4 2.6.9-55.0.6.EL (RHEL 4),
5 2.6.16.46-0.14 (SLES 10),
6 2.6.18-8.1.14.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.
15 Severity : enhancement
17 Description: self-adjustable client's lru lists
18 Details : use adaptive algorithm for managing client cached locks lru
19 lists according to current server load, other client's work
20 pattern, memory activities, etc. Both, server and client
21 side namespaces provide number of proc tunables for controlling
24 Severity : enhancement
26 Description: light-weight GSS support
27 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
28 and reduce performance overhead.
30 Severity : enhancement
32 Description: Linux keyring support
33 Details : Support using service of Linux keyring for Lustre GSS internal
34 context refresh/cache mechanism.
38 Description: Fix errors in lfs documentation
39 Details : Fixes man pages
43 Description: don't use GFP_* in generic Lustre code.
44 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
48 Description: obdclass is limited by single OBD_ALLOC(idarray)
49 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
53 Description: warnings with build patchless client with vanila 2.6.19 and up
54 Details : change old ctl_table style and replace ctl_table/ctl_table_header
55 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
59 Frequency : Only for SLES
60 Description: sanity_quota fail test_1
61 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
62 file, which makes TSTID[2] inunique.
66 Frequency : Only for SLES9
67 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
68 Details : Update target/ChangeLog/which_patch .
72 Frequency : Only for test_10 in sanity-quota.sh
73 Description: a bug in quota test code
74 Details : checking a test flag in a wrong place
78 Frequency : Only for test_13 in sanity-quota.sh
79 Description: a bug in quota test script
80 Details : shouldn't check overfull allocated quota
84 Description: jbd statistics
85 Details : Port older jbd statistics patch for sles10
87 Severity : enhancement
89 Description: Add printing inode info into message about error in writepage.
94 Description: reply_lock_interpret crash due to race with it and lock cancel.
95 Details : Do not replay locks that are being cancelled. Do not reference
96 locks by their address during replay, just by their handle.
100 Frequency : When flocks are used.
101 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
102 Details : Flock locks might destroy just granted lock if it could be merged
103 with another existing flock, this is done in completion handler,
104 so teach ldlm_cli_enquque_fini that this is a valid case for
110 Description: Oops in read and write path when failing to allocate lock.
111 Details : Check if lock allocation failed and return error back.
115 Description: lstripe command fails for valid OST index
116 Details : The stripe offset is compared to lov->desc.ld_tgt_count
117 instead of lov->desc.ld_active_tgt_count.
121 Description: sanity.sh failed test 103
122 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
127 Description: df doesn't work properly if diskfs blocksize != 4K
128 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
132 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
133 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
137 Description: obd_health_check_timeout too short
138 Details : set obd_health_check_timeout as 1.5x of obd_timeout
142 Description: llapi_file_create() does not allow some changes
143 Details : add llapi_file_open() that allows specifying the mode and
144 open flags, and also returns an open file handle.
148 Description: merge and cleanup kernel patches.
149 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
153 Description: Add journal checksum support.(Kernel part)
154 Details : The journal checksum feature adds two new flags i.e
155 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
156 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
157 indicates that the commit block contains the checksum for
158 the blocks described by the descriptor blocks. Now commit
159 record can be sent to disk without waiting for descriptor
160 blocks to be written to disk. This behavior is controlled
161 using JBD2_FEATURE_ASYNC_COMMIT flag.
165 Description: OSS needs mutliple precreate threads
166 Details : Add ability to start more than one create thread per OSS.
170 Description: Sanity.sh test_65k failed: setstripe should have succeeded
171 Details : execute lfs setstripe on client
175 Description: mds_obd_create error creating tmp object
176 Details : When the user sets quota on root, llog will be affected and can't
177 create files and write files.
180 Frequency : Always on ia64 patchless client, and possibly others.
182 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
183 Details : This allows the patchless client to be loaded on architectures
188 Description: RedHat Update kernel for RHEL5
189 Details : Add the kernel config file for RHEL5.
193 Description: RedHat Update kernel for RHEL5
194 Details : Modify the kernel config file more closer RHEL5.
198 Description: Build failure against Centos5 (RHEL5)
199 Details : Define PAGE_SIZE when it isn't present.
203 Description: test_7 on sanity-quota.sh can't be used on separate machines
204 Details : add facet to handle it
208 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
209 Details : with NFS, the anon dentry's parent was set to itself in
210 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
211 corresponding dentry other than use rec->ur_name.
213 Severity : enhancement
215 Description: lfs setstripe enhancement
216 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
220 Description: testing performance impact of enabling checksumming
221 Details : enable checksum by default, allow --disable-checksum
222 configure option and "-o nochecksum" mount option
226 Description: lustre support for RHEL5
227 Details : Add support for RHEL5.
231 Description: Client eviction due to failover config
232 Details : after a connection loss, the lustre client should attempt to
233 reconnect to the last active server first before trying the
234 other potential connections.
238 Description: when mds and osts use different quota unit(32bit and 64bit),
239 quota will be released repeatly.
240 Details : void sending multiple quota reqs to mds, which will keep the
241 status between the reqs.
245 Description: osts not allocated evenly to files
246 Details : change the condition to increase offset_idx
249 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
251 Description: Data corruption for OSTs that are formatted larger than 2TB
253 Details : When generating the bio request for lustre file writes the
254 sector number would overflow a temporary variable before being
255 used for the IO. The data reads correctly from Lustre (which
256 will overflow in a similar manner) but other file data or
257 filesystem metadata may be corrupted in some cases.
261 Description: Tune the kernel for good SCSI performance.
262 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
263 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
268 Description: rewrite ext2-derived code in obdclass/uuid.c
269 Details : rewrite inherited code (uuid parsing code from ext2 utils)
270 from scratch preserving functionality.
274 Description: "lfs find -obd UUID" prints directories
275 Details : "lfs find -obd UUID" will return all directory names instead
276 of just file names. It is incorrect because the directories
277 do not reside on the OSTs.
281 Description: use i_size_read and i_size_write in 2.6 port
282 Details : replace inode->i_size access with i_size_read/write()
285 Frequency : when removing large files
287 Description: scheduling issue during removal of large Lustre files
288 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
289 It causes scheduling issues when removing large files (17TB in the
293 Frequency : only with liblustre clients on XT3
295 Description: evictions taking too long
296 Details : allow llrd to evict clients directly on OSTs
299 Frequency : only on ppc
301 Description: /proc/fs/lustre/devices broken on ppc
302 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
303 We should make sure the type of 'index' is loff_t and then cast
304 explicitly as needed below. Do not assign an explicitly cast
308 Frequency : only for rhel5
310 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
311 Details : Modify the target file & which_kernel.
313 Severity : enhancement
315 Description: omit set fsid for export NFS
316 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
317 as FS_REQUIRES_DEV for avoid problems with generate fsid.
321 Frequency : Always, for kernels after 2.6.16
322 Description: Fix warning idr_remove called for id=.. which is not
324 Details : Last kernels save old s_dev before kill super and not allow
325 to restore from callback - restore it before call kill_anon_super.
329 Description: buffer overruns could theoretically occur
330 Details : llapi_semantic_traverse() modifies the "path" argument by
331 appending values to the end of the origin string, and a
332 overrun may occur. Adding buffer overrun check in liblustreapi.
336 Description: client sometimes tried revalidate root dentry.
337 Details : not revalidate dentry if it root dentry.
341 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
342 Details : Modify vm-tunables-rhel4.patch.
346 Description: Kernel config for 2.6.18-vanilla.
347 Details : Modify targets/2.6-vanilla.target.in.
348 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
349 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
350 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
351 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
355 Description: improve handling recoverable errors
356 Details : if request processig with error which can be recoverable on server
357 request should be resend, otherwise page released from cache and
362 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
363 Details : Modify target file & which_patch.
364 A flaw was found in the IA32 system call emulation provided
365 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
366 value could be stored in the %RAX register, which could trigger an
367 out-of-bounds system call table access. An untrusted local user
368 could exploit this flaw to run code in the kernel
369 (ie a root privilege escalation). (CVE-2007-4573).
373 Description: Remove 2.4 kernel checks lustre 1.8
374 Details : Remove 2.5.0 check from quota_check.c, quota_ctl.c,
375 quota_interface.c in quota subsystem.
376 Remove 2.4 kernel check from obd_mount.c, linux/linux-sysctl.c in
378 Remove lock_24kernel from fsfilt_ext3.c fsfilt_reiserfs.c
379 lvfs_linux.c upcall_cache.c in lvfs subsystem.
383 Description: O_DIRECT bypasses client statistics.
384 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
385 read_ahead_stats were updated. copied the stats section from
386 osc_send_oap_rpc() into async_internal().
390 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
393 --------------------------------------------------------------------------------
395 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
397 * Support for kernels:
398 2.6.5-7.283 (SLES 9),
399 2.6.9-55.EL (RHEL 4),
400 2.6.16.46-0.14 (SLES 10),
401 2.6.18.8 vanilla (kernel.org)
402 * Client support for unpatched kernels:
403 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
404 2.6.16 - 2.6.22 vanilla (kernel.org)
405 * Due to recently discovered recovery problems, we do not recommend
406 using patchless RHEL 4 clients with this or any earlier release.
407 * Recommended e2fsprogs version: 1.39.cfs8
408 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
409 * Starting with this release, the ldiskfs backing filesystem required
410 by Lustre is now in its own package, lustre-ldiskfs. This package
411 should be installed. It is versioned separately from Lustre and
412 may be released separately in future.
416 Description: block reactivating mgc import until all deactivates complete
417 Details : Fix race when failing back MDT/MGS to itself (testing)
419 Severity : enhancement
421 Description: add optional extra BUILD_VERSION info
422 Details : add a new environment variable (namely LUSTRE_VERS) which allows
423 to override the lustre version.
426 Frequency : 2.6.18 servers only
428 Description: ll_kern_mount() doesn't release the module reference
429 Details : The ldiskfs module reference count never drops down to 0
430 because ll_kern_mount() doesn't release the module reference.
435 Description: server LBUG when using old ost_num_threads parameter
436 Details : Accept the old ost_num_threads parameter but warn that it
437 is deprecated, and fix an off-by-one error that caused an LBUG.
442 Description: Transient SCSI error results in persistent IO issue
443 Details : iobuf->dr_error is not reinitialized to 0 between two
447 Frequency : sometimes when underlying device returns I/O errors
449 Description: OSTs not going read-only during write failures
450 Details : OSTs are not remounted read-only when the journal commit threads
451 get I/O errors because fsfilt_ext3 calls journal_start/stop()
452 instead of the ext3 wrappers.
456 Description: poor connect scaling with increasing client count
457 Details : Don't run filter_grant_sanity_check for more than 100 exports
458 to improve scaling for large numbers of clients.
461 Frequency : SLES10 only
463 Description: sanity-quota.sh quotacheck failed: rc = -22
464 Details : Quotas cannot be enabled on SLES10.
467 Frequency : liblustre clients only
469 Description: getdirentries does not give error when run on compute nodes
470 Details : getdirentries does not fail when the size specified as an argument
471 is too small to contain at least one entry
473 Severity : enhancement
475 Description: Add LNET router traceability for debug purposes
476 Details : If a checksum failure occurs with a router as part of the
477 IO path, the NID of the last router that forwarded the bulk data
478 is printed so it can be identified.
483 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
484 Details : Due to a race condition, liblustre clients were occasionally
487 Severity : enhancement
489 Description: lfs setstripe use optional parameters instead of postional
492 Severity : enhancement
494 Description: Nanosecond timestamp support for ldiskfs
495 Details : The on-disk ldiskfs filesystem has added support for nanosecond
496 resolution timestamps. There is not yet support for this at
497 the Lustre filesystem level.
500 Frequency : during server recovery
502 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
503 Details : request with rq_no_resend flag not awake l_wait_event if they get a
507 Frequency : nfs export on patchless client
509 Description: connectathon hang when test nfs export over patchless client
510 Details : Disconnected dentry cannot be found with lookup, so we do not need
511 to unhash it or make it invalid
514 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
515 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
516 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
521 Description: open req refcounting wrong on reconnect
522 Details : If reconnect happened between getting open reply from server and
523 call to mdc_set_replay_data in ll_file_open, we will schedule
524 replay for unreferenced request that we are about to free.
525 Subsequent close will crash in variety of ways.
526 Check that request is still eligible for replay in
527 mdc_set_replay_data().
532 Description: disable writes to filesystem when reading health_check file
533 Details : the default for reading the health_check proc file has changed
534 to NOT do a journal transaction and write to disk, because this
535 can cause reads of the /proc file to hang and block HA state
536 checking on a healthy but otherwise heavily loaded system. It
537 is possible to return to the previous behaviour during configure
538 with --enable-health-write.
540 Severity : enhancement
542 Description: 64-bit inode version
543 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
544 to the inode. This will be required for version-based recovery.
549 Description: MDS fails to start if a duplicate client export is detected
550 Details : in some rare cases it was possible for a client to connect to
551 an MDS multiple times. Upon recovery the MDS would detect this
552 and fail during startup. Handle this more gracefully.
554 Severity : enhancement
556 Description: Add -o localflock option to simulate old noflock
558 Details : This will achieve local-only flock/fcntl locks
564 Description: log_commit_thread vs filter_destroy race leads to crash
565 Details : Take import reference before releasing llog record semaphore
570 Description: Wrong request locking in request set processing
571 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
572 imp_delayed_list, in this place should be used imp_lock.
575 Frequency : when reconnection
577 Description: Grant Leak when osc reconnect to OST
578 Details : When osc reconnect ost, OST(filter) should check whether it
579 should grant more space to client by comparing fed_grant and
580 cl_avail_grant, and return the granted space to client instead
581 of "new granted" space, because client will call osc_init_grant
582 to update the client grant space info.
585 Frequency : when client reconnect to OST
587 Description: Grant Leak when osc do resend and replay bulk write
588 Details : When osc reconnect to OST, OST(filter)should clear grant info of
589 bulk write request, because the grant info while be sync between
590 OSC and OST when reconnect, and we should ignore the grant info
591 these of resend/replay write req.
596 Description: Grant space more than avaiable left space sometimes.
597 Details : When then OST is about to be full, if two bulk writing from
598 different clients came to OST. Accord the avaliable space of the
599 OST, the first req should be permitted, and the second one
600 should be denied by ENOSPC. But if the seconde arrived before
601 the first one is commited. The OST might wrongly permit second
602 writing, which will cause grant space > avaiable space.
605 Frequency : when client is evicted
607 Description: Grant might be wrongly erased when osc is evicted by OST
608 Details : when the import is evicted by server, it will fork another
609 thread ptlrpc_invalidate_import_thread to invalidate the
610 import, where the grant will be set to 0. While the original
611 thread will update the grant it got when connecting. So if
612 the former happened latter, the grant will be wrongly errased
613 because of this race.
618 Description: Checking Stale with correct fid
619 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
620 to check whether it is stale, because sometimes, we want the
621 enqueue happened anyway, and op_data.fid2 will not be initialized.
623 Severity : enhancement
625 Description: update patchless client
626 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
629 Frequency : only with 2.4 kernel
631 Description: random memory corruption
632 Details : size of struct ll_inode_info is to big for union inode.u and this
633 can be cause of random memory corruption.
638 Description: Memory leak in recovery
639 Details : Lov_mds_md was not free in an error handler in mds_create_object.
640 It should also check obd_fail before fsfilt_start, otherwise if
641 fsfilt_start return -EROFS,(failover mds during mds recovery).
642 then the req will return with repmsg->transno = 0 and rc = EROFS.
643 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
644 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
645 be freed no matter whether fsfilt_commit success or not.
648 Frequency : only with huge count clients
650 Description: Prevents from taking the superblock lock in llap_from_page for
652 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
653 for llap_from_page call in ll_removepage prevents from taking the
654 superblock lock for a soon died page.
659 Description: Not check open intent error before release open handle
660 Details : in some rare cases, the open intent error is not checked before
661 release open handle, which may cause
662 ASSERTION(open_req->rq_transno != 0), because it tries to release
663 the failed open handle.
668 Description: Set cat log bitmap only after create log success.
669 Details : in some rare cases, the cat log bitmap is set too early. and it
670 should be set only after create log success.
674 Description: Accessing a block bevice can re-enable I/O when Lustre is
675 tearing down a device.
676 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
681 Description: service threads may hog cpus when there are a lot of requests
683 Details : Insert cond_resched to give other threads a chance to use some of
689 Description: the cat log was not initialized in recovery
690 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
691 unlink log on mds will not be initialized until mds post
692 recovery. And also in mds post recovery, the unlink log will
693 initialization will be done asynchronausly, so there will be race
694 between add unlink log and unlink log initialization.
698 Description: brw_stats were being printed incorrectly
699 Details : brw_stats were being printed as log2 but all of them were not
700 recorded as log2. Also remove some code duplication arising from
701 filter_tally_{read,write}.
705 Frequency : rare, only in recovery.
706 Description: ASSERTION(req->rq_type != LI_POISON) failed
707 Details : imp_lock should be held while iterating over imp_sending_list for
708 prevent destroy request after get timeout in ptlrpc_queue_wait.
712 Description: replay-single.sh test 52 fails
713 Details : A lock's skiplist need to be cleanup when it being unlinked
714 from its resource list.
718 Description: Short directio read returns full requested size rather than
720 Details : Direct I/O operations should return actual amount of bytes
721 transferred rather than requested size.
723 Severity : enhancement
725 Description: metadata RPC reduction (e.g. for rm performance)
726 Details : decrease the amount of synchronous RPC between clients and servers
727 by canceling conflicing lock before the operation on the client side
728 and packing thier handles into the main operation RPC to server.
730 Severity : enhancement
732 Description: Async OSC create to avoid the blocking unnecessarily.
733 Details : If a OST has no remain object, system will block on the creating
734 when need to create a new object on this OST. Now, ways use
735 pre-created objects when available, instead of blocking on an
736 empty osc while others are not empty. If we must block, we block
737 for the shortest possible period of time.
741 Description: improve handling recoverable errors
742 Details : if request processig with error which can be recoverable on server
743 request should be resend, otherwise page released from cache and
746 Severity : enhancement
748 Description: refine locking for avoid write wrong info into lov_objid file
749 Details : fix possible races with add new target and write/update data in
752 --------------------------------------------------------------------------------
754 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
759 Frequency : on some architectures
761 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
762 Details : Uninitialized flags sometimes cause configuration commands to
766 Frequency : patchless clients only
768 Description: missing __iget() symbol export
769 Details : The __iget() symbol export is missing. To avoid the need for
770 this on patchless clients the deathrow inode reaper is turned
771 off, and we depend on the VM to clean up old inodes. This
772 dependency was during via the fix for bug 12181.
774 --------------------------------------------------------------------------------
776 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
778 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
779 INTEROPERATE with older versions automatically. In many cases a
780 special upgrade step is needed. Please read the
781 user documentation before upgrading any part of a live system.
782 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
783 version will not interoperate with 1.6 betas before beta5 (1.5.95).
784 * WARNING: Lustre configuration and startup changes are required with
785 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
791 Severity : enhancement
793 Description: MountConf
794 Details : Lustre configuration is now managed via mkfs and mount
795 commands instead of lmc and lconf. New obd types (MGS, MGC)
796 are added for dynamic configuration management. See
797 https://mail.clusterfs.com/wikis/lustre/MountConf for
800 Severity : enhancement
802 Description: dynamic OST addition
803 Details : OSTs can now be added to a live filesystem
805 Severity : enhancement
807 Description: startup order invariance
808 Details : MDTs and OSTs can be started in any order. Clients only
809 require the MDT to complete startup.
811 Severity : enhancement
813 Description: parallel, asynchronous orphan cleanup
814 Details : orphan cleanup is now performed in separate threads for each
815 OST, allowing parallel non-blocking operation.
817 Severity : enhancement
819 Description: optimized stripe assignment
820 Details : stripe assignments are now made based on ost space available,
821 ost previous usage, and OSS previous usage, in order to try
822 to optimize storage space and networking resources.
824 Severity : enhancement
826 Description: Permanently set tunables
827 Details : All writable /proc/fs/lustre tunables can now be permanently
828 set on a per-server basis, at mkfs time or on a live
831 Severity : enhancement
833 Description: Lustre message v2
834 Details : Add lustre message format v2.
836 Severity : enhancement
838 Description: client OST exclusion list
839 Details : Clients can be started with a list of OSTs that should be
840 declared "inactive" for known non-responsive OSTs.
844 Description: SPEC SFS validation failure on NFS v2 over lustre.
845 Details : Changes the blocksize for regular files to be 2x RPC size,
846 and not depend on stripe size.
848 Severity : enhancement
850 Description: Multiple MD RPCs in flight.
851 Details : Further unserialise some read-only MDS RPCs - learn about intents.
852 To avoid overly-overloading MDS, introduce a limit on number of
853 MDS RPCs in flight for a single client and add /proc controls
854 to adjust this limit.
856 Severity : enhancement
858 Description: client read/write statistics
859 Details : Add client read/write call usage stats for performance
860 analysis of user processes.
861 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
862 file access. extents_stats shows chunk size distribution.
863 extents_stats_per_process show chunk size distribution per
866 Severity : enhancement
868 Description: mds statistics
869 Details : Add detailed mds operations statistics in
870 /proc/fs/lustre/mds/*/stats.
874 Description: Failure of copying files with lustre special EAs.
875 Details : Client side always return success for setxattr call for lustre
876 special xattr (currently only "trusted.lov").
881 Description: Refcount LNET uuids
882 Details : The global LNET uuid list grew linearly with every startup;
883 refcount repeated list entries instead of always adding to
886 Severity : enhancement
888 Description: Dynamic service threads
889 Details : Within a small range, start extra service threads
890 automatically when the request queue builds up.
893 Frequency : mixed-endian client/server environments
895 Description: mixed-endian crashes
896 Details : The new msg_v2 system had some failures in mixed-endian
899 Severity : enhancement
901 Description: Easy OST removal
902 Details : OSTs can be permanently deactivated with e.g. 'lctl
903 conf_param lustre-OST0001.osc.active=0'
905 Severity : enhancement
907 Description: MGS proc entries
908 Details : Added basic proc entries for the MGS showing what filesystems
911 Severity : enhancement
913 Description: provide MGS failover
914 Details : Added config lock reacquisition after MGS server failover.
916 Severity : enhancement
918 Description: add Linux 2.4 support
919 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
923 Description: a large application tries to do I/O to the same resource and dies
925 Details : Check the req->rq_arrival time after the call to
926 ost_brw_lock_get(), but before we do anything about
927 processing it & sending the BULK transfer request. This
928 should help move old stale pending locks off the queue as
929 quickly as obd_timeout.
932 Frequency : when an incorrect nid is specified during startup
934 Description: ptlrpc connect to non-existant node causes kernel crash
935 Details : LNET can't be re-entered from an event callback, which
936 happened when we expire a message after the export has been
937 cleaned up. Instead, hand the zombie cleanup off to another
940 Severity : enhancement
942 Description: plain/inodebits lock performance improvement
943 Details : Grouping plain/inodebits in granted list by their request modes
944 and bits policy, thus improving the performance of search through
948 Frequency : only if OST filesystem is corrupted
950 Description: client incorrectly hits assertion in ptlrpc_replay_req()
951 Details : for a short time RPCs with bulk IO are in the replay list,
952 but replay of bulk IOs is unimplemented. If the OST filesystem
953 is corrupted due to disk cache incoherency and then replay is
954 started it is possible to trip an assertion. Avoid putting
955 committed RPCs into the replay list at all to avoid this issue.
958 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
960 Description: System hang on startup
961 Details : This bug allowed the liblustre (e.g. catamount) client to
962 return to the app before handling all startup RPCs. This
963 could leave the node unresponsive to lustre network traffic
964 and manifested as a server ptllnd timeout.
966 Severity : enhancement
968 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
969 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
970 boolean to control whether to print peer debug info when a
971 client's RPC times out.
974 Frequency : only for kernels with patches from Lustre below 1.4.3
976 Description: Remove old rdonly API
977 Details : Remove old rdonly API which unused from at least lustre 1.4.3
980 Frequency : only for devices with external journals
982 Description: Set external device read-only also
983 Details : During a commanded failover stop, we set the disk device
984 read-only while the server shuts down. We now also set any
985 external journal device read-only at the same time.
988 Frequency : when upgrading from 1.4 while trying to change parameters
990 Description: The wrong (new) MDC name was used when setting parameters for
991 upgraded MDT's. Also allows changing of OSC (and MDC)
992 parameters if --writeconf is specified at tunefs upgrade time.
995 Frequency : when setting specific ost indicies
997 Description: QOS code breaks on skipped indicies
998 Details : Add checks for missing OST indicies in the QOS code, so OSTs
999 created with --index need not be sequential.
1001 Severity : enhancement
1003 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1004 Details : The uninit_groups feature works in conjunction with the kernel
1005 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1006 up the pass1 processing of e2fsck. This is a read-only feature
1007 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1008 mount filesystems that have had this feature enabled.
1010 Severity : enhancement
1012 Description: Improve multi-block allocation algorithm to avoid fragmentation
1013 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1014 allocation locality and avoid filesystem fragmentation.
1016 ------------------------------------------------------------------------------
1018 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1020 * Support for kernels:
1021 2.4.21-47.0.1.EL (RHEL 3)
1022 2.6.5-7.283 (SLES 9)
1023 2.6.9-42.0.10.EL (RHEL 4)
1024 2.6.12.6 vanilla (kernel.org)
1025 2.6.16.27-0.9 (SLES 10)
1026 * Recommended e2fsprogs version: 1.39.cfs5
1028 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1032 Frequency : occasional, depends on client load and configuration
1033 Bugzilla : 12181, 12203
1034 Description: data loss for recently-modified files
1036 Details : In some cases it is possible that recently written or created
1037 files may not be written to disk in a timely manner (this should
1038 normally be within 30s unless client IO load is very high).
1039 The problem appears as zero-length files or files that are a
1040 multiple of 1MB in size after a client crash or client eviction
1041 that are missing data at the end of the file.
1043 This problem is more likely to be hit on clients where files are
1044 repeatedly created and unlinked in the same directory, clients
1045 have a large amount of RAM, have many CPUs, the filesystem has
1046 many OSTs, the clients are rebooted frequently, and/or the files
1047 are not accessed by other nodes after being written.
1049 The presence of the problem can be detected by looking at
1050 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1051 smaller than the second (nr_unused) then dirty files will not
1052 be flushed automatically to disk. "sync; sleep 10" should be
1053 run several times on the node before unmounting it to update
1054 Lustre (this is also safe to run on nodes without this problem).
1056 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1057 that can cause this same problem, so customers using that kernel
1058 also need to update the kernel in addition to Lustre. In order
1059 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1061 It is normal that files written just before a client crash (less
1062 than 30s) may not yet have been flushed to disk, even for local
1066 Frequency : frequent on thin XT3 nodes
1068 Description: UUID collision on thin XT3 Linux nodes
1069 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1070 randomly, since we relied on an insufficiently-seeded PRNG.
1075 Description: OSS hangs after "All ost request buffers busy"
1076 Details : A deadlock between quota and journal operations caused OSS
1077 hangs after printing "All ost request buffers busy."
1080 Frequency : always on liblustre builds
1082 Description: Cleanup compiler warnings on liblustre
1085 Frequency : always on liblustre builds on XT3
1087 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1091 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1092 > 32000 subdirectories
1093 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1094 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1095 subdirectory count crosses 32000. This will aid e2fsck to
1096 correctly handle more than 32000 subdirectories.
1099 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1101 Description: System hang on startup
1102 Details : This bug allowed the liblustre (e.g. catamount) client to
1103 return to the app before handling all startup RPCs. This
1104 could leave the node unresponsive to lustre network traffic
1105 and manifested as a server ptllnd timeout.
1107 Severity : enhancement
1109 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1110 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1111 boolean to control whether to print peer debug info when a
1112 client's RPC times out.
1117 Description: make O_SYNC working on 2.6 kernels
1118 Details : 2.6 kernels use different method for mark pages for write,
1119 so need add a code to lustre for O_SYNC work.
1124 Description: Failure to close file and release space on NFS
1125 Details : Put inode details into lock acquired in ll_intent_file_open.
1126 Use mdc_intent_lock in ll_intent_open to properly
1127 detect all kind of errors unhandled by mdc_enqueue.
1132 Description: proc file read during shutdown sometimes raced obd removal,
1134 Details : Add lock to prevent obd access after proc file removal.
1137 Frequency : Only for files larger than 4GB on 32-bit clients.
1139 Description: improperly doing page alignment of locks
1140 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1141 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1146 Description: RPCs being resent when they shouldn't be
1147 Details : Some RPCs that should not be resent are being resent. This
1148 can cause inconsistencies in the RPC state machine. Do not
1149 resend such requests.
1152 Frequency : rare, only with NFS export
1154 Description: Crash on NFS re-export node
1155 Details : under very unusual load conditions an assertion is hit in
1156 ll_intent_file_open()
1159 Frequency : only if OST filesystem is corrupted
1161 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1162 Details : for a short time RPCs with bulk IO are in the replay list,
1163 but replay of bulk IOs is unimplemented. If the OST filesystem
1164 is corrupted due to disk cache incoherency and then replay is
1165 started it is possible to trip an assertion. Avoid putting
1166 committed RPCs into the replay list at all to avoid this issue.
1171 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1172 Details : Large single O_DIRECT read and write calls can fail to allocate
1173 a sufficiently large buffer to process the request. In case of
1174 allocation failure the allocation is retried with a smaller
1175 buffer and broken into smaller requests.
1177 Severity : enhancement
1179 Description: Add -o localflock option to simulate old noflock behaviour.
1180 Details : This will achieve local-only flock/fcntl locks coherentness.
1185 Description: versioning check is incomplete
1186 Details : Checking the version difference of client vs. server, report
1187 error if the gap is too big.
1192 Description: add support PG_writeback bit
1193 Details : add support for PG_writeback bit for Lustre, for more carefull
1194 work with page cache in 2.6 kernel. This also fix some deadlocks
1195 and remove hack for work O_SYNC with 2.6 kernel.
1197 Severity : enhancement
1199 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1200 Details : The uninit_groups feature works in conjunction with the kernel
1201 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1202 up the pass1 processing of e2fsck. This is a read-only feature
1203 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1204 mount filesystems that have had this feature enabled.
1206 Severity : enhancement
1208 Description: Improve multi-block allocation algorithm to avoid fragmentation
1209 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1210 allocation locality and avoid filesystem fragmentation.
1212 ------------------------------------------------------------------------------
1214 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1216 * Support for kernels:
1217 2.6.9-42.0.3.EL (RHEL 4)
1218 2.6.5-7.276 (SLES 9)
1219 2.4.21-47.0.1.EL (RHEL 3)
1220 2.6.12.6 vanilla (kernel.org)
1221 2.6.16.21-0.8 (SLES10)
1222 * Recommended e2fsprogs version: 1.39.cfs2-0
1224 * The backwards-compatible /proc/sys/portals symlink has been removed
1225 in this release. Before upgrading, please ensure that you change
1226 any configuration scripts or /etc/sysctl.conf files that access
1227 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1228 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1229 in advance of the upgrade on any system running Lustre 1.4.6 or
1230 newer, since /proc/sys/lnet was added in that version.
1231 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1235 Frequency : only when quota is used
1237 Description: avoid scanning export list for quota master
1238 Details : Change the algorithms to avoid scanning export list in order
1239 to improve the efficiency.
1242 Frequency : MDS failover only, very rarely
1244 Description: "went back in time" messages on mds failover
1245 Details : The greatest transno may be lost when the current operation
1246 finishes with an error (transno==0) and the client's last_rcvd
1247 record is over-written. Save the greatest transno in the
1248 mds_last_transno for this case.
1251 Frequency : always for specific kernels and striping counts
1253 Description: client may get "Matching packet too big" without ACL support
1254 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1255 when trying to access files in certain configurations. The
1256 clients should in fact be denied when mounting because they do
1257 not understand ACLs.
1260 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1262 Description: many clients connecting with IO in progress causes connect timeouts
1263 Details : Avoid synchronous journal commits to avoid delays caused by many
1264 clients connecting/disconnecting when bulk IO is in progress.
1265 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1266 OST_IO_PORTAL to avoid delays behind potentially many pending
1270 Frequency : occasionally with multiple writers to a single file
1272 Description: shared writes to file may result in wrong size reported by stat()
1273 Details : Allow growing of kms when extent lock is cancelled
1276 Frequency : always with random mmap IO to multi-striped file
1278 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1279 Details : Only if the hole is at the end of OST object so that kms is too
1280 small. Fix is to increase kms accordingly in ll_nopage.
1283 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1285 Description: writes to a missing object would leak memory on the OST
1286 Details : If there is an inconsistency between the MDS and OST filesystems,
1287 such that the MDS references an object that doesn't exist, writes
1288 to that object will leak memory due to incorrect cleanup in the
1289 error handling path, eventually running out of memory on the OST.
1294 Description: Creating too long symlink causes lustre errors
1295 Details : Check symlink and name lengths before sending requests to MDS.
1298 Frequency : only if flock is enabled (not on by default)
1300 Description: posix locks not released on fd closure on 2.6.9+
1301 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1302 kernels, this caused such locks not to be released on fd close and
1303 then assertions on fs unmount about still used locks.
1306 Frequency : MDS failover only, very rarely
1308 Description: clients may get ASSERTION(granted_lock != NULL)
1309 Details : When request was taking a long time, and a client was resending
1310 a getattr by name lock request. The were multiple lock requests
1311 with the same client lock handle and
1312 mds_getattr_name->fixup_handle_for_resent_request found one of the
1313 lock handles but later failed with ASSERTION(granted_lock != NULL).
1318 Description: handle->h_buffer_credits > 0, assertion failure
1319 Details : h_buffer_credits is zero after truncate, causing assertion
1320 failure. This patch extends the transaction or creates a new
1324 Frequency : NFS re-export or patchless client
1325 Bugzilla : 11179, 10796
1326 Description: Crash on NFS re-export node (__d_move)
1327 Details : We do not want to hash the dentry if we don't have a lock.
1328 But if this dentry is later used in d_move, we'd hit uninitialised
1329 list head d_hash, so we just do this to init d_hash field but
1330 leave dentry unhashed.
1333 Frequency : NFS re-export or patchless client
1335 Description: NFS exports has problem with symbolic link
1336 Details : lustre client didn't properly install dentry when re-exported
1337 to NFS or running patchless client.
1340 Frequency : NFS re-export or patchless client
1342 Description: Various nfs/patchless fixes.
1343 Details : fixes reuse disconected alias for lookup process - this fixes
1344 warning "find_exported_dentry: npd != pd",
1345 fix permission error with open files at nfs.
1346 fix apply umask when do revalidate.
1349 Frequency : occasional
1351 Description: Crash on NFS re-export node
1352 Details : calling clear_page() on the wrong pointer triggered oops in
1353 generic_mapping_read().
1356 Frequency : rarely, using O_DIRECT IO
1358 Description: unaligned directio crashes client with LASSERT
1359 Details : check for unaligned buffers before trying any requests.
1362 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1364 Description: stale data returned from RAID cache
1365 Details : If only a small amount of IO is done to the RAID device before
1366 reading it again it is possible to get stale data from the RAID
1367 cache instead of reading it from disk.
1370 Frequency : always for sles10 kernel
1372 Description: sles10 support
1373 Details : ll_follow_link: compile fixes and using of nd_set_link
1374 under newer kernels.
1377 Frequency : depends on arch, kernel and compiler version, always on sles10
1380 Description: recursive or deep enough symlinks cause stack overflow
1381 Details : getting rid of large stack-allocated variable in
1385 Frequency : depends on hardware
1387 Description: lustre write performance loss in the SLES10 kernel
1388 Details : the performance loss is caused by using of write barriers in the
1389 ext3 code. The SLES10 kernel turns barrier support on by
1390 default. The fix is to undo that change for ldiskfs.
1392 ------------------------------------------------------------------------------
1394 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1396 * Support for kernels:
1397 2.6.9-42.0.3EL (RHEL 4)
1398 2.6.5-7.276 (SLES 9)
1399 2.4.21-47.0.1.EL (RHEL 3)
1400 2.6.12.6 vanilla (kernel.org)
1404 Frequency : quota enabled and large files being deleted
1406 Description: releasing more than 4GB of quota at once hangs OST
1407 Details : If a user deletes more than 4GB of files on a single OST it
1408 will cause the OST to spin in an infinite loop. Release
1409 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1414 Description: statfs data retrieved from /proc may be stale or zero
1415 Details : When reading per-device statfs data from /proc, in the
1416 {kbytes,files}_{total,free,avail} files, it may appear
1417 as zero or be out of date.
1420 Frequency : systems with MD RAID1 external journal devices
1422 Description: lconf's call to blkid is confused by RAID1 journal devices
1423 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1424 of returning all block devices that match the journal UUID.
1427 Frequency : always, for aggregate stripe size over 4GB
1429 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1430 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1431 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1434 Frequency : always if "lfs find" used on a local file/directory
1436 Description: "lfs find" segfaults if used on a local file/directory
1437 Details : The case where a directory component was not specified wasn't
1438 handled correctly. Handle this properly.
1441 Frequency : always on ppc64
1443 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1444 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1445 perform endianness conversion.
1448 Frequency : rarely (truncate to non-zero file size after write under load)
1449 Bugzilla : 10730, 10687
1450 Description: Files padded with zeros to next 4K multiple
1451 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1452 it is possible that files that are truncated to a non-zero size
1453 immediately after being written are filled with zero bytes beyond
1454 the truncated size. No file data is lost.
1456 Severity : enhancement
1458 Description: Allow recovery/failover for liblustre clients.
1459 Details : liblustre clients were unaware of failover configurations until
1462 Severity : enhancement
1464 Description: user file locks should fail when not mounting with flock option
1465 Details : Set up an error-returning stub in ll_file_operations.lock field
1466 to prevent incorrect behaviour when client is mounted without
1467 flock option. Also, set up properly f_op->flock field for
1471 Frequency : always on ia64
1473 Description: "lfs df" loops on printing out MDS statfs information
1474 Details : The obd_ioctl_data was not initialized and in some systems
1475 this caused a failure during the ioctl that did not return
1476 an error. Initialize the struct and return an error on failure.
1479 Frequency : SLES 9 only
1481 Description: Error of copying files with lustre special EAs as root
1482 Details : Client side always return success for setxattr call for lustre
1483 special xattr (currently only "trusted.lov").
1486 Frequency : rarely on clusters with both ia64+i386 clients
1488 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1489 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1490 clients are doing concurrent writes to the same file it is
1491 possible that the ia64 clients may LASSERT because the OST
1492 extent locks are not PAGE_SIZE aligned. Ensure that grown
1493 locks are always aligned on the request boundary.
1496 Frequency : specific use, occasional
1498 Description: Overwriting in use executable truncates on-disk binary image
1499 Details : If one node attempts to overwrite an executable in use by
1500 another node, we now correctly return ETXTBSY instead of
1501 truncating the file.
1503 Severity : enhancement
1505 Description: Async OSC create to avoid the blocking unnecessarily.
1506 Details : If a OST has no remain object, system will block on the creating
1507 when need to create a new object on this OST. Now, ways use
1508 pre-created objects when available, instead of blocking on an
1509 empty osc while others are not empty. If we must block, we block
1510 for the shortest possible period of time.
1515 Description: chmod on Lustre root is propagated to other clients
1516 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1517 invalid by the follow_mount time.
1520 Frequency : liblustre clients only
1522 Description: Race in 'instant cancel' lock handling could lead to such locks
1523 never to be granted in case of SMP MDS
1524 Details : Do not destroy not yet granted but cbpending locks in
1528 Frequency : replay/resend of open
1530 Description: non null lock assetion failure in mds_intent_policy
1531 Details : Trying to replay/resend lockless open requests resulted in
1532 mds_open() returning 0 with no lock. Now it sets a flag if
1533 a lock is going to be returned.
1535 Severity : enhancement
1537 Description: Checksum enhancements
1538 Details : New checksum enhancements allow for resending RPCs that failed
1541 Severity : enhancement
1543 Description: Tunables on number of dirty pages in cacche
1544 Details : Allow to set limit on number of dirty pages cached.
1549 Description: client crash on unmount - lock still has references
1550 Details : In some error handling cases it was possible to leak a lock
1551 reference on a client while accessing a file. This was not
1552 harmful to the client during operation, but would cause the
1553 client to crash when the filesystem is unmounted.
1556 Frequency : specific case, rare
1558 Description: ETXTBSY on mds though file not in use by client
1559 Details : ETXTBSY is no longer incorrectly returned when attempting to
1560 chmod or chown a directory that the user previously tried to
1561 execute or a currently-executing binary.
1564 Frequency : extremely rare except on liblustre-based clients
1566 Description: Lustre space not freed when files are deleted
1567 Details : Clean up open-unlinked files after client eviction. Previously
1568 the unlink was skipped and the files remained as orphans.
1573 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1574 Details : In some cases it was possible to send a blocking callback to
1575 a client doing a glimpse, even though that client didn't get
1576 a lock granted. When the glimpse lock is cancelled on the OST
1577 the freed lock is left on the waiting list and corrupted the list.
1580 Frequency : all core dumps
1582 Description: Broke core dumps to lustre
1583 Details : Negative dentry may be unhashed if parent does not have UPDATE
1584 lock, but some callers, e.g. do_coredump, expect dentry to be
1585 hashed after successful create, hash it in ll_create_it.
1587 ------------------------------------------------------------------------------
1589 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1591 * Support for kernels:
1592 2.6.9-42.0.2.EL (RHEL 4)
1593 2.6.5-7.276 (SLES 9)
1594 2.4.21-40.EL (RHEL 3)
1595 2.6.12.6 vanilla (kernel.org)
1599 Frequency : always on RHEL 3
1601 Description: Number of open files grows over time
1602 Details : The number of open files grows over time, whether or not
1603 Lustre is started. This was due to a filp leak introduced
1604 by one of our kernel patches.
1606 ------------------------------------------------------------------------------
1608 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1610 * Support for kernels:
1611 2.6.9-42.EL (RHEL 4)
1612 2.6.5-7.276 (SLES 9)
1613 2.4.21-40.EL (RHEL 3)
1614 2.6.12.6 vanilla (kernel.org)
1619 Bugzilla : 5719, 9635, 9792, 9684
1620 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1621 Details : If a server is under heavy load and cannot reply to new
1622 connection requests before the client resends the (re)connect,
1623 the connection handling code can behave badly if two service
1624 threads are concurrently handing separate (re)connections from
1625 the same client. Add better locking to the connection handling
1626 code, and ensure that only a single connection will be processed
1627 for a given client UUID, even if the lock is dropped.
1629 Severity : enhancement
1631 Description: add TCP zero-copy support to kernel
1632 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1633 sends if the hardware supports scatter-gather and checksumming.
1634 This allows socklnd to do client-write and server-read more
1635 efficiently and reduce CPU utilization from skbuf copying.
1638 Frequency : only if NFS exporting from client
1640 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1641 Details : If setting attributes on a file created under NFS that had
1642 never been opened it would be possible to oops the client
1643 if the file had no objects.
1646 Frequency : always for liblustre
1648 Description: liblustre client does MDS+OSTs setattr RPC for each write
1649 Details : When doing a write from a liblustre client, the client
1650 incorrectly issued an RPC to the MDS and each OST the file was
1651 striped over in order to update the timestamps. When writing
1652 with small chunks and many clients this could overwhelm the MDS
1653 with RPCs. In all cases it would slow down the write because
1654 these RPCs are unnecessary.
1656 Severity : enhancement
1658 Description: allow number of MDS service threads to be changed at module load
1659 Details : It is now possible to change the number of MDS service threads
1660 running. Adding "options mds mds_num_threads={N}" to the MDS's
1661 /etc/modprobe.conf will set the number of threads for the next
1662 time Lustre is restarted (assuming the "mds" module is also
1663 reloaded at that time). The default number of threads will
1664 stay the same, 32 for most systems.
1669 Description: OST crash if filesystem is unformatted or corrupt
1670 Details : If an OST is started on a device that has never been formatted
1671 or if the filesystem is corrupt and cannot even mount then the
1672 error handling cleanup routines would dereference a NULL pointer.
1677 Description: NULL pointer deref in llap_from_page.
1678 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1679 incorrect) mapping if the page was truncated/reclaimed while it was
1680 searched for. Check for this condition and skip such pages when
1681 doing readahead. Introduce extra check to llap_from_page() to
1682 verify page->mapping->host is non-NULL (so page is not anonymous).
1685 Frequency : Sometimes when using sys_sendfile
1687 Description: "page not covered by a lock" warnings from ll_readpage
1688 Details : sendfile called ll_readpage without right page locks present.
1689 Now we introduced ll_file_sendfile that does necessary locking
1690 around call to generic_file_sendfile() much like we do in
1694 Frequency : with certain MDS communication failures at client mount time
1696 Description: NULL pointer deref after failed client mount
1697 Details : a client connection request may delayed by the network layer
1698 and not be sent until after the PTLRPC layer has timed out the
1699 request. If the client fails the mount immediately it will try
1700 to clean up before the network times out the request. Add a
1701 reference from the request import to the obd device and delay
1702 the cleanup until the network drops the request.
1705 Frequency : occasionally during client (re)connect
1707 Description: assertion failure during client (re)connect
1708 Details : processing a client connection request may be delayed by the
1709 client or server longer than the client connect timeout. This
1710 causes the client to resend the connection request. If the
1711 original connection request is replied in this interval, the
1712 client may trip an assertion failure in ptlrpc_connect_interpret()
1713 which thought it would be the only running connect process.
1716 Frequency : only with obd_echo servers and clients that are rebooted
1718 Description: kernel BUG accessing uninitialized data structure
1719 Details : When running an obd_echo server it did not start the ping_evictor
1720 thread, and when a client was evicted an uninitialized data
1721 structure was accessed. Start the ping_evictor in the RPC
1722 service startup instead of the OBD startup.
1724 Severity : enhancement
1725 Bugzilla : 10193 (patchless)
1726 Description: Remove dependency on various unexported kernel interfaces.
1727 Details : No longer need reparent_to_init, exit_mm, exit_files,
1728 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1731 Frequency : rare (only users of deprecated and unsupported LDAP config)
1733 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1734 Details : LDAP apparently contains 'lustreName' attributes instead of
1735 'name'. A simple remapping of the name is sufficient.
1738 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1740 Description: waiting_locks_callback trips kernel BUG if client is evicted
1741 Details : Running with the dump_on_timeout debug flag turned on makes
1742 it possible that the waiting_locks_callback() can try to dump
1743 the Lustre kernel debug logs from an interrupt handler. Defer
1744 this log dumping to the expired_lock_main() thread.
1746 Severity : enhancement
1748 Description: Support NFS exporting on 2.6 kernels.
1749 Details : Implement non-rawops metadata methods for NFS server to use without
1750 changing NFS server code.
1753 Frequency : very rare (synthetic metadata workload only)
1755 Description: two racing renames might cause an MDS thread to deadlock
1756 Details : Running the "racer" program may cause one MDS thread to rename
1757 a file from being the source of a rename to being the target of
1758 a rename at exactly the same time that another thread is doing
1759 so, and the second thread has already enqueued these locks after
1760 doing a lookup of the target and is trying to relock them in
1761 order. Ensure that we don't try to re-lock the same resource.
1764 Frequency : only very large systems with liblustre clients
1766 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1767 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1768 all OSTs in parallel. This allows the eviction of stale liblustre
1769 clients to proceed much faster than if they were done in series,
1770 and also offers similar improvements for other set_info RPCs.
1775 Description: excessive CPU usage during initial read phase on client
1776 Details : During the initial read phase on a client, it would agressively
1777 retry readahead on the file, consuming too much CPU and impacting
1778 performance (since 1.4.5.8). Improve the readahead algorithm
1779 to avoid this, and also improve some other common cases (read
1780 of small files in particular, where "small" is files smaller than
1781 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1786 Description: MDS crash when receiving packet with unknown intent.
1787 Details : Do not LBUG in unknown intent case, just return -EFAULT
1789 Severity : enhancement
1790 Bugzilla : 9293, 9385
1791 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1792 Details : Do not serialize getattr (non-intent version) and statfs.
1795 Frequency : occasional, when OST network is overloaded/intermittent
1797 Description: client evicted by OST after bulk IO timeout
1798 Details : If a client sends a bulk IO request (read or write) the OST
1799 may evict the client if it is unresposive to its data GET/PUT
1800 request. This is incorrect if the network is overloaded (takes
1801 too long to transfer the RPC data) or dropped the OST GET/PUT
1802 request. There is no need to evict the client at all, since
1803 the pinger and/or lock callbacks will handle this, and the
1804 client can restart the bulk request.
1807 Frequency : Always when mmapping file with no objects
1809 Description: client crashes when mmapping file with no objects
1810 Details : Check that we actually have objects in a file before doing any
1811 operations on objects in ll_vm_open, ll_vm_close and
1817 Description: Request leak when working with deleted CWD
1818 Details : Introduce advanced request refcount tracking for requests
1819 referenced from lustre intent.
1821 Severity : Enhancement
1823 Description: Cache open file handles on client.
1824 Details : MDS now will return special lock along with openhandle, if
1825 requested and client is allowed to hold openhandle, even if unused,
1826 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1827 closing files for every read/write openration.
1829 Severity : Enhancement
1831 Description: Cache open negative dentries on client when possible.
1832 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1833 negative dentries on lock revocation.
1838 Description: Remounting a client read-only wasn't possible with a zconf mount
1839 Details : It wasn't possible to remount a client read-only with llmount.
1841 Severity : enhancement
1842 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1843 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1844 MPICH is included to add Lustre-specific ADIO interfaces.
1845 This is based closely on the UFS ADIO layer and only differs
1846 in file creation, in order to allow the OST striping to be set.
1847 This is user-contributed code and not supported by CFS.
1852 Description: extended inode attributes (immutable, append-only) work improperly
1853 when 2.4 and 2.6 kernels are used on client/server or vice versa
1854 Details : Introduce kernel-independent values for these flags.
1856 Severity : enhancement
1859 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1860 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1861 granularity. Now, Fractional MB granularity is be supported,
1862 this is very useful for low memory system.
1864 Severity : enhancement
1866 Description: Getattr by fid
1867 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1868 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1869 have locks and avoids getattr rpc requests for it on every stat.
1872 Frequency : Always, for filesystems larger than 2TB
1874 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1875 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1876 filesystems larger than 2TB to be mounted with mballoc enabled.
1879 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1881 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1882 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1883 filesystems larger than 2TB is mounted then the kernel will
1884 silently corrupt the start of the filesystem. CONFIG_LBD is
1885 enabled for all CFS-supported kernels, but the possibility of
1886 this happening with a modified kernel config exists.
1888 Severity : enhancement
1890 Description: add client O_DIRECT support for 2.6 kernels
1891 Details : It is now possible to do O_DIRECT reads and writes to files
1892 in the Lustre client mountpoint on 2.6 kernel clients.
1894 Severity : enhancement
1896 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1897 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1898 parallel, not waiting for response from every OST before sending
1899 a rpc to the next OST.
1904 Description: setattr vs write race when updating file timestamps
1905 Details : Client processes that update a file timestamp into the past
1906 right after writing to the file (e.g. tar) it is possible that
1907 the updated file modification time can be reset to the current
1908 time due to a race between processing the setattr and write RPC.
1910 Severity : enhancement
1912 Description: Bring 'lfs find' closer in line with regular Linux find.
1913 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1914 -print0 options and obtains all the needed info through the lustre
1917 Severity : enhancement
1919 Description: support up to 1024 configured devices on one node
1920 Details : change obd_dev array from statically allocated to dynamically
1921 allocated structs as they are first used to reduce memory usage
1926 Description: Flush dirty partially truncated pages during truncate
1927 Details : Immediatelly flush partially truncated pages in filter_setattr,
1928 this way we completely avoid having any pages in page cache on OST
1929 and can retire ugly workarounds during writes to flush such pages.
1934 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1935 Details : protect inode from truncation within vfs_unlink() context
1936 just take a reference before calling vfs_unlink() and release it
1937 when parent's i_sem is free.
1942 Description: last_id value checked outside lock on OST caused LASSERT failure
1943 Details : If there were multiple MDS->OST object precreate requests in
1944 flight, it was possible that the OST's last object id was checked
1945 outside a lock and incorrectly tripped an assertion. Move checks
1946 inside locks, and discard old precreate requests.
1949 Frequency : always, if extents are used on OSTs
1951 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1952 Details : OSTs using the extents format would not zero the high 16 bits of
1953 the index physical block number. This is not a problem for any
1954 OST filesystems smaller than 16TB, and no kernels support ext3
1955 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1956 new/modified files) and can be fixed for existing filesystems
1957 with e2fsprogs-1.39-cfs1.
1962 Description: import connection selection may be incorrect if timer wraps
1963 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1964 import connections to be ignored if the 32-bit jiffies counter
1965 wraps. Use a 64-bit jiffies counter.
1968 Frequency : very large clusters immediately after boot
1970 Description: LNET request buffers exhausted under heavy short-term load
1971 Details : If a large number of client requests are generated on a service
1972 that has previously never seen so many requests it is possible
1973 that the request buffer growth cannot keep up with the spike in
1974 demand. Instead of dropping incoming requests, they are held in
1975 the LND until the RPC service can accept more requests.
1978 Frequency : Sometimes during replay
1980 Description: Assertion failure in ll_local_open after replay.
1981 Details : If replay happened on an open request reply before we were able
1982 to set replay handler, reply will become not swabbed tripping the
1983 assertion in ll_local_open. Now we set the handler right after
1984 recognising of open request
1987 Frequency : very rare
1989 Description: kernel reports "badness in vsnprintf"
1990 Details : Reading from the "recovery_status" /proc file in small chunks
1991 may cause a negative length in lprocfs_obd_rd_recovery_status()
1992 call to vsnprintf() (which is otherwise harmless). Exit early
1993 if there is no more space in the output buffer.
1995 Severity : enhancement
1997 Description: clear OBD RPC statistics by writing to them
1998 Details : It is now possible to clear the OBD RPC statistics by writing
1999 to the "stats" file.
2004 Description: Client mtime is not the same on different clients after utimes
2005 Details : In some cases, the client was using the utimes() syscall on
2006 a file cached on another node. The clients now validate the
2007 ctime from the MDS + OSTs to determine which one is right.
2012 Description: Inability to activate failout mode
2013 Details : lconf script incorrectly assumed that in pythong string's numeric
2014 value is used in comparisons.
2017 Frequency : always with multiple stripes per file
2019 Description: Inefficient object allocation for mutli-stripe files
2020 Details : When selecting which OSTs to stripe files over, for files with
2021 a stripe count that divides evenly into the number of OSTs,
2022 the MDS is always picking the same starting OST for each file.
2023 Return the OST selection heuristic to the original design.
2028 Description: mount failures may take full timeout to return an error
2029 Details : Under some heavy load conditions it is possible that a
2030 failed mount can wait for the full obd_timeout interval,
2031 possibly several minutes, before reporting an error.
2032 Instead return an error as soon as the status is known.
2034 Frequency : quota enabled and large files being deleted
2036 Description: releasing more than 4GB of quota at once hangs OST
2037 Details : If a user deletes more than 4GB of files on a single OST it
2038 will cause the OST to spin in an infinite loop. Release
2039 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2044 Description: statfs data retrieved from /proc may be stale or zero
2045 Details : When reading per-device statfs data from /proc, in the
2046 {kbytes,files}_{total,free,avail} files, it may appear
2047 as zero or be out of date.
2050 Frequency : systems with MD RAID1 external journal devices
2052 Description: lconf's call to blkid is confused by RAID1 journal devices
2053 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2054 of returning all block devices that match the journal UUID.
2057 Frequency : always, for aggregate stripe size over 4GB
2059 Description: assertion fails when trying to use 4GB stripe size
2060 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2061 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2064 Frequency : always on ppc64
2066 Description: the first write on an ext3 filesystem with mballoc got stuck
2067 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2068 perform endianness conversion.
2070 ------------------------------------------------------------------------------
2072 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2074 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2075 INTEROPERATE with older versions automatically. Please read the
2076 user documentation before upgrading any part of a live system.
2077 * WARNING: Lustre networking configuration changes are required with
2078 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2081 * Support for kernels:
2082 2.6.9-22.0.2.EL (RHEL 4)
2083 2.6.5-7.244 (SLES 9)
2084 2.6.12.6 vanilla (kernel.org)
2087 Severity : enhancement
2088 Bugzilla : 7981/8208
2089 Description: Introduced Lustre Networking (LNET)
2090 Details : LNET is new networking infrastructure for Lustre, it includes
2091 a reorganized network configuration mode (see the user
2092 documentation for full details) as well as support for routing
2093 between different network fabrics. Lustre Networking Devices
2094 (LNDs) for the supported network fabrics have also been
2095 created for this new infrastructure.
2097 Severity : enhancement
2098 Description: Introduced Access control lists
2099 Details : clients can set ACLs on files and directories in order to have
2100 more fine-grained permissions than the standard Unix UGO+RWX.
2101 The MDS must be started with the "-o acl" mount option.
2103 Severity : enhancement
2104 Description: Introduced filesystem quotas
2105 Details : Administrators may now establish per-user quotas on the
2108 Severity : enhancement
2110 Description: Configuration change for the XT3
2111 The PTLLND is now used to run Lustre over Portals on the XT3
2112 The configure option(s) --with-cray-portals are no longer used.
2113 Rather --with-portals=<path-to-portals-includes> is used to
2114 enable building on the XT3. In addition to enable XT3 specific
2115 features the option --enable-cray-xt3 must be used.
2120 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2121 Details : A race between allocating a new llcd and re-getting the llcd_lock
2122 allowed another thread to grab newly-allocated llcd.
2124 Severity : enhancement
2126 Description: 2.6 OST async journal commit and locking fix to improve performance
2127 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2128 2.6 kernels are now async as they already were in 2.4 kernels so
2129 that they can commit concurrently with the network bulk transfer.
2130 For block-allocated files the filter allocation semaphore is held
2131 to avoid filesystem fragmentation during allocation. BKL lock
2132 removed for 2.6 xattr operations where it is no longer needed.
2137 Description: lconf incorrectly determined whether two IP networks could talk
2138 Details : In some more complicated routing and multiple-network
2139 configurations, lconf will avoid trying to make a network
2140 connection to a disjoint part of the IP space. It was doing the
2141 math incorrectly for one set of cases.
2146 Description: Fix for potential infinite loop processing records in an llog.
2147 Details : If an llog record is corrupted/zeroed, it is possible to loop
2148 forever in llog_process(). Validate the llog record length
2149 and skip the remainder of the block on error.
2152 Frequency : occasional (liblustre only)
2154 Description: liblustre could not open files whose last component is a symlink
2155 Details : sysio_path_walk() would incorrectly pass the open intent to
2156 intermediate path components.
2159 Frequency : rare (liblustre only with non-standard tuning)
2160 Bugzilla : 7201 (7350)
2161 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2162 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2163 after it has been released, only do this for referenced locks
2165 Severity : enhancement
2167 Description: specifying an (invalid) directory default stripe_size of -1
2168 would reset the directory default striping
2169 Details : stripe_size -1 was used internally to signal directory stripe
2170 removal, now use "all default" to signal dir stripe removal
2171 as a directory striping of "all default" is not useful
2174 Frequency : common for large clusters running liblustre clients
2176 Description: doing an ls when liblustre clients are running is slow
2177 Details : sending a glimpse AST to a liblustre client waits for every AST
2178 to time out, as liblustre clients will not respond. Since they
2179 cannot cache data we refresh the OST lock LVB from disk instead.
2181 Severity : enhancement
2183 Description: doing an ls at the same time as file IO can be slow
2184 Details : enqueue and other "small" requests can be blocked behind many
2185 large IO requests. Create a new OST IO portal for non-IO
2186 requests so they can be processed faster.
2189 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2191 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2192 created file with mode 000
2193 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2194 allow an ftruncate() as a non-root user to a file with mode 000.
2195 Lustre now respects this flag to disable mode checks when
2196 truncating a file owned by the user
2199 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2202 Description: Revoking locks from clients that went dead or catatonic might take
2204 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2205 cancellation of such locks instant on servers without waiting for
2206 any reply from clients. Clients drops these locks when cancel
2207 notification from server is received without replying.
2210 Frequency : liblustre-only, when liblustre client dies or becomes busy
2212 Description: Doing ls on Linux clients can take a long time with active
2214 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2215 granting such locks to it in the first place if possible. Locks
2216 are taken by proxy on the OST during the read or write and
2217 dropped immediately afterward. Add connect flags handling, do
2218 not grant locks to liblustre clients for glimpse ASTs.
2220 Severity : enhancement
2222 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2223 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2224 sequential reads and behaves badly with applications doing random
2225 reads. Improve it by reading ahead at least read region, and
2226 avoiding excessive large RPC for small reads.
2228 Severity : enhancement
2230 Description: Creating more than 1000 files for a single job may cause a load
2231 imbalance on the OSTs if there are also a large number of OSTs.
2232 Details : qos_prep_create() uses an OST index reseed value that is an
2233 even multiple of the number of available OSTs so that if the
2234 reseed happens in the middle of the object allocation it will
2235 still utilize the OSTs as uniformly as possible.
2240 Description: OST or MDS may oops in ping_evictor_main()
2241 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2242 but doesn't restart at beginning of obd_exports_timed list
2245 Severity : enhancement
2247 Description: improve by-nid export eviction on the MDS and OST
2248 Details : allow multiple exports with the same NID to be evicted at one
2249 time without re-searching the exports list.
2252 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2254 Description: MDS may oops in groups_free()
2255 Details : in rare race conditions a newly allocated group_info struct is
2256 freed again, and this can be NULL. The 2.4 compatibility code
2257 for groups_free() checked for a NULL pointer, but 2.6 did not.
2260 Frequency : common for liblustre clients doing little filesystem IO
2261 Bugzilla : 9352, 7313
2262 Description: server may evict liblustre clients accessing contended locks
2263 Details : if a client is granted a lock or receives a completion AST
2264 with a blocking AST already set it would not reply to the AST
2265 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2268 Frequency : lfs setstripe, only systems with more than 160 OSTs
2270 Description: unable to set striping with a starting offset beyond OST 160
2271 Details : llapi_create_file() incorrectly limited the starting stripe
2272 index to the maximum single-file stripe count.
2275 Frequency : LDAP users only
2277 Description: lconf did not handle in-kernel recovery with LDAP properly
2278 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2280 Severity : enhancement
2282 Description: bind OST threads to NUMA nodes to improve performance
2283 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2284 node and do their allocations there to localize memory access
2286 Severity : enhancement
2288 Description: llmount can determine client NID directly from Myrinet (GM)
2289 Details : the client NID code from gmnalnid was moved directly into
2290 llmount, removing the need to use this or specifying the
2291 client NID explicitly when mounting GM clients with zeroconf
2294 Frequency : if client is started with down MDS
2296 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2297 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2298 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2299 ptlrpc_import_delayed_req() because we didn't check if the
2300 request was interrupted, and we also didn't break out of the
2301 event loop if there was no timeout
2306 Description: data loss during non-page-aligned writes to a single file from
2307 both multiple nodes and multiple threads on one node at same time
2308 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2309 inconsistency led to false short-reads, that were cached and later
2310 used by ->prepare_write() to fill in partially written page,
2311 leading to data loss.
2314 Frequency : always, if lconf --abort_recovery used
2316 Description: lconf --abort_recovery fails with 'Operation not supported'
2317 Details : lconf was attempting to abort recovery on the MDT device and not
2320 Severity : enhancement
2322 Description: remove cleanup logs
2323 Details : replace lconf-generated cleanup logs with lustre internal
2324 cleanup routines. Eliminates the need for client-cleanup and
2327 Severity : enhancement
2329 Description: add support for EAs (user and system) on lustre filesystems
2330 Details : it is now possible to store extended attributes in the Lustre
2331 client filesystem, and with the user_xattr mount option it
2332 is possible to allow users to store EAs on their files also
2334 Severity : enhancement
2336 Description: Add possibility (config option) to show minimal available OST free
2338 Details : When compiled with --enable-mindf configure option, statfs(2)
2339 (and so, df) will return least minimal free space available from
2340 all OSTs as amount of free space on FS, instead of summary of
2341 free spaces of all OSTs.
2343 Severity : enhancement
2345 Description: do not expand extent locks acquired on OST-side
2346 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2347 by server: they are not cached anyway.
2350 Frequency : when mmap is used/binaries executed from Lustre
2352 Description: Unmmap pages before throwing them away from read cache.
2353 Details : llap_shrink cache now attempts to unmap pages before discarding
2354 them (if unmapping failed - do not discard). SLES9 kernel has
2355 extra checks that trigger if this unmapping is not done first.
2360 Description: lconf didn't resolve symlinks before checking to see whether a
2361 given mountpoint was already in use
2364 Frequency : when migrating failover services
2365 Bugzilla : 6395, 9514
2366 Description: When migrating a subset of services from a node (e.g. failback
2367 from a failover service node) the remaining services would
2368 time out and evict clients.
2369 Details : lconf --force (implied by --failover) sets the global obd_timeout
2370 to 5 seconds in order to quickly disconnect, but this caused
2371 other RPCs to time out too quickly. Do not change the global
2372 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2374 Severity : enhancement
2375 Frequency : if MDS is started with down OST
2376 Bugzilla : 9439,5706
2377 Description: Allow startup/shutdown of an MDS without depending on the
2378 availability of the OSTs.
2379 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2380 Add appropriate locking and lov-osc refcounts for safe
2381 cleaning. Add osc abort_inflight calls in case the
2382 synchronize never started.
2385 Frequency : occasional (Cray XT3 only)
2387 Description: root not authorized to access files in CRAY_PORTALS environment
2388 Details : The client process capabilities were not honoured on the MDS in
2389 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2390 been accessed by an authorized user then root was able to access
2391 the file on the local client also. The root user capabilities
2392 are now allowed on the MDS, as this environment has secure UID.
2395 Frequency : occasional
2397 Description: ldiskfs "too long searching" message happens too often
2398 Details : A debugging message (otherwise harmless) prints too often on
2399 the OST console. This has been reduced to only happen when
2400 there are fragmentation problems on the filesystem.
2405 Description: Division by zero in statfs when all OSCs are inactive
2406 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2407 lov_statfs divides by value returned by lov_get_stripecnt().
2411 Bugzilla : 9489, 3273
2412 Description: First write from each client to each OST was only 4kB in size,
2413 to initialize client writeback cache, which caused sub-optimal
2414 RPCs and poor layout on disk for the first writen file.
2415 Details : Clients now request an initial cache grant at (re)connect time
2416 and so that they can start streaming writes to the cache right
2417 away and always do full-sized RPCs if there is enough data.
2418 If the OST is rebooted the client also re-establishes its grant
2419 so that client cached writes will be honoured under the grant.
2424 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2425 Details : Now I/O RPCs go to different portal number and (presumably) fast
2426 lock requests (and glimses) and other RPCs get their own service
2427 threads pool that should be able to service those RPCs
2430 Severity : enhancement
2432 Description: Ability to exchange lustre version between client and servers and
2433 issue warnings at client side if client is too old. Also for
2434 liblustre clients there is ability to refuse connection of too old
2436 Details : New 'version' field is added to connect data structure that is
2437 filled with version info. That info is later checked by server and
2441 Frequency : rare, liblustre only.
2442 Bugzilla : 9296, 9581
2443 Description: Two simultaneous writes from liblustre at offset within same page
2444 might proceed at the same time overwriting eachother with stale
2446 Details : I/O lock withing llu_file_prwv was released too early, before data
2447 actually was hitting the wire. Extended lock-holding time until
2448 server acknowledges receiving data.
2451 Frequency : extremely rare. Never observed in practice.
2453 Description: avoid generating lustre_handle cookie of 0.
2454 Details : class_handle_hash() generates handle cookies by incrementing
2455 global counter, and can hit 0 occasionaly (this is unlikely, but
2456 not impossible, because initial value of cookie counter is
2457 selected randonly). Value of 0 is used as a sentinel meaning
2458 "unassigned handle" --- avoid it. Also coalesce two critical
2459 sections in this function into one.
2461 Severity : enhancement
2463 Description: allow liblustre clients to delegate truncate locking to OST
2464 Details : To avoid overhead of locking, liblustre client instructs OST to
2465 take extent lock in ost_punch() on client's behalf. New connection
2466 flag is added to handle backward compatibility.
2468 Severity : enhancement
2469 Bugzilla : 4928, 7341, 9758
2470 Description: allow number of OST service threads to be specified
2471 Details : a module parameter allows the number of OST service threads
2472 to be specified via "options ost ost_num_threads={N}" in the
2473 OSS's /etc/modules.conf or /etc/modprobe.conf.
2477 Bugzilla : 6146, 9635, 9895
2478 Description: servers crash with bad pointer in target_handle_connect()
2479 Details : In rare cases when a client is reconnecting it was possible that
2480 the connection request was the last reference for that export.
2481 We would temporarily drop the export reference and get a new
2482 one, but this may have been the last reference and the export
2483 was just destroyed. Get new reference before dropping old one.
2485 Severity : enhancement
2486 Frequency : if client is started with failover MDS
2488 Description: Allow multiple MDS hostnames in the mount command
2489 Details : Try to read the configuration from all specified MDS
2490 hostnames during a client mount in case the "primary"
2493 Severity : enhancement
2495 Description: Stop sending data to evicted clients as soon as possible.
2496 Details : Check if the client we are about to send or are sending data to
2497 was evicted already. (Check is done every second of waiting,
2498 for which l_wait_event interface was extended to allow checking
2499 of exit condition at specified intervals).
2502 Frequency : rare, normally only when NFS exporting is done from client
2504 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2506 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2509 Frequency : always, if srand() is called before liblustre initialization
2511 Description: Liblustre uses system PRNG disturbing its usage by user application
2512 Details : Introduce internal to lustre fast and high-quality PRNG for
2513 lustre usage and make liblustre and some other places in generic
2514 lustre code to use it.
2516 Severity : enhancement
2517 Bugzilla : 9477, 9557, 9870
2518 Description: Verify that the MDS configuration logs are updated when xml is
2519 Details : Check if the .xml configuration logs are newer than the config
2520 logs stored on the MDS and report an error if this is the case.
2521 Request --write-conf, or allow starting with --old_conf.
2523 Severity : enhancement
2525 Description: Handle symlinks in the path when checking if Lustre is mounted.
2526 Details : Resolve intermediate symlinks when checking if a client has
2527 mounted a filesystem to avoid duplicate client mounts.
2532 Description: lconf can hit an error exception but still return success.
2533 Details : The lconf command catches the Command error exception at the top
2534 level script context and will exit with the associated exit
2535 status, but doesn't ensure that this exit status is non-zero.
2540 Description: failure of ptlrpc thread startup can cause oops
2541 Details : Starting a ptlrpc service thread can fail if there are a large
2542 number of threads or the server memory is very fragmented.
2543 Handle this without oopsing.
2546 Frequency : always, only if liblustre and non-default acceptor port was used
2548 Description: liblustre cannot connect to servers with non-default acceptor port
2549 Details : tcpnal_set_default_params() was not called and was therefore
2550 ignoring the environment varaible TCPNAL_PORT, as well as other
2551 TCPNAL_ environment variables
2556 Description: two objects could be created on the same OST for a single file
2557 Details : If an OST is down, in some cases it was possible to create two
2558 objects on a single OST for a single file. No problems other
2559 than potential performance impact and spurious error messages.
2563 Bugzilla : 5681, 9562
2564 Description: Client may oops in ll_unhash_aliases
2565 Details : Client dcache may become inconsistent in race condition.
2566 In some cases "getcwd" can fail if the current directory is
2572 Description: Inode refcounting problems in NFS export code
2573 Details : link_raw functions used to call d_instantiate without obtaining
2574 extra inode reference first.
2578 Bugzilla : 9942, 9903
2579 Description: Referencing freed requests leading to crash, memleaks with NFS.
2580 Details : We used to require that call to ll_revalidate_it was always
2581 followed by ll_lookup_it. Also with revalidate_special() it is
2582 possible to call ll_revalidate_it() twice for the same dentry
2583 even if first occurence returned success. This fix changes semantic
2584 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2585 reference on a request referred from the intent.
2586 ll_intent_release() then releases such a request.
2589 Frequency : rare, normally benchmark loads only
2591 Description: unlinked inodes were kept in memory on the client
2592 Details : If a client is repeatedly creating and unlinking files it
2593 can accumulate a lot of stale inodes in the inode slab cache.
2594 If there is no other client load running this can cause the
2595 client node to run out of memory. Instead flush old inodes
2596 from client cache that have the same inode number as a new inode.
2599 Frequency : SLES9 2.6.5 kernel and long filenames only
2600 Bugzilla : 9969, 10379
2601 Description: utime reports stale NFS file handle
2602 Details : SLES9 uses out-of-dentry names in some cases, which confused
2603 the lustre dentry revalidation. Change it to always use the
2607 Frequency : rare, unless heavy write-truncate concurrency is continuous
2608 Bugzilla : 4180, 6984, 7171, 9963, 9331
2609 Description: OST becomes very slow and/or deadlocked during object unlink
2610 Details : filter_destroy() was holding onto the parent directory lock
2611 while truncating+unlinking objects. For very large objects this
2612 may block other threads for a long time and slow overall OST
2613 responsiveness. It may also be possible to get a lock ordering
2614 deadlock in this case, or run out of journal credits because of
2615 the combined truncate+unlink. Solution is to do object truncate
2616 first in one transaction without parent lock, and then do the
2617 final unlink in a new transaction with the parent lock. This
2618 reduces the lock hold time dramatically.
2621 Frequency : rare, 2.4 kernels only
2623 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2624 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2625 before unmounting their filesystem in order to avoid deadlock.
2626 The kernel_locked() function in 2.4 kernels only checks whether
2627 the kernel lock is held, not whether it is this process that is
2628 holding it as 2.6 kernels do.
2633 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2634 Details : The client ptlrpc code may be trying to reconnect to a down
2635 server before a previous connection attempt has timed out.
2636 Increase the reconnect interval to be longer than the connection
2637 timeout interval to avoid sending duplicate connections to
2641 Frequency : echo_client brw_test command
2643 Description: fix echo_client to work with OST preallocated code
2644 Details : OST preallocation code (5137) didn't take echo_client IO path
2645 into account: echo_client calls filter methods outside of any
2646 OST thread and, hence, there is no per-thread preallocated
2647 pages and buffers to use. Solution: hijack pga pages for IO. As
2648 a byproduct, this avoids unnecessary data copying.
2652 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2653 Description: Client can oops in mdc_commit_close() after open replay
2654 Details : It was possible for the MDS to return an open request with no
2655 transaction number in mds_finish_transno() if the client was
2656 evicted, but without actually returning an error. Clients
2657 would later try to replay that open and may trip an assertion
2658 Simplify the client close codepath, and always return an error
2659 from the MDS in case the open is not successful.
2662 Frequency : rare, 2.6 OSTs only
2664 Description: OST may deadlock under high load on fragmented files
2665 Details : If there was a heavy load and highly-fragmented OST filesystems
2666 it was possible to have all the OST threads deadlock waiting on
2667 allocation of biovecs, because the biovecs were not released
2668 until the entire RPC IO was completed. Instead, release biovecs
2669 as soon as they are complete to ensure forward IO progress.
2671 Severity : enhancement
2673 Description: Support for specifying external journal device at mount
2674 Details : If an OST or MDS device is formatted with an external journal
2675 device, this device major/minor is stored in the ext3 superblock
2676 and may not be valid for failover. Allow detecting and
2677 specifying the external journal at mount time.
2682 Description: Mounting an MDS with pending unlinked files may cause oops
2683 Details : target_finish_recovery() calls mds_postrecov() which returned
2684 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2685 considers this an error and immediately begins cleaning up the
2686 lov, just after starting the mds_lov process
2688 Severity : enhancement
2690 Description: Implement 'lfs df' to report actual free space on per-OST basis
2691 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2692 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2693 usage of objects; '-h' to report in human readable format.
2695 ------------------------------------------------------------------------------
2697 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2704 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2705 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2706 at startup during orphan unlinking. Instead, skip buddy update
2707 before setup, as it will be regenerated after recovery is complete.
2712 Description: If an OST is inactive, its locks might reference stale inodes.
2713 Details : lov_change_cbdata() must iterate over all namespaces, even if
2714 they are inactive to clear inode references from the lock.
2716 Severity : enhancement
2717 Frequency : occasional, if non-standard max_dirty_mb used
2719 Description: Client will block write RPCs if not enough grant
2720 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2721 then the client will block writes while waiting for another RPC
2722 to complete instead of consuming its dirty limit. With change
2723 we get improved performance when max_dirty_mb is small.
2725 Severity : enhancement
2726 Bugzilla : 3389, 6253
2727 Description: Add support for supplementary groups on the MDS.
2728 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2729 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2730 lookups for user supplementary groups into a cache.
2734 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2735 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2736 for liblustre O_EXCL is enforced by MDS.
2740 Description: readdir never returns NULL in liblustre.
2741 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2742 dentry in struct dirent.
2746 Frequency : liblustre only, depends on application IO pattern
2747 Description: liblustre clients evicted if not contacting servers
2748 Details : Don't put liblustre clients into the ping_evictor list, so
2749 they will not be evicted by the pinger ever.
2751 Severity : enhancement
2753 Description: Add ability to evict clients by NID from MDS.
2754 Details : By echoing "nid:$NID" string into
2755 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2756 $NID would be instantly evicted from this MDS and from all active
2757 OSTs connected to it.
2761 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2762 Details : lookup_it_finish() used to query file size from OSTs that was not
2767 Description: service threads change working directory to that of init
2768 Details : Starting lustre service threads may pin the working directory
2769 of the parent thread, making that filesystem busy. Threads
2770 now change to the working directory of init to avoid this.
2774 Frequency : during shutdown only
2775 Description: shutdown with a failed MDS or OST can cause unmount to hang
2776 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2781 Frequency : 2.6 only
2782 Description: chmod/chown may include an extra supplementary group
2783 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2784 supplementary group and if none is specified this is used.
2787 Bugzilla : 5479 (6816)
2789 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2790 Details : If lookup is in progress on a file that is unlinked we might try
2791 to revalidate the inode and fail in revalidate after lookup is
2792 complete and ll_file_open() enqueues the open again but
2793 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2796 Frequency : always, if lconf --abort_recovery used
2798 Description: lconf --abort_recovery fails with 'Operation not supported'
2799 Details : lconf was attempting to abort recovery on the MDT device and not
2802 ------------------------------------------------------------------------------
2804 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2809 Frequency : rare (only unsupported configurations with a node running as an
2811 Bugzilla : 6514, 5137
2812 Description: Mounting a Lustre file system on a node running as an OST could
2814 Details : OSTs now preallocates memory needed to write out data at
2815 startup, instead of when needed, to avoid having to
2816 allocate memory in possibly low memory situations.
2817 Specifically, if the file system is mounted on on OST,
2818 memory pressure could force it to try to write out data,
2819 which it needed to allocate memory to do. Due to the low
2820 memory, it would be unable to do so and the node would
2821 become unresponsive.
2823 Severity : enhancement
2825 Description: Addition of lconf --service command line option
2826 Details : lconf now accepts a '--service <arg>' option, which is
2827 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2829 Severity : enhancement
2831 Description: Failover mode is now the default for OSTs.
2832 Details : By default, OSTs will now run in failover mode. To return to
2833 the old behaviour, add '--failout' to the lmc line for OSTs.
2835 Severity : enhancement
2837 Description: Health checks are now provided for MDS and OSTs
2838 Details : Additional detailed health check information on MSD and OSTs
2839 is now provided through the procfs health_check value.
2842 Frequency : occasional, depends on IO load
2844 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2845 after numerous create/delete cycles
2846 Details : The ext3 inode allocation policy would not allocate new inodes
2847 very well on the OSTs because there are no new directories
2848 being created. Instead we look for groups with free space if
2849 the parent directories are nearly full.
2854 Description: Network or server problems during mount may cause partially
2855 mounted clients instead of returning an error.
2856 Details : The config llog parsing code may overwrite the error return
2857 code during mount error handling, returning success instead
2863 Description: MDS can fail to allocate large reply buffers
2864 Details : After long uptimes the MDS can fail to allocate large reply
2865 buffers (e.g. zconf client mount config records) due to memory
2866 fragmentation or consumption by the buffer cache. Preallocate
2867 some large reply buffers so that these replies can be sent even
2868 under memory pressure.
2872 Frequency : rare (liblustre)
2873 Description: fsx running with liblustre complained that using truncate() to
2874 extend the file doesn't work. This patch corrects that issue.
2875 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2876 ATTR_SIZE and lsm use in llu_setattr_raw.
2880 Frequency : rare, only 2.6 kernels
2881 Description: Unusual file access patterns on the MDS may result in inode
2882 data being lost in very rare circumstances.
2883 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2884 code in the 2.6 kernel caused the inode and/or EA data not to
2885 be read from disk, causing single-file corruption.
2889 Frequency : rare, only 2.6 filesystems using extents
2890 Description: Heavy concurrent write and delete load may cause data corruption.
2891 Details : It was possible under high-load situations to have an extent
2892 metadata block in the block device cache from a just-unlinked
2893 file overwrite a newly-allocated data block. We now unmap any
2894 metadata buffers that alias just-allocated data blocks.
2898 Frequency : filesystems with default stripe_count larger than 77
2899 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2900 Details : lconf specifies an inode size of 4096 bytes when the default
2901 stripe_count is larger than 77. This conflicts with the default
2902 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2903 this case to avoid pinning too much memory for large EAs.
2905 ------------------------------------------------------------------------------
2907 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2912 Frequency : rare (extremely heavy IO load with hundreds of clients)
2914 Description: Client is evicted, gets IO error writing to file
2915 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2916 caused clients to be evicted for AST timeouts. The fixes in
2917 bug 5192 mean we no longer need to have such short AST timeouts
2918 so ldlm_timeout has been increased.
2921 Frequency : occasional during --force or --failover shutdown under load
2922 Bugzilla : 5949, 4834
2923 Description: Server oops/LBUG if stopped with --force or --failover under load
2924 Details : a collection of import/export refcount and cleanup ordering
2925 issues fixed for safer force cleanup
2928 Frequency : only filesystems larger than 120 OSTs
2929 Bugzilla : 5990, 6223
2930 Description: lfs getstripe would oops on a very large filesystem
2931 Details : lov_getconfig used kfree on vmalloc'd memory
2934 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2935 Bugzilla : 6242, 6243
2936 Description: reading from files that had been truncated to a non-zero size
2937 but never opened returned no data
2938 Details : ll_file_read() reads zeros from no-object files to EOF
2943 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2944 Details : There's little harm in aborting MDS/OSS recovery and letting it
2945 try again, so I removed the LASSERT and return an error instead.
2947 Severity : enhancement
2949 Description: New debugging infrastructure for tracking down data corruption
2950 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2951 (b) cover more of the client-side code path, and (c) try to narrow
2952 down where problems occurred
2956 Bugzilla : 3819, 4364, 4397, 6313
2957 Description: Racing close and eviction MDS could cause assertion in mds_close
2958 Details : It was possible to get multiple mfd references during close and
2959 client eviction, leading to one thread referencing a freed mfd.
2961 Severity: : enhancement
2962 Bugzilla : 3262, 6359
2963 Description: Attempts to reconnect to servers are now more aggressive.
2964 Details : This builds on the enhanced upcall-less recovery that was added
2965 in 1.4.2. When trying to reconnect to servers, clients will
2966 now try each server in the failover group every 10 seconds. By
2967 default, clients would previously try one server every 25 seconds.
2972 Description: After recovery, certain operations trigger a failed
2973 assertion on a client.
2974 Details : Failing over an mds, using lconf -d --failover, while a
2975 client was doing a readdir() call would cause the client to
2976 LBUG after recovery completed and the readdir() was resent.
2978 Severity : enhancement
2980 Description: Default groups are now added by lconf
2981 Details : You can now run lconf --group <servicename> without having to
2982 manually add groups with lmc.
2985 Frequency : occasional
2987 Description: Nodes with an elan id of 0 trigger a failed assertion
2990 Frequency : always when accessing e.g. tty/console device nodes
2992 Description: tty and some other devices nodes cannot be used on lustre
2993 Details : file's private_data field is used by device data and lustre
2994 values in there got lost. New field was added to struct file to
2995 store fs-specific private data.
2998 Frequency : when exporting Lustre via NFS
3000 Description: NFSD failed occasionally when looking up a path component
3001 Details : NFSD is looking up ".." which was broken in ext3 directories
3002 that had grown large enough to become hashed.
3005 Frequency : Clusters with multiple interfaces not on the same subnet
3007 Description: Nodes will repeatedly try to reconnect to an interface which it
3008 cannot reach and report an error to the log.
3009 Details : Extra peer list entries will be created by lconf with some peers
3010 unreachable. lconf now validates the peer before adding it.
3013 Frequency : Only if a default stripe is set on the filesystem root.
3015 Description: Setting a default stripe on the filesystem root prevented the
3016 filesystem from being remounted.
3017 Details : The client was sending extra request flags in the root getattr
3018 request and did not allocate a reply buffer for the dir EA.
3021 Frequency : occasional, higher if lots of files are accessed by one client
3022 Bugzilla : 6159, 6097
3023 Description: Client trips assertion regarding lsm mismatch/magic
3024 Details : While revalidating inodes the VFS looks up inodes with ifind()
3025 and in rare cases can find an inode that is being freed.
3026 The ll_test_inode() code will free the lsm during ifind()
3027 when it finds an existing inode and then the VFS later attaches
3028 this free lsm to a new inode.
3032 Bugzilla : 6422, 7030
3033 Description: MDS deadlock between mkdir and client eviction
3034 Details : Creating a new file via mkdir or mknod (starting a transaction
3035 and getting the ns lock) can deadlock with client eviction
3036 (gets ns lock and trying to finish a synchronous transaction).
3039 Frequency : occasional
3040 Description: While starting a server, the fsfilt_ext3 module could not be
3042 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3043 kernels. Previously, lconf would still use the ext3 name
3044 when trying to load modules. Now, it will correctly use
3045 ext3 on 2.4 and ldiskfs on 2.6.
3047 Severity : enhancement
3048 Description: The default stripe count has been changed to 1
3049 Details : The interpretation of the default stripe count (0, to lfs
3050 or lmc) has been changed to mean striping across a single
3051 OST, rather than all available. For general usage we have
3052 found a stripe count of 1 or 2 works best.
3054 Severity : enhancement
3055 Description: Add support for compiling against Cray portals.
3056 Details : Conditional compiling for some areas that are different
3060 Frequency : occasional
3061 Bugzilla : 6409, 6834
3062 Description: Creating files with an explicit stripe count may lead to
3063 a failed assertion on the MDS
3064 Details : If some OSTs are full or unavailable, creating files may
3065 trigger a failed assertion on the MDS. Now, Lustre will
3066 try to use other servers or return an error to the
3070 Frequency : occasional
3072 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3073 caused lock timeout during readahead (since 1.4.2).
3074 Details : Processes doing readahead might match a lock that hasn't been
3075 granted yet if there are overlapping and conflicting lock
3076 requests. The readahead process waits on ungranted lock
3077 (original lock is CBPENDING), while OST waits for that process
3078 to cancel CBPENDING read lock and eventually evicts client.
3080 Severity : enhancement
3082 Description: Initial enabling of flock support for clients
3083 Details : Implements fcntl advisory locking and file status functions.
3084 This feature is provided as an optional mount flag (default
3085 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3086 locking are implemented yet, and those that are are not guaranteed
3087 to be completely correct in production environments.
3088 mount -t lustre -o [flock|noflock] ...
3091 Frequency : occasional
3093 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3094 assertion in the ext3 JBD (journaling) layer.
3095 Details : The b_committed_data struct is protected by the big kernel lock
3096 in 2.4 kernels, serializing journal_commit_transaction() and
3097 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3098 access to this struct. In 2.6 kernels there is finer grained
3099 locking to improve SMP performance of the JBD layer.
3103 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3108 Description: merge and cleanup kernel patches.
3111 -----------------------------------------------------------------------------
3113 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3115 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3116 versions of Lustre. Please update all servers and clients to
3117 version 1.4.2 or later at the same time. You must also run
3118 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3119 to update the configuration logs.
3121 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3122 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3123 - swab llog records of type '0' so we get proper header size/idx (5861)
3124 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3125 - fix rename of one directory over another leaking an inode (5953)
3126 - avoid SetPageDirty on 2.6 (5981)
3127 - don't re-add just-being-destroyed locks to the waiting list (5653)
3128 - when creating new directories, inherit the parent's custom
3129 striping settings if present parent (3048)
3130 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3131 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3132 - don't hold client locks on temporary worklist from l_lru (5666)
3133 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3134 - automatically evict dead clients (5921)
3135 - Update file size properly in create+truncate+fstat case (6196)
3136 - Do not unhash mountpoint dentries, do not allow removal of
3138 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3139 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3140 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3141 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3142 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3143 - Correctly handle reads of files with no objects (6243)
3144 - lctl recover will also mark a device active if deactivate used (5933)
3146 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3147 - allow --write-conf on an MDS with different nettype than client (5619)
3148 - don't write config llogs to MDS for mounts not from that MDS (5617)
3149 - lconf should create multiple TCP connections from a client (5201)
3150 - init scripts are now turned off by default; run chkconfig --on
3151 lustre and chkconfig --on lustrefs to use them
3152 - upcalls are no longer needed for clients to recover to failover
3154 - add --abort-recovery option to lconf to abort recovery on device
3156 - add support for an arbitrary number of OSTs (3026)
3157 - Quota support protocol changes.
3158 - forward compatibility changes to wire structs (6007)
3159 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3160 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3161 - improved "lustre status" script
3162 - initialize blocksize for non-regular files (6062)
3163 - added --disable-server and --disable-client configure options (5782)
3164 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3165 - Vanilla 2.4.29 support
3166 - increase maximum number of obd devices to 520 (6242)
3167 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3168 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3169 SLES 9 2.6.5 kernels
3170 - limit stripes per file to 160 (the maximum EA size) (6093)
3172 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3175 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3176 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3177 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3178 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3179 - small ext3 extents cleanups and fixes (5733)
3180 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3181 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3182 - avoid needless client->OST connect, fix handle mismatch (5317)
3183 - fix DLM error path that led to out-of-sync client, long delays (5779)
3184 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3185 - fix several locking issues related to i_size (5492,5624,5654,5672)
3186 - don't move pending lock onto export if it is already evicted (5683)
3187 - fix kernel oops when creating .foo in unlinked directory (5548)
3188 - fix deadlock in obdfilter statistics vs. object create (5811)
3189 - use time_{before,after} to avoid timer jiffies wrap (5882)
3190 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3191 - Do not leak request if server was not able to process it (5154)
3192 - If mds_open unable to find parent dir, make that negative lookup(5154)
3193 - don't create new directories with extent-mapping (5909, 5936)
3195 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3196 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3197 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3198 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3199 - support for RHEL4 kernel on i686 (5773)
3200 - provide error messages when incompatible logs are encountered (5898)
3202 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3203 * version 1.4.0.10 (1.4.1 release candidate 1)
3205 - don't keep a lock reference when lock is not granted (4238)
3206 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3207 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3208 - drop import inflight refcount on signal_completed_replay error (5255)
3209 - unlock page after async write error during send (3677)
3210 - handle missing objects in filter_preprw_read properly (5265)
3211 - no transno return for symlink open, don't save no-trasno open (3440)
3212 - don't try to complete elan receive that already failed (4012)
3213 - free RPC server reply state on error (5406)
3214 - clean up thread from ptlrpc_start_thread() on error (5160)
3215 - readahead could read extra page into cache that wasn't ejected (5388)
3216 - prevent races in class_attach/setup/cleanup/detach (5260)
3217 - don't dereference de->d_inode after l_dput of de (5458)
3218 - use "int" for stripe value returned from lock_to_stripe (5544)
3219 - mballoc allocation and error-checking fixes in 2.6 (5504)
3220 - block device patches to fix I/O request sizes in 2.6 (5482)
3221 - look up hostnames for IB nals (5602)
3222 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3223 - don't start multiple acceptors for the same port (5277)
3224 - fix incorrect LASSERT in mds_getattr_name (5635)
3225 - export a proc file for general "ping" checking (5628)
3226 - fix "lfs check" to not block when the MDS is down (5628)
3228 - service request history (4965)
3229 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3230 - create an "evict_client" /proc entry on OSTs, like the MDS has
3231 - fix mount usage message, return errors per mount(8) (5168)
3232 - change grep [] to grep "[]" in tests so they work in more UMLs
3233 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3234 - remove ancient LOV_MAGIC_V0 EA support (5047)
3235 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3236 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3237 - allow readahead allocations to fail when low on memory (5383)
3238 - mmap locking landed again, after considerable improvement (2828)
3239 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3241 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3244 - send OST transaction number in read/write reply to free req (4966)
3245 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3246 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3247 - the watchdog thread now runs as interruptible (5246)
3248 - flock/lockf fixes (but it's still disabled, pending 5135)
3249 - don't use EXT3 constants in llite code (5094)
3250 - memory shortage at startup could cause assertion (5176)
3252 - reorganization of lov code
3253 - single portals codebase
3255 - add extents/mballoc support (5025)
3256 - direct I/O reads in the obdfilter (4048)
3257 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3259 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3262 - send OST transaction number in read/write reply to free req (4966)
3263 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3264 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3265 - the watchdog thread now runs as interruptible (5246)
3266 - handle missing objects in filter_preprw_read properly (5265)
3267 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3268 - drop import inflight refcount on signal_completed_replay error (5255)
3269 - unlock page after async write error during send (3677)
3270 - return original error code on reconstructed replies (3761)
3271 - no transno return for symlink open, don't save no-trasno open (3440)
3273 - add pid to ldlm debugging output (4922)
3274 - bump the watchdog timeouts -- we can't handle 30sec yet
3275 - extra debugging for orphan dentry/inode bug (5259)
3277 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3280 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3281 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3282 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3283 - fix NULL dereference in /proc/sys/portals/routes (4827)
3284 - allow failed mdc_close() operations to be interrupted (4561)
3285 - stop precreate on OST before MDS would time out on it (4778)
3286 - don't send partial-page writes before EOF from client (4410)
3287 - discard client grant for sub-page writes on large-page clients (4520)
3288 - don't free dentries not owned by NFS code, check generation (4806)
3289 - fix lsm leak if mds_create_objects() fails (4801)
3290 - limit debug_daemon file size, always print CERROR messages (4789)
3291 - use transno after validating reply (3892)
3292 - process timed out requests if import state changes (3754)
3293 - update mtime on OST during writes, return in glimpse (4829)
3294 - add mkfsoptions to LDAP (4679)
3295 - use ->max_readahead method instead of zapping global ra (5039)
3296 - don't interrupt __l_wait_event() during strace
3298 - add software watchdogs to catch hung threads quickly (4941)
3299 - make lustrefs init script start after nfs is mounted
3300 - fix CWARN/ERROR duplication (4930)
3301 - return async write errors to application if possible (2248)
3302 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3303 - print NAL number in %x format (4645)
3304 - update barely-supported suse-2.4.21-171 series (4842)
3305 - support for sles 9 %post scripts
3306 - support for building 2.6 kernel-source packages
3307 - support for sles km_* packages
3309 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3312 - ignore -ENOENT errors in osc_destroy (3639)
3313 - notify osc create thread that OSC is being cleaned up (4600)
3314 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3315 - reconstruct ost_handle() like mds_handle() (4657)
3316 - create a new thread to do import eviction to avoid deadlock (3969)
3317 - let lconf resolve symlinked-to devices (4629)
3318 - don't unlink "objects" from directory with default EA (4554)
3319 - hold socknal file ref over connect in case target is down (4394)
3320 - allow more than 32000 subdirectories in a single directory (3244)
3321 - fix blocks count for O_DIRECT writes (3751)
3322 - OST returns ENOSPC from object create when no space left (4539)
3323 - don't send truncate RPC if file size isn't changing (4410)
3324 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3325 - bind to privileged port in socknal and tcpnal (3689)
3327 - rate limit CERROR/CWARN console message to avoid overload (4519)
3328 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3329 - basic mmap support (3918)
3330 - kernel patch series update from b1_4 (4711)
3332 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3335 - avoid crash during MDS cleanup with OST shut down (2775)
3336 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3337 - don't use bad inodes on the MDS (3744)
3338 - dynamic object preallocation to improve recovery speed (4236)
3339 - don't hold spinlock over lock dumping or change debug flags (4401)
3340 - don't zero obd_dev when it is force cleaned (3651)
3341 - print grants to console if they go negative (4431)
3342 - "lctl deactivate" will stop automatic recovery attempts (3406)
3343 - look for existing locks in ldlm_handle_enqueue() (3764)
3344 - don't resolve lock handle twice in recovery avoiding race (4401)
3345 - revalidate should check working dir is a directory (4134)
3347 - don't always mark "slow" obdfilter messages as errors (4418)
3349 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3352 - don't close LustreDB during write_conf until it is done (3860)
3353 - fix typo in lconf for_each_profile (3821)
3354 - allow dumping logs from multiple threads at one time (3820)
3355 - don't allow multiple threads in OSC recovery (3812)
3356 - fix debug_size parameters (3864)
3357 - fix mds_postrecov to initialize import for llog ctxt (3121)
3358 - replace config semaphore with spinlock (3306)
3359 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3360 - don't allow enqueue to complete on a destroyed export (3822)
3361 - down write_lock before checking llog header bitmap (3825)
3362 - recover from lock replay timeout (3764)
3363 - up llog sem before sending rpc (3652)
3364 - reduce ns lock hold times when setting kms (3267)
3365 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3366 - fix NULL deref and obd_dev leak on setup error (3312)
3367 - replace some LBUG about llog ops with error handling (3841)
3368 - don't match INVALID dentries from d_lookup and spin (3784)
3369 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3370 - fix invalid assertion in ptlrpc_set_wait (3880)
3372 - add libwrap support for the TCP acceptor (3996)
3373 - add /proc/sys/portals/routes for non-root route listing (3994)
3374 - allow setting MDS UUID in .xml (2580)
3375 - print the stack of a process that LBUGs (4228)
3377 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3380 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3381 - make sure to unset replay flag from failed open requests (3440)
3382 - if default stripe count is 0, use OST count for inode size (3636)
3383 - update parent mtime/ctime on client for create/unlink (2611)
3384 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3385 - free recovery state on server during a forced cleanup (3571)
3386 - unregister_reply for resent reqs (3063)
3387 - loop back devices mounting and status check on 2.6 (3563)
3388 - fix resource-creation race that can provoke i_size == 0 (3513)
3389 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3390 - more debugging for page-accounting assertion (3746)
3391 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3392 - avoid deadlock after precreation failure (3758)
3393 - fix race and lock order deadlock in orphan handling (3450, 3750)
3394 - add validity checks when grabbing inodes from l_ast_data (3599)
3396 - add /proc/.../recovery_status to obdfilter (3428)
3397 - lightweight CDEBUG infrastructure, debug daemon (3668)
3398 - change default OSC RPC parameters to be better on small clusters
3399 - turn off OST read cache for files smaller than 32MB
3400 - install man pages and include them in rpms (3100)
3401 - add new init script for (un)mounting lustre filesystems (2593)
3402 - run chkconfig in %post for init scripts (3701)
3403 - drop scimac NAL (unmaintained)
3405 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3408 - clean kiobufs before and after use (3485)
3409 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3410 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3411 - change init script to fail more clearly if not run as root (1528)
3412 - allow clients to reconnect during replay (1742)
3413 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3414 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3415 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3416 - fix cond_resched (3554)
3417 - only evict unfinished clients after recovery (3515)
3418 - allow bulk resend, prevent data loss (3570)
3419 - dynamic ptlrpc request buffer allocation (2102)
3420 - don't allow unlinking open directory if it isn't empty (2904)
3421 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3422 - remove extraneous obd dereference causing LASSERT failure (3334)
3423 - don't use get_cycles() when creating temp. files on the mds (3156)
3424 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3425 - handle EEXIST for set-stripe, set proper directory name (3336)
3427 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3428 - fix ksocknal_fmb_callback() error messages (2918)
3430 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3433 - don't copy lvb into (possibly NULL) reply on error (2983)
3434 - don't deref dentry after dput, don't free lvb on error (2922)
3435 - use the kms to determine writeback rpc length (2947)
3436 - increment oti_logcookies when osc is inactive (2948)
3437 - update client's i_blocks count via lvb messages (2543)
3438 - handle intent open/close of special files properly (1557)
3439 - mount MDS with errors=remount-ro, like obdfilter (2009)
3440 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3441 - don't use cancelling-locks' kms values (2947)
3442 - use highest lock extent for kms, not last one (2925)
3443 - don't dereference ERR_PTR() dentry in error handling path (3107)
3444 - fix thread race in portals_debug_dumplog() (3122)
3445 - create lprocfs device entries at setup instead of at attach (1519)
3446 - common AST error handler, don't evict client on completion race (3145)
3447 - zero nameidata in detach_mnt in 2.6 (3118)
3448 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3449 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3450 - initialize RPC timeout timer earlier for 2.6 (3219)
3451 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3452 - print nal/nid for unknown nid (3258)
3453 - additional checks for oscc recovery before doing precreate (3284)
3454 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3455 - don't crash in mdc_close for bad permissions on open (3285)
3456 - zero i_rdev for non-device files (3147)
3457 - clear page->private before handing to FS, better assertion (3119)
3458 - tune the read pipeline (3236)
3459 - fix incorrect decref of invalidated dentry (2350)
3460 - provide read-ahead stats and refine rpc in flight stats (3328)
3461 - don't hold journal transaction open across create RPC (3313)
3462 - update atime on MDS at close time (3265)
3463 - close LDAP connection when recovering to avoid server load (3315)
3464 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3465 - don't leak open file on MDS after open resend (3325)
3466 - serialize filter_precreate and filter_destroy_precreated (3329)
3467 - loop device shouldn't call sync_dev() for nul device (3092)
3468 - clear page cache after eviction (2766)
3469 - resynchronize MDS->OST in background (2824)
3470 - refuse to mount the same filesystem twice on same mountpoint (3394)
3471 - allow llmount to create routes for mounting behind routers (3320)
3472 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3473 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3474 - fix rare problem with rename on htree directories (3417)
3476 - allow default OST striping configuration per directory (1414)
3477 - fix compilation for qswnal for 2.6 kernels (3125)
3478 - increase maximum number of MDS request buffers for large systems
3479 - change liblustreapi to be useful for external progs like lfsck (3098)
3480 - increase local configuration timeout for slow disks (3353)
3481 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3483 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3486 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3487 - don't overwrite extent policy data in reply if lock was blocked (2901)
3488 - drop filter export grants atomically with removal from device (2663)
3489 - del obd_self_export from work_list in class_disconnect_exports (2908)
3490 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3491 - swab reply message in mdc_close, other PPC fixes (2464)
3492 - fix destroying of named logs (2325)
3493 - overwrite old logs when running lconf --write_conf (2264)
3494 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3495 - fix race in target_handle_connect (2898)
3496 - mds_reint_create() should take same inode create lock (2926)
3497 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3498 - don't close files for self_export to avoid uninitialized obd (2936)
3499 - allow MDS with the same name as client node (2939)
3500 - hold dentry reference for closed log files for unlink (2325)
3501 - reserve space for all logs during transactions (2059)
3502 - don't evict page beyond end of stripe extent (2925)
3503 - don't oops on a deleted current working directory (2399)
3504 - handle hard links to targets without a parent properly (2517)
3505 - don't dereference NULL lock when racing during eviction (2867)
3506 - don't grow lock extents when lots of conflicting locks (2919)
3508 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3511 - account for cache space usage on clients to avoid data loss (974)
3512 - lfsck support in lustre kernel code (2349)
3513 - reduce journal credits needed for BRW writes (2370)
3514 - orphan handling to avoid losing space on client/server crashes
3515 - ptlrpcd can be blocked, stopping ALL progress (2477)
3516 - use lock value blocks to assist in proper KMS, faster stat (1021)
3517 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3518 - recovery for initial connections (2355)
3519 - fixes for mds_cleanup_orphans (1934)
3520 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3521 - block all file creations until orphan recovery completes (1901)
3522 - client remove rq_connection from request struct (2423)
3523 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3524 - recovery timer race (2670)
3525 - mdc_close recovey bug (2532)
3526 - ptlrpc cleanup bug (2710)
3527 - mds timeout on local locks (2588)
3528 - namespace lock held during RPCs (2431)
3529 - handle interrupted sync write properly (2503)
3530 - don't try to handle a message that hasn't been replied to (2699)
3531 - client assert failure during cleanup after abort recovery (2701)
3532 - leak mdc device after failed mount (2712)
3533 - ptlrpc_check_set allows timedout requests to complete (2714)
3534 - wait for inflight reqs when ptlrpcd finishes (2710)
3535 - make sure unregistered services are removed from the srv_list
3536 - reset bulk XID's when resending them (caught by 1138 test)
3537 - unregister_bulk after timeout
3538 - fix lconf error (2694)
3539 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3540 - readahead locks pages, leaves pending causing memory pressure (2673)
3541 - increase OST request buffers to 4096 on large machines (2729)
3542 - fix up permission of existing directories in simple_mkdir (2661)
3543 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3544 - don't assign transno to errored transactions (2742)
3545 - don't delete objects on OST if given a bogus objid from MDS (2751)
3546 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3547 - if rq_no_resend, then timeout request after recovery (2432)
3548 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3549 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3550 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3551 - MDS recovery shouldn't delete objects that it has given out (2730)
3552 - if enqueue arrives after completion, don't clobber LVB (2819)
3553 - don't unlock pages twice when trigger_group_io returns error (2814)
3554 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3555 - don't write pages to disk if there was an error (1450)
3556 - don't ping imports that have recovery disabled (2676)
3557 - take buffered bytes into account when balancing socknal conn (2817)
3558 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3559 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3560 - remove little-endian swabbing of llog records (1987)
3561 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3562 - retry reposting request buffers if they fail (1191)
3563 - grow extent at grant time to avoid granting a revoked lock (2809)
3564 - lock revoke doesn't evict page if covered by a second lock (2765)
3565 - disable VM readahead to avoid reading outside lock extents (2805)
3567 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3568 - updated kernel patches for hp-2.4.20 kernel (2681)
3570 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3573 - fix truncated write corruption (2366)
3574 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3576 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3577 - ensure len(uuid) < 37 in lmc (1171)
3578 - fix ia64 OOPS in llog_test (2255)
3579 - zero end of page at obdfilter for partial page writes (2648)
3580 - don't leave stale dentries around after renames (bug 2428)
3581 - fix timeouts when evicting a client with a single lock held (2642)
3582 - set deadline for the initial HELLO message to drain (2634)
3583 - print out dotted-quad IP addresses in the socknal (2302)
3585 - additional debugging for MDS client eviction problem (2443)
3586 - fix mkfsoptions support for osts (2603, 2604)
3588 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3591 - add series for the vanilla 2.6.0 kernel
3592 - add series for the vanilla 2.4.24 kernel
3593 - add series for a cray x86/64 UL kernel drop
3594 - fix xattr patches for the vanilla 2.4.19 series
3596 - generate true UUIDs in lmc (1171)
3597 - have portals stack dumping break in UML (2466)
3598 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3599 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3600 - wake callback waiting threads on client eviction (2460)
3601 - Add --ptldebug and --subsystem to lmc (1719)
3602 - update assertion to allow safe interrupt allocation
3603 - set rq_no_resend for cancel requests (2432)
3604 - recalculate ptlrpcd timeout after resend (2494)
3605 - call vfs_rmdir when removing pending directories (2368)
3606 - fix renaming a file to itself (2429)
3607 - lmc creates a default one-stripe lov (2454)
3608 - expand procfs space to handle large clusters (2326)
3609 - increase UML stack to avoid overflow
3610 - update lconf's list of debug and subsystem masks
3611 - fix lfs find --obd (2510)
3612 - /proc tunable for disabling filter read caching (2591)
3613 - stop rpm packages from altering slapd.conf (2301)
3614 - disable nagle in the socknal under 0conf (2578)
3615 - choose mds inode size based on stripe count (2572)
3616 - fix kernel-source rpm problems (2516)
3618 - add --disable-doc to avoid pdf generation (2421)
3619 - update documentation, tests, type-os, comments
3620 - avoid format warnings on ia64
3621 - remove the TOE NAL
3622 - tiny code cleanups by removing unused fields
3624 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3627 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3628 - remove the most blatant lies from BUILDING file (2371)
3629 - change default debug level to reasonable production setting
3630 - reduce client side cache size to reduce cache flush time
3631 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3632 - make TCP zerocopy and pinger support enabled by default (2476)
3633 - sync writes completed after process exits caused crashes (2319)
3634 - maintain correct mount count on the MDS (2356)
3635 - backout 1557, because 2316 wasn't really fixed
3636 - better file I/O statistics gathering in /proc
3637 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3638 - another kernel patch to fix zero-copy TCP function export
3639 - don't take duplicate lock when processing re-sent getattr (2420)
3640 - lctl uses obd_self_export instead of creating new conn (2353)
3641 - MDS/OST recovery case which requires object creation asserted (2425)
3642 - move lfs from /usr/sbin to /usr/bin in packages
3643 - fix race between mds_client_add and mds_client_free (2417)
3644 - use kmalloc instead of slabs in portals (2430)
3645 - don't create duplicate records when a failover MDS is present (2442)
3646 - remove unnecessary mount age check (2332)
3647 - don't remove directory inodes from locks prematurely (2451)
3648 - don't break if MDS service name is the same as hostname (2103)
3649 - fix races in client write RPC generation when cache full (2482)
3651 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3654 - remove now-unused request->rq_obd (278)
3655 - if an allocation fails, print out how much memory we've used (1933)
3656 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3657 - add the "configurable stack size" patch to most series files (1256)
3658 - ability to write large log records, for 100+ OST configs (2306)
3659 - fix NULL deref when filter_prep fails (2314)
3660 - fix operator precedence error in filter_sync
3661 - dynamic allocation of socknal TX descriptors (2315)
3662 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3663 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3664 - ensure that CWARN messages in Portals always get to the syslog
3665 - __init/__exit are not for prototype decls (ldlm_init/exit)
3666 - x86-64 compile warning fixes
3667 - fix gateway LMC keyword conflict (2318)
3668 - fix MDS lock inversions in getattr/reint paths (1844)
3669 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3670 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3671 - DLM race condition prevented some lock evictions (2328)
3672 - ENOMEM detection and retry on socknal sends (2230)
3673 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3674 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3675 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3676 - better error messages when a client is rejected during recovery (1505)
3677 - avoid cancelling locks which were never granted, after failure (2330)
3678 - fix i_sem/journal inversion in mds_client_add (2333)
3679 - fix truncate/getattr lock cycle deadlock (2334)
3680 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3681 - fix two rare exit paths which could leak an l_lock() ref (2321)
3682 - fencepost error in MDS/OST orphan recovery (2226)
3683 - make log record alignment 8 bytes (1988)
3684 - lstripe now fails when requested offset > ost_count (2237)
3685 - ensure that all kernel series have a complete list.h (1607)
3686 - fix crashes in special-file operations (2316)
3687 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3689 - allow configurable automake binary, for testing new versions
3690 - small update to the lfs documentation
3692 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3694 * fix negative export reference count in fsfilt_sync (2312)
3696 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3697 * release candidate 0.9.1
3699 - orphans are moved into the PENDING directory for possible recovery
3700 - replayed opens now open by fid for orphan/rename safety (1042)
3701 - last close of an orphan inode generates a transno (683)
3702 - chdir() and mount() now pin the directory entry (1020)
3703 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3704 - discard very old requests without processing them (1502)
3705 - remove some common, well-understood CERRORs (1505)
3706 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3707 - clear "grant" flags in OST replies until OST grant code lands (1644)
3708 - fix read performance by not clobbering i_blksize on client (1598)
3709 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3710 - make LRU size a /proc tunable, clears locks when reduced (707)
3711 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3712 - server locks take references on exports now (1558)
3713 - build fixes for 2.4.20-rh trees (1663)
3714 - return an error from lov_create if all OSCs are inactive (1751)
3715 - fix import levels when a reconnect happens without a timeout (1597)
3716 - exit early from mds_open if we get a lookup error (1749)
3717 - partial page read at EOF wouldn't wait for disk before sending (1642)
3718 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3719 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3720 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3721 - write/truncate lock inversion (1639)
3722 - Don't auto-load obdclass, portals modules during cleanup (1495)
3723 - fix timestamps from jumping to "now" (1763)
3724 - extra journal assertions (1648)
3725 - add an extra multiunlink test (1771)
3726 - fix read_record/write_record API (1776)
3727 - fix leak of offset_extent, possible incorrect i_size later (1772)
3728 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3729 - Debugging for the kqswnal_get_idle_tx problems (1820)
3730 - Allow recovery to be attempted multiple times (1536)
3731 - Write out MDS last_rcvd file after it is first created (1600)
3732 - Fix tx_descriptor leak in failed transmit situations (1827)
3733 - ext3 journaling fixes for assertion failure after IO error (1871)
3734 - class_export_put() on freed export after completion AST error (1896)
3735 - Fix revalidate looping in VFS (1322)
3736 - Don't access a freed export during MDS_REINT timeout (1521)
3737 - Add open-unlink recovery support on the MDS (1673,1764)
3738 - Return an error if no MDS data was read from last_rcvd (1946)
3739 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3740 - Don't setup a disk device that doesn't match exported UUID (317)
3741 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3742 - avoid committing NULL handle in force close
3743 - local.sh is now a one-stripe LOV configuration
3744 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3745 - fix ext3 htree duplicate directory entry corruption (1516)
3746 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3747 - update to new LOV EA format (2097)
3748 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3750 2003-06-15 Phil Schwan <phil@clusterfs.com>
3753 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3754 - per-import/export recovery handling (958, 931, 959)
3755 - multiple last-rcvd slots, for serving multiple FSes (949)
3756 - connections are again shared between multiple imp/exports (963, 964)
3757 - "umount -f" would hang if any requests needed to be sent (393, 978)
3758 - avoid pinning large req buffer by copying for queued messages (989)
3759 - add "uuid" to "lctl device" command to help upcalls (991)
3760 - "open" RPCs with transnos would confuse recovery counters (1037)
3761 - do proper endian conversion of all wire messages (288, 340, 891)
3762 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3763 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3764 - fix ptlrpc_connection leak in target_handle_connect (1174)
3765 - fix import refcounting bug in OST and MDS cleanup (1134)
3766 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3767 - fix very unlikely obd_types race condition (501)
3768 - remove osc_open hack for echo_client (1187)
3769 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3770 - a failure in read_inode2 leads to deadlock (1139)
3771 - cancel ack-locks as soon as transaction is committed (1072)
3772 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3773 - make sure to commitrw after any preprw to avoid deadlock (1162)
3774 - failing to execute a file in a lustre FS would lock inode (1203)
3775 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3776 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3777 - obd_uuid2tgt would walk off the end of the list (1255)
3778 - on IA64 the osc would give portals incorrect bulk size (1258)
3779 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3780 - fix lock inversion caused by new llite matching code (1282)
3781 - limit the number of dirty pages on a client to 10MB (1286)
3782 - timed out locks were not being corrected cancelled (1289)
3783 - fix O_DIRECT above 4GB on IA-32 (1292)
3784 * major user-visible changes
3785 - fail out/fail over policy now controlled by the upcall (993)
3787 - add OBD_PING to check server availability and failure (954)
3788 - lustre messages are now sent in sending host order (288, 340, 891)
3789 - add eadatalen to MDS getattr reply (340)
3790 - OST read replies may contain second buffer, with per-page status (593)
3792 2003-03-11 Phil Schwan <phil@clusterfs.com>
3795 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3796 - failed open()s could cause deadlock; fixed (867, 869)
3797 - stop cancelling OST locks when files are closed (481)
3798 - overlapping XID spaces caused network corruption (851, 853)
3799 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3800 - setattr_raw added, to do single-RPC, server-side setattrs
3801 - lmc/lconf syntax change for OST UUIDs
3802 - fix crashy race condition between ptlrpc_free_req and osc_close
3803 - don't use request in mdc_enqueue if we hit a timeout (889)
3804 - don't set the inode i_size for regular files from the MDS (896)
3805 - handle out of order completion AST (842)
3806 - don't LBUG if a lock request times out after receiving AST (913)
3807 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3808 - if a bad lock AST arrives, send an error instead of dropping entirely
3809 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3810 - fix leak in bulk IO when only partially completed (899, 900, 926)
3811 - fix O_DIRECT for ia64 (55)
3812 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3813 - C-z after timeout could hang a process forever; fixed (977)
3815 - client-side I/O cache (678, 924, 929, 941, 970)
3817 - READPAGE and SETATTRs which don't take server-side locks get
3820 2003-02-11 Phil Schwan <phil@clusterfs.com>
3823 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3824 - Fix fsfilt_extN_readpage() to read a full page of directory
3825 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3826 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3827 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3828 - Most intent operations were removed, in favour of a new RPC mode
3829 that does a single RPC to the server and bypasses most of the VFS
3830 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3831 - Aggressively cancel local locks on DLM servers
3832 - mds_reint_unlink sends EA to the client if it's the last nlink.
3833 client uses that EA to unlink OST objects.
3834 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3835 - recursive symlinks were fixed (439)
3836 - fixed NULL deref in DEBUG_REQ
3837 - filter_update_lastobjid no longer calls sync, which annoyed extN
3838 - fixed multi-client small-writes to a single file problem (445)
3839 - fixed mtime updates during file writes (607)
3840 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3841 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3842 - fixed Linux OST size reporting problem (444, 656)
3843 - OST now updates object mtime with writes or setattr (607, 619)
3844 - client verifies file size before zeroing page past EOF (445)
3845 - OST now writes last allocated objid to disk with allocation (108)
3846 - LOV on echo now works (409)
3848 - mds_reint_unlink sends a new buffer, with the EA included. this
3849 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3850 set if a regular file was being unlinked, and it was the last link
3851 - use PtlGet from the target for bulk writes (315)
3852 - OST now updates object mtime with writes or setattr (607, 619)
3853 - LDLM now has a grant-time callback to revalidate locked items, if
3855 - Many MDS operations were reorganized to combat race conditions
3857 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3858 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3860 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3863 - Fully reactivate OST imports after reconnection (512, others)
3864 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3865 - More graceful error handling for truncating on dead OST (515)
3866 - Don't error out unless we're actually accessing dead stripes (474)
3867 - Fix garbage sizes when stripes are missing (410)
3868 - LRU counters were broken, causing constant lock purge (433, 432)
3869 - garbage on read from stripes with failed OSTs (441)
3870 - mark OSCs as active before reconnecting during recovery (438)
3871 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3872 - lfind did not preserve OST order in output (443)
3873 - symlinks cause hung clients, incorrect data (439)
3874 - stop dereferencing request after dropping refcount (457)
3875 - don't LASSERT(spin_is_locked) on non-SMP (455)
3876 - fixes for many rename() bugs
3877 - fstat didn't correctly synchronize attributes (399)
3878 - server must handle lock cancellation during blocking AST prep (487)
3879 - bulk descriptors were free()d too soon (511)
3880 - fix paths in lconf, which would load incorrect modules (451, 507)
3881 - fix confusing lconf 'host not found' error message (386)
3882 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3883 - fix race condition in mdc_blocking_ast() for inode access (526)
3884 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3885 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3886 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3887 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3888 - fix lstripe in several ways (kernel side) (527)
3889 - fix request leak in ldlm_cli_enqueue (262)
3890 - incorrect OSC was marked inactive after OST failure
3891 - call mds_fs_cleanup before unmounting filesystem (524)
3892 - fix races between taking ns_lock and ldlm_lock_change_resource
3893 - fix races updating LOV export open file list
3894 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3895 - fix recovery NULL deref in ldlm_cli_cancel_unused
3896 - fix some DLM races by using new hash table for lock handles (419)
3897 - permit the client to specify desired inodes, at replay
3898 - duplicate requests when we queue them for replay reintegration
3899 - fix last_rcvd offset calculation
3900 - sync after each recovered transaction, so we always make progress
3901 - never, not always, ERESTART requests without transnos
3902 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3904 - skip replay if the MDS says that the client is already connected
3905 - don't check for a recovery-enabled export to match lctl's UUID
3906 - don't INC_USE_COUNT for phantom exports
3907 - don't crash when cleaning up phantom exports (567)
3908 - don't double-finish or set replay data for errored mdc_open requests
3909 - abort requests when they time out, so we don't get old replies
3910 - send/receive replies for AST messages again
3911 - if the client says that it doesn't have the lock, cancel it on the
3913 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3914 mark it as destroyed, it will all work out when decref is called
3915 - fix module use counts (22, 581)
3917 - ASTs now expect a reply (server cancels lock on error reply)
3919 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3922 - fix many simultaneous client startup (392)
3923 - fix dentry->d_it clobbering
3924 - credentials weren't being shipped for readdir/getattr operations
3925 - remove invalid assertions triggered during some concurrent MD
3927 - proper Lustre versions added (336, 389)
3928 - fix memory leak for create error case (398)
3929 - fix LOV locking bug that would get cli/srv out of sync
3930 - fix echo client over LOV (409)
3931 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3932 - fix double-O_EXCL intent crash (424)
3933 - avoid sending multiple lock CANCELs (352)
3935 - MDS can do multi-client recovery (modulo bugs in new code)
3937 - many updates, edits, cleanups
3939 2002-11-18 Phil Schwan <phil@clusterfs.com>
3942 - fix null d_it dereference (346)
3943 - fix full OST/dbench hang (333)
3944 - fix permission problem with file removal (286)
3945 - fix removal of OSCs from LOV when they fail
3946 - fix NULL deref during bulk timeout (214)
3947 - fix problems related to multiple filesystems on one MDS (241)
3948 - fixed serious subtle metadata locking bugs
3949 - free locks on clients when inodes are removed due to memory
3951 - fix inode pointer in lock data (285)
3952 - partial support for multiple MDS on a single host (241)
3953 - data locks weren't cancelled at clear_inode time (290, 311)
3954 - intent locks could lead to unbounded lock growth (205)
3955 - added a maximum lock count, an LRU list, and a flusher
3956 - fix multiple rename (365)
3957 - properly abstracted the echo client
3958 - OSC locked 1 byte too many; fixed
3959 - rewrote brw callback code:
3960 - fixed recovery bugs related to LOVs (306)
3961 - fixed too-many-pages-in-one-write crash (191)
3962 - fixed (again) crash in sync_io_timeout (214)
3963 - probably fixed callback-related race (385)
3965 - Add capability to MDS protocol
3966 - LDLM cancellations and callbacks on different portals
3968 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3971 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3972 - defer open object destruction to close time (601981, 138)
3973 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3974 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3975 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3976 - don't call obd_setattr in ll_file_release for destroyed objects
3977 * protocol change to lustre_msg: move |version| and add |flags|
3978 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3979 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3980 * added replay of create, unlink, link and rename operations during
3981 MDS failover; recovery should be much more robust now
3982 * remove failed OSCs from LOVs (only lov_create uses this so far)
3983 * the lustre-HOWTO was brought (more) up to date (582544)
3985 2002-10-23 Phil Schwan <phil@clusterfs.com>
3988 - in-use dentries weren't being reused properly (617851)
3989 - prevent multiple LDLM setup (599178)
3990 - fix LOV size calculations for truncate (617853)
3991 - fix client handling of MDS intent errors (POSIX)
3992 - fix permission bug in lovstripe.c test (624321)
3993 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3994 - truncate past end of file could corrupt data
3995 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3996 - a race in recovery could return ETIMEDOUT to apps (623947)
3997 - building outside the source directory was fixed
3998 * the lustre-HOWTO was brought (more) up to date (582544)
3999 * major progress was made on recovery functionality
4001 2002-10-10 Phil Schwan <phil@clusterfs.com>
4004 - recovery deadlock fix
4005 - rm -rf causes LBUG fix (617817)
4006 - file open by multiple tasks fix (618962)
4007 - directory permissions bugs (602707 and 620007)
4008 - journal_stop fixed with locking (611313)
4009 - O_APPEND failures resolved (618273, perhaps 614459)
4010 - lconf PATH fix (619770)
4011 - IA64 build fix (621450)
4012 - RPC buffer sizes scale with amount of memory
4014 2002-10-01 Phil Schwan <phil@clusterfs.com>
4017 - locks would be cancelled without throwing away data pages,
4018 resulting in inconsistent data (605627)
4019 - inode attributes were not always being refreshed (605627, 612449)
4020 - lconf now continues to cleanup after lctl reports an error
4021 - MDS now enforces user permissions (602707)
4022 - lprocfs cleanup fixed, but not yet enabled (614157)
4023 - fixed infinite server hang, should a client not respond to an AST
4024 - avoid going into recovery if user calls readlink() with a buffer
4025 that's too small (613941)
4026 - AST RPCs no longer require replies (614867) -- this may be changed
4027 - don't crash server if client sends an IOV that's too big (611336)
4028 - fixed lock conversion deadlock (611892)
4029 - fixed the following of symlinks (614622)
4030 * recovery: the server can remove locks from a client that dies, other
4031 clients can make progress
4032 * more extN patch fixes
4033 * compile-time configurable ptlrpc buffer allocations
4035 - collaborative read cache document
4036 - Lustre Lite Performance CDR document-in-progress
4038 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4041 - fix typo in patch-2.4.18
4043 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4046 - clear ptlrpc request each time in handle_incoming_request()
4047 - unlink of files now destroys the object on the OST
4049 2002-09-19 Peter Braam <braam@clusterfs.com>
4051 * add hard link support
4052 * change obdfile creation method
4053 * kernel patch changed
4055 2002-09-19 Peter Braam <braam@clusterfs.com>
4058 - stack overflow bug in extN fixed
4060 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4062 * documentation updates
4063 - add man pages for config tools
4064 - update tests/README to describe testing with new config tools
4065 - finish metadata API descriptions
4066 * bug fixes and cleanups
4067 - statfs workaround for 16TB limit
4068 - LOV stripe allocation improved, can stripe on subset of OSTs
4069 - LOV file size/IO offset was wrong for files > 4GB in size
4070 - object EA data was being dropped, caused files to be unreadable
4071 - memory overflow with non-LOV OST caused memory corruption
4072 - fixed regression tests to work with new config tools, obdfilter
4073 - fixed bug when directory size became larger than 1 block
4074 - fixed bug (for single client case) when PWD was deleted
4075 - invalidate local directory pages when doing intent-based ops
4076 - avoid LDLM oops when lock callback contained bad data
4078 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4080 * documentation updates
4081 * bug fixes and cleanups
4082 - configuration tools
4085 - 64-bit compile warnings
4086 - 64-bit internal statfs data
4088 * test_brw on persistent OST devices
4090 * lprocfs (disabled)
4092 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4094 * documentation updates
4095 * bug fixes and cleanups
4096 * configuration tools
4098 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4101 * many small fixes to 0_5_4
4102 * io/network handling
4103 * thinkos in MDS operations
4105 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4108 * crucial basic fixes to 0.5.3
4109 * IOR, Iozone work over Elan
4112 2002-08-07 Phil Schwan <phil@clusterfs.com>
4113 * version 0_5_3, our first alpha
4114 * we use the new Portals iovs
4115 * documentation updates
4116 * bug fixes and cleanups
4117 * small changes in the DLM wire protocol
4119 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4120 * version 0_5_1 with some initial stability,
4121 * locking on MD and file I/O.
4122 * documentation updates
4123 * several bug fixes since 0.5.0
4124 * small changes in wire protocol
4126 2002-07-18 Phil Schwan <phil@clusterfs.com>
4128 * delivered as Lustre Light Alpha
4129 * fixed a crash after handling invalid MDS requests
4130 * fixed directory pages for architectures with non-4k pages sizes
4132 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4133 * release version v0_4_4
4134 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4136 * Fixes a number of interruption problems with OST operations.
4137 * Update documentation for portals header changes
4138 * Move all wire protocol structs/defines to lustre_idl.h
4139 * Fixes symlink length bug.
4140 * Add tcpdump to repository.
4142 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4143 * release version v0_4_3
4144 * Fixes statfs for inodes on extN.
4145 * Fixes bug in runtests which would delete /etc/hosts.
4146 * Use 64-bit object IDs wherever possible (not into VFS though)
4147 Remove ost_get_info, which is unused by lustre, and out of date.
4149 2002-07-03 Peter Braam <braam@clusterfs.com>
4150 * release version v0_4_2 Fixes a lookup error (type not passed)
4151 * move forward to head of Portals
4152 * move forward to latest Lustre kernel
4154 2002-06-25 Peter Braam <braam@clusterfs.com>
4155 * release version v0_4_1. Hopefully stable on single node use.