devices
1940 (LNDs) for the supported network fabrics have also been
1941 created for this new infrastructure.
1943 Severity : enhancement
1944 Description: Introduced Access control lists
1945 Details : clients can set ACLs on files and directories in order to have
1946 more fine-grained permissions than the standard Unix UGO+RWX.
1947 The MDS must be started with the "-o acl" mount option.
1949 Severity : enhancement
1950 Description: Introduced filesystem quotas
1951 Details : Administrators may now establish per-user quotas on the
1954 Severity : enhancement
1956 Description: Configuration change for the XT3
1957 The PTLLND is now used to run Lustre over Portals on the XT3
1958 The configure option(s) --with-cray-portals are no longer used.
1959 Rather --with-portals=<path-to-portals-includes> is used to
1960 enable building on the XT3. In addition to enable XT3 specific
1961 features the option --enable-cray-xt3 must be used.
1966 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1967 Details : A race between allocating a new llcd and re-getting the llcd_lock
1968 allowed another thread to grab newly-allocated llcd.
1970 Severity : enhancement
1972 Description: 2.6 OST async journal commit and locking fix to improve performance
1973 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1974 2.6 kernels are now async as they already were in 2.4 kernels so
1975 that they can commit concurrently with the network bulk transfer.
1976 For block-allocated files the filter allocation semaphore is held
1977 to avoid filesystem fragmentation during allocation. BKL lock
1978 removed for 2.6 xattr operations where it is no longer needed.
1983 Description: lconf incorrectly determined whether two IP networks could talk
1984 Details : In some more complicated routing and multiple-network
1985 configurations, lconf will avoid trying to make a network
1986 connection to a disjoint part of the IP space. It was doing the
1987 math incorrectly for one set of cases.
1992 Description: Fix for potential infinite loop processing records in an llog.
1993 Details : If an llog record is corrupted/zeroed, it is possible to loop
1994 forever in llog_process(). Validate the llog record length
1995 and skip the remainder of the block on error.
1998 Frequency : occasional (liblustre only)
2000 Description: liblustre could not open files whose last component is a symlink
2001 Details : sysio_path_walk() would incorrectly pass the open intent to
2002 intermediate path components.
2005 Frequency : rare (liblustre only with non-standard tuning)
2006 Bugzilla : 7201 (7350)
2007 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2008 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2009 after it has been released, only do this for referenced locks
2011 Severity : enhancement
2013 Description: specifying an (invalid) directory default stripe_size of -1
2014 would reset the directory default striping
2015 Details : stripe_size -1 was used internally to signal directory stripe
2016 removal, now use "all default" to signal dir stripe removal
2017 as a directory striping of "all default" is not useful
2020 Frequency : common for large clusters running liblustre clients
2022 Description: doing an ls when liblustre clients are running is slow
2023 Details : sending a glimpse AST to a liblustre client waits for every AST
2024 to time out, as liblustre clients will not respond. Since they
2025 cannot cache data we refresh the OST lock LVB from disk instead.
2027 Severity : enhancement
2029 Description: doing an ls at the same time as file IO can be slow
2030 Details : enqueue and other "small" requests can be blocked behind many
2031 large IO requests. Create a new OST IO portal for non-IO
2032 requests so they can be processed faster.
2035 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2037 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2038 created file with mode 000
2039 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2040 allow an ftruncate() as a non-root user to a file with mode 000.
2041 Lustre now respects this flag to disable mode checks when
2042 truncating a file owned by the user
2045 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2048 Description: Revoking locks from clients that went dead or catatonic might take
2050 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2051 cancellation of such locks instant on servers without waiting for
2052 any reply from clients. Clients drops these locks when cancel
2053 notification from server is received without replying.
2056 Frequency : liblustre-only, when liblustre client dies or becomes busy
2058 Description: Doing ls on Linux clients can take a long time with active
2060 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2061 granting such locks to it in the first place if possible. Locks
2062 are taken by proxy on the OST during the read or write and
2063 dropped immediately afterward. Add connect flags handling, do
2064 not grant locks to liblustre clients for glimpse ASTs.
2066 Severity : enhancement
2068 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2069 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2070 sequential reads and behaves badly with applications doing random
2071 reads. Improve it by reading ahead at least read region, and
2072 avoiding excessive large RPC for small reads.
2074 Severity : enhancement
2076 Description: Creating more than 1000 files for a single job may cause a load
2077 imbalance on the OSTs if there are also a large number of OSTs.
2078 Details : qos_prep_create() uses an OST index reseed value that is an
2079 even multiple of the number of available OSTs so that if the
2080 reseed happens in the middle of the object allocation it will
2081 still utilize the OSTs as uniformly as possible.
2086 Description: OST or MDS may oops in ping_evictor_main()
2087 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2088 but doesn't restart at beginning of obd_exports_timed list
2091 Severity : enhancement
2093 Description: improve by-nid export eviction on the MDS and OST
2094 Details : allow multiple exports with the same NID to be evicted at one
2095 time without re-searching the exports list.
2098 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2100 Description: MDS may oops in groups_free()
2101 Details : in rare race conditions a newly allocated group_info struct is
2102 freed again, and this can be NULL. The 2.4 compatibility code
2103 for groups_free() checked for a NULL pointer, but 2.6 did not.
2106 Frequency : common for liblustre clients doing little filesystem IO
2107 Bugzilla : 9352, 7313
2108 Description: server may evict liblustre clients accessing contended locks
2109 Details : if a client is granted a lock or receives a completion AST
2110 with a blocking AST already set it would not reply to the AST
2111 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2114 Frequency : lfs setstripe, only systems with more than 160 OSTs
2116 Description: unable to set striping with a starting offset beyond OST 160
2117 Details : llapi_create_file() incorrectly limited the starting stripe
2118 index to the maximum single-file stripe count.
2121 Frequency : LDAP users only
2123 Description: lconf did not handle in-kernel recovery with LDAP properly
2124 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2126 Severity : enhancement
2128 Description: bind OST threads to NUMA nodes to improve performance
2129 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2130 node and do their allocations there to localize memory access
2132 Severity : enhancement
2134 Description: llmount can determine client NID directly from Myrinet (GM)
2135 Details : the client NID code from gmnalnid was moved directly into
2136 llmount, removing the need to use this or specifying the
2137 client NID explicitly when mounting GM clients with zeroconf
2140 Frequency : if client is started with down MDS
2142 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2143 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2144 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2145 ptlrpc_import_delayed_req() because we didn't check if the
2146 request was interrupted, and we also didn't break out of the
2147 event loop if there was no timeout
2152 Description: data loss during non-page-aligned writes to a single file from
2153 both multiple nodes and multiple threads on one node at same time
2154 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2155 inconsistency led to false short-reads, that were cached and later
2156 used by ->prepare_write() to fill in partially written page,
2157 leading to data loss.
2160 Frequency : always, if lconf --abort_recovery used
2162 Description: lconf --abort_recovery fails with 'Operation not supported'
2163 Details : lconf was attempting to abort recovery on the MDT device and not
2166 Severity : enhancement
2168 Description: remove cleanup logs
2169 Details : replace lconf-generated cleanup logs with lustre internal
2170 cleanup routines. Eliminates the need for client-cleanup and
2173 Severity : enhancement
2175 Description: add support for EAs (user and system) on lustre filesystems
2176 Details : it is now possible to store extended attributes in the Lustre
2177 client filesystem, and with the user_xattr mount option it
2178 is possible to allow users to store EAs on their files also
2180 Severity : enhancement
2182 Description: Add possibility (config option) to show minimal available OST free
2184 Details : When compiled with --enable-mindf configure option, statfs(2)
2185 (and so, df) will return least minimal free space available from
2186 all OSTs as amount of free space on FS, instead of summary of
2187 free spaces of all OSTs.
2189 Severity : enhancement
2191 Description: do not expand extent locks acquired on OST-side
2192 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2193 by server: they are not cached anyway.
2196 Frequency : when mmap is used/binaries executed from Lustre
2198 Description: Unmmap pages before throwing them away from read cache.
2199 Details : llap_shrink cache now attempts to unmap pages before discarding
2200 them (if unmapping failed - do not discard). SLES9 kernel has
2201 extra checks that trigger if this unmapping is not done first.
2206 Description: lconf didn't resolve symlinks before checking to see whether a
2207 given mountpoint was already in use
2210 Frequency : when migrating failover services
2211 Bugzilla : 6395, 9514
2212 Description: When migrating a subset of services from a node (e.g. failback
2213 from a failover service node) the remaining services would
2214 time out and evict clients.
2215 Details : lconf --force (implied by --failover) sets the global obd_timeout
2216 to 5 seconds in order to quickly disconnect, but this caused
2217 other RPCs to time out too quickly. Do not change the global
2218 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2220 Severity : enhancement
2221 Frequency : if MDS is started with down OST
2222 Bugzilla : 9439,5706
2223 Description: Allow startup/shutdown of an MDS without depending on the
2224 availability of the OSTs.
2225 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2226 Add appropriate locking and lov-osc refcounts for safe
2227 cleaning. Add osc abort_inflight calls in case the
2228 synchronize never started.
2231 Frequency : occasional (Cray XT3 only)
2233 Description: root not authorized to access files in CRAY_PORTALS environment
2234 Details : The client process capabilities were not honoured on the MDS in
2235 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2236 been accessed by an authorized user then root was able to access
2237 the file on the local client also. The root user capabilities
2238 are now allowed on the MDS, as this environment has secure UID.
2241 Frequency : occasional
2243 Description: ldiskfs "too long searching" message happens too often
2244 Details : A debugging message (otherwise harmless) prints too often on
2245 the OST console. This has been reduced to only happen when
2246 there are fragmentation problems on the filesystem.
2251 Description: Division by zero in statfs when all OSCs are inactive
2252 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2253 lov_statfs divides by value returned by lov_get_stripecnt().
2257 Bugzilla : 9489, 3273
2258 Description: First write from each client to each OST was only 4kB in size,
2259 to initialize client writeback cache, which caused sub-optimal
2260 RPCs and poor layout on disk for the first writen file.
2261 Details : Clients now request an initial cache grant at (re)connect time
2262 and so that they can start streaming writes to the cache right
2263 away and always do full-sized RPCs if there is enough data.
2264 If the OST is rebooted the client also re-establishes its grant
2265 so that client cached writes will be honoured under the grant.
2270 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2271 Details : Now I/O RPCs go to different portal number and (presumably) fast
2272 lock requests (and glimses) and other RPCs get their own service
2273 threads pool that should be able to service those RPCs
2276 Severity : enhancement
2278 Description: Ability to exchange lustre version between client and servers and
2279 issue warnings at client side if client is too old. Also for
2280 liblustre clients there is ability to refuse connection of too old
2282 Details : New 'version' field is added to connect data structure that is
2283 filled with version info. That info is later checked by server and
2287 Frequency : rare, liblustre only.
2288 Bugzilla : 9296, 9581
2289 Description: Two simultaneous writes from liblustre at offset within same page
2290 might proceed at the same time overwriting eachother with stale
2292 Details : I/O lock withing llu_file_prwv was released too early, before data
2293 actually was hitting the wire. Extended lock-holding time until
2294 server acknowledges receiving data.
2297 Frequency : extremely rare. Never observed in practice.
2299 Description: avoid generating lustre_handle cookie of 0.
2300 Details : class_handle_hash() generates handle cookies by incrementing
2301 global counter, and can hit 0 occasionaly (this is unlikely, but
2302 not impossible, because initial value of cookie counter is
2303 selected randonly). Value of 0 is used as a sentinel meaning
2304 "unassigned handle" --- avoid it. Also coalesce two critical
2305 sections in this function into one.
2307 Severity : enhancement
2309 Description: allow liblustre clients to delegate truncate locking to OST
2310 Details : To avoid overhead of locking, liblustre client instructs OST to
2311 take extent lock in ost_punch() on client's behalf. New connection
2312 flag is added to handle backward compatibility.
2314 Severity : enhancement
2315 Bugzilla : 4928, 7341, 9758
2316 Description: allow number of OST service threads to be specified
2317 Details : a module parameter allows the number of OST service threads
2318 to be specified via "options ost ost_num_threads={N}" in the
2319 OSS's /etc/modules.conf or /etc/modprobe.conf.
2323 Bugzilla : 6146, 9635, 9895
2324 Description: servers crash with bad pointer in target_handle_connect()
2325 Details : In rare cases when a client is reconnecting it was possible that
2326 the connection request was the last reference for that export.
2327 We would temporarily drop the export reference and get a new
2328 one, but this may have been the last reference and the export
2329 was just destroyed. Get new reference before dropping old one.
2331 Severity : enhancement
2332 Frequency : if client is started with failover MDS
2334 Description: Allow multiple MDS hostnames in the mount command
2335 Details : Try to read the configuration from all specified MDS
2336 hostnames during a client mount in case the "primary"
2339 Severity : enhancement
2341 Description: Stop sending data to evicted clients as soon as possible.
2342 Details : Check if the client we are about to send or are sending data to
2343 was evicted already. (Check is done every second of waiting,
2344 for which l_wait_event interface was extended to allow checking
2345 of exit condition at specified intervals).
2348 Frequency : rare, normally only when NFS exporting is done from client
2350 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2352 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2355 Frequency : always, if srand() is called before liblustre initialization
2357 Description: Liblustre uses system PRNG disturbing its usage by user application
2358 Details : Introduce internal to lustre fast and high-quality PRNG for
2359 lustre usage and make liblustre and some other places in generic
2360 lustre code to use it.
2362 Severity : enhancement
2363 Bugzilla : 9477, 9557, 9870
2364 Description: Verify that the MDS configuration logs are updated when xml is
2365 Details : Check if the .xml configuration logs are newer than the config
2366 logs stored on the MDS and report an error if this is the case.
2367 Request --write-conf, or allow starting with --old_conf.
2369 Severity : enhancement
2371 Description: Handle symlinks in the path when checking if Lustre is mounted.
2372 Details : Resolve intermediate symlinks when checking if a client has
2373 mounted a filesystem to avoid duplicate client mounts.
2378 Description: lconf can hit an error exception but still return success.
2379 Details : The lconf command catches the Command error exception at the top
2380 level script context and will exit with the associated exit
2381 status, but doesn't ensure that this exit status is non-zero.
2386 Description: failure of ptlrpc thread startup can cause oops
2387 Details : Starting a ptlrpc service thread can fail if there are a large
2388 number of threads or the server memory is very fragmented.
2389 Handle this without oopsing.
2392 Frequency : always, only if liblustre and non-default acceptor port was used
2394 Description: liblustre cannot connect to servers with non-default acceptor port
2395 Details : tcpnal_set_default_params() was not called and was therefore
2396 ignoring the environment varaible TCPNAL_PORT, as well as other
2397 TCPNAL_ environment variables
2402 Description: two objects could be created on the same OST for a single file
2403 Details : If an OST is down, in some cases it was possible to create two
2404 objects on a single OST for a single file. No problems other
2405 than potential performance impact and spurious error messages.
2409 Bugzilla : 5681, 9562
2410 Description: Client may oops in ll_unhash_aliases
2411 Details : Client dcache may become inconsistent in race condition.
2412 In some cases "getcwd" can fail if the current directory is
2418 Description: Inode refcounting problems in NFS export code
2419 Details : link_raw functions used to call d_instantiate without obtaining
2420 extra inode reference first.
2424 Bugzilla : 9942, 9903
2425 Description: Referencing freed requests leading to crash, memleaks with NFS.
2426 Details : We used to require that call to ll_revalidate_it was always
2427 followed by ll_lookup_it. Also with revalidate_special() it is
2428 possible to call ll_revalidate_it() twice for the same dentry
2429 even if first occurence returned success. This fix changes semantic
2430 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2431 reference on a request referred from the intent.
2432 ll_intent_release() then releases such a request.
2435 Frequency : rare, normally benchmark loads only
2437 Description: unlinked inodes were kept in memory on the client
2438 Details : If a client is repeatedly creating and unlinking files it
2439 can accumulate a lot of stale inodes in the inode slab cache.
2440 If there is no other client load running this can cause the
2441 client node to run out of memory. Instead flush old inodes
2442 from client cache that have the same inode number as a new inode.
2445 Frequency : SLES9 2.6.5 kernel and long filenames only
2446 Bugzilla : 9969, 10379
2447 Description: utime reports stale NFS file handle
2448 Details : SLES9 uses out-of-dentry names in some cases, which confused
2449 the lustre dentry revalidation. Change it to always use the
2453 Frequency : rare, unless heavy write-truncate concurrency is continuous
2454 Bugzilla : 4180, 6984, 7171, 9963, 9331
2455 Description: OST becomes very slow and/or deadlocked during object unlink
2456 Details : filter_destroy() was holding onto the parent directory lock
2457 while truncating+unlinking objects. For very large objects this
2458 may block other threads for a long time and slow overall OST
2459 responsiveness. It may also be possible to get a lock ordering
2460 deadlock in this case, or run out of journal credits because of
2461 the combined truncate+unlink. Solution is to do object truncate
2462 first in one transaction without parent lock, and then do the
2463 final unlink in a new transaction with the parent lock. This
2464 reduces the lock hold time dramatically.
2467 Frequency : rare, 2.4 kernels only
2469 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2470 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2471 before unmounting their filesystem in order to avoid deadlock.
2472 The kernel_locked() function in 2.4 kernels only checks whether
2473 the kernel lock is held, not whether it is this process that is
2474 holding it as 2.6 kernels do.
2479 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2480 Details : The client ptlrpc code may be trying to reconnect to a down
2481 server before a previous connection attempt has timed out.
2482 Increase the reconnect interval to be longer than the connection
2483 timeout interval to avoid sending duplicate connections to
2487 Frequency : echo_client brw_test command
2489 Description: fix echo_client to work with OST preallocated code
2490 Details : OST preallocation code (5137) didn't take echo_client IO path
2491 into account: echo_client calls filter methods outside of any
2492 OST thread and, hence, there is no per-thread preallocated
2493 pages and buffers to use. Solution: hijack pga pages for IO. As
2494 a byproduct, this avoids unnecessary data copying.
2498 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2499 Description: Client can oops in mdc_commit_close() after open replay
2500 Details : It was possible for the MDS to return an open request with no
2501 transaction number in mds_finish_transno() if the client was
2502 evicted, but without actually returning an error. Clients
2503 would later try to replay that open and may trip an assertion
2504 Simplify the client close codepath, and always return an error
2505 from the MDS in case the open is not successful.
2508 Frequency : rare, 2.6 OSTs only
2510 Description: OST may deadlock under high load on fragmented files
2511 Details : If there was a heavy load and highly-fragmented OST filesystems
2512 it was possible to have all the OST threads deadlock waiting on
2513 allocation of biovecs, because the biovecs were not released
2514 until the entire RPC IO was completed. Instead, release biovecs
2515 as soon as they are complete to ensure forward IO progress.
2517 Severity : enhancement
2519 Description: Support for specifying external journal device at mount
2520 Details : If an OST or MDS device is formatted with an external journal
2521 device, this device major/minor is stored in the ext3 superblock
2522 and may not be valid for failover. Allow detecting and
2523 specifying the external journal at mount time.
2528 Description: Mounting an MDS with pending unlinked files may cause oops
2529 Details : target_finish_recovery() calls mds_postrecov() which returned
2530 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2531 considers this an error and immediately begins cleaning up the
2532 lov, just after starting the mds_lov process
2534 Severity : enhancement
2536 Description: Implement 'lfs df' to report actual free space on per-OST basis
2537 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2538 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2539 usage of objects; '-h' to report in human readable format.
2541 ------------------------------------------------------------------------------
2543 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2550 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2551 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2552 at startup during orphan unlinking. Instead, skip buddy update
2553 before setup, as it will be regenerated after recovery is complete.
2558 Description: If an OST is inactive, its locks might reference stale inodes.
2559 Details : lov_change_cbdata() must iterate over all namespaces, even if
2560 they are inactive to clear inode references from the lock.
2562 Severity : enhancement
2563 Frequency : occasional, if non-standard max_dirty_mb used
2565 Description: Client will block write RPCs if not enough grant
2566 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2567 then the client will block writes while waiting for another RPC
2568 to complete instead of consuming its dirty limit. With change
2569 we get improved performance when max_dirty_mb is small.
2571 Severity : enhancement
2572 Bugzilla : 3389, 6253
2573 Description: Add support for supplementary groups on the MDS.
2574 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2575 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2576 lookups for user supplementary groups into a cache.
2580 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2581 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2582 for liblustre O_EXCL is enforced by MDS.
2586 Description: readdir never returns NULL in liblustre.
2587 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2588 dentry in struct dirent.
2592 Frequency : liblustre only, depends on application IO pattern
2593 Description: liblustre clients evicted if not contacting servers
2594 Details : Don't put liblustre clients into the ping_evictor list, so
2595 they will not be evicted by the pinger ever.
2597 Severity : enhancement
2599 Description: Add ability to evict clients by NID from MDS.
2600 Details : By echoing "nid:$NID" string into
2601 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2602 $NID would be instantly evicted from this MDS and from all active
2603 OSTs connected to it.
2607 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2608 Details : lookup_it_finish() used to query file size from OSTs that was not
2613 Description: service threads change working directory to that of init
2614 Details : Starting lustre service threads may pin the working directory
2615 of the parent thread, making that filesystem busy. Threads
2616 now change to the working directory of init to avoid this.
2620 Frequency : during shutdown only
2621 Description: shutdown with a failed MDS or OST can cause unmount to hang
2622 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2627 Frequency : 2.6 only
2628 Description: chmod/chown may include an extra supplementary group
2629 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2630 supplementary group and if none is specified this is used.
2633 Bugzilla : 5479 (6816)
2635 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2636 Details : If lookup is in progress on a file that is unlinked we might try
2637 to revalidate the inode and fail in revalidate after lookup is
2638 complete and ll_file_open() enqueues the open again but
2639 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2642 Frequency : always, if lconf --abort_recovery used
2644 Description: lconf --abort_recovery fails with 'Operation not supported'
2645 Details : lconf was attempting to abort recovery on the MDT device and not
2648 ------------------------------------------------------------------------------
2650 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2655 Frequency : rare (only unsupported configurations with a node running as an
2657 Bugzilla : 6514, 5137
2658 Description: Mounting a Lustre file system on a node running as an OST could
2660 Details : OSTs now preallocates memory needed to write out data at
2661 startup, instead of when needed, to avoid having to
2662 allocate memory in possibly low memory situations.
2663 Specifically, if the file system is mounted on on OST,
2664 memory pressure could force it to try to write out data,
2665 which it needed to allocate memory to do. Due to the low
2666 memory, it would be unable to do so and the node would
2667 become unresponsive.
2669 Severity : enhancement
2671 Description: Addition of lconf --service command line option
2672 Details : lconf now accepts a '--service <arg>' option, which is
2673 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2675 Severity : enhancement
2677 Description: Failover mode is now the default for OSTs.
2678 Details : By default, OSTs will now run in failover mode. To return to
2679 the old behaviour, add '--failout' to the lmc line for OSTs.
2681 Severity : enhancement
2683 Description: Health checks are now provided for MDS and OSTs
2684 Details : Additional detailed health check information on MSD and OSTs
2685 is now provided through the procfs health_check value.
2688 Frequency : occasional, depends on IO load
2690 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2691 after numerous create/delete cycles
2692 Details : The ext3 inode allocation policy would not allocate new inodes
2693 very well on the OSTs because there are no new directories
2694 being created. Instead we look for groups with free space if
2695 the parent directories are nearly full.
2700 Description: Network or server problems during mount may cause partially
2701 mounted clients instead of returning an error.
2702 Details : The config llog parsing code may overwrite the error return
2703 code during mount error handling, returning success instead
2709 Description: MDS can fail to allocate large reply buffers
2710 Details : After long uptimes the MDS can fail to allocate large reply
2711 buffers (e.g. zconf client mount config records) due to memory
2712 fragmentation or consumption by the buffer cache. Preallocate
2713 some large reply buffers so that these replies can be sent even
2714 under memory pressure.
2718 Frequency : rare (liblustre)
2719 Description: fsx running with liblustre complained that using truncate() to
2720 extend the file doesn't work. This patch corrects that issue.
2721 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2722 ATTR_SIZE and lsm use in llu_setattr_raw.
2726 Frequency : rare, only 2.6 kernels
2727 Description: Unusual file access patterns on the MDS may result in inode
2728 data being lost in very rare circumstances.
2729 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2730 code in the 2.6 kernel caused the inode and/or EA data not to
2731 be read from disk, causing single-file corruption.
2735 Frequency : rare, only 2.6 filesystems using extents
2736 Description: Heavy concurrent write and delete load may cause data corruption.
2737 Details : It was possible under high-load situations to have an extent
2738 metadata block in the block device cache from a just-unlinked
2739 file overwrite a newly-allocated data block. We now unmap any
2740 metadata buffers that alias just-allocated data blocks.
2744 Frequency : filesystems with default stripe_count larger than 77
2745 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2746 Details : lconf specifies an inode size of 4096 bytes when the default
2747 stripe_count is larger than 77. This conflicts with the default
2748 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2749 this case to avoid pinning too much memory for large EAs.
2751 ------------------------------------------------------------------------------
2753 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2758 Frequency : rare (extremely heavy IO load with hundreds of clients)
2760 Description: Client is evicted, gets IO error writing to file
2761 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2762 caused clients to be evicted for AST timeouts. The fixes in
2763 bug 5192 mean we no longer need to have such short AST timeouts
2764 so ldlm_timeout has been increased.
2767 Frequency : occasional during --force or --failover shutdown under load
2768 Bugzilla : 5949, 4834
2769 Description: Server oops/LBUG if stopped with --force or --failover under load
2770 Details : a collection of import/export refcount and cleanup ordering
2771 issues fixed for safer force cleanup
2774 Frequency : only filesystems larger than 120 OSTs
2775 Bugzilla : 5990, 6223
2776 Description: lfs getstripe would oops on a very large filesystem
2777 Details : lov_getconfig used kfree on vmalloc'd memory
2780 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2781 Bugzilla : 6242, 6243
2782 Description: reading from files that had been truncated to a non-zero size
2783 but never opened returned no data
2784 Details : ll_file_read() reads zeros from no-object files to EOF
2789 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2790 Details : There's little harm in aborting MDS/OSS recovery and letting it
2791 try again, so I removed the LASSERT and return an error instead.
2793 Severity : enhancement
2795 Description: New debugging infrastructure for tracking down data corruption
2796 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2797 (b) cover more of the client-side code path, and (c) try to narrow
2798 down where problems occurred
2802 Bugzilla : 3819, 4364, 4397, 6313
2803 Description: Racing close and eviction MDS could cause assertion in mds_close
2804 Details : It was possible to get multiple mfd references during close and
2805 client eviction, leading to one thread referencing a freed mfd.
2807 Severity: : enhancement
2808 Bugzilla : 3262, 6359
2809 Description: Attempts to reconnect to servers are now more aggressive.
2810 Details : This builds on the enhanced upcall-less recovery that was added
2811 in 1.4.2. When trying to reconnect to servers, clients will
2812 now try each server in the failover group every 10 seconds. By
2813 default, clients would previously try one server every 25 seconds.
2818 Description: After recovery, certain operations trigger a failed
2819 assertion on a client.
2820 Details : Failing over an mds, using lconf -d --failover, while a
2821 client was doing a readdir() call would cause the client to
2822 LBUG after recovery completed and the readdir() was resent.
2824 Severity : enhancement
2826 Description: Default groups are now added by lconf
2827 Details : You can now run lconf --group <servicename> without having to
2828 manually add groups with lmc.
2831 Frequency : occasional
2833 Description: Nodes with an elan id of 0 trigger a failed assertion
2836 Frequency : always when accessing e.g. tty/console device nodes
2838 Description: tty and some other devices nodes cannot be used on lustre
2839 Details : file's private_data field is used by device data and lustre
2840 values in there got lost. New field was added to struct file to
2841 store fs-specific private data.
2844 Frequency : when exporting Lustre via NFS
2846 Description: NFSD failed occasionally when looking up a path component
2847 Details : NFSD is looking up ".." which was broken in ext3 directories
2848 that had grown large enough to become hashed.
2851 Frequency : Clusters with multiple interfaces not on the same subnet
2853 Description: Nodes will repeatedly try to reconnect to an interface which it
2854 cannot reach and report an error to the log.
2855 Details : Extra peer list entries will be created by lconf with some peers
2856 unreachable. lconf now validates the peer before adding it.
2859 Frequency : Only if a default stripe is set on the filesystem root.
2861 Description: Setting a default stripe on the filesystem root prevented the
2862 filesystem from being remounted.
2863 Details : The client was sending extra request flags in the root getattr
2864 request and did not allocate a reply buffer for the dir EA.
2867 Frequency : occasional, higher if lots of files are accessed by one client
2868 Bugzilla : 6159, 6097
2869 Description: Client trips assertion regarding lsm mismatch/magic
2870 Details : While revalidating inodes the VFS looks up inodes with ifind()
2871 and in rare cases can find an inode that is being freed.
2872 The ll_test_inode() code will free the lsm during ifind()
2873 when it finds an existing inode and then the VFS later attaches
2874 this free lsm to a new inode.
2878 Bugzilla : 6422, 7030
2879 Description: MDS deadlock between mkdir and client eviction
2880 Details : Creating a new file via mkdir or mknod (starting a transaction
2881 and getting the ns lock) can deadlock with client eviction
2882 (gets ns lock and trying to finish a synchronous transaction).
2885 Frequency : occasional
2886 Description: While starting a server, the fsfilt_ext3 module could not be
2888 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2889 kernels. Previously, lconf would still use the ext3 name
2890 when trying to load modules. Now, it will correctly use
2891 ext3 on 2.4 and ldiskfs on 2.6.
2893 Severity : enhancement
2894 Description: The default stripe count has been changed to 1
2895 Details : The interpretation of the default stripe count (0, to lfs
2896 or lmc) has been changed to mean striping across a single
2897 OST, rather than all available. For general usage we have
2898 found a stripe count of 1 or 2 works best.
2900 Severity : enhancement
2901 Description: Add support for compiling against Cray portals.
2902 Details : Conditional compiling for some areas that are different
2906 Frequency : occasional
2907 Bugzilla : 6409, 6834
2908 Description: Creating files with an explicit stripe count may lead to
2909 a failed assertion on the MDS
2910 Details : If some OSTs are full or unavailable, creating files may
2911 trigger a failed assertion on the MDS. Now, Lustre will
2912 try to use other servers or return an error to the
2916 Frequency : occasional
2918 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2919 caused lock timeout during readahead (since 1.4.2).
2920 Details : Processes doing readahead might match a lock that hasn't been
2921 granted yet if there are overlapping and conflicting lock
2922 requests. The readahead process waits on ungranted lock
2923 (original lock is CBPENDING), while OST waits for that process
2924 to cancel CBPENDING read lock and eventually evicts client.
2926 Severity : enhancement
2928 Description: Initial enabling of flock support for clients
2929 Details : Implements fcntl advisory locking and file status functions.
2930 This feature is provided as an optional mount flag (default
2931 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2932 locking are implemented yet, and those that are are not guaranteed
2933 to be completely correct in production environments.
2934 mount -t lustre -o [flock|noflock] ...
2937 Frequency : occasional
2939 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2940 assertion in the ext3 JBD (journaling) layer.
2941 Details : The b_committed_data struct is protected by the big kernel lock
2942 in 2.4 kernels, serializing journal_commit_transaction() and
2943 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2944 access to this struct. In 2.6 kernels there is finer grained
2945 locking to improve SMP performance of the JBD layer.
2949 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2954 Description: merge and cleanup kernel patches.
2957 -----------------------------------------------------------------------------
2959 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2961 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2962 versions of Lustre. Please update all servers and clients to
2963 version 1.4.2 or later at the same time. You must also run
2964 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2965 to update the configuration logs.
2967 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2968 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2969 - swab llog records of type '0' so we get proper header size/idx (5861)
2970 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2971 - fix rename of one directory over another leaking an inode (5953)
2972 - avoid SetPageDirty on 2.6 (5981)
2973 - don't re-add just-being-destroyed locks to the waiting list (5653)
2974 - when creating new directories, inherit the parent's custom
2975 striping settings if present parent (3048)
2976 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2977 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2978 - don't hold client locks on temporary worklist from l_lru (5666)
2979 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2980 - automatically evict dead clients (5921)
2981 - Update file size properly in create+truncate+fstat case (6196)
2982 - Do not unhash mountpoint dentries, do not allow removal of
2984 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2985 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2986 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2987 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2988 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2989 - Correctly handle reads of files with no objects (6243)
2990 - lctl recover will also mark a device active if deactivate used (5933)
2992 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2993 - allow --write-conf on an MDS with different nettype than client (5619)
2994 - don't write config llogs to MDS for mounts not from that MDS (5617)
2995 - lconf should create multiple TCP connections from a client (5201)
2996 - init scripts are now turned off by default; run chkconfig --on
2997 lustre and chkconfig --on lustrefs to use them
2998 - upcalls are no longer needed for clients to recover to failover
3000 - add --abort-recovery option to lconf to abort recovery on device
3002 - add support for an arbitrary number of OSTs (3026)
3003 - Quota support protocol changes.
3004 - forward compatibility changes to wire structs (6007)
3005 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3006 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3007 - improved "lustre status" script
3008 - initialize blocksize for non-regular files (6062)
3009 - added --disable-server and --disable-client configure options (5782)
3010 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3011 - Vanilla 2.4.29 support
3012 - increase maximum number of obd devices to 520 (6242)
3013 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3014 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3015 SLES 9 2.6.5 kernels
3016 - limit stripes per file to 160 (the maximum EA size) (6093)
3018 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3021 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3022 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3023 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3024 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3025 - small ext3 extents cleanups and fixes (5733)
3026 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3027 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3028 - avoid needless client->OST connect, fix handle mismatch (5317)
3029 - fix DLM error path that led to out-of-sync client, long delays (5779)
3030 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3031 - fix several locking issues related to i_size (5492,5624,5654,5672)
3032 - don't move pending lock onto export if it is already evicted (5683)
3033 - fix kernel oops when creating .foo in unlinked directory (5548)
3034 - fix deadlock in obdfilter statistics vs. object create (5811)
3035 - use time_{before,after} to avoid timer jiffies wrap (5882)
3036 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3037 - Do not leak request if server was not able to process it (5154)
3038 - If mds_open unable to find parent dir, make that negative lookup(5154)
3039 - don't create new directories with extent-mapping (5909, 5936)
3041 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3042 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3043 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3044 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3045 - support for RHEL4 kernel on i686 (5773)
3046 - provide error messages when incompatible logs are encountered (5898)
3048 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3049 * version 1.4.0.10 (1.4.1 release candidate 1)
3051 - don't keep a lock reference when lock is not granted (4238)
3052 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3053 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3054 - drop import inflight refcount on signal_completed_replay error (5255)
3055 - unlock page after async write error during send (3677)
3056 - handle missing objects in filter_preprw_read properly (5265)
3057 - no transno return for symlink open, don't save no-trasno open (3440)
3058 - don't try to complete elan receive that already failed (4012)
3059 - free RPC server reply state on error (5406)
3060 - clean up thread from ptlrpc_start_thread() on error (5160)
3061 - readahead could read extra page into cache that wasn't ejected (5388)
3062 - prevent races in class_attach/setup/cleanup/detach (5260)
3063 - don't dereference de->d_inode after l_dput of de (5458)
3064 - use "int" for stripe value returned from lock_to_stripe (5544)
3065 - mballoc allocation and error-checking fixes in 2.6 (5504)
3066 - block device patches to fix I/O request sizes in 2.6 (5482)
3067 - look up hostnames for IB nals (5602)
3068 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3069 - don't start multiple acceptors for the same port (5277)
3070 - fix incorrect LASSERT in mds_getattr_name (5635)
3071 - export a proc file for general "ping" checking (5628)
3072 - fix "lfs check" to not block when the MDS is down (5628)
3074 - service request history (4965)
3075 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3076 - create an "evict_client" /proc entry on OSTs, like the MDS has
3077 - fix mount usage message, return errors per mount(8) (5168)
3078 - change grep [] to grep "[]" in tests so they work in more UMLs
3079 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3080 - remove ancient LOV_MAGIC_V0 EA support (5047)
3081 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3082 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3083 - allow readahead allocations to fail when low on memory (5383)
3084 - mmap locking landed again, after considerable improvement (2828)
3085 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3087 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3090 - send OST transaction number in read/write reply to free req (4966)
3091 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3092 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3093 - the watchdog thread now runs as interruptible (5246)
3094 - flock/lockf fixes (but it's still disabled, pending 5135)
3095 - don't use EXT3 constants in llite code (5094)
3096 - memory shortage at startup could cause assertion (5176)
3098 - reorganization of lov code
3099 - single portals codebase
3101 - add extents/mballoc support (5025)
3102 - direct I/O reads in the obdfilter (4048)
3103 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3105 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3108 - send OST transaction number in read/write reply to free req (4966)
3109 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3110 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3111 - the watchdog thread now runs as interruptible (5246)
3112 - handle missing objects in filter_preprw_read properly (5265)
3113 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3114 - drop import inflight refcount on signal_completed_replay error (5255)
3115 - unlock page after async write error during send (3677)
3116 - return original error code on reconstructed replies (3761)
3117 - no transno return for symlink open, don't save no-trasno open (3440)
3119 - add pid to ldlm debugging output (4922)
3120 - bump the watchdog timeouts -- we can't handle 30sec yet
3121 - extra debugging for orphan dentry/inode bug (5259)
3123 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3126 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3127 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3128 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3129 - fix NULL dereference in /proc/sys/portals/routes (4827)
3130 - allow failed mdc_close() operations to be interrupted (4561)
3131 - stop precreate on OST before MDS would time out on it (4778)
3132 - don't send partial-page writes before EOF from client (4410)
3133 - discard client grant for sub-page writes on large-page clients (4520)
3134 - don't free dentries not owned by NFS code, check generation (4806)
3135 - fix lsm leak if mds_create_objects() fails (4801)
3136 - limit debug_daemon file size, always print CERROR messages (4789)
3137 - use transno after validating reply (3892)
3138 - process timed out requests if import state changes (3754)
3139 - update mtime on OST during writes, return in glimpse (4829)
3140 - add mkfsoptions to LDAP (4679)
3141 - use ->max_readahead method instead of zapping global ra (5039)
3142 - don't interrupt __l_wait_event() during strace
3144 - add software watchdogs to catch hung threads quickly (4941)
3145 - make lustrefs init script start after nfs is mounted
3146 - fix CWARN/ERROR duplication (4930)
3147 - return async write errors to application if possible (2248)
3148 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3149 - print NAL number in %x format (4645)
3150 - update barely-supported suse-2.4.21-171 series (4842)
3151 - support for sles 9 %post scripts
3152 - support for building 2.6 kernel-source packages
3153 - support for sles km_* packages
3155 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3158 - ignore -ENOENT errors in osc_destroy (3639)
3159 - notify osc create thread that OSC is being cleaned up (4600)
3160 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3161 - reconstruct ost_handle() like mds_handle() (4657)
3162 - create a new thread to do import eviction to avoid deadlock (3969)
3163 - let lconf resolve symlinked-to devices (4629)
3164 - don't unlink "objects" from directory with default EA (4554)
3165 - hold socknal file ref over connect in case target is down (4394)
3166 - allow more than 32000 subdirectories in a single directory (3244)
3167 - fix blocks count for O_DIRECT writes (3751)
3168 - OST returns ENOSPC from object create when no space left (4539)
3169 - don't send truncate RPC if file size isn't changing (4410)
3170 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3171 - bind to privileged port in socknal and tcpnal (3689)
3173 - rate limit CERROR/CWARN console message to avoid overload (4519)
3174 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3175 - basic mmap support (3918)
3176 - kernel patch series update from b1_4 (4711)
3178 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3181 - avoid crash during MDS cleanup with OST shut down (2775)
3182 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3183 - don't use bad inodes on the MDS (3744)
3184 - dynamic object preallocation to improve recovery speed (4236)
3185 - don't hold spinlock over lock dumping or change debug flags (4401)
3186 - don't zero obd_dev when it is force cleaned (3651)
3187 - print grants to console if they go negative (4431)
3188 - "lctl deactivate" will stop automatic recovery attempts (3406)
3189 - look for existing locks in ldlm_handle_enqueue() (3764)
3190 - don't resolve lock handle twice in recovery avoiding race (4401)
3191 - revalidate should check working dir is a directory (4134)
3193 - don't always mark "slow" obdfilter messages as errors (4418)
3195 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3198 - don't close LustreDB during write_conf until it is done (3860)
3199 - fix typo in lconf for_each_profile (3821)
3200 - allow dumping logs from multiple threads at one time (3820)
3201 - don't allow multiple threads in OSC recovery (3812)
3202 - fix debug_size parameters (3864)
3203 - fix mds_postrecov to initialize import for llog ctxt (3121)
3204 - replace config semaphore with spinlock (3306)
3205 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3206 - don't allow enqueue to complete on a destroyed export (3822)
3207 - down write_lock before checking llog header bitmap (3825)
3208 - recover from lock replay timeout (3764)
3209 - up llog sem before sending rpc (3652)
3210 - reduce ns lock hold times when setting kms (3267)
3211 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3212 - fix NULL deref and obd_dev leak on setup error (3312)
3213 - replace some LBUG about llog ops with error handling (3841)
3214 - don't match INVALID dentries from d_lookup and spin (3784)
3215 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3216 - fix invalid assertion in ptlrpc_set_wait (3880)
3218 - add libwrap support for the TCP acceptor (3996)
3219 - add /proc/sys/portals/routes for non-root route listing (3994)
3220 - allow setting MDS UUID in .xml (2580)
3221 - print the stack of a process that LBUGs (4228)
3223 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3226 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3227 - make sure to unset replay flag from failed open requests (3440)
3228 - if default stripe count is 0, use OST count for inode size (3636)
3229 - update parent mtime/ctime on client for create/unlink (2611)
3230 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3231 - free recovery state on server during a forced cleanup (3571)
3232 - unregister_reply for resent reqs (3063)
3233 - loop back devices mounting and status check on 2.6 (3563)
3234 - fix resource-creation race that can provoke i_size == 0 (3513)
3235 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3236 - more debugging for page-accounting assertion (3746)
3237 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3238 - avoid deadlock after precreation failure (3758)
3239 - fix race and lock order deadlock in orphan handling (3450, 3750)
3240 - add validity checks when grabbing inodes from l_ast_data (3599)
3242 - add /proc/.../recovery_status to obdfilter (3428)
3243 - lightweight CDEBUG infrastructure, debug daemon (3668)
3244 - change default OSC RPC parameters to be better on small clusters
3245 - turn off OST read cache for files smaller than 32MB
3246 - install man pages and include them in rpms (3100)
3247 - add new init script for (un)mounting lustre filesystems (2593)
3248 - run chkconfig in %post for init scripts (3701)
3249 - drop scimac NAL (unmaintained)
3251 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3254 - clean kiobufs before and after use (3485)
3255 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3256 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3257 - change init script to fail more clearly if not run as root (1528)
3258 - allow clients to reconnect during replay (1742)
3259 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3260 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3261 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3262 - fix cond_resched (3554)
3263 - only evict unfinished clients after recovery (3515)
3264 - allow bulk resend, prevent data loss (3570)
3265 - dynamic ptlrpc request buffer allocation (2102)
3266 - don't allow unlinking open directory if it isn't empty (2904)
3267 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3268 - remove extraneous obd dereference causing LASSERT failure (3334)
3269 - don't use get_cycles() when creating temp. files on the mds (3156)
3270 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3271 - handle EEXIST for set-stripe, set proper directory name (3336)
3273 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3274 - fix ksocknal_fmb_callback() error messages (2918)
3276 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3279 - don't copy lvb into (possibly NULL) reply on error (2983)
3280 - don't deref dentry after dput, don't free lvb on error (2922)
3281 - use the kms to determine writeback rpc length (2947)
3282 - increment oti_logcookies when osc is inactive (2948)
3283 - update client's i_blocks count via lvb messages (2543)
3284 - handle intent open/close of special files properly (1557)
3285 - mount MDS with errors=remount-ro, like obdfilter (2009)
3286 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3287 - don't use cancelling-locks' kms values (2947)
3288 - use highest lock extent for kms, not last one (2925)
3289 - don't dereference ERR_PTR() dentry in error handling path (3107)
3290 - fix thread race in portals_debug_dumplog() (3122)
3291 - create lprocfs device entries at setup instead of at attach (1519)
3292 - common AST error handler, don't evict client on completion race (3145)
3293 - zero nameidata in detach_mnt in 2.6 (3118)
3294 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3295 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3296 - initialize RPC timeout timer earlier for 2.6 (3219)
3297 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3298 - print nal/nid for unknown nid (3258)
3299 - additional checks for oscc recovery before doing precreate (3284)
3300 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3301 - don't crash in mdc_close for bad permissions on open (3285)
3302 - zero i_rdev for non-device files (3147)
3303 - clear page->private before handing to FS, better assertion (3119)
3304 - tune the read pipeline (3236)
3305 - fix incorrect decref of invalidated dentry (2350)
3306 - provide read-ahead stats and refine rpc in flight stats (3328)
3307 - don't hold journal transaction open across create RPC (3313)
3308 - update atime on MDS at close time (3265)
3309 - close LDAP connection when recovering to avoid server load (3315)
3310 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3311 - don't leak open file on MDS after open resend (3325)
3312 - serialize filter_precreate and filter_destroy_precreated (3329)
3313 - loop device shouldn't call sync_dev() for nul device (3092)
3314 - clear page cache after eviction (2766)
3315 - resynchronize MDS->OST in background (2824)
3316 - refuse to mount the same filesystem twice on same mountpoint (3394)
3317 - allow llmount to create routes for mounting behind routers (3320)
3318 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3319 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3320 - fix rare problem with rename on htree directories (3417)
3322 - allow default OST striping configuration per directory (1414)
3323 - fix compilation for qswnal for 2.6 kernels (3125)
3324 - increase maximum number of MDS request buffers for large systems
3325 - change liblustreapi to be useful for external progs like lfsck (3098)
3326 - increase local configuration timeout for slow disks (3353)
3327 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3329 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3332 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3333 - don't overwrite extent policy data in reply if lock was blocked (2901)
3334 - drop filter export grants atomically with removal from device (2663)
3335 - del obd_self_export from work_list in class_disconnect_exports (2908)
3336 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3337 - swab reply message in mdc_close, other PPC fixes (2464)
3338 - fix destroying of named logs (2325)
3339 - overwrite old logs when running lconf --write_conf (2264)
3340 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3341 - fix race in target_handle_connect (2898)
3342 - mds_reint_create() should take same inode create lock (2926)
3343 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3344 - don't close files for self_export to avoid uninitialized obd (2936)
3345 - allow MDS with the same name as client node (2939)
3346 - hold dentry reference for closed log files for unlink (2325)
3347 - reserve space for all logs during transactions (2059)
3348 - don't evict page beyond end of stripe extent (2925)
3349 - don't oops on a deleted current working directory (2399)
3350 - handle hard links to targets without a parent properly (2517)
3351 - don't dereference NULL lock when racing during eviction (2867)
3352 - don't grow lock extents when lots of conflicting locks (2919)
3354 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3357 - account for cache space usage on clients to avoid data loss (974)
3358 - lfsck support in lustre kernel code (2349)
3359 - reduce journal credits needed for BRW writes (2370)
3360 - orphan handling to avoid losing space on client/server crashes
3361 - ptlrpcd can be blocked, stopping ALL progress (2477)
3362 - use lock value blocks to assist in proper KMS, faster stat (1021)
3363 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3364 - recovery for initial connections (2355)
3365 - fixes for mds_cleanup_orphans (1934)
3366 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3367 - block all file creations until orphan recovery completes (1901)
3368 - client remove rq_connection from request struct (2423)
3369 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3370 - recovery timer race (2670)
3371 - mdc_close recovey bug (2532)
3372 - ptlrpc cleanup bug (2710)
3373 - mds timeout on local locks (2588)
3374 - namespace lock held during RPCs (2431)
3375 - handle interrupted sync write properly (2503)
3376 - don't try to handle a message that hasn't been replied to (2699)
3377 - client assert failure during cleanup after abort recovery (2701)
3378 - leak mdc device after failed mount (2712)
3379 - ptlrpc_check_set allows timedout requests to complete (2714)
3380 - wait for inflight reqs when ptlrpcd finishes (2710)
3381 - make sure unregistered services are removed from the srv_list
3382 - reset bulk XID's when resending them (caught by 1138 test)
3383 - unregister_bulk after timeout
3384 - fix lconf error (2694)
3385 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3386 - readahead locks pages, leaves pending causing memory pressure (2673)
3387 - increase OST request buffers to 4096 on large machines (2729)
3388 - fix up permission of existing directories in simple_mkdir (2661)
3389 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3390 - don't assign transno to errored transactions (2742)
3391 - don't delete objects on OST if given a bogus objid from MDS (2751)
3392 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3393 - if rq_no_resend, then timeout request after recovery (2432)
3394 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3395 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3396 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3397 - MDS recovery shouldn't delete objects that it has given out (2730)
3398 - if enqueue arrives after completion, don't clobber LVB (2819)
3399 - don't unlock pages twice when trigger_group_io returns error (2814)
3400 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3401 - don't write pages to disk if there was an error (1450)
3402 - don't ping imports that have recovery disabled (2676)
3403 - take buffered bytes into account when balancing socknal conn (2817)
3404 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3405 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3406 - remove little-endian swabbing of llog records (1987)
3407 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3408 - retry reposting request buffers if they fail (1191)
3409 - grow extent at grant time to avoid granting a revoked lock (2809)
3410 - lock revoke doesn't evict page if covered by a second lock (2765)
3411 - disable VM readahead to avoid reading outside lock extents (2805)
3413 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3414 - updated kernel patches for hp-2.4.20 kernel (2681)
3416 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3419 - fix truncated write corruption (2366)
3420 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3422 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3423 - ensure len(uuid) < 37 in lmc (1171)
3424 - fix ia64 OOPS in llog_test (2255)
3425 - zero end of page at obdfilter for partial page writes (2648)
3426 - don't leave stale dentries around after renames (bug 2428)
3427 - fix timeouts when evicting a client with a single lock held (2642)
3428 - set deadline for the initial HELLO message to drain (2634)
3429 - print out dotted-quad IP addresses in the socknal (2302)
3431 - additional debugging for MDS client eviction problem (2443)
3432 - fix mkfsoptions support for osts (2603, 2604)
3434 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3437 - add series for the vanilla 2.6.0 kernel
3438 - add series for the vanilla 2.4.24 kernel
3439 - add series for a cray x86/64 UL kernel drop
3440 - fix xattr patches for the vanilla 2.4.19 series
3442 - generate true UUIDs in lmc (1171)
3443 - have portals stack dumping break in UML (2466)
3444 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3445 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3446 - wake callback waiting threads on client eviction (2460)
3447 - Add --ptldebug and --subsystem to lmc (1719)
3448 - update assertion to allow safe interrupt allocation
3449 - set rq_no_resend for cancel requests (2432)
3450 - recalculate ptlrpcd timeout after resend (2494)
3451 - call vfs_rmdir when removing pending directories (2368)
3452 - fix renaming a file to itself (2429)
3453 - lmc creates a default one-stripe lov (2454)
3454 - expand procfs space to handle large clusters (2326)
3455 - increase UML stack to avoid overflow
3456 - update lconf's list of debug and subsystem masks
3457 - fix lfs find --obd (2510)
3458 - /proc tunable for disabling filter read caching (2591)
3459 - stop rpm packages from altering slapd.conf (2301)
3460 - disable nagle in the socknal under 0conf (2578)
3461 - choose mds inode size based on stripe count (2572)
3462 - fix kernel-source rpm problems (2516)
3464 - add --disable-doc to avoid pdf generation (2421)
3465 - update documentation, tests, type-os, comments
3466 - avoid format warnings on ia64
3467 - remove the TOE NAL
3468 - tiny code cleanups by removing unused fields
3470 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3473 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3474 - remove the most blatant lies from BUILDING file (2371)
3475 - change default debug level to reasonable production setting
3476 - reduce client side cache size to reduce cache flush time
3477 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3478 - make TCP zerocopy and pinger support enabled by default (2476)
3479 - sync writes completed after process exits caused crashes (2319)
3480 - maintain correct mount count on the MDS (2356)
3481 - backout 1557, because 2316 wasn't really fixed
3482 - better file I/O statistics gathering in /proc
3483 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3484 - another kernel patch to fix zero-copy TCP function export
3485 - don't take duplicate lock when processing re-sent getattr (2420)
3486 - lctl uses obd_self_export instead of creating new conn (2353)
3487 - MDS/OST recovery case which requires object creation asserted (2425)
3488 - move lfs from /usr/sbin to /usr/bin in packages
3489 - fix race between mds_client_add and mds_client_free (2417)
3490 - use kmalloc instead of slabs in portals (2430)
3491 - don't create duplicate records when a failover MDS is present (2442)
3492 - remove unnecessary mount age check (2332)
3493 - don't remove directory inodes from locks prematurely (2451)
3494 - don't break if MDS service name is the same as hostname (2103)
3495 - fix races in client write RPC generation when cache full (2482)
3497 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3500 - remove now-unused request->rq_obd (278)
3501 - if an allocation fails, print out how much memory we've used (1933)
3502 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3503 - add the "configurable stack size" patch to most series files (1256)
3504 - ability to write large log records, for 100+ OST configs (2306)
3505 - fix NULL deref when filter_prep fails (2314)
3506 - fix operator precedence error in filter_sync
3507 - dynamic allocation of socknal TX descriptors (2315)
3508 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3509 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3510 - ensure that CWARN messages in Portals always get to the syslog
3511 - __init/__exit are not for prototype decls (ldlm_init/exit)
3512 - x86-64 compile warning fixes
3513 - fix gateway LMC keyword conflict (2318)
3514 - fix MDS lock inversions in getattr/reint paths (1844)
3515 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3516 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3517 - DLM race condition prevented some lock evictions (2328)
3518 - ENOMEM detection and retry on socknal sends (2230)
3519 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3520 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3521 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3522 - better error messages when a client is rejected during recovery (1505)
3523 - avoid cancelling locks which were never granted, after failure (2330)
3524 - fix i_sem/journal inversion in mds_client_add (2333)
3525 - fix truncate/getattr lock cycle deadlock (2334)
3526 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3527 - fix two rare exit paths which could leak an l_lock() ref (2321)
3528 - fencepost error in MDS/OST orphan recovery (2226)
3529 - make log record alignment 8 bytes (1988)
3530 - lstripe now fails when requested offset > ost_count (2237)
3531 - ensure that all kernel series have a complete list.h (1607)
3532 - fix crashes in special-file operations (2316)
3533 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3535 - allow configurable automake binary, for testing new versions
3536 - small update to the lfs documentation
3538 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3540 * fix negative export reference count in fsfilt_sync (2312)
3542 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3543 * release candidate 0.9.1
3545 - orphans are moved into the PENDING directory for possible recovery
3546 - replayed opens now open by fid for orphan/rename safety (1042)
3547 - last close of an orphan inode generates a transno (683)
3548 - chdir() and mount() now pin the directory entry (1020)
3549 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3550 - discard very old requests without processing them (1502)
3551 - remove some common, well-understood CERRORs (1505)
3552 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3553 - clear "grant" flags in OST replies until OST grant code lands (1644)
3554 - fix read performance by not clobbering i_blksize on client (1598)
3555 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3556 - make LRU size a /proc tunable, clears locks when reduced (707)
3557 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3558 - server locks take references on exports now (1558)
3559 - build fixes for 2.4.20-rh trees (1663)
3560 - return an error from lov_create if all OSCs are inactive (1751)
3561 - fix import levels when a reconnect happens without a timeout (1597)
3562 - exit early from mds_open if we get a lookup error (1749)
3563 - partial page read at EOF wouldn't wait for disk before sending (1642)
3564 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3565 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3566 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3567 - write/truncate lock inversion (1639)
3568 - Don't auto-load obdclass, portals modules during cleanup (1495)
3569 - fix timestamps from jumping to "now" (1763)
3570 - extra journal assertions (1648)
3571 - add an extra multiunlink test (1771)
3572 - fix read_record/write_record API (1776)
3573 - fix leak of offset_extent, possible incorrect i_size later (1772)
3574 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3575 - Debugging for the kqswnal_get_idle_tx problems (1820)
3576 - Allow recovery to be attempted multiple times (1536)
3577 - Write out MDS last_rcvd file after it is first created (1600)
3578 - Fix tx_descriptor leak in failed transmit situations (1827)
3579 - ext3 journaling fixes for assertion failure after IO error (1871)
3580 - class_export_put() on freed export after completion AST error (1896)
3581 - Fix revalidate looping in VFS (1322)
3582 - Don't access a freed export during MDS_REINT timeout (1521)
3583 - Add open-unlink recovery support on the MDS (1673,1764)
3584 - Return an error if no MDS data was read from last_rcvd (1946)
3585 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3586 - Don't setup a disk device that doesn't match exported UUID (317)
3587 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3588 - avoid committing NULL handle in force close
3589 - local.sh is now a one-stripe LOV configuration
3590 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3591 - fix ext3 htree duplicate directory entry corruption (1516)
3592 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3593 - update to new LOV EA format (2097)
3594 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3596 2003-06-15 Phil Schwan <phil@clusterfs.com>
3599 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3600 - per-import/export recovery handling (958, 931, 959)
3601 - multiple last-rcvd slots, for serving multiple FSes (949)
3602 - connections are again shared between multiple imp/exports (963, 964)
3603 - "umount -f" would hang if any requests needed to be sent (393, 978)
3604 - avoid pinning large req buffer by copying for queued messages (989)
3605 - add "uuid" to "lctl device" command to help upcalls (991)
3606 - "open" RPCs with transnos would confuse recovery counters (1037)
3607 - do proper endian conversion of all wire messages (288, 340, 891)
3608 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3609 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3610 - fix ptlrpc_connection leak in target_handle_connect (1174)
3611 - fix import refcounting bug in OST and MDS cleanup (1134)
3612 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3613 - fix very unlikely obd_types race condition (501)
3614 - remove osc_open hack for echo_client (1187)
3615 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3616 - a failure in read_inode2 leads to deadlock (1139)
3617 - cancel ack-locks as soon as transaction is committed (1072)
3618 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3619 - make sure to commitrw after any preprw to avoid deadlock (1162)
3620 - failing to execute a file in a lustre FS would lock inode (1203)
3621 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3622 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3623 - obd_uuid2tgt would walk off the end of the list (1255)
3624 - on IA64 the osc would give portals incorrect bulk size (1258)
3625 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3626 - fix lock inversion caused by new llite matching code (1282)
3627 - limit the number of dirty pages on a client to 10MB (1286)
3628 - timed out locks were not being corrected cancelled (1289)
3629 - fix O_DIRECT above 4GB on IA-32 (1292)
3630 * major user-visible changes
3631 - fail out/fail over policy now controlled by the upcall (993)
3633 - add OBD_PING to check server availability and failure (954)
3634 - lustre messages are now sent in sending host order (288, 340, 891)
3635 - add eadatalen to MDS getattr reply (340)
3636 - OST read replies may contain second buffer, with per-page status (593)
3638 2003-03-11 Phil Schwan <phil@clusterfs.com>
3641 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3642 - failed open()s could cause deadlock; fixed (867, 869)
3643 - stop cancelling OST locks when files are closed (481)
3644 - overlapping XID spaces caused network corruption (851, 853)
3645 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3646 - setattr_raw added, to do single-RPC, server-side setattrs
3647 - lmc/lconf syntax change for OST UUIDs
3648 - fix crashy race condition between ptlrpc_free_req and osc_close
3649 - don't use request in mdc_enqueue if we hit a timeout (889)
3650 - don't set the inode i_size for regular files from the MDS (896)
3651 - handle out of order completion AST (842)
3652 - don't LBUG if a lock request times out after receiving AST (913)
3653 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3654 - if a bad lock AST arrives, send an error instead of dropping entirely
3655 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3656 - fix leak in bulk IO when only partially completed (899, 900, 926)
3657 - fix O_DIRECT for ia64 (55)
3658 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3659 - C-z after timeout could hang a process forever; fixed (977)
3661 - client-side I/O cache (678, 924, 929, 941, 970)
3663 - READPAGE and SETATTRs which don't take server-side locks get
3666 2003-02-11 Phil Schwan <phil@clusterfs.com>
3669 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3670 - Fix fsfilt_extN_readpage() to read a full page of directory
3671 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3672 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3673 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3674 - Most intent operations were removed, in favour of a new RPC mode
3675 that does a single RPC to the server and bypasses most of the VFS
3676 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3677 - Aggressively cancel local locks on DLM servers
3678 - mds_reint_unlink sends EA to the client if it's the last nlink.
3679 client uses that EA to unlink OST objects.
3680 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3681 - recursive symlinks were fixed (439)
3682 - fixed NULL deref in DEBUG_REQ
3683 - filter_update_lastobjid no longer calls sync, which annoyed extN
3684 - fixed multi-client small-writes to a single file problem (445)
3685 - fixed mtime updates during file writes (607)
3686 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3687 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3688 - fixed Linux OST size reporting problem (444, 656)
3689 - OST now updates object mtime with writes or setattr (607, 619)
3690 - client verifies file size before zeroing page past EOF (445)
3691 - OST now writes last allocated objid to disk with allocation (108)
3692 - LOV on echo now works (409)
3694 - mds_reint_unlink sends a new buffer, with the EA included. this
3695 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3696 set if a regular file was being unlinked, and it was the last link
3697 - use PtlGet from the target for bulk writes (315)
3698 - OST now updates object mtime with writes or setattr (607, 619)
3699 - LDLM now has a grant-time callback to revalidate locked items, if
3701 - Many MDS operations were reorganized to combat race conditions
3703 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3704 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3706 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3709 - Fully reactivate OST imports after reconnection (512, others)
3710 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3711 - More graceful error handling for truncating on dead OST (515)
3712 - Don't error out unless we're actually accessing dead stripes (474)
3713 - Fix garbage sizes when stripes are missing (410)
3714 - LRU counters were broken, causing constant lock purge (433, 432)
3715 - garbage on read from stripes with failed OSTs (441)
3716 - mark OSCs as active before reconnecting during recovery (438)
3717 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3718 - lfind did not preserve OST order in output (443)
3719 - symlinks cause hung clients, incorrect data (439)
3720 - stop dereferencing request after dropping refcount (457)
3721 - don't LASSERT(spin_is_locked) on non-SMP (455)
3722 - fixes for many rename() bugs
3723 - fstat didn't correctly synchronize attributes (399)
3724 - server must handle lock cancellation during blocking AST prep (487)
3725 - bulk descriptors were free()d too soon (511)
3726 - fix paths in lconf, which would load incorrect modules (451, 507)
3727 - fix confusing lconf 'host not found' error message (386)
3728 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3729 - fix race condition in mdc_blocking_ast() for inode access (526)
3730 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3731 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3732 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3733 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3734 - fix lstripe in several ways (kernel side) (527)
3735 - fix request leak in ldlm_cli_enqueue (262)
3736 - incorrect OSC was marked inactive after OST failure
3737 - call mds_fs_cleanup before unmounting filesystem (524)
3738 - fix races between taking ns_lock and ldlm_lock_change_resource
3739 - fix races updating LOV export open file list
3740 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3741 - fix recovery NULL deref in ldlm_cli_cancel_unused
3742 - fix some DLM races by using new hash table for lock handles (419)
3743 - permit the client to specify desired inodes, at replay
3744 - duplicate requests when we queue them for replay reintegration
3745 - fix last_rcvd offset calculation
3746 - sync after each recovered transaction, so we always make progress
3747 - never, not always, ERESTART requests without transnos
3748 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3750 - skip replay if the MDS says that the client is already connected
3751 - don't check for a recovery-enabled export to match lctl's UUID
3752 - don't INC_USE_COUNT for phantom exports
3753 - don't crash when cleaning up phantom exports (567)
3754 - don't double-finish or set replay data for errored mdc_open requests
3755 - abort requests when they time out, so we don't get old replies
3756 - send/receive replies for AST messages again
3757 - if the client says that it doesn't have the lock, cancel it on the
3759 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3760 mark it as destroyed, it will all work out when decref is called
3761 - fix module use counts (22, 581)
3763 - ASTs now expect a reply (server cancels lock on error reply)
3765 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3768 - fix many simultaneous client startup (392)
3769 - fix dentry->d_it clobbering
3770 - credentials weren't being shipped for readdir/getattr operations
3771 - remove invalid assertions triggered during some concurrent MD
3773 - proper Lustre versions added (336, 389)
3774 - fix memory leak for create error case (398)
3775 - fix LOV locking bug that would get cli/srv out of sync
3776 - fix echo client over LOV (409)
3777 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3778 - fix double-O_EXCL intent crash (424)
3779 - avoid sending multiple lock CANCELs (352)
3781 - MDS can do multi-client recovery (modulo bugs in new code)
3783 - many updates, edits, cleanups
3785 2002-11-18 Phil Schwan <phil@clusterfs.com>
3788 - fix null d_it dereference (346)
3789 - fix full OST/dbench hang (333)
3790 - fix permission problem with file removal (286)
3791 - fix removal of OSCs from LOV when they fail
3792 - fix NULL deref during bulk timeout (214)
3793 - fix problems related to multiple filesystems on one MDS (241)
3794 - fixed serious subtle metadata locking bugs
3795 - free locks on clients when inodes are removed due to memory
3797 - fix inode pointer in lock data (285)
3798 - partial support for multiple MDS on a single host (241)
3799 - data locks weren't cancelled at clear_inode time (290, 311)
3800 - intent locks could lead to unbounded lock growth (205)
3801 - added a maximum lock count, an LRU list, and a flusher
3802 - fix multiple rename (365)
3803 - properly abstracted the echo client
3804 - OSC locked 1 byte too many; fixed
3805 - rewrote brw callback code:
3806 - fixed recovery bugs related to LOVs (306)
3807 - fixed too-many-pages-in-one-write crash (191)
3808 - fixed (again) crash in sync_io_timeout (214)
3809 - probably fixed callback-related race (385)
3811 - Add capability to MDS protocol
3812 - LDLM cancellations and callbacks on different portals
3814 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3817 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3818 - defer open object destruction to close time (601981, 138)
3819 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3820 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3821 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3822 - don't call obd_setattr in ll_file_release for destroyed objects
3823 * protocol change to lustre_msg: move |version| and add |flags|
3824 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3825 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3826 * added replay of create, unlink, link and rename operations during
3827 MDS failover; recovery should be much more robust now
3828 * remove failed OSCs from LOVs (only lov_create uses this so far)
3829 * the lustre-HOWTO was brought (more) up to date (582544)
3831 2002-10-23 Phil Schwan <phil@clusterfs.com>
3834 - in-use dentries weren't being reused properly (617851)
3835 - prevent multiple LDLM setup (599178)
3836 - fix LOV size calculations for truncate (617853)
3837 - fix client handling of MDS intent errors (POSIX)
3838 - fix permission bug in lovstripe.c test (624321)
3839 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3840 - truncate past end of file could corrupt data
3841 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3842 - a race in recovery could return ETIMEDOUT to apps (623947)
3843 - building outside the source directory was fixed
3844 * the lustre-HOWTO was brought (more) up to date (582544)
3845 * major progress was made on recovery functionality
3847 2002-10-10 Phil Schwan <phil@clusterfs.com>
3850 - recovery deadlock fix
3851 - rm -rf causes LBUG fix (617817)
3852 - file open by multiple tasks fix (618962)
3853 - directory permissions bugs (602707 and 620007)
3854 - journal_stop fixed with locking (611313)
3855 - O_APPEND failures resolved (618273, perhaps 614459)
3856 - lconf PATH fix (619770)
3857 - IA64 build fix (621450)
3858 - RPC buffer sizes scale with amount of memory
3860 2002-10-01 Phil Schwan <phil@clusterfs.com>
3863 - locks would be cancelled without throwing away data pages,
3864 resulting in inconsistent data (605627)
3865 - inode attributes were not always being refreshed (605627, 612449)
3866 - lconf now continues to cleanup after lctl reports an error
3867 - MDS now enforces user permissions (602707)
3868 - lprocfs cleanup fixed, but not yet enabled (614157)
3869 - fixed infinite server hang, should a client not respond to an AST
3870 - avoid going into recovery if user calls readlink() with a buffer
3871 that's too small (613941)
3872 - AST RPCs no longer require replies (614867) -- this may be changed
3873 - don't crash server if client sends an IOV that's too big (611336)
3874 - fixed lock conversion deadlock (611892)
3875 - fixed the following of symlinks (614622)
3876 * recovery: the server can remove locks from a client that dies, other
3877 clients can make progress
3878 * more extN patch fixes
3879 * compile-time configurable ptlrpc buffer allocations
3881 - collaborative read cache document
3882 - Lustre Lite Performance CDR document-in-progress
3884 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3887 - fix typo in patch-2.4.18
3889 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3892 - clear ptlrpc request each time in handle_incoming_request()
3893 - unlink of files now destroys the object on the OST
3895 2002-09-19 Peter Braam <braam@clusterfs.com>
3897 * add hard link support
3898 * change obdfile creation method
3899 * kernel patch changed
3901 2002-09-19 Peter Braam <braam@clusterfs.com>
3904 - stack overflow bug in extN fixed
3906 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3908 * documentation updates
3909 - add man pages for config tools
3910 - update tests/README to describe testing with new config tools
3911 - finish metadata API descriptions
3912 * bug fixes and cleanups
3913 - statfs workaround for 16TB limit
3914 - LOV stripe allocation improved, can stripe on subset of OSTs
3915 - LOV file size/IO offset was wrong for files > 4GB in size
3916 - object EA data was being dropped, caused files to be unreadable
3917 - memory overflow with non-LOV OST caused memory corruption
3918 - fixed regression tests to work with new config tools, obdfilter
3919 - fixed bug when directory size became larger than 1 block
3920 - fixed bug (for single client case) when PWD was deleted
3921 - invalidate local directory pages when doing intent-based ops
3922 - avoid LDLM oops when lock callback contained bad data
3924 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3926 * documentation updates
3927 * bug fixes and cleanups
3928 - configuration tools
3931 - 64-bit compile warnings
3932 - 64-bit internal statfs data
3934 * test_brw on persistent OST devices
3936 * lprocfs (disabled)
3938 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3940 * documentation updates
3941 * bug fixes and cleanups
3942 * configuration tools
3944 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3947 * many small fixes to 0_5_4
3948 * io/network handling
3949 * thinkos in MDS operations
3951 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3954 * crucial basic fixes to 0.5.3
3955 * IOR, Iozone work over Elan
3958 2002-08-07 Phil Schwan <phil@clusterfs.com>
3959 * version 0_5_3, our first alpha
3960 * we use the new Portals iovs
3961 * documentation updates
3962 * bug fixes and cleanups
3963 * small changes in the DLM wire protocol
3965 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3966 * version 0_5_1 with some initial stability,
3967 * locking on MD and file I/O.
3968 * documentation updates
3969 * several bug fixes since 0.5.0
3970 * small changes in wire protocol
3972 2002-07-18 Phil Schwan <phil@clusterfs.com>
3974 * delivered as Lustre Light Alpha
3975 * fixed a crash after handling invalid MDS requests
3976 * fixed directory pages for architectures with non-4k pages sizes
3978 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3979 * release version v0_4_4
3980 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3982 * Fixes a number of interruption problems with OST operations.
3983 * Update documentation for portals header changes
3984 * Move all wire protocol structs/defines to lustre_idl.h
3985 * Fixes symlink length bug.
3986 * Add tcpdump to repository.
3988 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3989 * release version v0_4_3
3990 * Fixes statfs for inodes on extN.
3991 * Fixes bug in runtests which would delete /etc/hosts.
3992 * Use 64-bit object IDs wherever possible (not into VFS though)
3993 Remove ost_get_info, which is unused by lustre, and out of date.
3995 2002-07-03 Peter Braam <braam@clusterfs.com>
3996 * release version v0_4_2 Fixes a lookup error (type not passed)
3997 * move forward to head of Portals
3998 * move forward to latest Lustre kernel
4000 2002-06-25 Peter Braam <braam@clusterfs.com>
4001 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob