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.
377 --------------------------------------------------------------------------------
379 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
381 * Support for kernels:
382 2.6.5-7.283 (SLES 9),
383 2.6.9-55.EL (RHEL 4),
384 2.6.16.46-0.14 (SLES 10),
385 2.6.18.8 vanilla (kernel.org)
386 * Client support for unpatched kernels:
387 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
388 2.6.16 - 2.6.22 vanilla (kernel.org)
389 * Due to recently discovered recovery problems, we do not recommend
390 using patchless RHEL 4 clients with this or any earlier release.
391 * Recommended e2fsprogs version: 1.39.cfs8
392 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
393 * Starting with this release, the ldiskfs backing filesystem required
394 by Lustre is now in its own package, lustre-ldiskfs. This package
395 should be installed. It is versioned separately from Lustre and
396 may be released separately in future.
400 Description: block reactivating mgc import until all deactivates complete
401 Details : Fix race when failing back MDT/MGS to itself (testing)
403 Severity : enhancement
405 Description: add optional extra BUILD_VERSION info
406 Details : add a new environment variable (namely LUSTRE_VERS) which allows
407 to override the lustre version.
410 Frequency : 2.6.18 servers only
412 Description: ll_kern_mount() doesn't release the module reference
413 Details : The ldiskfs module reference count never drops down to 0
414 because ll_kern_mount() doesn't release the module reference.
419 Description: server LBUG when using old ost_num_threads parameter
420 Details : Accept the old ost_num_threads parameter but warn that it
421 is deprecated, and fix an off-by-one error that caused an LBUG.
426 Description: Transient SCSI error results in persistent IO issue
427 Details : iobuf->dr_error is not reinitialized to 0 between two
431 Frequency : sometimes when underlying device returns I/O errors
433 Description: OSTs not going read-only during write failures
434 Details : OSTs are not remounted read-only when the journal commit threads
435 get I/O errors because fsfilt_ext3 calls journal_start/stop()
436 instead of the ext3 wrappers.
440 Description: poor connect scaling with increasing client count
441 Details : Don't run filter_grant_sanity_check for more than 100 exports
442 to improve scaling for large numbers of clients.
445 Frequency : SLES10 only
447 Description: sanity-quota.sh quotacheck failed: rc = -22
448 Details : Quotas cannot be enabled on SLES10.
451 Frequency : liblustre clients only
453 Description: getdirentries does not give error when run on compute nodes
454 Details : getdirentries does not fail when the size specified as an argument
455 is too small to contain at least one entry
457 Severity : enhancement
459 Description: Add LNET router traceability for debug purposes
460 Details : If a checksum failure occurs with a router as part of the
461 IO path, the NID of the last router that forwarded the bulk data
462 is printed so it can be identified.
467 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
468 Details : Due to a race condition, liblustre clients were occasionally
471 Severity : enhancement
473 Description: lfs setstripe use optional parameters instead of postional
476 Severity : enhancement
478 Description: Nanosecond timestamp support for ldiskfs
479 Details : The on-disk ldiskfs filesystem has added support for nanosecond
480 resolution timestamps. There is not yet support for this at
481 the Lustre filesystem level.
484 Frequency : during server recovery
486 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
487 Details : request with rq_no_resend flag not awake l_wait_event if they get a
491 Frequency : nfs export on patchless client
493 Description: connectathon hang when test nfs export over patchless client
494 Details : Disconnected dentry cannot be found with lookup, so we do not need
495 to unhash it or make it invalid
498 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
499 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
500 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
505 Description: open req refcounting wrong on reconnect
506 Details : If reconnect happened between getting open reply from server and
507 call to mdc_set_replay_data in ll_file_open, we will schedule
508 replay for unreferenced request that we are about to free.
509 Subsequent close will crash in variety of ways.
510 Check that request is still eligible for replay in
511 mdc_set_replay_data().
516 Description: disable writes to filesystem when reading health_check file
517 Details : the default for reading the health_check proc file has changed
518 to NOT do a journal transaction and write to disk, because this
519 can cause reads of the /proc file to hang and block HA state
520 checking on a healthy but otherwise heavily loaded system. It
521 is possible to return to the previous behaviour during configure
522 with --enable-health-write.
524 Severity : enhancement
526 Description: 64-bit inode version
527 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
528 to the inode. This will be required for version-based recovery.
533 Description: MDS fails to start if a duplicate client export is detected
534 Details : in some rare cases it was possible for a client to connect to
535 an MDS multiple times. Upon recovery the MDS would detect this
536 and fail during startup. Handle this more gracefully.
538 Severity : enhancement
540 Description: Add -o localflock option to simulate old noflock
542 Details : This will achieve local-only flock/fcntl locks
548 Description: log_commit_thread vs filter_destroy race leads to crash
549 Details : Take import reference before releasing llog record semaphore
554 Description: Wrong request locking in request set processing
555 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
556 imp_delayed_list, in this place should be used imp_lock.
559 Frequency : when reconnection
561 Description: Grant Leak when osc reconnect to OST
562 Details : When osc reconnect ost, OST(filter) should check whether it
563 should grant more space to client by comparing fed_grant and
564 cl_avail_grant, and return the granted space to client instead
565 of "new granted" space, because client will call osc_init_grant
566 to update the client grant space info.
569 Frequency : when client reconnect to OST
571 Description: Grant Leak when osc do resend and replay bulk write
572 Details : When osc reconnect to OST, OST(filter)should clear grant info of
573 bulk write request, because the grant info while be sync between
574 OSC and OST when reconnect, and we should ignore the grant info
575 these of resend/replay write req.
580 Description: Grant space more than avaiable left space sometimes.
581 Details : When then OST is about to be full, if two bulk writing from
582 different clients came to OST. Accord the avaliable space of the
583 OST, the first req should be permitted, and the second one
584 should be denied by ENOSPC. But if the seconde arrived before
585 the first one is commited. The OST might wrongly permit second
586 writing, which will cause grant space > avaiable space.
589 Frequency : when client is evicted
591 Description: Grant might be wrongly erased when osc is evicted by OST
592 Details : when the import is evicted by server, it will fork another
593 thread ptlrpc_invalidate_import_thread to invalidate the
594 import, where the grant will be set to 0. While the original
595 thread will update the grant it got when connecting. So if
596 the former happened latter, the grant will be wrongly errased
597 because of this race.
602 Description: Checking Stale with correct fid
603 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
604 to check whether it is stale, because sometimes, we want the
605 enqueue happened anyway, and op_data.fid2 will not be initialized.
607 Severity : enhancement
609 Description: update patchless client
610 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
613 Frequency : only with 2.4 kernel
615 Description: random memory corruption
616 Details : size of struct ll_inode_info is to big for union inode.u and this
617 can be cause of random memory corruption.
622 Description: Memory leak in recovery
623 Details : Lov_mds_md was not free in an error handler in mds_create_object.
624 It should also check obd_fail before fsfilt_start, otherwise if
625 fsfilt_start return -EROFS,(failover mds during mds recovery).
626 then the req will return with repmsg->transno = 0 and rc = EROFS.
627 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
628 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
629 be freed no matter whether fsfilt_commit success or not.
632 Frequency : only with huge count clients
634 Description: Prevents from taking the superblock lock in llap_from_page for
636 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
637 for llap_from_page call in ll_removepage prevents from taking the
638 superblock lock for a soon died page.
643 Description: Not check open intent error before release open handle
644 Details : in some rare cases, the open intent error is not checked before
645 release open handle, which may cause
646 ASSERTION(open_req->rq_transno != 0), because it tries to release
647 the failed open handle.
652 Description: Set cat log bitmap only after create log success.
653 Details : in some rare cases, the cat log bitmap is set too early. and it
654 should be set only after create log success.
658 Description: Accessing a block bevice can re-enable I/O when Lustre is
659 tearing down a device.
660 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
665 Description: service threads may hog cpus when there are a lot of requests
667 Details : Insert cond_resched to give other threads a chance to use some of
673 Description: the cat log was not initialized in recovery
674 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
675 unlink log on mds will not be initialized until mds post
676 recovery. And also in mds post recovery, the unlink log will
677 initialization will be done asynchronausly, so there will be race
678 between add unlink log and unlink log initialization.
682 Description: brw_stats were being printed incorrectly
683 Details : brw_stats were being printed as log2 but all of them were not
684 recorded as log2. Also remove some code duplication arising from
685 filter_tally_{read,write}.
689 Frequency : rare, only in recovery.
690 Description: ASSERTION(req->rq_type != LI_POISON) failed
691 Details : imp_lock should be held while iterating over imp_sending_list for
692 prevent destroy request after get timeout in ptlrpc_queue_wait.
696 Description: replay-single.sh test 52 fails
697 Details : A lock's skiplist need to be cleanup when it being unlinked
698 from its resource list.
702 Description: Short directio read returns full requested size rather than
704 Details : Direct I/O operations should return actual amount of bytes
705 transferred rather than requested size.
707 Severity : enhancement
709 Description: metadata RPC reduction (e.g. for rm performance)
710 Details : decrease the amount of synchronous RPC between clients and servers
711 by canceling conflicing lock before the operation on the client side
712 and packing thier handles into the main operation RPC to server.
714 Severity : enhancement
716 Description: Async OSC create to avoid the blocking unnecessarily.
717 Details : If a OST has no remain object, system will block on the creating
718 when need to create a new object on this OST. Now, ways use
719 pre-created objects when available, instead of blocking on an
720 empty osc while others are not empty. If we must block, we block
721 for the shortest possible period of time.
725 Description: improve handling recoverable errors
726 Details : if request processig with error which can be recoverable on server
727 request should be resend, otherwise page released from cache and
730 Severity : enhancement
732 Description: refine locking for avoid write wrong info into lov_objid file
733 Details : fix possible races with add new target and write/update data in
736 --------------------------------------------------------------------------------
738 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
743 Frequency : on some architectures
745 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
746 Details : Uninitialized flags sometimes cause configuration commands to
750 Frequency : patchless clients only
752 Description: missing __iget() symbol export
753 Details : The __iget() symbol export is missing. To avoid the need for
754 this on patchless clients the deathrow inode reaper is turned
755 off, and we depend on the VM to clean up old inodes. This
756 dependency was during via the fix for bug 12181.
758 --------------------------------------------------------------------------------
760 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
762 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
763 INTEROPERATE with older versions automatically. In many cases a
764 special upgrade step is needed. Please read the
765 user documentation before upgrading any part of a live system.
766 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
767 version will not interoperate with 1.6 betas before beta5 (1.5.95).
768 * WARNING: Lustre configuration and startup changes are required with
769 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
775 Severity : enhancement
777 Description: MountConf
778 Details : Lustre configuration is now managed via mkfs and mount
779 commands instead of lmc and lconf. New obd types (MGS, MGC)
780 are added for dynamic configuration management. See
781 https://mail.clusterfs.com/wikis/lustre/MountConf for
784 Severity : enhancement
786 Description: dynamic OST addition
787 Details : OSTs can now be added to a live filesystem
789 Severity : enhancement
791 Description: startup order invariance
792 Details : MDTs and OSTs can be started in any order. Clients only
793 require the MDT to complete startup.
795 Severity : enhancement
797 Description: parallel, asynchronous orphan cleanup
798 Details : orphan cleanup is now performed in separate threads for each
799 OST, allowing parallel non-blocking operation.
801 Severity : enhancement
803 Description: optimized stripe assignment
804 Details : stripe assignments are now made based on ost space available,
805 ost previous usage, and OSS previous usage, in order to try
806 to optimize storage space and networking resources.
808 Severity : enhancement
810 Description: Permanently set tunables
811 Details : All writable /proc/fs/lustre tunables can now be permanently
812 set on a per-server basis, at mkfs time or on a live
815 Severity : enhancement
817 Description: Lustre message v2
818 Details : Add lustre message format v2.
820 Severity : enhancement
822 Description: client OST exclusion list
823 Details : Clients can be started with a list of OSTs that should be
824 declared "inactive" for known non-responsive OSTs.
828 Description: SPEC SFS validation failure on NFS v2 over lustre.
829 Details : Changes the blocksize for regular files to be 2x RPC size,
830 and not depend on stripe size.
832 Severity : enhancement
834 Description: Multiple MD RPCs in flight.
835 Details : Further unserialise some read-only MDS RPCs - learn about intents.
836 To avoid overly-overloading MDS, introduce a limit on number of
837 MDS RPCs in flight for a single client and add /proc controls
838 to adjust this limit.
840 Severity : enhancement
842 Description: client read/write statistics
843 Details : Add client read/write call usage stats for performance
844 analysis of user processes.
845 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
846 file access. extents_stats shows chunk size distribution.
847 extents_stats_per_process show chunk size distribution per
850 Severity : enhancement
852 Description: mds statistics
853 Details : Add detailed mds operations statistics in
854 /proc/fs/lustre/mds/*/stats.
858 Description: Failure of copying files with lustre special EAs.
859 Details : Client side always return success for setxattr call for lustre
860 special xattr (currently only "trusted.lov").
865 Description: Refcount LNET uuids
866 Details : The global LNET uuid list grew linearly with every startup;
867 refcount repeated list entries instead of always adding to
870 Severity : enhancement
872 Description: Dynamic service threads
873 Details : Within a small range, start extra service threads
874 automatically when the request queue builds up.
877 Frequency : mixed-endian client/server environments
879 Description: mixed-endian crashes
880 Details : The new msg_v2 system had some failures in mixed-endian
883 Severity : enhancement
885 Description: Easy OST removal
886 Details : OSTs can be permanently deactivated with e.g. 'lctl
887 conf_param lustre-OST0001.osc.active=0'
889 Severity : enhancement
891 Description: MGS proc entries
892 Details : Added basic proc entries for the MGS showing what filesystems
895 Severity : enhancement
897 Description: provide MGS failover
898 Details : Added config lock reacquisition after MGS server failover.
900 Severity : enhancement
902 Description: add Linux 2.4 support
903 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
907 Description: a large application tries to do I/O to the same resource and dies
909 Details : Check the req->rq_arrival time after the call to
910 ost_brw_lock_get(), but before we do anything about
911 processing it & sending the BULK transfer request. This
912 should help move old stale pending locks off the queue as
913 quickly as obd_timeout.
916 Frequency : when an incorrect nid is specified during startup
918 Description: ptlrpc connect to non-existant node causes kernel crash
919 Details : LNET can't be re-entered from an event callback, which
920 happened when we expire a message after the export has been
921 cleaned up. Instead, hand the zombie cleanup off to another
924 Severity : enhancement
926 Description: plain/inodebits lock performance improvement
927 Details : Grouping plain/inodebits in granted list by their request modes
928 and bits policy, thus improving the performance of search through
932 Frequency : only if OST filesystem is corrupted
934 Description: client incorrectly hits assertion in ptlrpc_replay_req()
935 Details : for a short time RPCs with bulk IO are in the replay list,
936 but replay of bulk IOs is unimplemented. If the OST filesystem
937 is corrupted due to disk cache incoherency and then replay is
938 started it is possible to trip an assertion. Avoid putting
939 committed RPCs into the replay list at all to avoid this issue.
942 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
944 Description: System hang on startup
945 Details : This bug allowed the liblustre (e.g. catamount) client to
946 return to the app before handling all startup RPCs. This
947 could leave the node unresponsive to lustre network traffic
948 and manifested as a server ptllnd timeout.
950 Severity : enhancement
952 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
953 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
954 boolean to control whether to print peer debug info when a
955 client's RPC times out.
958 Frequency : only for kernels with patches from Lustre below 1.4.3
960 Description: Remove old rdonly API
961 Details : Remove old rdonly API which unused from at least lustre 1.4.3
964 Frequency : only for devices with external journals
966 Description: Set external device read-only also
967 Details : During a commanded failover stop, we set the disk device
968 read-only while the server shuts down. We now also set any
969 external journal device read-only at the same time.
972 Frequency : when upgrading from 1.4 while trying to change parameters
974 Description: The wrong (new) MDC name was used when setting parameters for
975 upgraded MDT's. Also allows changing of OSC (and MDC)
976 parameters if --writeconf is specified at tunefs upgrade time.
979 Frequency : when setting specific ost indicies
981 Description: QOS code breaks on skipped indicies
982 Details : Add checks for missing OST indicies in the QOS code, so OSTs
983 created with --index need not be sequential.
985 Severity : enhancement
987 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
988 Details : The uninit_groups feature works in conjunction with the kernel
989 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
990 up the pass1 processing of e2fsck. This is a read-only feature
991 in ldiskfs2 only, so older kernels and current ldiskfs cannot
992 mount filesystems that have had this feature enabled.
994 Severity : enhancement
996 Description: Improve multi-block allocation algorithm to avoid fragmentation
997 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
998 allocation locality and avoid filesystem fragmentation.
1000 ------------------------------------------------------------------------------
1002 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1004 * Support for kernels:
1005 2.4.21-47.0.1.EL (RHEL 3)
1006 2.6.5-7.283 (SLES 9)
1007 2.6.9-42.0.10.EL (RHEL 4)
1008 2.6.12.6 vanilla (kernel.org)
1009 2.6.16.27-0.9 (SLES 10)
1010 * Recommended e2fsprogs version: 1.39.cfs5
1012 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1016 Frequency : occasional, depends on client load and configuration
1017 Bugzilla : 12181, 12203
1018 Description: data loss for recently-modified files
1020 Details : In some cases it is possible that recently written or created
1021 files may not be written to disk in a timely manner (this should
1022 normally be within 30s unless client IO load is very high).
1023 The problem appears as zero-length files or files that are a
1024 multiple of 1MB in size after a client crash or client eviction
1025 that are missing data at the end of the file.
1027 This problem is more likely to be hit on clients where files are
1028 repeatedly created and unlinked in the same directory, clients
1029 have a large amount of RAM, have many CPUs, the filesystem has
1030 many OSTs, the clients are rebooted frequently, and/or the files
1031 are not accessed by other nodes after being written.
1033 The presence of the problem can be detected by looking at
1034 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1035 smaller than the second (nr_unused) then dirty files will not
1036 be flushed automatically to disk. "sync; sleep 10" should be
1037 run several times on the node before unmounting it to update
1038 Lustre (this is also safe to run on nodes without this problem).
1040 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1041 that can cause this same problem, so customers using that kernel
1042 also need to update the kernel in addition to Lustre. In order
1043 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1045 It is normal that files written just before a client crash (less
1046 than 30s) may not yet have been flushed to disk, even for local
1050 Frequency : frequent on thin XT3 nodes
1052 Description: UUID collision on thin XT3 Linux nodes
1053 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1054 randomly, since we relied on an insufficiently-seeded PRNG.
1059 Description: OSS hangs after "All ost request buffers busy"
1060 Details : A deadlock between quota and journal operations caused OSS
1061 hangs after printing "All ost request buffers busy."
1064 Frequency : always on liblustre builds
1066 Description: Cleanup compiler warnings on liblustre
1069 Frequency : always on liblustre builds on XT3
1071 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1075 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1076 > 32000 subdirectories
1077 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1078 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1079 subdirectory count crosses 32000. This will aid e2fsck to
1080 correctly handle more than 32000 subdirectories.
1083 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1085 Description: System hang on startup
1086 Details : This bug allowed the liblustre (e.g. catamount) client to
1087 return to the app before handling all startup RPCs. This
1088 could leave the node unresponsive to lustre network traffic
1089 and manifested as a server ptllnd timeout.
1091 Severity : enhancement
1093 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1094 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1095 boolean to control whether to print peer debug info when a
1096 client's RPC times out.
1101 Description: make O_SYNC working on 2.6 kernels
1102 Details : 2.6 kernels use different method for mark pages for write,
1103 so need add a code to lustre for O_SYNC work.
1108 Description: Failure to close file and release space on NFS
1109 Details : Put inode details into lock acquired in ll_intent_file_open.
1110 Use mdc_intent_lock in ll_intent_open to properly
1111 detect all kind of errors unhandled by mdc_enqueue.
1116 Description: proc file read during shutdown sometimes raced obd removal,
1118 Details : Add lock to prevent obd access after proc file removal.
1121 Frequency : Only for files larger than 4GB on 32-bit clients.
1123 Description: improperly doing page alignment of locks
1124 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1125 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1130 Description: RPCs being resent when they shouldn't be
1131 Details : Some RPCs that should not be resent are being resent. This
1132 can cause inconsistencies in the RPC state machine. Do not
1133 resend such requests.
1136 Frequency : rare, only with NFS export
1138 Description: Crash on NFS re-export node
1139 Details : under very unusual load conditions an assertion is hit in
1140 ll_intent_file_open()
1143 Frequency : only if OST filesystem is corrupted
1145 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1146 Details : for a short time RPCs with bulk IO are in the replay list,
1147 but replay of bulk IOs is unimplemented. If the OST filesystem
1148 is corrupted due to disk cache incoherency and then replay is
1149 started it is possible to trip an assertion. Avoid putting
1150 committed RPCs into the replay list at all to avoid this issue.
1155 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1156 Details : Large single O_DIRECT read and write calls can fail to allocate
1157 a sufficiently large buffer to process the request. In case of
1158 allocation failure the allocation is retried with a smaller
1159 buffer and broken into smaller requests.
1161 Severity : enhancement
1163 Description: Add -o localflock option to simulate old noflock behaviour.
1164 Details : This will achieve local-only flock/fcntl locks coherentness.
1169 Description: versioning check is incomplete
1170 Details : Checking the version difference of client vs. server, report
1171 error if the gap is too big.
1176 Description: add support PG_writeback bit
1177 Details : add support for PG_writeback bit for Lustre, for more carefull
1178 work with page cache in 2.6 kernel. This also fix some deadlocks
1179 and remove hack for work O_SYNC with 2.6 kernel.
1181 Severity : enhancement
1183 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1184 Details : The uninit_groups feature works in conjunction with the kernel
1185 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1186 up the pass1 processing of e2fsck. This is a read-only feature
1187 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1188 mount filesystems that have had this feature enabled.
1190 Severity : enhancement
1192 Description: Improve multi-block allocation algorithm to avoid fragmentation
1193 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1194 allocation locality and avoid filesystem fragmentation.
1196 ------------------------------------------------------------------------------
1198 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1200 * Support for kernels:
1201 2.6.9-42.0.3.EL (RHEL 4)
1202 2.6.5-7.276 (SLES 9)
1203 2.4.21-47.0.1.EL (RHEL 3)
1204 2.6.12.6 vanilla (kernel.org)
1205 2.6.16.21-0.8 (SLES10)
1206 * Recommended e2fsprogs version: 1.39.cfs2-0
1208 * The backwards-compatible /proc/sys/portals symlink has been removed
1209 in this release. Before upgrading, please ensure that you change
1210 any configuration scripts or /etc/sysctl.conf files that access
1211 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1212 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1213 in advance of the upgrade on any system running Lustre 1.4.6 or
1214 newer, since /proc/sys/lnet was added in that version.
1215 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1219 Frequency : only when quota is used
1221 Description: avoid scanning export list for quota master
1222 Details : Change the algorithms to avoid scanning export list in order
1223 to improve the efficiency.
1226 Frequency : MDS failover only, very rarely
1228 Description: "went back in time" messages on mds failover
1229 Details : The greatest transno may be lost when the current operation
1230 finishes with an error (transno==0) and the client's last_rcvd
1231 record is over-written. Save the greatest transno in the
1232 mds_last_transno for this case.
1235 Frequency : always for specific kernels and striping counts
1237 Description: client may get "Matching packet too big" without ACL support
1238 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1239 when trying to access files in certain configurations. The
1240 clients should in fact be denied when mounting because they do
1241 not understand ACLs.
1244 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1246 Description: many clients connecting with IO in progress causes connect timeouts
1247 Details : Avoid synchronous journal commits to avoid delays caused by many
1248 clients connecting/disconnecting when bulk IO is in progress.
1249 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1250 OST_IO_PORTAL to avoid delays behind potentially many pending
1254 Frequency : occasionally with multiple writers to a single file
1256 Description: shared writes to file may result in wrong size reported by stat()
1257 Details : Allow growing of kms when extent lock is cancelled
1260 Frequency : always with random mmap IO to multi-striped file
1262 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1263 Details : Only if the hole is at the end of OST object so that kms is too
1264 small. Fix is to increase kms accordingly in ll_nopage.
1267 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1269 Description: writes to a missing object would leak memory on the OST
1270 Details : If there is an inconsistency between the MDS and OST filesystems,
1271 such that the MDS references an object that doesn't exist, writes
1272 to that object will leak memory due to incorrect cleanup in the
1273 error handling path, eventually running out of memory on the OST.
1278 Description: Creating too long symlink causes lustre errors
1279 Details : Check symlink and name lengths before sending requests to MDS.
1282 Frequency : only if flock is enabled (not on by default)
1284 Description: posix locks not released on fd closure on 2.6.9+
1285 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1286 kernels, this caused such locks not to be released on fd close and
1287 then assertions on fs unmount about still used locks.
1290 Frequency : MDS failover only, very rarely
1292 Description: clients may get ASSERTION(granted_lock != NULL)
1293 Details : When request was taking a long time, and a client was resending
1294 a getattr by name lock request. The were multiple lock requests
1295 with the same client lock handle and
1296 mds_getattr_name->fixup_handle_for_resent_request found one of the
1297 lock handles but later failed with ASSERTION(granted_lock != NULL).
1302 Description: handle->h_buffer_credits > 0, assertion failure
1303 Details : h_buffer_credits is zero after truncate, causing assertion
1304 failure. This patch extends the transaction or creates a new
1308 Frequency : NFS re-export or patchless client
1309 Bugzilla : 11179, 10796
1310 Description: Crash on NFS re-export node (__d_move)
1311 Details : We do not want to hash the dentry if we don't have a lock.
1312 But if this dentry is later used in d_move, we'd hit uninitialised
1313 list head d_hash, so we just do this to init d_hash field but
1314 leave dentry unhashed.
1317 Frequency : NFS re-export or patchless client
1319 Description: NFS exports has problem with symbolic link
1320 Details : lustre client didn't properly install dentry when re-exported
1321 to NFS or running patchless client.
1324 Frequency : NFS re-export or patchless client
1326 Description: Various nfs/patchless fixes.
1327 Details : fixes reuse disconected alias for lookup process - this fixes
1328 warning "find_exported_dentry: npd != pd",
1329 fix permission error with open files at nfs.
1330 fix apply umask when do revalidate.
1333 Frequency : occasional
1335 Description: Crash on NFS re-export node
1336 Details : calling clear_page() on the wrong pointer triggered oops in
1337 generic_mapping_read().
1340 Frequency : rarely, using O_DIRECT IO
1342 Description: unaligned directio crashes client with LASSERT
1343 Details : check for unaligned buffers before trying any requests.
1346 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1348 Description: stale data returned from RAID cache
1349 Details : If only a small amount of IO is done to the RAID device before
1350 reading it again it is possible to get stale data from the RAID
1351 cache instead of reading it from disk.
1354 Frequency : always for sles10 kernel
1356 Description: sles10 support
1357 Details : ll_follow_link: compile fixes and using of nd_set_link
1358 under newer kernels.
1361 Frequency : depends on arch, kernel and compiler version, always on sles10
1364 Description: recursive or deep enough symlinks cause stack overflow
1365 Details : getting rid of large stack-allocated variable in
1369 Frequency : depends on hardware
1371 Description: lustre write performance loss in the SLES10 kernel
1372 Details : the performance loss is caused by using of write barriers in the
1373 ext3 code. The SLES10 kernel turns barrier support on by
1374 default. The fix is to undo that change for ldiskfs.
1376 ------------------------------------------------------------------------------
1378 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1380 * Support for kernels:
1381 2.6.9-42.0.3EL (RHEL 4)
1382 2.6.5-7.276 (SLES 9)
1383 2.4.21-47.0.1.EL (RHEL 3)
1384 2.6.12.6 vanilla (kernel.org)
1388 Frequency : quota enabled and large files being deleted
1390 Description: releasing more than 4GB of quota at once hangs OST
1391 Details : If a user deletes more than 4GB of files on a single OST it
1392 will cause the OST to spin in an infinite loop. Release
1393 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1398 Description: statfs data retrieved from /proc may be stale or zero
1399 Details : When reading per-device statfs data from /proc, in the
1400 {kbytes,files}_{total,free,avail} files, it may appear
1401 as zero or be out of date.
1404 Frequency : systems with MD RAID1 external journal devices
1406 Description: lconf's call to blkid is confused by RAID1 journal devices
1407 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1408 of returning all block devices that match the journal UUID.
1411 Frequency : always, for aggregate stripe size over 4GB
1413 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1414 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1415 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1418 Frequency : always if "lfs find" used on a local file/directory
1420 Description: "lfs find" segfaults if used on a local file/directory
1421 Details : The case where a directory component was not specified wasn't
1422 handled correctly. Handle this properly.
1425 Frequency : always on ppc64
1427 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1428 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1429 perform endianness conversion.
1432 Frequency : rarely (truncate to non-zero file size after write under load)
1433 Bugzilla : 10730, 10687
1434 Description: Files padded with zeros to next 4K multiple
1435 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1436 it is possible that files that are truncated to a non-zero size
1437 immediately after being written are filled with zero bytes beyond
1438 the truncated size. No file data is lost.
1440 Severity : enhancement
1442 Description: Allow recovery/failover for liblustre clients.
1443 Details : liblustre clients were unaware of failover configurations until
1446 Severity : enhancement
1448 Description: user file locks should fail when not mounting with flock option
1449 Details : Set up an error-returning stub in ll_file_operations.lock field
1450 to prevent incorrect behaviour when client is mounted without
1451 flock option. Also, set up properly f_op->flock field for
1455 Frequency : always on ia64
1457 Description: "lfs df" loops on printing out MDS statfs information
1458 Details : The obd_ioctl_data was not initialized and in some systems
1459 this caused a failure during the ioctl that did not return
1460 an error. Initialize the struct and return an error on failure.
1463 Frequency : SLES 9 only
1465 Description: Error of copying files with lustre special EAs as root
1466 Details : Client side always return success for setxattr call for lustre
1467 special xattr (currently only "trusted.lov").
1470 Frequency : rarely on clusters with both ia64+i386 clients
1472 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1473 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1474 clients are doing concurrent writes to the same file it is
1475 possible that the ia64 clients may LASSERT because the OST
1476 extent locks are not PAGE_SIZE aligned. Ensure that grown
1477 locks are always aligned on the request boundary.
1480 Frequency : specific use, occasional
1482 Description: Overwriting in use executable truncates on-disk binary image
1483 Details : If one node attempts to overwrite an executable in use by
1484 another node, we now correctly return ETXTBSY instead of
1485 truncating the file.
1487 Severity : enhancement
1489 Description: Async OSC create to avoid the blocking unnecessarily.
1490 Details : If a OST has no remain object, system will block on the creating
1491 when need to create a new object on this OST. Now, ways use
1492 pre-created objects when available, instead of blocking on an
1493 empty osc while others are not empty. If we must block, we block
1494 for the shortest possible period of time.
1499 Description: chmod on Lustre root is propagated to other clients
1500 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1501 invalid by the follow_mount time.
1504 Frequency : liblustre clients only
1506 Description: Race in 'instant cancel' lock handling could lead to such locks
1507 never to be granted in case of SMP MDS
1508 Details : Do not destroy not yet granted but cbpending locks in
1512 Frequency : replay/resend of open
1514 Description: non null lock assetion failure in mds_intent_policy
1515 Details : Trying to replay/resend lockless open requests resulted in
1516 mds_open() returning 0 with no lock. Now it sets a flag if
1517 a lock is going to be returned.
1519 Severity : enhancement
1521 Description: Checksum enhancements
1522 Details : New checksum enhancements allow for resending RPCs that failed
1525 Severity : enhancement
1527 Description: Tunables on number of dirty pages in cacche
1528 Details : Allow to set limit on number of dirty pages cached.
1533 Description: client crash on unmount - lock still has references
1534 Details : In some error handling cases it was possible to leak a lock
1535 reference on a client while accessing a file. This was not
1536 harmful to the client during operation, but would cause the
1537 client to crash when the filesystem is unmounted.
1540 Frequency : specific case, rare
1542 Description: ETXTBSY on mds though file not in use by client
1543 Details : ETXTBSY is no longer incorrectly returned when attempting to
1544 chmod or chown a directory that the user previously tried to
1545 execute or a currently-executing binary.
1548 Frequency : extremely rare except on liblustre-based clients
1550 Description: Lustre space not freed when files are deleted
1551 Details : Clean up open-unlinked files after client eviction. Previously
1552 the unlink was skipped and the files remained as orphans.
1557 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1558 Details : In some cases it was possible to send a blocking callback to
1559 a client doing a glimpse, even though that client didn't get
1560 a lock granted. When the glimpse lock is cancelled on the OST
1561 the freed lock is left on the waiting list and corrupted the list.
1564 Frequency : all core dumps
1566 Description: Broke core dumps to lustre
1567 Details : Negative dentry may be unhashed if parent does not have UPDATE
1568 lock, but some callers, e.g. do_coredump, expect dentry to be
1569 hashed after successful create, hash it in ll_create_it.
1571 ------------------------------------------------------------------------------
1573 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1575 * Support for kernels:
1576 2.6.9-42.0.2.EL (RHEL 4)
1577 2.6.5-7.276 (SLES 9)
1578 2.4.21-40.EL (RHEL 3)
1579 2.6.12.6 vanilla (kernel.org)
1583 Frequency : always on RHEL 3
1585 Description: Number of open files grows over time
1586 Details : The number of open files grows over time, whether or not
1587 Lustre is started. This was due to a filp leak introduced
1588 by one of our kernel patches.
1590 ------------------------------------------------------------------------------
1592 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1594 * Support for kernels:
1595 2.6.9-42.EL (RHEL 4)
1596 2.6.5-7.276 (SLES 9)
1597 2.4.21-40.EL (RHEL 3)
1598 2.6.12.6 vanilla (kernel.org)
1603 Bugzilla : 5719, 9635, 9792, 9684
1604 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1605 Details : If a server is under heavy load and cannot reply to new
1606 connection requests before the client resends the (re)connect,
1607 the connection handling code can behave badly if two service
1608 threads are concurrently handing separate (re)connections from
1609 the same client. Add better locking to the connection handling
1610 code, and ensure that only a single connection will be processed
1611 for a given client UUID, even if the lock is dropped.
1613 Severity : enhancement
1615 Description: add TCP zero-copy support to kernel
1616 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1617 sends if the hardware supports scatter-gather and checksumming.
1618 This allows socklnd to do client-write and server-read more
1619 efficiently and reduce CPU utilization from skbuf copying.
1622 Frequency : only if NFS exporting from client
1624 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1625 Details : If setting attributes on a file created under NFS that had
1626 never been opened it would be possible to oops the client
1627 if the file had no objects.
1630 Frequency : always for liblustre
1632 Description: liblustre client does MDS+OSTs setattr RPC for each write
1633 Details : When doing a write from a liblustre client, the client
1634 incorrectly issued an RPC to the MDS and each OST the file was
1635 striped over in order to update the timestamps. When writing
1636 with small chunks and many clients this could overwhelm the MDS
1637 with RPCs. In all cases it would slow down the write because
1638 these RPCs are unnecessary.
1640 Severity : enhancement
1642 Description: allow number of MDS service threads to be changed at module load
1643 Details : It is now possible to change the number of MDS service threads
1644 running. Adding "options mds mds_num_threads={N}" to the MDS's
1645 /etc/modprobe.conf will set the number of threads for the next
1646 time Lustre is restarted (assuming the "mds" module is also
1647 reloaded at that time). The default number of threads will
1648 stay the same, 32 for most systems.
1653 Description: OST crash if filesystem is unformatted or corrupt
1654 Details : If an OST is started on a device that has never been formatted
1655 or if the filesystem is corrupt and cannot even mount then the
1656 error handling cleanup routines would dereference a NULL pointer.
1661 Description: NULL pointer deref in llap_from_page.
1662 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1663 incorrect) mapping if the page was truncated/reclaimed while it was
1664 searched for. Check for this condition and skip such pages when
1665 doing readahead. Introduce extra check to llap_from_page() to
1666 verify page->mapping->host is non-NULL (so page is not anonymous).
1669 Frequency : Sometimes when using sys_sendfile
1671 Description: "page not covered by a lock" warnings from ll_readpage
1672 Details : sendfile called ll_readpage without right page locks present.
1673 Now we introduced ll_file_sendfile that does necessary locking
1674 around call to generic_file_sendfile() much like we do in
1678 Frequency : with certain MDS communication failures at client mount time
1680 Description: NULL pointer deref after failed client mount
1681 Details : a client connection request may delayed by the network layer
1682 and not be sent until after the PTLRPC layer has timed out the
1683 request. If the client fails the mount immediately it will try
1684 to clean up before the network times out the request. Add a
1685 reference from the request import to the obd device and delay
1686 the cleanup until the network drops the request.
1689 Frequency : occasionally during client (re)connect
1691 Description: assertion failure during client (re)connect
1692 Details : processing a client connection request may be delayed by the
1693 client or server longer than the client connect timeout. This
1694 causes the client to resend the connection request. If the
1695 original connection request is replied in this interval, the
1696 client may trip an assertion failure in ptlrpc_connect_interpret()
1697 which thought it would be the only running connect process.
1700 Frequency : only with obd_echo servers and clients that are rebooted
1702 Description: kernel BUG accessing uninitialized data structure
1703 Details : When running an obd_echo server it did not start the ping_evictor
1704 thread, and when a client was evicted an uninitialized data
1705 structure was accessed. Start the ping_evictor in the RPC
1706 service startup instead of the OBD startup.
1708 Severity : enhancement
1709 Bugzilla : 10193 (patchless)
1710 Description: Remove dependency on various unexported kernel interfaces.
1711 Details : No longer need reparent_to_init, exit_mm, exit_files,
1712 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1715 Frequency : rare (only users of deprecated and unsupported LDAP config)
1717 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1718 Details : LDAP apparently contains 'lustreName' attributes instead of
1719 'name'. A simple remapping of the name is sufficient.
1722 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1724 Description: waiting_locks_callback trips kernel BUG if client is evicted
1725 Details : Running with the dump_on_timeout debug flag turned on makes
1726 it possible that the waiting_locks_callback() can try to dump
1727 the Lustre kernel debug logs from an interrupt handler. Defer
1728 this log dumping to the expired_lock_main() thread.
1730 Severity : enhancement
1732 Description: Support NFS exporting on 2.6 kernels.
1733 Details : Implement non-rawops metadata methods for NFS server to use without
1734 changing NFS server code.
1737 Frequency : very rare (synthetic metadata workload only)
1739 Description: two racing renames might cause an MDS thread to deadlock
1740 Details : Running the "racer" program may cause one MDS thread to rename
1741 a file from being the source of a rename to being the target of
1742 a rename at exactly the same time that another thread is doing
1743 so, and the second thread has already enqueued these locks after
1744 doing a lookup of the target and is trying to relock them in
1745 order. Ensure that we don't try to re-lock the same resource.
1748 Frequency : only very large systems with liblustre clients
1750 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1751 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1752 all OSTs in parallel. This allows the eviction of stale liblustre
1753 clients to proceed much faster than if they were done in series,
1754 and also offers similar improvements for other set_info RPCs.
1759 Description: excessive CPU usage during initial read phase on client
1760 Details : During the initial read phase on a client, it would agressively
1761 retry readahead on the file, consuming too much CPU and impacting
1762 performance (since 1.4.5.8). Improve the readahead algorithm
1763 to avoid this, and also improve some other common cases (read
1764 of small files in particular, where "small" is files smaller than
1765 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1770 Description: MDS crash when receiving packet with unknown intent.
1771 Details : Do not LBUG in unknown intent case, just return -EFAULT
1773 Severity : enhancement
1774 Bugzilla : 9293, 9385
1775 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1776 Details : Do not serialize getattr (non-intent version) and statfs.
1779 Frequency : occasional, when OST network is overloaded/intermittent
1781 Description: client evicted by OST after bulk IO timeout
1782 Details : If a client sends a bulk IO request (read or write) the OST
1783 may evict the client if it is unresposive to its data GET/PUT
1784 request. This is incorrect if the network is overloaded (takes
1785 too long to transfer the RPC data) or dropped the OST GET/PUT
1786 request. There is no need to evict the client at all, since
1787 the pinger and/or lock callbacks will handle this, and the
1788 client can restart the bulk request.
1791 Frequency : Always when mmapping file with no objects
1793 Description: client crashes when mmapping file with no objects
1794 Details : Check that we actually have objects in a file before doing any
1795 operations on objects in ll_vm_open, ll_vm_close and
1801 Description: Request leak when working with deleted CWD
1802 Details : Introduce advanced request refcount tracking for requests
1803 referenced from lustre intent.
1805 Severity : Enhancement
1807 Description: Cache open file handles on client.
1808 Details : MDS now will return special lock along with openhandle, if
1809 requested and client is allowed to hold openhandle, even if unused,
1810 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1811 closing files for every read/write openration.
1813 Severity : Enhancement
1815 Description: Cache open negative dentries on client when possible.
1816 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1817 negative dentries on lock revocation.
1822 Description: Remounting a client read-only wasn't possible with a zconf mount
1823 Details : It wasn't possible to remount a client read-only with llmount.
1825 Severity : enhancement
1826 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1827 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1828 MPICH is included to add Lustre-specific ADIO interfaces.
1829 This is based closely on the UFS ADIO layer and only differs
1830 in file creation, in order to allow the OST striping to be set.
1831 This is user-contributed code and not supported by CFS.
1836 Description: extended inode attributes (immutable, append-only) work improperly
1837 when 2.4 and 2.6 kernels are used on client/server or vice versa
1838 Details : Introduce kernel-independent values for these flags.
1840 Severity : enhancement
1843 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1844 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1845 granularity. Now, Fractional MB granularity is be supported,
1846 this is very useful for low memory system.
1848 Severity : enhancement
1850 Description: Getattr by fid
1851 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1852 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1853 have locks and avoids getattr rpc requests for it on every stat.
1856 Frequency : Always, for filesystems larger than 2TB
1858 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1859 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1860 filesystems larger than 2TB to be mounted with mballoc enabled.
1863 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1865 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1866 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1867 filesystems larger than 2TB is mounted then the kernel will
1868 silently corrupt the start of the filesystem. CONFIG_LBD is
1869 enabled for all CFS-supported kernels, but the possibility of
1870 this happening with a modified kernel config exists.
1872 Severity : enhancement
1874 Description: add client O_DIRECT support for 2.6 kernels
1875 Details : It is now possible to do O_DIRECT reads and writes to files
1876 in the Lustre client mountpoint on 2.6 kernel clients.
1878 Severity : enhancement
1880 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1881 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1882 parallel, not waiting for response from every OST before sending
1883 a rpc to the next OST.
1888 Description: setattr vs write race when updating file timestamps
1889 Details : Client processes that update a file timestamp into the past
1890 right after writing to the file (e.g. tar) it is possible that
1891 the updated file modification time can be reset to the current
1892 time due to a race between processing the setattr and write RPC.
1894 Severity : enhancement
1896 Description: Bring 'lfs find' closer in line with regular Linux find.
1897 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1898 -print0 options and obtains all the needed info through the lustre
1901 Severity : enhancement
1903 Description: support up to 1024 configured devices on one node
1904 Details : change obd_dev array from statically allocated to dynamically
1905 allocated structs as they are first used to reduce memory usage
1910 Description: Flush dirty partially truncated pages during truncate
1911 Details : Immediatelly flush partially truncated pages in filter_setattr,
1912 this way we completely avoid having any pages in page cache on OST
1913 and can retire ugly workarounds during writes to flush such pages.
1918 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1919 Details : protect inode from truncation within vfs_unlink() context
1920 just take a reference before calling vfs_unlink() and release it
1921 when parent's i_sem is free.
1926 Description: last_id value checked outside lock on OST caused LASSERT failure
1927 Details : If there were multiple MDS->OST object precreate requests in
1928 flight, it was possible that the OST's last object id was checked
1929 outside a lock and incorrectly tripped an assertion. Move checks
1930 inside locks, and discard old precreate requests.
1933 Frequency : always, if extents are used on OSTs
1935 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1936 Details : OSTs using the extents format would not zero the high 16 bits of
1937 the index physical block number. This is not a problem for any
1938 OST filesystems smaller than 16TB, and no kernels support ext3
1939 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1940 new/modified files) and can be fixed for existing filesystems
1941 with e2fsprogs-1.39-cfs1.
1946 Description: import connection selection may be incorrect if timer wraps
1947 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1948 import connections to be ignored if the 32-bit jiffies counter
1949 wraps. Use a 64-bit jiffies counter.
1952 Frequency : very large clusters immediately after boot
1954 Description: LNET request buffers exhausted under heavy short-term load
1955 Details : If a large number of client requests are generated on a service
1956 that has previously never seen so many requests it is possible
1957 that the request buffer growth cannot keep up with the spike in
1958 demand. Instead of dropping incoming requests, they are held in
1959 the LND until the RPC service can accept more requests.
1962 Frequency : Sometimes during replay
1964 Description: Assertion failure in ll_local_open after replay.
1965 Details : If replay happened on an open request reply before we were able
1966 to set replay handler, reply will become not swabbed tripping the
1967 assertion in ll_local_open. Now we set the handler right after
1968 recognising of open request
1971 Frequency : very rare
1973 Description: kernel reports "badness in vsnprintf"
1974 Details : Reading from the "recovery_status" /proc file in small chunks
1975 may cause a negative length in lprocfs_obd_rd_recovery_status()
1976 call to vsnprintf() (which is otherwise harmless). Exit early
1977 if there is no more space in the output buffer.
1979 Severity : enhancement
1981 Description: clear OBD RPC statistics by writing to them
1982 Details : It is now possible to clear the OBD RPC statistics by writing
1983 to the "stats" file.
1988 Description: Client mtime is not the same on different clients after utimes
1989 Details : In some cases, the client was using the utimes() syscall on
1990 a file cached on another node. The clients now validate the
1991 ctime from the MDS + OSTs to determine which one is right.
1996 Description: Inability to activate failout mode
1997 Details : lconf script incorrectly assumed that in pythong string's numeric
1998 value is used in comparisons.
2001 Frequency : always with multiple stripes per file
2003 Description: Inefficient object allocation for mutli-stripe files
2004 Details : When selecting which OSTs to stripe files over, for files with
2005 a stripe count that divides evenly into the number of OSTs,
2006 the MDS is always picking the same starting OST for each file.
2007 Return the OST selection heuristic to the original design.
2012 Description: mount failures may take full timeout to return an error
2013 Details : Under some heavy load conditions it is possible that a
2014 failed mount can wait for the full obd_timeout interval,
2015 possibly several minutes, before reporting an error.
2016 Instead return an error as soon as the status is known.
2018 Frequency : quota enabled and large files being deleted
2020 Description: releasing more than 4GB of quota at once hangs OST
2021 Details : If a user deletes more than 4GB of files on a single OST it
2022 will cause the OST to spin in an infinite loop. Release
2023 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2028 Description: statfs data retrieved from /proc may be stale or zero
2029 Details : When reading per-device statfs data from /proc, in the
2030 {kbytes,files}_{total,free,avail} files, it may appear
2031 as zero or be out of date.
2034 Frequency : systems with MD RAID1 external journal devices
2036 Description: lconf's call to blkid is confused by RAID1 journal devices
2037 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2038 of returning all block devices that match the journal UUID.
2041 Frequency : always, for aggregate stripe size over 4GB
2043 Description: assertion fails when trying to use 4GB stripe size
2044 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2045 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2048 Frequency : always on ppc64
2050 Description: the first write on an ext3 filesystem with mballoc got stuck
2051 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2052 perform endianness conversion.
2054 ------------------------------------------------------------------------------
2056 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2058 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2059 INTEROPERATE with older versions automatically. Please read the
2060 user documentation before upgrading any part of a live system.
2061 * WARNING: Lustre networking configuration changes are required with
2062 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2065 * Support for kernels:
2066 2.6.9-22.0.2.EL (RHEL 4)
2067 2.6.5-7.244 (SLES 9)
2068 2.6.12.6 vanilla (kernel.org)
2071 Severity : enhancement
2072 Bugzilla : 7981/8208
2073 Description: Introduced Lustre Networking (LNET)
2074 Details : LNET is new networking infrastructure for Lustre, it includes
2075 a reorganized network configuration mode (see the user
2076 documentation for full details) as well as support for routing
2077 between different network fabrics. Lustre Networking Devices
2078 (LNDs) for the supported network fabrics have also been
2079 created for this new infrastructure.
2081 Severity : enhancement
2082 Description: Introduced Access control lists
2083 Details : clients can set ACLs on files and directories in order to have
2084 more fine-grained permissions than the standard Unix UGO+RWX.
2085 The MDS must be started with the "-o acl" mount option.
2087 Severity : enhancement
2088 Description: Introduced filesystem quotas
2089 Details : Administrators may now establish per-user quotas on the
2092 Severity : enhancement
2094 Description: Configuration change for the XT3
2095 The PTLLND is now used to run Lustre over Portals on the XT3
2096 The configure option(s) --with-cray-portals are no longer used.
2097 Rather --with-portals=<path-to-portals-includes> is used to
2098 enable building on the XT3. In addition to enable XT3 specific
2099 features the option --enable-cray-xt3 must be used.
2104 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2105 Details : A race between allocating a new llcd and re-getting the llcd_lock
2106 allowed another thread to grab newly-allocated llcd.
2108 Severity : enhancement
2110 Description: 2.6 OST async journal commit and locking fix to improve performance
2111 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2112 2.6 kernels are now async as they already were in 2.4 kernels so
2113 that they can commit concurrently with the network bulk transfer.
2114 For block-allocated files the filter allocation semaphore is held
2115 to avoid filesystem fragmentation during allocation. BKL lock
2116 removed for 2.6 xattr operations where it is no longer needed.
2121 Description: lconf incorrectly determined whether two IP networks could talk
2122 Details : In some more complicated routing and multiple-network
2123 configurations, lconf will avoid trying to make a network
2124 connection to a disjoint part of the IP space. It was doing the
2125 math incorrectly for one set of cases.
2130 Description: Fix for potential infinite loop processing records in an llog.
2131 Details : If an llog record is corrupted/zeroed, it is possible to loop
2132 forever in llog_process(). Validate the llog record length
2133 and skip the remainder of the block on error.
2136 Frequency : occasional (liblustre only)
2138 Description: liblustre could not open files whose last component is a symlink
2139 Details : sysio_path_walk() would incorrectly pass the open intent to
2140 intermediate path components.
2143 Frequency : rare (liblustre only with non-standard tuning)
2144 Bugzilla : 7201 (7350)
2145 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2146 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2147 after it has been released, only do this for referenced locks
2149 Severity : enhancement
2151 Description: specifying an (invalid) directory default stripe_size of -1
2152 would reset the directory default striping
2153 Details : stripe_size -1 was used internally to signal directory stripe
2154 removal, now use "all default" to signal dir stripe removal
2155 as a directory striping of "all default" is not useful
2158 Frequency : common for large clusters running liblustre clients
2160 Description: doing an ls when liblustre clients are running is slow
2161 Details : sending a glimpse AST to a liblustre client waits for every AST
2162 to time out, as liblustre clients will not respond. Since they
2163 cannot cache data we refresh the OST lock LVB from disk instead.
2165 Severity : enhancement
2167 Description: doing an ls at the same time as file IO can be slow
2168 Details : enqueue and other "small" requests can be blocked behind many
2169 large IO requests. Create a new OST IO portal for non-IO
2170 requests so they can be processed faster.
2173 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2175 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2176 created file with mode 000
2177 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2178 allow an ftruncate() as a non-root user to a file with mode 000.
2179 Lustre now respects this flag to disable mode checks when
2180 truncating a file owned by the user
2183 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2186 Description: Revoking locks from clients that went dead or catatonic might take
2188 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2189 cancellation of such locks instant on servers without waiting for
2190 any reply from clients. Clients drops these locks when cancel
2191 notification from server is received without replying.
2194 Frequency : liblustre-only, when liblustre client dies or becomes busy
2196 Description: Doing ls on Linux clients can take a long time with active
2198 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2199 granting such locks to it in the first place if possible. Locks
2200 are taken by proxy on the OST during the read or write and
2201 dropped immediately afterward. Add connect flags handling, do
2202 not grant locks to liblustre clients for glimpse ASTs.
2204 Severity : enhancement
2206 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2207 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2208 sequential reads and behaves badly with applications doing random
2209 reads. Improve it by reading ahead at least read region, and
2210 avoiding excessive large RPC for small reads.
2212 Severity : enhancement
2214 Description: Creating more than 1000 files for a single job may cause a load
2215 imbalance on the OSTs if there are also a large number of OSTs.
2216 Details : qos_prep_create() uses an OST index reseed value that is an
2217 even multiple of the number of available OSTs so that if the
2218 reseed happens in the middle of the object allocation it will
2219 still utilize the OSTs as uniformly as possible.
2224 Description: OST or MDS may oops in ping_evictor_main()
2225 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2226 but doesn't restart at beginning of obd_exports_timed list
2229 Severity : enhancement
2231 Description: improve by-nid export eviction on the MDS and OST
2232 Details : allow multiple exports with the same NID to be evicted at one
2233 time without re-searching the exports list.
2236 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2238 Description: MDS may oops in groups_free()
2239 Details : in rare race conditions a newly allocated group_info struct is
2240 freed again, and this can be NULL. The 2.4 compatibility code
2241 for groups_free() checked for a NULL pointer, but 2.6 did not.
2244 Frequency : common for liblustre clients doing little filesystem IO
2245 Bugzilla : 9352, 7313
2246 Description: server may evict liblustre clients accessing contended locks
2247 Details : if a client is granted a lock or receives a completion AST
2248 with a blocking AST already set it would not reply to the AST
2249 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2252 Frequency : lfs setstripe, only systems with more than 160 OSTs
2254 Description: unable to set striping with a starting offset beyond OST 160
2255 Details : llapi_create_file() incorrectly limited the starting stripe
2256 index to the maximum single-file stripe count.
2259 Frequency : LDAP users only
2261 Description: lconf did not handle in-kernel recovery with LDAP properly
2262 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2264 Severity : enhancement
2266 Description: bind OST threads to NUMA nodes to improve performance
2267 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2268 node and do their allocations there to localize memory access
2270 Severity : enhancement
2272 Description: llmount can determine client NID directly from Myrinet (GM)
2273 Details : the client NID code from gmnalnid was moved directly into
2274 llmount, removing the need to use this or specifying the
2275 client NID explicitly when mounting GM clients with zeroconf
2278 Frequency : if client is started with down MDS
2280 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2281 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2282 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2283 ptlrpc_import_delayed_req() because we didn't check if the
2284 request was interrupted, and we also didn't break out of the
2285 event loop if there was no timeout
2290 Description: data loss during non-page-aligned writes to a single file from
2291 both multiple nodes and multiple threads on one node at same time
2292 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2293 inconsistency led to false short-reads, that were cached and later
2294 used by ->prepare_write() to fill in partially written page,
2295 leading to data loss.
2298 Frequency : always, if lconf --abort_recovery used
2300 Description: lconf --abort_recovery fails with 'Operation not supported'
2301 Details : lconf was attempting to abort recovery on the MDT device and not
2304 Severity : enhancement
2306 Description: remove cleanup logs
2307 Details : replace lconf-generated cleanup logs with lustre internal
2308 cleanup routines. Eliminates the need for client-cleanup and
2311 Severity : enhancement
2313 Description: add support for EAs (user and system) on lustre filesystems
2314 Details : it is now possible to store extended attributes in the Lustre
2315 client filesystem, and with the user_xattr mount option it
2316 is possible to allow users to store EAs on their files also
2318 Severity : enhancement
2320 Description: Add possibility (config option) to show minimal available OST free
2322 Details : When compiled with --enable-mindf configure option, statfs(2)
2323 (and so, df) will return least minimal free space available from
2324 all OSTs as amount of free space on FS, instead of summary of
2325 free spaces of all OSTs.
2327 Severity : enhancement
2329 Description: do not expand extent locks acquired on OST-side
2330 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2331 by server: they are not cached anyway.
2334 Frequency : when mmap is used/binaries executed from Lustre
2336 Description: Unmmap pages before throwing them away from read cache.
2337 Details : llap_shrink cache now attempts to unmap pages before discarding
2338 them (if unmapping failed - do not discard). SLES9 kernel has
2339 extra checks that trigger if this unmapping is not done first.
2344 Description: lconf didn't resolve symlinks before checking to see whether a
2345 given mountpoint was already in use
2348 Frequency : when migrating failover services
2349 Bugzilla : 6395, 9514
2350 Description: When migrating a subset of services from a node (e.g. failback
2351 from a failover service node) the remaining services would
2352 time out and evict clients.
2353 Details : lconf --force (implied by --failover) sets the global obd_timeout
2354 to 5 seconds in order to quickly disconnect, but this caused
2355 other RPCs to time out too quickly. Do not change the global
2356 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2358 Severity : enhancement
2359 Frequency : if MDS is started with down OST
2360 Bugzilla : 9439,5706
2361 Description: Allow startup/shutdown of an MDS without depending on the
2362 availability of the OSTs.
2363 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2364 Add appropriate locking and lov-osc refcounts for safe
2365 cleaning. Add osc abort_inflight calls in case the
2366 synchronize never started.
2369 Frequency : occasional (Cray XT3 only)
2371 Description: root not authorized to access files in CRAY_PORTALS environment
2372 Details : The client process capabilities were not honoured on the MDS in
2373 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2374 been accessed by an authorized user then root was able to access
2375 the file on the local client also. The root user capabilities
2376 are now allowed on the MDS, as this environment has secure UID.
2379 Frequency : occasional
2381 Description: ldiskfs "too long searching" message happens too often
2382 Details : A debugging message (otherwise harmless) prints too often on
2383 the OST console. This has been reduced to only happen when
2384 there are fragmentation problems on the filesystem.
2389 Description: Division by zero in statfs when all OSCs are inactive
2390 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2391 lov_statfs divides by value returned by lov_get_stripecnt().
2395 Bugzilla : 9489, 3273
2396 Description: First write from each client to each OST was only 4kB in size,
2397 to initialize client writeback cache, which caused sub-optimal
2398 RPCs and poor layout on disk for the first writen file.
2399 Details : Clients now request an initial cache grant at (re)connect time
2400 and so that they can start streaming writes to the cache right
2401 away and always do full-sized RPCs if there is enough data.
2402 If the OST is rebooted the client also re-establishes its grant
2403 so that client cached writes will be honoured under the grant.
2408 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2409 Details : Now I/O RPCs go to different portal number and (presumably) fast
2410 lock requests (and glimses) and other RPCs get their own service
2411 threads pool that should be able to service those RPCs
2414 Severity : enhancement
2416 Description: Ability to exchange lustre version between client and servers and
2417 issue warnings at client side if client is too old. Also for
2418 liblustre clients there is ability to refuse connection of too old
2420 Details : New 'version' field is added to connect data structure that is
2421 filled with version info. That info is later checked by server and
2425 Frequency : rare, liblustre only.
2426 Bugzilla : 9296, 9581
2427 Description: Two simultaneous writes from liblustre at offset within same page
2428 might proceed at the same time overwriting eachother with stale
2430 Details : I/O lock withing llu_file_prwv was released too early, before data
2431 actually was hitting the wire. Extended lock-holding time until
2432 server acknowledges receiving data.
2435 Frequency : extremely rare. Never observed in practice.
2437 Description: avoid generating lustre_handle cookie of 0.
2438 Details : class_handle_hash() generates handle cookies by incrementing
2439 global counter, and can hit 0 occasionaly (this is unlikely, but
2440 not impossible, because initial value of cookie counter is
2441 selected randonly). Value of 0 is used as a sentinel meaning
2442 "unassigned handle" --- avoid it. Also coalesce two critical
2443 sections in this function into one.
2445 Severity : enhancement
2447 Description: allow liblustre clients to delegate truncate locking to OST
2448 Details : To avoid overhead of locking, liblustre client instructs OST to
2449 take extent lock in ost_punch() on client's behalf. New connection
2450 flag is added to handle backward compatibility.
2452 Severity : enhancement
2453 Bugzilla : 4928, 7341, 9758
2454 Description: allow number of OST service threads to be specified
2455 Details : a module parameter allows the number of OST service threads
2456 to be specified via "options ost ost_num_threads={N}" in the
2457 OSS's /etc/modules.conf or /etc/modprobe.conf.
2461 Bugzilla : 6146, 9635, 9895
2462 Description: servers crash with bad pointer in target_handle_connect()
2463 Details : In rare cases when a client is reconnecting it was possible that
2464 the connection request was the last reference for that export.
2465 We would temporarily drop the export reference and get a new
2466 one, but this may have been the last reference and the export
2467 was just destroyed. Get new reference before dropping old one.
2469 Severity : enhancement
2470 Frequency : if client is started with failover MDS
2472 Description: Allow multiple MDS hostnames in the mount command
2473 Details : Try to read the configuration from all specified MDS
2474 hostnames during a client mount in case the "primary"
2477 Severity : enhancement
2479 Description: Stop sending data to evicted clients as soon as possible.
2480 Details : Check if the client we are about to send or are sending data to
2481 was evicted already. (Check is done every second of waiting,
2482 for which l_wait_event interface was extended to allow checking
2483 of exit condition at specified intervals).
2486 Frequency : rare, normally only when NFS exporting is done from client
2488 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2490 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2493 Frequency : always, if srand() is called before liblustre initialization
2495 Description: Liblustre uses system PRNG disturbing its usage by user application
2496 Details : Introduce internal to lustre fast and high-quality PRNG for
2497 lustre usage and make liblustre and some other places in generic
2498 lustre code to use it.
2500 Severity : enhancement
2501 Bugzilla : 9477, 9557, 9870
2502 Description: Verify that the MDS configuration logs are updated when xml is
2503 Details : Check if the .xml configuration logs are newer than the config
2504 logs stored on the MDS and report an error if this is the case.
2505 Request --write-conf, or allow starting with --old_conf.
2507 Severity : enhancement
2509 Description: Handle symlinks in the path when checking if Lustre is mounted.
2510 Details : Resolve intermediate symlinks when checking if a client has
2511 mounted a filesystem to avoid duplicate client mounts.
2516 Description: lconf can hit an error exception but still return success.
2517 Details : The lconf command catches the Command error exception at the top
2518 level script context and will exit with the associated exit
2519 status, but doesn't ensure that this exit status is non-zero.
2524 Description: failure of ptlrpc thread startup can cause oops
2525 Details : Starting a ptlrpc service thread can fail if there are a large
2526 number of threads or the server memory is very fragmented.
2527 Handle this without oopsing.
2530 Frequency : always, only if liblustre and non-default acceptor port was used
2532 Description: liblustre cannot connect to servers with non-default acceptor port
2533 Details : tcpnal_set_default_params() was not called and was therefore
2534 ignoring the environment varaible TCPNAL_PORT, as well as other
2535 TCPNAL_ environment variables
2540 Description: two objects could be created on the same OST for a single file
2541 Details : If an OST is down, in some cases it was possible to create two
2542 objects on a single OST for a single file. No problems other
2543 than potential performance impact and spurious error messages.
2547 Bugzilla : 5681, 9562
2548 Description: Client may oops in ll_unhash_aliases
2549 Details : Client dcache may become inconsistent in race condition.
2550 In some cases "getcwd" can fail if the current directory is
2556 Description: Inode refcounting problems in NFS export code
2557 Details : link_raw functions used to call d_instantiate without obtaining
2558 extra inode reference first.
2562 Bugzilla : 9942, 9903
2563 Description: Referencing freed requests leading to crash, memleaks with NFS.
2564 Details : We used to require that call to ll_revalidate_it was always
2565 followed by ll_lookup_it. Also with revalidate_special() it is
2566 possible to call ll_revalidate_it() twice for the same dentry
2567 even if first occurence returned success. This fix changes semantic
2568 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2569 reference on a request referred from the intent.
2570 ll_intent_release() then releases such a request.
2573 Frequency : rare, normally benchmark loads only
2575 Description: unlinked inodes were kept in memory on the client
2576 Details : If a client is repeatedly creating and unlinking files it
2577 can accumulate a lot of stale inodes in the inode slab cache.
2578 If there is no other client load running this can cause the
2579 client node to run out of memory. Instead flush old inodes
2580 from client cache that have the same inode number as a new inode.
2583 Frequency : SLES9 2.6.5 kernel and long filenames only
2584 Bugzilla : 9969, 10379
2585 Description: utime reports stale NFS file handle
2586 Details : SLES9 uses out-of-dentry names in some cases, which confused
2587 the lustre dentry revalidation. Change it to always use the
2591 Frequency : rare, unless heavy write-truncate concurrency is continuous
2592 Bugzilla : 4180, 6984, 7171, 9963, 9331
2593 Description: OST becomes very slow and/or deadlocked during object unlink
2594 Details : filter_destroy() was holding onto the parent directory lock
2595 while truncating+unlinking objects. For very large objects this
2596 may block other threads for a long time and slow overall OST
2597 responsiveness. It may also be possible to get a lock ordering
2598 deadlock in this case, or run out of journal credits because of
2599 the combined truncate+unlink. Solution is to do object truncate
2600 first in one transaction without parent lock, and then do the
2601 final unlink in a new transaction with the parent lock. This
2602 reduces the lock hold time dramatically.
2605 Frequency : rare, 2.4 kernels only
2607 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2608 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2609 before unmounting their filesystem in order to avoid deadlock.
2610 The kernel_locked() function in 2.4 kernels only checks whether
2611 the kernel lock is held, not whether it is this process that is
2612 holding it as 2.6 kernels do.
2617 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2618 Details : The client ptlrpc code may be trying to reconnect to a down
2619 server before a previous connection attempt has timed out.
2620 Increase the reconnect interval to be longer than the connection
2621 timeout interval to avoid sending duplicate connections to
2625 Frequency : echo_client brw_test command
2627 Description: fix echo_client to work with OST preallocated code
2628 Details : OST preallocation code (5137) didn't take echo_client IO path
2629 into account: echo_client calls filter methods outside of any
2630 OST thread and, hence, there is no per-thread preallocated
2631 pages and buffers to use. Solution: hijack pga pages for IO. As
2632 a byproduct, this avoids unnecessary data copying.
2636 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2637 Description: Client can oops in mdc_commit_close() after open replay
2638 Details : It was possible for the MDS to return an open request with no
2639 transaction number in mds_finish_transno() if the client was
2640 evicted, but without actually returning an error. Clients
2641 would later try to replay that open and may trip an assertion
2642 Simplify the client close codepath, and always return an error
2643 from the MDS in case the open is not successful.
2646 Frequency : rare, 2.6 OSTs only
2648 Description: OST may deadlock under high load on fragmented files
2649 Details : If there was a heavy load and highly-fragmented OST filesystems
2650 it was possible to have all the OST threads deadlock waiting on
2651 allocation of biovecs, because the biovecs were not released
2652 until the entire RPC IO was completed. Instead, release biovecs
2653 as soon as they are complete to ensure forward IO progress.
2655 Severity : enhancement
2657 Description: Support for specifying external journal device at mount
2658 Details : If an OST or MDS device is formatted with an external journal
2659 device, this device major/minor is stored in the ext3 superblock
2660 and may not be valid for failover. Allow detecting and
2661 specifying the external journal at mount time.
2666 Description: Mounting an MDS with pending unlinked files may cause oops
2667 Details : target_finish_recovery() calls mds_postrecov() which returned
2668 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2669 considers this an error and immediately begins cleaning up the
2670 lov, just after starting the mds_lov process
2672 Severity : enhancement
2674 Description: Implement 'lfs df' to report actual free space on per-OST basis
2675 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2676 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2677 usage of objects; '-h' to report in human readable format.
2679 ------------------------------------------------------------------------------
2681 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2688 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2689 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2690 at startup during orphan unlinking. Instead, skip buddy update
2691 before setup, as it will be regenerated after recovery is complete.
2696 Description: If an OST is inactive, its locks might reference stale inodes.
2697 Details : lov_change_cbdata() must iterate over all namespaces, even if
2698 they are inactive to clear inode references from the lock.
2700 Severity : enhancement
2701 Frequency : occasional, if non-standard max_dirty_mb used
2703 Description: Client will block write RPCs if not enough grant
2704 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2705 then the client will block writes while waiting for another RPC
2706 to complete instead of consuming its dirty limit. With change
2707 we get improved performance when max_dirty_mb is small.
2709 Severity : enhancement
2710 Bugzilla : 3389, 6253
2711 Description: Add support for supplementary groups on the MDS.
2712 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2713 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2714 lookups for user supplementary groups into a cache.
2718 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2719 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2720 for liblustre O_EXCL is enforced by MDS.
2724 Description: readdir never returns NULL in liblustre.
2725 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2726 dentry in struct dirent.
2730 Frequency : liblustre only, depends on application IO pattern
2731 Description: liblustre clients evicted if not contacting servers
2732 Details : Don't put liblustre clients into the ping_evictor list, so
2733 they will not be evicted by the pinger ever.
2735 Severity : enhancement
2737 Description: Add ability to evict clients by NID from MDS.
2738 Details : By echoing "nid:$NID" string into
2739 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2740 $NID would be instantly evicted from this MDS and from all active
2741 OSTs connected to it.
2745 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2746 Details : lookup_it_finish() used to query file size from OSTs that was not
2751 Description: service threads change working directory to that of init
2752 Details : Starting lustre service threads may pin the working directory
2753 of the parent thread, making that filesystem busy. Threads
2754 now change to the working directory of init to avoid this.
2758 Frequency : during shutdown only
2759 Description: shutdown with a failed MDS or OST can cause unmount to hang
2760 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2765 Frequency : 2.6 only
2766 Description: chmod/chown may include an extra supplementary group
2767 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2768 supplementary group and if none is specified this is used.
2771 Bugzilla : 5479 (6816)
2773 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2774 Details : If lookup is in progress on a file that is unlinked we might try
2775 to revalidate the inode and fail in revalidate after lookup is
2776 complete and ll_file_open() enqueues the open again but
2777 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2780 Frequency : always, if lconf --abort_recovery used
2782 Description: lconf --abort_recovery fails with 'Operation not supported'
2783 Details : lconf was attempting to abort recovery on the MDT device and not
2786 ------------------------------------------------------------------------------
2788 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2793 Frequency : rare (only unsupported configurations with a node running as an
2795 Bugzilla : 6514, 5137
2796 Description: Mounting a Lustre file system on a node running as an OST could
2798 Details : OSTs now preallocates memory needed to write out data at
2799 startup, instead of when needed, to avoid having to
2800 allocate memory in possibly low memory situations.
2801 Specifically, if the file system is mounted on on OST,
2802 memory pressure could force it to try to write out data,
2803 which it needed to allocate memory to do. Due to the low
2804 memory, it would be unable to do so and the node would
2805 become unresponsive.
2807 Severity : enhancement
2809 Description: Addition of lconf --service command line option
2810 Details : lconf now accepts a '--service <arg>' option, which is
2811 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2813 Severity : enhancement
2815 Description: Failover mode is now the default for OSTs.
2816 Details : By default, OSTs will now run in failover mode. To return to
2817 the old behaviour, add '--failout' to the lmc line for OSTs.
2819 Severity : enhancement
2821 Description: Health checks are now provided for MDS and OSTs
2822 Details : Additional detailed health check information on MSD and OSTs
2823 is now provided through the procfs health_check value.
2826 Frequency : occasional, depends on IO load
2828 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2829 after numerous create/delete cycles
2830 Details : The ext3 inode allocation policy would not allocate new inodes
2831 very well on the OSTs because there are no new directories
2832 being created. Instead we look for groups with free space if
2833 the parent directories are nearly full.
2838 Description: Network or server problems during mount may cause partially
2839 mounted clients instead of returning an error.
2840 Details : The config llog parsing code may overwrite the error return
2841 code during mount error handling, returning success instead
2847 Description: MDS can fail to allocate large reply buffers
2848 Details : After long uptimes the MDS can fail to allocate large reply
2849 buffers (e.g. zconf client mount config records) due to memory
2850 fragmentation or consumption by the buffer cache. Preallocate
2851 some large reply buffers so that these replies can be sent even
2852 under memory pressure.
2856 Frequency : rare (liblustre)
2857 Description: fsx running with liblustre complained that using truncate() to
2858 extend the file doesn't work. This patch corrects that issue.
2859 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2860 ATTR_SIZE and lsm use in llu_setattr_raw.
2864 Frequency : rare, only 2.6 kernels
2865 Description: Unusual file access patterns on the MDS may result in inode
2866 data being lost in very rare circumstances.
2867 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2868 code in the 2.6 kernel caused the inode and/or EA data not to
2869 be read from disk, causing single-file corruption.
2873 Frequency : rare, only 2.6 filesystems using extents
2874 Description: Heavy concurrent write and delete load may cause data corruption.
2875 Details : It was possible under high-load situations to have an extent
2876 metadata block in the block device cache from a just-unlinked
2877 file overwrite a newly-allocated data block. We now unmap any
2878 metadata buffers that alias just-allocated data blocks.
2882 Frequency : filesystems with default stripe_count larger than 77
2883 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2884 Details : lconf specifies an inode size of 4096 bytes when the default
2885 stripe_count is larger than 77. This conflicts with the default
2886 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2887 this case to avoid pinning too much memory for large EAs.
2889 ------------------------------------------------------------------------------
2891 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2896 Frequency : rare (extremely heavy IO load with hundreds of clients)
2898 Description: Client is evicted, gets IO error writing to file
2899 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2900 caused clients to be evicted for AST timeouts. The fixes in
2901 bug 5192 mean we no longer need to have such short AST timeouts
2902 so ldlm_timeout has been increased.
2905 Frequency : occasional during --force or --failover shutdown under load
2906 Bugzilla : 5949, 4834
2907 Description: Server oops/LBUG if stopped with --force or --failover under load
2908 Details : a collection of import/export refcount and cleanup ordering
2909 issues fixed for safer force cleanup
2912 Frequency : only filesystems larger than 120 OSTs
2913 Bugzilla : 5990, 6223
2914 Description: lfs getstripe would oops on a very large filesystem
2915 Details : lov_getconfig used kfree on vmalloc'd memory
2918 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2919 Bugzilla : 6242, 6243
2920 Description: reading from files that had been truncated to a non-zero size
2921 but never opened returned no data
2922 Details : ll_file_read() reads zeros from no-object files to EOF
2927 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2928 Details : There's little harm in aborting MDS/OSS recovery and letting it
2929 try again, so I removed the LASSERT and return an error instead.
2931 Severity : enhancement
2933 Description: New debugging infrastructure for tracking down data corruption
2934 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2935 (b) cover more of the client-side code path, and (c) try to narrow
2936 down where problems occurred
2940 Bugzilla : 3819, 4364, 4397, 6313
2941 Description: Racing close and eviction MDS could cause assertion in mds_close
2942 Details : It was possible to get multiple mfd references during close and
2943 client eviction, leading to one thread referencing a freed mfd.
2945 Severity: : enhancement
2946 Bugzilla : 3262, 6359
2947 Description: Attempts to reconnect to servers are now more aggressive.
2948 Details : This builds on the enhanced upcall-less recovery that was added
2949 in 1.4.2. When trying to reconnect to servers, clients will
2950 now try each server in the failover group every 10 seconds. By
2951 default, clients would previously try one server every 25 seconds.
2956 Description: After recovery, certain operations trigger a failed
2957 assertion on a client.
2958 Details : Failing over an mds, using lconf -d --failover, while a
2959 client was doing a readdir() call would cause the client to
2960 LBUG after recovery completed and the readdir() was resent.
2962 Severity : enhancement
2964 Description: Default groups are now added by lconf
2965 Details : You can now run lconf --group <servicename> without having to
2966 manually add groups with lmc.
2969 Frequency : occasional
2971 Description: Nodes with an elan id of 0 trigger a failed assertion
2974 Frequency : always when accessing e.g. tty/console device nodes
2976 Description: tty and some other devices nodes cannot be used on lustre
2977 Details : file's private_data field is used by device data and lustre
2978 values in there got lost. New field was added to struct file to
2979 store fs-specific private data.
2982 Frequency : when exporting Lustre via NFS
2984 Description: NFSD failed occasionally when looking up a path component
2985 Details : NFSD is looking up ".." which was broken in ext3 directories
2986 that had grown large enough to become hashed.
2989 Frequency : Clusters with multiple interfaces not on the same subnet
2991 Description: Nodes will repeatedly try to reconnect to an interface which it
2992 cannot reach and report an error to the log.
2993 Details : Extra peer list entries will be created by lconf with some peers
2994 unreachable. lconf now validates the peer before adding it.
2997 Frequency : Only if a default stripe is set on the filesystem root.
2999 Description: Setting a default stripe on the filesystem root prevented the
3000 filesystem from being remounted.
3001 Details : The client was sending extra request flags in the root getattr
3002 request and did not allocate a reply buffer for the dir EA.
3005 Frequency : occasional, higher if lots of files are accessed by one client
3006 Bugzilla : 6159, 6097
3007 Description: Client trips assertion regarding lsm mismatch/magic
3008 Details : While revalidating inodes the VFS looks up inodes with ifind()
3009 and in rare cases can find an inode that is being freed.
3010 The ll_test_inode() code will free the lsm during ifind()
3011 when it finds an existing inode and then the VFS later attaches
3012 this free lsm to a new inode.
3016 Bugzilla : 6422, 7030
3017 Description: MDS deadlock between mkdir and client eviction
3018 Details : Creating a new file via mkdir or mknod (starting a transaction
3019 and getting the ns lock) can deadlock with client eviction
3020 (gets ns lock and trying to finish a synchronous transaction).
3023 Frequency : occasional
3024 Description: While starting a server, the fsfilt_ext3 module could not be
3026 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3027 kernels. Previously, lconf would still use the ext3 name
3028 when trying to load modules. Now, it will correctly use
3029 ext3 on 2.4 and ldiskfs on 2.6.
3031 Severity : enhancement
3032 Description: The default stripe count has been changed to 1
3033 Details : The interpretation of the default stripe count (0, to lfs
3034 or lmc) has been changed to mean striping across a single
3035 OST, rather than all available. For general usage we have
3036 found a stripe count of 1 or 2 works best.
3038 Severity : enhancement
3039 Description: Add support for compiling against Cray portals.
3040 Details : Conditional compiling for some areas that are different
3044 Frequency : occasional
3045 Bugzilla : 6409, 6834
3046 Description: Creating files with an explicit stripe count may lead to
3047 a failed assertion on the MDS
3048 Details : If some OSTs are full or unavailable, creating files may
3049 trigger a failed assertion on the MDS. Now, Lustre will
3050 try to use other servers or return an error to the
3054 Frequency : occasional
3056 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3057 caused lock timeout during readahead (since 1.4.2).
3058 Details : Processes doing readahead might match a lock that hasn't been
3059 granted yet if there are overlapping and conflicting lock
3060 requests. The readahead process waits on ungranted lock
3061 (original lock is CBPENDING), while OST waits for that process
3062 to cancel CBPENDING read lock and eventually evicts client.
3064 Severity : enhancement
3066 Description: Initial enabling of flock support for clients
3067 Details : Implements fcntl advisory locking and file status functions.
3068 This feature is provided as an optional mount flag (default
3069 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3070 locking are implemented yet, and those that are are not guaranteed
3071 to be completely correct in production environments.
3072 mount -t lustre -o [flock|noflock] ...
3075 Frequency : occasional
3077 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3078 assertion in the ext3 JBD (journaling) layer.
3079 Details : The b_committed_data struct is protected by the big kernel lock
3080 in 2.4 kernels, serializing journal_commit_transaction() and
3081 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3082 access to this struct. In 2.6 kernels there is finer grained
3083 locking to improve SMP performance of the JBD layer.
3087 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3092 Description: merge and cleanup kernel patches.
3095 -----------------------------------------------------------------------------
3097 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3099 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3100 versions of Lustre. Please update all servers and clients to
3101 version 1.4.2 or later at the same time. You must also run
3102 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3103 to update the configuration logs.
3105 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3106 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3107 - swab llog records of type '0' so we get proper header size/idx (5861)
3108 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3109 - fix rename of one directory over another leaking an inode (5953)
3110 - avoid SetPageDirty on 2.6 (5981)
3111 - don't re-add just-being-destroyed locks to the waiting list (5653)
3112 - when creating new directories, inherit the parent's custom
3113 striping settings if present parent (3048)
3114 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3115 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3116 - don't hold client locks on temporary worklist from l_lru (5666)
3117 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3118 - automatically evict dead clients (5921)
3119 - Update file size properly in create+truncate+fstat case (6196)
3120 - Do not unhash mountpoint dentries, do not allow removal of
3122 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3123 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3124 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3125 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3126 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3127 - Correctly handle reads of files with no objects (6243)
3128 - lctl recover will also mark a device active if deactivate used (5933)
3130 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3131 - allow --write-conf on an MDS with different nettype than client (5619)
3132 - don't write config llogs to MDS for mounts not from that MDS (5617)
3133 - lconf should create multiple TCP connections from a client (5201)
3134 - init scripts are now turned off by default; run chkconfig --on
3135 lustre and chkconfig --on lustrefs to use them
3136 - upcalls are no longer needed for clients to recover to failover
3138 - add --abort-recovery option to lconf to abort recovery on device
3140 - add support for an arbitrary number of OSTs (3026)
3141 - Quota support protocol changes.
3142 - forward compatibility changes to wire structs (6007)
3143 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3144 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3145 - improved "lustre status" script
3146 - initialize blocksize for non-regular files (6062)
3147 - added --disable-server and --disable-client configure options (5782)
3148 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3149 - Vanilla 2.4.29 support
3150 - increase maximum number of obd devices to 520 (6242)
3151 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3152 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3153 SLES 9 2.6.5 kernels
3154 - limit stripes per file to 160 (the maximum EA size) (6093)
3156 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3159 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3160 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3161 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3162 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3163 - small ext3 extents cleanups and fixes (5733)
3164 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3165 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3166 - avoid needless client->OST connect, fix handle mismatch (5317)
3167 - fix DLM error path that led to out-of-sync client, long delays (5779)
3168 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3169 - fix several locking issues related to i_size (5492,5624,5654,5672)
3170 - don't move pending lock onto export if it is already evicted (5683)
3171 - fix kernel oops when creating .foo in unlinked directory (5548)
3172 - fix deadlock in obdfilter statistics vs. object create (5811)
3173 - use time_{before,after} to avoid timer jiffies wrap (5882)
3174 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3175 - Do not leak request if server was not able to process it (5154)
3176 - If mds_open unable to find parent dir, make that negative lookup(5154)
3177 - don't create new directories with extent-mapping (5909, 5936)
3179 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3180 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3181 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3182 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3183 - support for RHEL4 kernel on i686 (5773)
3184 - provide error messages when incompatible logs are encountered (5898)
3186 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3187 * version 1.4.0.10 (1.4.1 release candidate 1)
3189 - don't keep a lock reference when lock is not granted (4238)
3190 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3191 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3192 - drop import inflight refcount on signal_completed_replay error (5255)
3193 - unlock page after async write error during send (3677)
3194 - handle missing objects in filter_preprw_read properly (5265)
3195 - no transno return for symlink open, don't save no-trasno open (3440)
3196 - don't try to complete elan receive that already failed (4012)
3197 - free RPC server reply state on error (5406)
3198 - clean up thread from ptlrpc_start_thread() on error (5160)
3199 - readahead could read extra page into cache that wasn't ejected (5388)
3200 - prevent races in class_attach/setup/cleanup/detach (5260)
3201 - don't dereference de->d_inode after l_dput of de (5458)
3202 - use "int" for stripe value returned from lock_to_stripe (5544)
3203 - mballoc allocation and error-checking fixes in 2.6 (5504)
3204 - block device patches to fix I/O request sizes in 2.6 (5482)
3205 - look up hostnames for IB nals (5602)
3206 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3207 - don't start multiple acceptors for the same port (5277)
3208 - fix incorrect LASSERT in mds_getattr_name (5635)
3209 - export a proc file for general "ping" checking (5628)
3210 - fix "lfs check" to not block when the MDS is down (5628)
3212 - service request history (4965)
3213 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3214 - create an "evict_client" /proc entry on OSTs, like the MDS has
3215 - fix mount usage message, return errors per mount(8) (5168)
3216 - change grep [] to grep "[]" in tests so they work in more UMLs
3217 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3218 - remove ancient LOV_MAGIC_V0 EA support (5047)
3219 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3220 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3221 - allow readahead allocations to fail when low on memory (5383)
3222 - mmap locking landed again, after considerable improvement (2828)
3223 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3225 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3228 - send OST transaction number in read/write reply to free req (4966)
3229 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3230 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3231 - the watchdog thread now runs as interruptible (5246)
3232 - flock/lockf fixes (but it's still disabled, pending 5135)
3233 - don't use EXT3 constants in llite code (5094)
3234 - memory shortage at startup could cause assertion (5176)
3236 - reorganization of lov code
3237 - single portals codebase
3239 - add extents/mballoc support (5025)
3240 - direct I/O reads in the obdfilter (4048)
3241 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3243 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3246 - send OST transaction number in read/write reply to free req (4966)
3247 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3248 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3249 - the watchdog thread now runs as interruptible (5246)
3250 - handle missing objects in filter_preprw_read properly (5265)
3251 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3252 - drop import inflight refcount on signal_completed_replay error (5255)
3253 - unlock page after async write error during send (3677)
3254 - return original error code on reconstructed replies (3761)
3255 - no transno return for symlink open, don't save no-trasno open (3440)
3257 - add pid to ldlm debugging output (4922)
3258 - bump the watchdog timeouts -- we can't handle 30sec yet
3259 - extra debugging for orphan dentry/inode bug (5259)
3261 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3264 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3265 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3266 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3267 - fix NULL dereference in /proc/sys/portals/routes (4827)
3268 - allow failed mdc_close() operations to be interrupted (4561)
3269 - stop precreate on OST before MDS would time out on it (4778)
3270 - don't send partial-page writes before EOF from client (4410)
3271 - discard client grant for sub-page writes on large-page clients (4520)
3272 - don't free dentries not owned by NFS code, check generation (4806)
3273 - fix lsm leak if mds_create_objects() fails (4801)
3274 - limit debug_daemon file size, always print CERROR messages (4789)
3275 - use transno after validating reply (3892)
3276 - process timed out requests if import state changes (3754)
3277 - update mtime on OST during writes, return in glimpse (4829)
3278 - add mkfsoptions to LDAP (4679)
3279 - use ->max_readahead method instead of zapping global ra (5039)
3280 - don't interrupt __l_wait_event() during strace
3282 - add software watchdogs to catch hung threads quickly (4941)
3283 - make lustrefs init script start after nfs is mounted
3284 - fix CWARN/ERROR duplication (4930)
3285 - return async write errors to application if possible (2248)
3286 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3287 - print NAL number in %x format (4645)
3288 - update barely-supported suse-2.4.21-171 series (4842)
3289 - support for sles 9 %post scripts
3290 - support for building 2.6 kernel-source packages
3291 - support for sles km_* packages
3293 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3296 - ignore -ENOENT errors in osc_destroy (3639)
3297 - notify osc create thread that OSC is being cleaned up (4600)
3298 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3299 - reconstruct ost_handle() like mds_handle() (4657)
3300 - create a new thread to do import eviction to avoid deadlock (3969)
3301 - let lconf resolve symlinked-to devices (4629)
3302 - don't unlink "objects" from directory with default EA (4554)
3303 - hold socknal file ref over connect in case target is down (4394)
3304 - allow more than 32000 subdirectories in a single directory (3244)
3305 - fix blocks count for O_DIRECT writes (3751)
3306 - OST returns ENOSPC from object create when no space left (4539)
3307 - don't send truncate RPC if file size isn't changing (4410)
3308 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3309 - bind to privileged port in socknal and tcpnal (3689)
3311 - rate limit CERROR/CWARN console message to avoid overload (4519)
3312 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3313 - basic mmap support (3918)
3314 - kernel patch series update from b1_4 (4711)
3316 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3319 - avoid crash during MDS cleanup with OST shut down (2775)
3320 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3321 - don't use bad inodes on the MDS (3744)
3322 - dynamic object preallocation to improve recovery speed (4236)
3323 - don't hold spinlock over lock dumping or change debug flags (4401)
3324 - don't zero obd_dev when it is force cleaned (3651)
3325 - print grants to console if they go negative (4431)
3326 - "lctl deactivate" will stop automatic recovery attempts (3406)
3327 - look for existing locks in ldlm_handle_enqueue() (3764)
3328 - don't resolve lock handle twice in recovery avoiding race (4401)
3329 - revalidate should check working dir is a directory (4134)
3331 - don't always mark "slow" obdfilter messages as errors (4418)
3333 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3336 - don't close LustreDB during write_conf until it is done (3860)
3337 - fix typo in lconf for_each_profile (3821)
3338 - allow dumping logs from multiple threads at one time (3820)
3339 - don't allow multiple threads in OSC recovery (3812)
3340 - fix debug_size parameters (3864)
3341 - fix mds_postrecov to initialize import for llog ctxt (3121)
3342 - replace config semaphore with spinlock (3306)
3343 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3344 - don't allow enqueue to complete on a destroyed export (3822)
3345 - down write_lock before checking llog header bitmap (3825)
3346 - recover from lock replay timeout (3764)
3347 - up llog sem before sending rpc (3652)
3348 - reduce ns lock hold times when setting kms (3267)
3349 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3350 - fix NULL deref and obd_dev leak on setup error (3312)
3351 - replace some LBUG about llog ops with error handling (3841)
3352 - don't match INVALID dentries from d_lookup and spin (3784)
3353 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3354 - fix invalid assertion in ptlrpc_set_wait (3880)
3356 - add libwrap support for the TCP acceptor (3996)
3357 - add /proc/sys/portals/routes for non-root route listing (3994)
3358 - allow setting MDS UUID in .xml (2580)
3359 - print the stack of a process that LBUGs (4228)
3361 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3364 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3365 - make sure to unset replay flag from failed open requests (3440)
3366 - if default stripe count is 0, use OST count for inode size (3636)
3367 - update parent mtime/ctime on client for create/unlink (2611)
3368 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3369 - free recovery state on server during a forced cleanup (3571)
3370 - unregister_reply for resent reqs (3063)
3371 - loop back devices mounting and status check on 2.6 (3563)
3372 - fix resource-creation race that can provoke i_size == 0 (3513)
3373 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3374 - more debugging for page-accounting assertion (3746)
3375 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3376 - avoid deadlock after precreation failure (3758)
3377 - fix race and lock order deadlock in orphan handling (3450, 3750)
3378 - add validity checks when grabbing inodes from l_ast_data (3599)
3380 - add /proc/.../recovery_status to obdfilter (3428)
3381 - lightweight CDEBUG infrastructure, debug daemon (3668)
3382 - change default OSC RPC parameters to be better on small clusters
3383 - turn off OST read cache for files smaller than 32MB
3384 - install man pages and include them in rpms (3100)
3385 - add new init script for (un)mounting lustre filesystems (2593)
3386 - run chkconfig in %post for init scripts (3701)
3387 - drop scimac NAL (unmaintained)
3389 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3392 - clean kiobufs before and after use (3485)
3393 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3394 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3395 - change init script to fail more clearly if not run as root (1528)
3396 - allow clients to reconnect during replay (1742)
3397 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3398 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3399 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3400 - fix cond_resched (3554)
3401 - only evict unfinished clients after recovery (3515)
3402 - allow bulk resend, prevent data loss (3570)
3403 - dynamic ptlrpc request buffer allocation (2102)
3404 - don't allow unlinking open directory if it isn't empty (2904)
3405 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3406 - remove extraneous obd dereference causing LASSERT failure (3334)
3407 - don't use get_cycles() when creating temp. files on the mds (3156)
3408 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3409 - handle EEXIST for set-stripe, set proper directory name (3336)
3411 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3412 - fix ksocknal_fmb_callback() error messages (2918)
3414 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3417 - don't copy lvb into (possibly NULL) reply on error (2983)
3418 - don't deref dentry after dput, don't free lvb on error (2922)
3419 - use the kms to determine writeback rpc length (2947)
3420 - increment oti_logcookies when osc is inactive (2948)
3421 - update client's i_blocks count via lvb messages (2543)
3422 - handle intent open/close of special files properly (1557)
3423 - mount MDS with errors=remount-ro, like obdfilter (2009)
3424 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3425 - don't use cancelling-locks' kms values (2947)
3426 - use highest lock extent for kms, not last one (2925)
3427 - don't dereference ERR_PTR() dentry in error handling path (3107)
3428 - fix thread race in portals_debug_dumplog() (3122)
3429 - create lprocfs device entries at setup instead of at attach (1519)
3430 - common AST error handler, don't evict client on completion race (3145)
3431 - zero nameidata in detach_mnt in 2.6 (3118)
3432 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3433 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3434 - initialize RPC timeout timer earlier for 2.6 (3219)
3435 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3436 - print nal/nid for unknown nid (3258)
3437 - additional checks for oscc recovery before doing precreate (3284)
3438 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3439 - don't crash in mdc_close for bad permissions on open (3285)
3440 - zero i_rdev for non-device files (3147)
3441 - clear page->private before handing to FS, better assertion (3119)
3442 - tune the read pipeline (3236)
3443 - fix incorrect decref of invalidated dentry (2350)
3444 - provide read-ahead stats and refine rpc in flight stats (3328)
3445 - don't hold journal transaction open across create RPC (3313)
3446 - update atime on MDS at close time (3265)
3447 - close LDAP connection when recovering to avoid server load (3315)
3448 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3449 - don't leak open file on MDS after open resend (3325)
3450 - serialize filter_precreate and filter_destroy_precreated (3329)
3451 - loop device shouldn't call sync_dev() for nul device (3092)
3452 - clear page cache after eviction (2766)
3453 - resynchronize MDS->OST in background (2824)
3454 - refuse to mount the same filesystem twice on same mountpoint (3394)
3455 - allow llmount to create routes for mounting behind routers (3320)
3456 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3457 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3458 - fix rare problem with rename on htree directories (3417)
3460 - allow default OST striping configuration per directory (1414)
3461 - fix compilation for qswnal for 2.6 kernels (3125)
3462 - increase maximum number of MDS request buffers for large systems
3463 - change liblustreapi to be useful for external progs like lfsck (3098)
3464 - increase local configuration timeout for slow disks (3353)
3465 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3467 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3470 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3471 - don't overwrite extent policy data in reply if lock was blocked (2901)
3472 - drop filter export grants atomically with removal from device (2663)
3473 - del obd_self_export from work_list in class_disconnect_exports (2908)
3474 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3475 - swab reply message in mdc_close, other PPC fixes (2464)
3476 - fix destroying of named logs (2325)
3477 - overwrite old logs when running lconf --write_conf (2264)
3478 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3479 - fix race in target_handle_connect (2898)
3480 - mds_reint_create() should take same inode create lock (2926)
3481 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3482 - don't close files for self_export to avoid uninitialized obd (2936)
3483 - allow MDS with the same name as client node (2939)
3484 - hold dentry reference for closed log files for unlink (2325)
3485 - reserve space for all logs during transactions (2059)
3486 - don't evict page beyond end of stripe extent (2925)
3487 - don't oops on a deleted current working directory (2399)
3488 - handle hard links to targets without a parent properly (2517)
3489 - don't dereference NULL lock when racing during eviction (2867)
3490 - don't grow lock extents when lots of conflicting locks (2919)
3492 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3495 - account for cache space usage on clients to avoid data loss (974)
3496 - lfsck support in lustre kernel code (2349)
3497 - reduce journal credits needed for BRW writes (2370)
3498 - orphan handling to avoid losing space on client/server crashes
3499 - ptlrpcd can be blocked, stopping ALL progress (2477)
3500 - use lock value blocks to assist in proper KMS, faster stat (1021)
3501 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3502 - recovery for initial connections (2355)
3503 - fixes for mds_cleanup_orphans (1934)
3504 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3505 - block all file creations until orphan recovery completes (1901)
3506 - client remove rq_connection from request struct (2423)
3507 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3508 - recovery timer race (2670)
3509 - mdc_close recovey bug (2532)
3510 - ptlrpc cleanup bug (2710)
3511 - mds timeout on local locks (2588)
3512 - namespace lock held during RPCs (2431)
3513 - handle interrupted sync write properly (2503)
3514 - don't try to handle a message that hasn't been replied to (2699)
3515 - client assert failure during cleanup after abort recovery (2701)
3516 - leak mdc device after failed mount (2712)
3517 - ptlrpc_check_set allows timedout requests to complete (2714)
3518 - wait for inflight reqs when ptlrpcd finishes (2710)
3519 - make sure unregistered services are removed from the srv_list
3520 - reset bulk XID's when resending them (caught by 1138 test)
3521 - unregister_bulk after timeout
3522 - fix lconf error (2694)
3523 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3524 - readahead locks pages, leaves pending causing memory pressure (2673)
3525 - increase OST request buffers to 4096 on large machines (2729)
3526 - fix up permission of existing directories in simple_mkdir (2661)
3527 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3528 - don't assign transno to errored transactions (2742)
3529 - don't delete objects on OST if given a bogus objid from MDS (2751)
3530 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3531 - if rq_no_resend, then timeout request after recovery (2432)
3532 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3533 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3534 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3535 - MDS recovery shouldn't delete objects that it has given out (2730)
3536 - if enqueue arrives after completion, don't clobber LVB (2819)
3537 - don't unlock pages twice when trigger_group_io returns error (2814)
3538 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3539 - don't write pages to disk if there was an error (1450)
3540 - don't ping imports that have recovery disabled (2676)
3541 - take buffered bytes into account when balancing socknal conn (2817)
3542 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3543 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3544 - remove little-endian swabbing of llog records (1987)
3545 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3546 - retry reposting request buffers if they fail (1191)
3547 - grow extent at grant time to avoid granting a revoked lock (2809)
3548 - lock revoke doesn't evict page if covered by a second lock (2765)
3549 - disable VM readahead to avoid reading outside lock extents (2805)
3551 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3552 - updated kernel patches for hp-2.4.20 kernel (2681)
3554 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3557 - fix truncated write corruption (2366)
3558 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3560 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3561 - ensure len(uuid) < 37 in lmc (1171)
3562 - fix ia64 OOPS in llog_test (2255)
3563 - zero end of page at obdfilter for partial page writes (2648)
3564 - don't leave stale dentries around after renames (bug 2428)
3565 - fix timeouts when evicting a client with a single lock held (2642)
3566 - set deadline for the initial HELLO message to drain (2634)
3567 - print out dotted-quad IP addresses in the socknal (2302)
3569 - additional debugging for MDS client eviction problem (2443)
3570 - fix mkfsoptions support for osts (2603, 2604)
3572 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3575 - add series for the vanilla 2.6.0 kernel
3576 - add series for the vanilla 2.4.24 kernel
3577 - add series for a cray x86/64 UL kernel drop
3578 - fix xattr patches for the vanilla 2.4.19 series
3580 - generate true UUIDs in lmc (1171)
3581 - have portals stack dumping break in UML (2466)
3582 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3583 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3584 - wake callback waiting threads on client eviction (2460)
3585 - Add --ptldebug and --subsystem to lmc (1719)
3586 - update assertion to allow safe interrupt allocation
3587 - set rq_no_resend for cancel requests (2432)
3588 - recalculate ptlrpcd timeout after resend (2494)
3589 - call vfs_rmdir when removing pending directories (2368)
3590 - fix renaming a file to itself (2429)
3591 - lmc creates a default one-stripe lov (2454)
3592 - expand procfs space to handle large clusters (2326)
3593 - increase UML stack to avoid overflow
3594 - update lconf's list of debug and subsystem masks
3595 - fix lfs find --obd (2510)
3596 - /proc tunable for disabling filter read caching (2591)
3597 - stop rpm packages from altering slapd.conf (2301)
3598 - disable nagle in the socknal under 0conf (2578)
3599 - choose mds inode size based on stripe count (2572)
3600 - fix kernel-source rpm problems (2516)
3602 - add --disable-doc to avoid pdf generation (2421)
3603 - update documentation, tests, type-os, comments
3604 - avoid format warnings on ia64
3605 - remove the TOE NAL
3606 - tiny code cleanups by removing unused fields
3608 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3611 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3612 - remove the most blatant lies from BUILDING file (2371)
3613 - change default debug level to reasonable production setting
3614 - reduce client side cache size to reduce cache flush time
3615 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3616 - make TCP zerocopy and pinger support enabled by default (2476)
3617 - sync writes completed after process exits caused crashes (2319)
3618 - maintain correct mount count on the MDS (2356)
3619 - backout 1557, because 2316 wasn't really fixed
3620 - better file I/O statistics gathering in /proc
3621 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3622 - another kernel patch to fix zero-copy TCP function export
3623 - don't take duplicate lock when processing re-sent getattr (2420)
3624 - lctl uses obd_self_export instead of creating new conn (2353)
3625 - MDS/OST recovery case which requires object creation asserted (2425)
3626 - move lfs from /usr/sbin to /usr/bin in packages
3627 - fix race between mds_client_add and mds_client_free (2417)
3628 - use kmalloc instead of slabs in portals (2430)
3629 - don't create duplicate records when a failover MDS is present (2442)
3630 - remove unnecessary mount age check (2332)
3631 - don't remove directory inodes from locks prematurely (2451)
3632 - don't break if MDS service name is the same as hostname (2103)
3633 - fix races in client write RPC generation when cache full (2482)
3635 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3638 - remove now-unused request->rq_obd (278)
3639 - if an allocation fails, print out how much memory we've used (1933)
3640 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3641 - add the "configurable stack size" patch to most series files (1256)
3642 - ability to write large log records, for 100+ OST configs (2306)
3643 - fix NULL deref when filter_prep fails (2314)
3644 - fix operator precedence error in filter_sync
3645 - dynamic allocation of socknal TX descriptors (2315)
3646 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3647 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3648 - ensure that CWARN messages in Portals always get to the syslog
3649 - __init/__exit are not for prototype decls (ldlm_init/exit)
3650 - x86-64 compile warning fixes
3651 - fix gateway LMC keyword conflict (2318)
3652 - fix MDS lock inversions in getattr/reint paths (1844)
3653 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3654 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3655 - DLM race condition prevented some lock evictions (2328)
3656 - ENOMEM detection and retry on socknal sends (2230)
3657 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3658 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3659 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3660 - better error messages when a client is rejected during recovery (1505)
3661 - avoid cancelling locks which were never granted, after failure (2330)
3662 - fix i_sem/journal inversion in mds_client_add (2333)
3663 - fix truncate/getattr lock cycle deadlock (2334)
3664 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3665 - fix two rare exit paths which could leak an l_lock() ref (2321)
3666 - fencepost error in MDS/OST orphan recovery (2226)
3667 - make log record alignment 8 bytes (1988)
3668 - lstripe now fails when requested offset > ost_count (2237)
3669 - ensure that all kernel series have a complete list.h (1607)
3670 - fix crashes in special-file operations (2316)
3671 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3673 - allow configurable automake binary, for testing new versions
3674 - small update to the lfs documentation
3676 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3678 * fix negative export reference count in fsfilt_sync (2312)
3680 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3681 * release candidate 0.9.1
3683 - orphans are moved into the PENDING directory for possible recovery
3684 - replayed opens now open by fid for orphan/rename safety (1042)
3685 - last close of an orphan inode generates a transno (683)
3686 - chdir() and mount() now pin the directory entry (1020)
3687 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3688 - discard very old requests without processing them (1502)
3689 - remove some common, well-understood CERRORs (1505)
3690 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3691 - clear "grant" flags in OST replies until OST grant code lands (1644)
3692 - fix read performance by not clobbering i_blksize on client (1598)
3693 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3694 - make LRU size a /proc tunable, clears locks when reduced (707)
3695 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3696 - server locks take references on exports now (1558)
3697 - build fixes for 2.4.20-rh trees (1663)
3698 - return an error from lov_create if all OSCs are inactive (1751)
3699 - fix import levels when a reconnect happens without a timeout (1597)
3700 - exit early from mds_open if we get a lookup error (1749)
3701 - partial page read at EOF wouldn't wait for disk before sending (1642)
3702 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3703 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3704 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3705 - write/truncate lock inversion (1639)
3706 - Don't auto-load obdclass, portals modules during cleanup (1495)
3707 - fix timestamps from jumping to "now" (1763)
3708 - extra journal assertions (1648)
3709 - add an extra multiunlink test (1771)
3710 - fix read_record/write_record API (1776)
3711 - fix leak of offset_extent, possible incorrect i_size later (1772)
3712 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3713 - Debugging for the kqswnal_get_idle_tx problems (1820)
3714 - Allow recovery to be attempted multiple times (1536)
3715 - Write out MDS last_rcvd file after it is first created (1600)
3716 - Fix tx_descriptor leak in failed transmit situations (1827)
3717 - ext3 journaling fixes for assertion failure after IO error (1871)
3718 - class_export_put() on freed export after completion AST error (1896)
3719 - Fix revalidate looping in VFS (1322)
3720 - Don't access a freed export during MDS_REINT timeout (1521)
3721 - Add open-unlink recovery support on the MDS (1673,1764)
3722 - Return an error if no MDS data was read from last_rcvd (1946)
3723 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3724 - Don't setup a disk device that doesn't match exported UUID (317)
3725 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3726 - avoid committing NULL handle in force close
3727 - local.sh is now a one-stripe LOV configuration
3728 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3729 - fix ext3 htree duplicate directory entry corruption (1516)
3730 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3731 - update to new LOV EA format (2097)
3732 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3734 2003-06-15 Phil Schwan <phil@clusterfs.com>
3737 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3738 - per-import/export recovery handling (958, 931, 959)
3739 - multiple last-rcvd slots, for serving multiple FSes (949)
3740 - connections are again shared between multiple imp/exports (963, 964)
3741 - "umount -f" would hang if any requests needed to be sent (393, 978)
3742 - avoid pinning large req buffer by copying for queued messages (989)
3743 - add "uuid" to "lctl device" command to help upcalls (991)
3744 - "open" RPCs with transnos would confuse recovery counters (1037)
3745 - do proper endian conversion of all wire messages (288, 340, 891)
3746 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3747 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3748 - fix ptlrpc_connection leak in target_handle_connect (1174)
3749 - fix import refcounting bug in OST and MDS cleanup (1134)
3750 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3751 - fix very unlikely obd_types race condition (501)
3752 - remove osc_open hack for echo_client (1187)
3753 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3754 - a failure in read_inode2 leads to deadlock (1139)
3755 - cancel ack-locks as soon as transaction is committed (1072)
3756 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3757 - make sure to commitrw after any preprw to avoid deadlock (1162)
3758 - failing to execute a file in a lustre FS would lock inode (1203)
3759 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3760 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3761 - obd_uuid2tgt would walk off the end of the list (1255)
3762 - on IA64 the osc would give portals incorrect bulk size (1258)
3763 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3764 - fix lock inversion caused by new llite matching code (1282)
3765 - limit the number of dirty pages on a client to 10MB (1286)
3766 - timed out locks were not being corrected cancelled (1289)
3767 - fix O_DIRECT above 4GB on IA-32 (1292)
3768 * major user-visible changes
3769 - fail out/fail over policy now controlled by the upcall (993)
3771 - add OBD_PING to check server availability and failure (954)
3772 - lustre messages are now sent in sending host order (288, 340, 891)
3773 - add eadatalen to MDS getattr reply (340)
3774 - OST read replies may contain second buffer, with per-page status (593)
3776 2003-03-11 Phil Schwan <phil@clusterfs.com>
3779 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3780 - failed open()s could cause deadlock; fixed (867, 869)
3781 - stop cancelling OST locks when files are closed (481)
3782 - overlapping XID spaces caused network corruption (851, 853)
3783 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3784 - setattr_raw added, to do single-RPC, server-side setattrs
3785 - lmc/lconf syntax change for OST UUIDs
3786 - fix crashy race condition between ptlrpc_free_req and osc_close
3787 - don't use request in mdc_enqueue if we hit a timeout (889)
3788 - don't set the inode i_size for regular files from the MDS (896)
3789 - handle out of order completion AST (842)
3790 - don't LBUG if a lock request times out after receiving AST (913)
3791 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3792 - if a bad lock AST arrives, send an error instead of dropping entirely
3793 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3794 - fix leak in bulk IO when only partially completed (899, 900, 926)
3795 - fix O_DIRECT for ia64 (55)
3796 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3797 - C-z after timeout could hang a process forever; fixed (977)
3799 - client-side I/O cache (678, 924, 929, 941, 970)
3801 - READPAGE and SETATTRs which don't take server-side locks get
3804 2003-02-11 Phil Schwan <phil@clusterfs.com>
3807 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3808 - Fix fsfilt_extN_readpage() to read a full page of directory
3809 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3810 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3811 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3812 - Most intent operations were removed, in favour of a new RPC mode
3813 that does a single RPC to the server and bypasses most of the VFS
3814 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3815 - Aggressively cancel local locks on DLM servers
3816 - mds_reint_unlink sends EA to the client if it's the last nlink.
3817 client uses that EA to unlink OST objects.
3818 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3819 - recursive symlinks were fixed (439)
3820 - fixed NULL deref in DEBUG_REQ
3821 - filter_update_lastobjid no longer calls sync, which annoyed extN
3822 - fixed multi-client small-writes to a single file problem (445)
3823 - fixed mtime updates during file writes (607)
3824 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3825 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3826 - fixed Linux OST size reporting problem (444, 656)
3827 - OST now updates object mtime with writes or setattr (607, 619)
3828 - client verifies file size before zeroing page past EOF (445)
3829 - OST now writes last allocated objid to disk with allocation (108)
3830 - LOV on echo now works (409)
3832 - mds_reint_unlink sends a new buffer, with the EA included. this
3833 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3834 set if a regular file was being unlinked, and it was the last link
3835 - use PtlGet from the target for bulk writes (315)
3836 - OST now updates object mtime with writes or setattr (607, 619)
3837 - LDLM now has a grant-time callback to revalidate locked items, if
3839 - Many MDS operations were reorganized to combat race conditions
3841 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3842 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3844 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3847 - Fully reactivate OST imports after reconnection (512, others)
3848 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3849 - More graceful error handling for truncating on dead OST (515)
3850 - Don't error out unless we're actually accessing dead stripes (474)
3851 - Fix garbage sizes when stripes are missing (410)
3852 - LRU counters were broken, causing constant lock purge (433, 432)
3853 - garbage on read from stripes with failed OSTs (441)
3854 - mark OSCs as active before reconnecting during recovery (438)
3855 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3856 - lfind did not preserve OST order in output (443)
3857 - symlinks cause hung clients, incorrect data (439)
3858 - stop dereferencing request after dropping refcount (457)
3859 - don't LASSERT(spin_is_locked) on non-SMP (455)
3860 - fixes for many rename() bugs
3861 - fstat didn't correctly synchronize attributes (399)
3862 - server must handle lock cancellation during blocking AST prep (487)
3863 - bulk descriptors were free()d too soon (511)
3864 - fix paths in lconf, which would load incorrect modules (451, 507)
3865 - fix confusing lconf 'host not found' error message (386)
3866 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3867 - fix race condition in mdc_blocking_ast() for inode access (526)
3868 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3869 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3870 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3871 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3872 - fix lstripe in several ways (kernel side) (527)
3873 - fix request leak in ldlm_cli_enqueue (262)
3874 - incorrect OSC was marked inactive after OST failure
3875 - call mds_fs_cleanup before unmounting filesystem (524)
3876 - fix races between taking ns_lock and ldlm_lock_change_resource
3877 - fix races updating LOV export open file list
3878 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3879 - fix recovery NULL deref in ldlm_cli_cancel_unused
3880 - fix some DLM races by using new hash table for lock handles (419)
3881 - permit the client to specify desired inodes, at replay
3882 - duplicate requests when we queue them for replay reintegration
3883 - fix last_rcvd offset calculation
3884 - sync after each recovered transaction, so we always make progress
3885 - never, not always, ERESTART requests without transnos
3886 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3888 - skip replay if the MDS says that the client is already connected
3889 - don't check for a recovery-enabled export to match lctl's UUID
3890 - don't INC_USE_COUNT for phantom exports
3891 - don't crash when cleaning up phantom exports (567)
3892 - don't double-finish or set replay data for errored mdc_open requests
3893 - abort requests when they time out, so we don't get old replies
3894 - send/receive replies for AST messages again
3895 - if the client says that it doesn't have the lock, cancel it on the
3897 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3898 mark it as destroyed, it will all work out when decref is called
3899 - fix module use counts (22, 581)
3901 - ASTs now expect a reply (server cancels lock on error reply)
3903 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3906 - fix many simultaneous client startup (392)
3907 - fix dentry->d_it clobbering
3908 - credentials weren't being shipped for readdir/getattr operations
3909 - remove invalid assertions triggered during some concurrent MD
3911 - proper Lustre versions added (336, 389)
3912 - fix memory leak for create error case (398)
3913 - fix LOV locking bug that would get cli/srv out of sync
3914 - fix echo client over LOV (409)
3915 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3916 - fix double-O_EXCL intent crash (424)
3917 - avoid sending multiple lock CANCELs (352)
3919 - MDS can do multi-client recovery (modulo bugs in new code)
3921 - many updates, edits, cleanups
3923 2002-11-18 Phil Schwan <phil@clusterfs.com>
3926 - fix null d_it dereference (346)
3927 - fix full OST/dbench hang (333)
3928 - fix permission problem with file removal (286)
3929 - fix removal of OSCs from LOV when they fail
3930 - fix NULL deref during bulk timeout (214)
3931 - fix problems related to multiple filesystems on one MDS (241)
3932 - fixed serious subtle metadata locking bugs
3933 - free locks on clients when inodes are removed due to memory
3935 - fix inode pointer in lock data (285)
3936 - partial support for multiple MDS on a single host (241)
3937 - data locks weren't cancelled at clear_inode time (290, 311)
3938 - intent locks could lead to unbounded lock growth (205)
3939 - added a maximum lock count, an LRU list, and a flusher
3940 - fix multiple rename (365)
3941 - properly abstracted the echo client
3942 - OSC locked 1 byte too many; fixed
3943 - rewrote brw callback code:
3944 - fixed recovery bugs related to LOVs (306)
3945 - fixed too-many-pages-in-one-write crash (191)
3946 - fixed (again) crash in sync_io_timeout (214)
3947 - probably fixed callback-related race (385)
3949 - Add capability to MDS protocol
3950 - LDLM cancellations and callbacks on different portals
3952 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3955 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3956 - defer open object destruction to close time (601981, 138)
3957 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3958 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3959 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3960 - don't call obd_setattr in ll_file_release for destroyed objects
3961 * protocol change to lustre_msg: move |version| and add |flags|
3962 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3963 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3964 * added replay of create, unlink, link and rename operations during
3965 MDS failover; recovery should be much more robust now
3966 * remove failed OSCs from LOVs (only lov_create uses this so far)
3967 * the lustre-HOWTO was brought (more) up to date (582544)
3969 2002-10-23 Phil Schwan <phil@clusterfs.com>
3972 - in-use dentries weren't being reused properly (617851)
3973 - prevent multiple LDLM setup (599178)
3974 - fix LOV size calculations for truncate (617853)
3975 - fix client handling of MDS intent errors (POSIX)
3976 - fix permission bug in lovstripe.c test (624321)
3977 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3978 - truncate past end of file could corrupt data
3979 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3980 - a race in recovery could return ETIMEDOUT to apps (623947)
3981 - building outside the source directory was fixed
3982 * the lustre-HOWTO was brought (more) up to date (582544)
3983 * major progress was made on recovery functionality
3985 2002-10-10 Phil Schwan <phil@clusterfs.com>
3988 - recovery deadlock fix
3989 - rm -rf causes LBUG fix (617817)
3990 - file open by multiple tasks fix (618962)
3991 - directory permissions bugs (602707 and 620007)
3992 - journal_stop fixed with locking (611313)
3993 - O_APPEND failures resolved (618273, perhaps 614459)
3994 - lconf PATH fix (619770)
3995 - IA64 build fix (621450)
3996 - RPC buffer sizes scale with amount of memory
3998 2002-10-01 Phil Schwan <phil@clusterfs.com>
4001 - locks would be cancelled without throwing away data pages,
4002 resulting in inconsistent data (605627)
4003 - inode attributes were not always being refreshed (605627, 612449)
4004 - lconf now continues to cleanup after lctl reports an error
4005 - MDS now enforces user permissions (602707)
4006 - lprocfs cleanup fixed, but not yet enabled (614157)
4007 - fixed infinite server hang, should a client not respond to an AST
4008 - avoid going into recovery if user calls readlink() with a buffer
4009 that's too small (613941)
4010 - AST RPCs no longer require replies (614867) -- this may be changed
4011 - don't crash server if client sends an IOV that's too big (611336)
4012 - fixed lock conversion deadlock (611892)
4013 - fixed the following of symlinks (614622)
4014 * recovery: the server can remove locks from a client that dies, other
4015 clients can make progress
4016 * more extN patch fixes
4017 * compile-time configurable ptlrpc buffer allocations
4019 - collaborative read cache document
4020 - Lustre Lite Performance CDR document-in-progress
4022 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4025 - fix typo in patch-2.4.18
4027 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4030 - clear ptlrpc request each time in handle_incoming_request()
4031 - unlink of files now destroys the object on the OST
4033 2002-09-19 Peter Braam <braam@clusterfs.com>
4035 * add hard link support
4036 * change obdfile creation method
4037 * kernel patch changed
4039 2002-09-19 Peter Braam <braam@clusterfs.com>
4042 - stack overflow bug in extN fixed
4044 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4046 * documentation updates
4047 - add man pages for config tools
4048 - update tests/README to describe testing with new config tools
4049 - finish metadata API descriptions
4050 * bug fixes and cleanups
4051 - statfs workaround for 16TB limit
4052 - LOV stripe allocation improved, can stripe on subset of OSTs
4053 - LOV file size/IO offset was wrong for files > 4GB in size
4054 - object EA data was being dropped, caused files to be unreadable
4055 - memory overflow with non-LOV OST caused memory corruption
4056 - fixed regression tests to work with new config tools, obdfilter
4057 - fixed bug when directory size became larger than 1 block
4058 - fixed bug (for single client case) when PWD was deleted
4059 - invalidate local directory pages when doing intent-based ops
4060 - avoid LDLM oops when lock callback contained bad data
4062 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4064 * documentation updates
4065 * bug fixes and cleanups
4066 - configuration tools
4069 - 64-bit compile warnings
4070 - 64-bit internal statfs data
4072 * test_brw on persistent OST devices
4074 * lprocfs (disabled)
4076 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4078 * documentation updates
4079 * bug fixes and cleanups
4080 * configuration tools
4082 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4085 * many small fixes to 0_5_4
4086 * io/network handling
4087 * thinkos in MDS operations
4089 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4092 * crucial basic fixes to 0.5.3
4093 * IOR, Iozone work over Elan
4096 2002-08-07 Phil Schwan <phil@clusterfs.com>
4097 * version 0_5_3, our first alpha
4098 * we use the new Portals iovs
4099 * documentation updates
4100 * bug fixes and cleanups
4101 * small changes in the DLM wire protocol
4103 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4104 * version 0_5_1 with some initial stability,
4105 * locking on MD and file I/O.
4106 * documentation updates
4107 * several bug fixes since 0.5.0
4108 * small changes in wire protocol
4110 2002-07-18 Phil Schwan <phil@clusterfs.com>
4112 * delivered as Lustre Light Alpha
4113 * fixed a crash after handling invalid MDS requests
4114 * fixed directory pages for architectures with non-4k pages sizes
4116 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4117 * release version v0_4_4
4118 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4120 * Fixes a number of interruption problems with OST operations.
4121 * Update documentation for portals header changes
4122 * Move all wire protocol structs/defines to lustre_idl.h
4123 * Fixes symlink length bug.
4124 * Add tcpdump to repository.
4126 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4127 * release version v0_4_3
4128 * Fixes statfs for inodes on extN.
4129 * Fixes bug in runtests which would delete /etc/hosts.
4130 * Use 64-bit object IDs wherever possible (not into VFS though)
4131 Remove ost_get_info, which is unused by lustre, and out of date.
4133 2002-07-03 Peter Braam <braam@clusterfs.com>
4134 * release version v0_4_2 Fixes a lookup error (type not passed)
4135 * move forward to head of Portals
4136 * move forward to latest Lustre kernel
4138 2002-06-25 Peter Braam <braam@clusterfs.com>
4139 * release version v0_4_1. Hopefully stable on single node use.