1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10),
7 2.6.18-8.1.8.EL5 (RHEL 5).
8 2.6.18.8 vanilla (kernel.org)
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.9-42.0.10.EL (RHEL 4),
12 2.6.16 - 2.6.21 vanilla (kernel.org)
13 * Recommended e2fsprogs version: 1.40.2-cfs1
14 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
18 Description: don't use GFP_* in generic Lustre code.
19 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
23 Description: obdclass is limited by single OBD_ALLOC(idarray)
24 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
28 Description: warnings with build patchless client with vanila 2.6.19 and up
29 Details : change old ctl_table style and replace ctl_table/ctl_table_header
30 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
34 Frequency : Only for SLES
35 Description: sanity_quota fail test_1
36 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
37 file, which makes TSTID[2] inunique.
41 Frequency : Only for SLES9
42 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
43 Details : Update target/ChangeLog/which_patch .
47 Frequency : Only for test_10 in sanity-quota.sh
48 Description: a bug in quota test code
49 Details : checking a test flag in a wrong place
53 Frequency : Only for test_13 in sanity-quota.sh
54 Description: a bug in quota test script
55 Details : shouldn't check overfull allocated quota
59 Description: jbd statistics
60 Details : Port older jbd statistics patch for sles10
62 Severity : enhancement
64 Description: Add printing inode info into message about error in writepage.
69 Description: reply_lock_interpret crash due to race with it and lock cancel.
70 Details : Do not replay locks that are being cancelled. Do not reference
71 locks by their address during replay, just by their handle.
75 Frequency : When flocks are used.
76 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
77 Details : Flock locks might destroy just granted lock if it could be merged
78 with another existing flock, this is done in completion handler,
79 so teach ldlm_cli_enquque_fini that this is a valid case for
85 Description: Oops in read and write path when failing to allocate lock.
86 Details : Check if lock allocation failed and return error back.
90 Description: lstripe command fails for valid OST index
91 Details : The stripe offset is compared to lov->desc.ld_tgt_count
92 instead of lov->desc.ld_active_tgt_count.
96 Description: sanity.sh failed test 103
97 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
102 Description: df doesn't work properly if diskfs blocksize != 4K
103 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
107 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
108 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
112 Description: obd_health_check_timeout too short
113 Details : set obd_health_check_timeout as 1.5x of obd_timeout
117 Description: llapi_file_create() does not allow some changes
118 Details : add llapi_file_open() that allows specifying the mode and
119 open flags, and also returns an open file handle.
123 Description: merge and cleanup kernel patches.
124 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
128 Description: Add journal checksum support.(Kernel part)
129 Details : The journal checksum feature adds two new flags i.e
130 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
131 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
132 indicates that the commit block contains the checksum for
133 the blocks described by the descriptor blocks. Now commit
134 record can be sent to disk without waiting for descriptor
135 blocks to be written to disk. This behavior is controlled
136 using JBD2_FEATURE_ASYNC_COMMIT flag.
140 Description: OSS needs mutliple precreate threads
141 Details : Add ability to start more than one create thread per OSS.
145 Description: Sanity.sh test_65k failed: setstripe should have succeeded
146 Details : execute lfs setstripe on client
150 Description: mds_obd_create error creating tmp object
151 Details : When the user sets quota on root, llog will be affected and can't
152 create files and write files.
155 Frequency : Always on ia64 patchless client, and possibly others.
157 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
158 Details : This allows the patchless client to be loaded on architectures
163 Description: RedHat Update kernel for RHEL5
164 Details : Add the kernel config file for RHEL5.
168 Description: RedHat Update kernel for RHEL5
169 Details : Modify the kernel config file more closer RHEL5.
173 Description: Build failure against Centos5 (RHEL5)
174 Details : Define PAGE_SIZE when it isn't present.
178 Description: test_7 on sanity-quota.sh can't be used on separate machines
179 Details : add facet to handle it
183 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
184 Details : with NFS, the anon dentry's parent was set to itself in
185 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
186 corresponding dentry other than use rec->ur_name.
188 Severity : enhancement
190 Description: lfs setstripe enhancement
191 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
195 Description: testing performance impact of enabling checksumming
196 Details : enable checksum by default, allow --disable-checksum
197 configure option and "-o nochecksum" mount option
201 Description: lustre support for RHEL5
202 Details : Add support for RHEL5.
204 --------------------------------------------------------------------------------
206 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
208 * Support for kernels:
209 2.6.5-7.283 (SLES 9),
210 2.6.9-55.EL (RHEL 4),
211 2.6.16.46-0.14 (SLES 10),
212 2.6.18.8 vanilla (kernel.org)
213 * Client support for unpatched kernels:
214 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
215 2.6.16 - 2.6.22 vanilla (kernel.org)
216 * Due to recently discovered recovery problems, we do not recommend
217 using patchless RHEL 4 clients with this or any earlier release.
218 * Recommended e2fsprogs version: 1.39.cfs8
219 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
220 * Starting with this release, the ldiskfs backing filesystem required
221 by Lustre is now in its own package, lustre-ldiskfs. This package
222 should be installed. It is versioned separately from Lustre and
223 may be released separately in future.
227 Description: block reactivating mgc import until all deactivates complete
228 Details : Fix race when failing back MDT/MGS to itself (testing)
230 Severity : enhancement
232 Description: add optional extra BUILD_VERSION info
233 Details : add a new environment variable (namely LUSTRE_VERS) which allows
234 to override the lustre version.
237 Frequency : 2.6.18 servers only
239 Description: ll_kern_mount() doesn't release the module reference
240 Details : The ldiskfs module reference count never drops down to 0
241 because ll_kern_mount() doesn't release the module reference.
246 Description: server LBUG when using old ost_num_threads parameter
247 Details : Accept the old ost_num_threads parameter but warn that it
248 is deprecated, and fix an off-by-one error that caused an LBUG.
253 Description: Transient SCSI error results in persistent IO issue
254 Details : iobuf->dr_error is not reinitialized to 0 between two
258 Frequency : sometimes when underlying device returns I/O errors
260 Description: OSTs not going read-only during write failures
261 Details : OSTs are not remounted read-only when the journal commit threads
262 get I/O errors because fsfilt_ext3 calls journal_start/stop()
263 instead of the ext3 wrappers.
267 Description: poor connect scaling with increasing client count
268 Details : Don't run filter_grant_sanity_check for more than 100 exports
269 to improve scaling for large numbers of clients.
272 Frequency : SLES10 only
274 Description: sanity-quota.sh quotacheck failed: rc = -22
275 Details : Quotas cannot be enabled on SLES10.
278 Frequency : liblustre clients only
280 Description: getdirentries does not give error when run on compute nodes
281 Details : getdirentries does not fail when the size specified as an argument
282 is too small to contain at least one entry
284 Severity : enhancement
286 Description: Add LNET router traceability for debug purposes
287 Details : If a checksum failure occurs with a router as part of the
288 IO path, the NID of the last router that forwarded the bulk data
289 is printed so it can be identified.
294 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
295 Details : Due to a race condition, liblustre clients were occasionally
298 Severity : enhancement
300 Description: lfs setstripe use optional parameters instead of postional
303 Severity : enhancement
305 Description: Nanosecond timestamp support for ldiskfs
306 Details : The on-disk ldiskfs filesystem has added support for nanosecond
307 resolution timestamps. There is not yet support for this at
308 the Lustre filesystem level.
311 Frequency : during server recovery
313 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
314 Details : request with rq_no_resend flag not awake l_wait_event if they get a
318 Frequency : nfs export on patchless client
320 Description: connectathon hang when test nfs export over patchless client
321 Details : Disconnected dentry cannot be found with lookup, so we do not need
322 to unhash it or make it invalid
325 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
326 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
327 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
332 Description: open req refcounting wrong on reconnect
333 Details : If reconnect happened between getting open reply from server and
334 call to mdc_set_replay_data in ll_file_open, we will schedule
335 replay for unreferenced request that we are about to free.
336 Subsequent close will crash in variety of ways.
337 Check that request is still eligible for replay in
338 mdc_set_replay_data().
343 Description: disable writes to filesystem when reading health_check file
344 Details : the default for reading the health_check proc file has changed
345 to NOT do a journal transaction and write to disk, because this
346 can cause reads of the /proc file to hang and block HA state
347 checking on a healthy but otherwise heavily loaded system. It
348 is possible to return to the previous behaviour during configure
349 with --enable-health-write.
351 Severity : enhancement
353 Description: 64-bit inode version
354 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
355 to the inode. This will be required for version-based recovery.
360 Description: MDS fails to start if a duplicate client export is detected
361 Details : in some rare cases it was possible for a client to connect to
362 an MDS multiple times. Upon recovery the MDS would detect this
363 and fail during startup. Handle this more gracefully.
365 Severity : enhancement
367 Description: Add -o localflock option to simulate old noflock
369 Details : This will achieve local-only flock/fcntl locks
375 Description: log_commit_thread vs filter_destroy race leads to crash
376 Details : Take import reference before releasing llog record semaphore
381 Description: Wrong request locking in request set processing
382 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
383 imp_delayed_list, in this place should be used imp_lock.
386 Frequency : when reconnection
388 Description: Grant Leak when osc reconnect to OST
389 Details : When osc reconnect ost, OST(filter) should check whether it
390 should grant more space to client by comparing fed_grant and
391 cl_avail_grant, and return the granted space to client instead
392 of "new granted" space, because client will call osc_init_grant
393 to update the client grant space info.
396 Frequency : when client reconnect to OST
398 Description: Grant Leak when osc do resend and replay bulk write
399 Details : When osc reconnect to OST, OST(filter)should clear grant info of
400 bulk write request, because the grant info while be sync between
401 OSC and OST when reconnect, and we should ignore the grant info
402 these of resend/replay write req.
407 Description: Grant space more than avaiable left space sometimes.
408 Details : When then OST is about to be full, if two bulk writing from
409 different clients came to OST. Accord the avaliable space of the
410 OST, the first req should be permitted, and the second one
411 should be denied by ENOSPC. But if the seconde arrived before
412 the first one is commited. The OST might wrongly permit second
413 writing, which will cause grant space > avaiable space.
416 Frequency : when client is evicted
418 Description: Grant might be wrongly erased when osc is evicted by OST
419 Details : when the import is evicted by server, it will fork another
420 thread ptlrpc_invalidate_import_thread to invalidate the
421 import, where the grant will be set to 0. While the original
422 thread will update the grant it got when connecting. So if
423 the former happened latter, the grant will be wrongly errased
424 because of this race.
429 Description: Checking Stale with correct fid
430 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
431 to check whether it is stale, because sometimes, we want the
432 enqueue happened anyway, and op_data.fid2 will not be initialized.
434 Severity : enhancement
436 Description: update patchless client
437 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
440 Frequency : only with 2.4 kernel
442 Description: random memory corruption
443 Details : size of struct ll_inode_info is to big for union inode.u and this
444 can be cause of random memory corruption.
449 Description: Memory leak in recovery
450 Details : Lov_mds_md was not free in an error handler in mds_create_object.
451 It should also check obd_fail before fsfilt_start, otherwise if
452 fsfilt_start return -EROFS,(failover mds during mds recovery).
453 then the req will return with repmsg->transno = 0 and rc = EROFS.
454 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
455 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
456 be freed no matter whether fsfilt_commit success or not.
459 Frequency : only with huge count clients
461 Description: Prevents from taking the superblock lock in llap_from_page for
463 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
464 for llap_from_page call in ll_removepage prevents from taking the
465 superblock lock for a soon died page.
470 Description: Not check open intent error before release open handle
471 Details : in some rare cases, the open intent error is not checked before
472 release open handle, which may cause
473 ASSERTION(open_req->rq_transno != 0), because it tries to release
474 the failed open handle.
479 Description: Set cat log bitmap only after create log success.
480 Details : in some rare cases, the cat log bitmap is set too early. and it
481 should be set only after create log success.
485 Description: Accessing a block bevice can re-enable I/O when Lustre is
486 tearing down a device.
487 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
492 Description: service threads may hog cpus when there are a lot of requests
494 Details : Insert cond_resched to give other threads a chance to use some of
500 Description: the cat log was not initialized in recovery
501 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
502 unlink log on mds will not be initialized until mds post
503 recovery. And also in mds post recovery, the unlink log will
504 initialization will be done asynchronausly, so there will be race
505 between add unlink log and unlink log initialization.
509 Description: brw_stats were being printed incorrectly
510 Details : brw_stats were being printed as log2 but all of them were not
511 recorded as log2. Also remove some code duplication arising from
512 filter_tally_{read,write}.
516 Frequency : rare, only in recovery.
517 Description: ASSERTION(req->rq_type != LI_POISON) failed
518 Details : imp_lock should be held while iterating over imp_sending_list for
519 prevent destroy request after get timeout in ptlrpc_queue_wait.
523 Description: replay-single.sh test 52 fails
524 Details : A lock's skiplist need to be cleanup when it being unlinked
525 from its resource list.
529 Description: Short directio read returns full requested size rather than
531 Details : Direct I/O operations should return actual amount of bytes
532 transferred rather than requested size.
534 Severity : enhancement
536 Description: metadata RPC reduction (e.g. for rm performance)
537 Details : decrease the amount of synchronous RPC between clients and servers
538 by canceling conflicing lock before the operation on the client side
539 and packing thier handles into the main operation RPC to server.
541 Severity : enhancement
543 Description: Async OSC create to avoid the blocking unnecessarily.
544 Details : If a OST has no remain object, system will block on the creating
545 when need to create a new object on this OST. Now, ways use
546 pre-created objects when available, instead of blocking on an
547 empty osc while others are not empty. If we must block, we block
548 for the shortest possible period of time.
550 --------------------------------------------------------------------------------
552 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
557 Frequency : on some architectures
559 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
560 Details : Uninitialized flags sometimes cause configuration commands to
564 Frequency : patchless clients only
566 Description: missing __iget() symbol export
567 Details : The __iget() symbol export is missing. To avoid the need for
568 this on patchless clients the deathrow inode reaper is turned
569 off, and we depend on the VM to clean up old inodes. This
570 dependency was during via the fix for bug 12181.
572 --------------------------------------------------------------------------------
574 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
576 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
577 INTEROPERATE with older versions automatically. In many cases a
578 special upgrade step is needed. Please read the
579 user documentation before upgrading any part of a live system.
580 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
581 version will not interoperate with 1.6 betas before beta5 (1.5.95).
582 * WARNING: Lustre configuration and startup changes are required with
583 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
589 Severity : enhancement
591 Description: MountConf
592 Details : Lustre configuration is now managed via mkfs and mount
593 commands instead of lmc and lconf. New obd types (MGS, MGC)
594 are added for dynamic configuration management. See
595 https://mail.clusterfs.com/wikis/lustre/MountConf for
598 Severity : enhancement
600 Description: dynamic OST addition
601 Details : OSTs can now be added to a live filesystem
603 Severity : enhancement
605 Description: startup order invariance
606 Details : MDTs and OSTs can be started in any order. Clients only
607 require the MDT to complete startup.
609 Severity : enhancement
611 Description: parallel, asynchronous orphan cleanup
612 Details : orphan cleanup is now performed in separate threads for each
613 OST, allowing parallel non-blocking operation.
615 Severity : enhancement
617 Description: optimized stripe assignment
618 Details : stripe assignments are now made based on ost space available,
619 ost previous usage, and OSS previous usage, in order to try
620 to optimize storage space and networking resources.
622 Severity : enhancement
624 Description: Permanently set tunables
625 Details : All writable /proc/fs/lustre tunables can now be permanently
626 set on a per-server basis, at mkfs time or on a live
629 Severity : enhancement
631 Description: Lustre message v2
632 Details : Add lustre message format v2.
634 Severity : enhancement
636 Description: client OST exclusion list
637 Details : Clients can be started with a list of OSTs that should be
638 declared "inactive" for known non-responsive OSTs.
642 Description: SPEC SFS validation failure on NFS v2 over lustre.
643 Details : Changes the blocksize for regular files to be 2x RPC size,
644 and not depend on stripe size.
646 Severity : enhancement
648 Description: Multiple MD RPCs in flight.
649 Details : Further unserialise some read-only MDS RPCs - learn about intents.
650 To avoid overly-overloading MDS, introduce a limit on number of
651 MDS RPCs in flight for a single client and add /proc controls
652 to adjust this limit.
654 Severity : enhancement
656 Description: client read/write statistics
657 Details : Add client read/write call usage stats for performance
658 analysis of user processes.
659 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
660 file access. extents_stats shows chunk size distribution.
661 extents_stats_per_process show chunk size distribution per
664 Severity : enhancement
666 Description: mds statistics
667 Details : Add detailed mds operations statistics in
668 /proc/fs/lustre/mds/*/stats.
672 Description: Failure of copying files with lustre special EAs.
673 Details : Client side always return success for setxattr call for lustre
674 special xattr (currently only "trusted.lov").
679 Description: Refcount LNET uuids
680 Details : The global LNET uuid list grew linearly with every startup;
681 refcount repeated list entries instead of always adding to
684 Severity : enhancement
686 Description: Dynamic service threads
687 Details : Within a small range, start extra service threads
688 automatically when the request queue builds up.
691 Frequency : mixed-endian client/server environments
693 Description: mixed-endian crashes
694 Details : The new msg_v2 system had some failures in mixed-endian
697 Severity : enhancement
699 Description: Easy OST removal
700 Details : OSTs can be permanently deactivated with e.g. 'lctl
701 conf_param lustre-OST0001.osc.active=0'
703 Severity : enhancement
705 Description: MGS proc entries
706 Details : Added basic proc entries for the MGS showing what filesystems
709 Severity : enhancement
711 Description: provide MGS failover
712 Details : Added config lock reacquisition after MGS server failover.
714 Severity : enhancement
716 Description: add Linux 2.4 support
717 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
721 Description: a large application tries to do I/O to the same resource and dies
723 Details : Check the req->rq_arrival time after the call to
724 ost_brw_lock_get(), but before we do anything about
725 processing it & sending the BULK transfer request. This
726 should help move old stale pending locks off the queue as
727 quickly as obd_timeout.
730 Frequency : when an incorrect nid is specified during startup
732 Description: ptlrpc connect to non-existant node causes kernel crash
733 Details : LNET can't be re-entered from an event callback, which
734 happened when we expire a message after the export has been
735 cleaned up. Instead, hand the zombie cleanup off to another
738 Severity : enhancement
740 Description: plain/inodebits lock performance improvement
741 Details : Grouping plain/inodebits in granted list by their request modes
742 and bits policy, thus improving the performance of search through
746 Frequency : only if OST filesystem is corrupted
748 Description: client incorrectly hits assertion in ptlrpc_replay_req()
749 Details : for a short time RPCs with bulk IO are in the replay list,
750 but replay of bulk IOs is unimplemented. If the OST filesystem
751 is corrupted due to disk cache incoherency and then replay is
752 started it is possible to trip an assertion. Avoid putting
753 committed RPCs into the replay list at all to avoid this issue.
756 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
758 Description: System hang on startup
759 Details : This bug allowed the liblustre (e.g. catamount) client to
760 return to the app before handling all startup RPCs. This
761 could leave the node unresponsive to lustre network traffic
762 and manifested as a server ptllnd timeout.
764 Severity : enhancement
766 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
767 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
768 boolean to control whether to print peer debug info when a
769 client's RPC times out.
772 Frequency : only for kernels with patches from Lustre below 1.4.3
774 Description: Remove old rdonly API
775 Details : Remove old rdonly API which unused from at least lustre 1.4.3
778 Frequency : only for devices with external journals
780 Description: Set external device read-only also
781 Details : During a commanded failover stop, we set the disk device
782 read-only while the server shuts down. We now also set any
783 external journal device read-only at the same time.
786 Frequency : when upgrading from 1.4 while trying to change parameters
788 Description: The wrong (new) MDC name was used when setting parameters for
789 upgraded MDT's. Also allows changing of OSC (and MDC)
790 parameters if --writeconf is specified at tunefs upgrade time.
793 Frequency : when setting specific ost indicies
795 Description: QOS code breaks on skipped indicies
796 Details : Add checks for missing OST indicies in the QOS code, so OSTs
797 created with --index need not be sequential.
799 Severity : enhancement
801 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
802 Details : The uninit_groups feature works in conjunction with the kernel
803 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
804 up the pass1 processing of e2fsck. This is a read-only feature
805 in ldiskfs2 only, so older kernels and current ldiskfs cannot
806 mount filesystems that have had this feature enabled.
808 Severity : enhancement
810 Description: Improve multi-block allocation algorithm to avoid fragmentation
811 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
812 allocation locality and avoid filesystem fragmentation.
814 ------------------------------------------------------------------------------
816 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
818 * Support for kernels:
819 2.4.21-47.0.1.EL (RHEL 3)
821 2.6.9-42.0.10.EL (RHEL 4)
822 2.6.12.6 vanilla (kernel.org)
823 2.6.16.27-0.9 (SLES 10)
824 * Recommended e2fsprogs version: 1.39.cfs5
826 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
830 Frequency : occasional, depends on client load and configuration
831 Bugzilla : 12181, 12203
832 Description: data loss for recently-modified files
834 Details : In some cases it is possible that recently written or created
835 files may not be written to disk in a timely manner (this should
836 normally be within 30s unless client IO load is very high).
837 The problem appears as zero-length files or files that are a
838 multiple of 1MB in size after a client crash or client eviction
839 that are missing data at the end of the file.
841 This problem is more likely to be hit on clients where files are
842 repeatedly created and unlinked in the same directory, clients
843 have a large amount of RAM, have many CPUs, the filesystem has
844 many OSTs, the clients are rebooted frequently, and/or the files
845 are not accessed by other nodes after being written.
847 The presence of the problem can be detected by looking at
848 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
849 smaller than the second (nr_unused) then dirty files will not
850 be flushed automatically to disk. "sync; sleep 10" should be
851 run several times on the node before unmounting it to update
852 Lustre (this is also safe to run on nodes without this problem).
854 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
855 that can cause this same problem, so customers using that kernel
856 also need to update the kernel in addition to Lustre. In order
857 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
859 It is normal that files written just before a client crash (less
860 than 30s) may not yet have been flushed to disk, even for local
864 Frequency : frequent on thin XT3 nodes
866 Description: UUID collision on thin XT3 Linux nodes
867 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
868 randomly, since we relied on an insufficiently-seeded PRNG.
873 Description: OSS hangs after "All ost request buffers busy"
874 Details : A deadlock between quota and journal operations caused OSS
875 hangs after printing "All ost request buffers busy."
878 Frequency : always on liblustre builds
880 Description: Cleanup compiler warnings on liblustre
883 Frequency : always on liblustre builds on XT3
885 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
889 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
890 > 32000 subdirectories
891 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
892 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
893 subdirectory count crosses 32000. This will aid e2fsck to
894 correctly handle more than 32000 subdirectories.
897 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
899 Description: System hang on startup
900 Details : This bug allowed the liblustre (e.g. catamount) client to
901 return to the app before handling all startup RPCs. This
902 could leave the node unresponsive to lustre network traffic
903 and manifested as a server ptllnd timeout.
905 Severity : enhancement
907 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
908 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
909 boolean to control whether to print peer debug info when a
910 client's RPC times out.
915 Description: make O_SYNC working on 2.6 kernels
916 Details : 2.6 kernels use different method for mark pages for write,
917 so need add a code to lustre for O_SYNC work.
922 Description: Failure to close file and release space on NFS
923 Details : Put inode details into lock acquired in ll_intent_file_open.
924 Use mdc_intent_lock in ll_intent_open to properly
925 detect all kind of errors unhandled by mdc_enqueue.
930 Description: proc file read during shutdown sometimes raced obd removal,
932 Details : Add lock to prevent obd access after proc file removal.
935 Frequency : Only for files larger than 4GB on 32-bit clients.
937 Description: improperly doing page alignment of locks
938 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
939 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
944 Description: RPCs being resent when they shouldn't be
945 Details : Some RPCs that should not be resent are being resent. This
946 can cause inconsistencies in the RPC state machine. Do not
947 resend such requests.
950 Frequency : rare, only with NFS export
952 Description: Crash on NFS re-export node
953 Details : under very unusual load conditions an assertion is hit in
954 ll_intent_file_open()
957 Frequency : only if OST filesystem is corrupted
959 Description: client incorrectly hits assertion in ptlrpc_replay_req()
960 Details : for a short time RPCs with bulk IO are in the replay list,
961 but replay of bulk IOs is unimplemented. If the OST filesystem
962 is corrupted due to disk cache incoherency and then replay is
963 started it is possible to trip an assertion. Avoid putting
964 committed RPCs into the replay list at all to avoid this issue.
969 Description: large O_DIRECT requests fail under memory pressure/fragmentation
970 Details : Large single O_DIRECT read and write calls can fail to allocate
971 a sufficiently large buffer to process the request. In case of
972 allocation failure the allocation is retried with a smaller
973 buffer and broken into smaller requests.
975 Severity : enhancement
977 Description: Add -o localflock option to simulate old noflock behaviour.
978 Details : This will achieve local-only flock/fcntl locks coherentness.
983 Description: versioning check is incomplete
984 Details : Checking the version difference of client vs. server, report
985 error if the gap is too big.
990 Description: add support PG_writeback bit
991 Details : add support for PG_writeback bit for Lustre, for more carefull
992 work with page cache in 2.6 kernel. This also fix some deadlocks
993 and remove hack for work O_SYNC with 2.6 kernel.
995 Severity : enhancement
997 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
998 Details : The uninit_groups feature works in conjunction with the kernel
999 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1000 up the pass1 processing of e2fsck. This is a read-only feature
1001 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1002 mount filesystems that have had this feature enabled.
1004 Severity : enhancement
1006 Description: Improve multi-block allocation algorithm to avoid fragmentation
1007 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1008 allocation locality and avoid filesystem fragmentation.
1010 ------------------------------------------------------------------------------
1012 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1014 * Support for kernels:
1015 2.6.9-42.0.3.EL (RHEL 4)
1016 2.6.5-7.276 (SLES 9)
1017 2.4.21-47.0.1.EL (RHEL 3)
1018 2.6.12.6 vanilla (kernel.org)
1019 2.6.16.21-0.8 (SLES10)
1020 * Recommended e2fsprogs version: 1.39.cfs2-0
1022 * The backwards-compatible /proc/sys/portals symlink has been removed
1023 in this release. Before upgrading, please ensure that you change
1024 any configuration scripts or /etc/sysctl.conf files that access
1025 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1026 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1027 in advance of the upgrade on any system running Lustre 1.4.6 or
1028 newer, since /proc/sys/lnet was added in that version.
1029 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1033 Frequency : only when quota is used
1035 Description: avoid scanning export list for quota master
1036 Details : Change the algorithms to avoid scanning export list in order
1037 to improve the efficiency.
1040 Frequency : MDS failover only, very rarely
1042 Description: "went back in time" messages on mds failover
1043 Details : The greatest transno may be lost when the current operation
1044 finishes with an error (transno==0) and the client's last_rcvd
1045 record is over-written. Save the greatest transno in the
1046 mds_last_transno for this case.
1049 Frequency : always for specific kernels and striping counts
1051 Description: client may get "Matching packet too big" without ACL support
1052 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1053 when trying to access files in certain configurations. The
1054 clients should in fact be denied when mounting because they do
1055 not understand ACLs.
1058 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1060 Description: many clients connecting with IO in progress causes connect timeouts
1061 Details : Avoid synchronous journal commits to avoid delays caused by many
1062 clients connecting/disconnecting when bulk IO is in progress.
1063 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1064 OST_IO_PORTAL to avoid delays behind potentially many pending
1068 Frequency : occasionally with multiple writers to a single file
1070 Description: shared writes to file may result in wrong size reported by stat()
1071 Details : Allow growing of kms when extent lock is cancelled
1074 Frequency : always with random mmap IO to multi-striped file
1076 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1077 Details : Only if the hole is at the end of OST object so that kms is too
1078 small. Fix is to increase kms accordingly in ll_nopage.
1081 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1083 Description: writes to a missing object would leak memory on the OST
1084 Details : If there is an inconsistency between the MDS and OST filesystems,
1085 such that the MDS references an object that doesn't exist, writes
1086 to that object will leak memory due to incorrect cleanup in the
1087 error handling path, eventually running out of memory on the OST.
1092 Description: Creating too long symlink causes lustre errors
1093 Details : Check symlink and name lengths before sending requests to MDS.
1096 Frequency : only if flock is enabled (not on by default)
1098 Description: posix locks not released on fd closure on 2.6.9+
1099 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1100 kernels, this caused such locks not to be released on fd close and
1101 then assertions on fs unmount about still used locks.
1104 Frequency : MDS failover only, very rarely
1106 Description: clients may get ASSERTION(granted_lock != NULL)
1107 Details : When request was taking a long time, and a client was resending
1108 a getattr by name lock request. The were multiple lock requests
1109 with the same client lock handle and
1110 mds_getattr_name->fixup_handle_for_resent_request found one of the
1111 lock handles but later failed with ASSERTION(granted_lock != NULL).
1116 Description: handle->h_buffer_credits > 0, assertion failure
1117 Details : h_buffer_credits is zero after truncate, causing assertion
1118 failure. This patch extends the transaction or creates a new
1122 Frequency : NFS re-export or patchless client
1123 Bugzilla : 11179, 10796
1124 Description: Crash on NFS re-export node (__d_move)
1125 Details : We do not want to hash the dentry if we don't have a lock.
1126 But if this dentry is later used in d_move, we'd hit uninitialised
1127 list head d_hash, so we just do this to init d_hash field but
1128 leave dentry unhashed.
1131 Frequency : NFS re-export or patchless client
1133 Description: NFS exports has problem with symbolic link
1134 Details : lustre client didn't properly install dentry when re-exported
1135 to NFS or running patchless client.
1138 Frequency : NFS re-export or patchless client
1140 Description: Various nfs/patchless fixes.
1141 Details : fixes reuse disconected alias for lookup process - this fixes
1142 warning "find_exported_dentry: npd != pd",
1143 fix permission error with open files at nfs.
1144 fix apply umask when do revalidate.
1147 Frequency : occasional
1149 Description: Crash on NFS re-export node
1150 Details : calling clear_page() on the wrong pointer triggered oops in
1151 generic_mapping_read().
1154 Frequency : rarely, using O_DIRECT IO
1156 Description: unaligned directio crashes client with LASSERT
1157 Details : check for unaligned buffers before trying any requests.
1160 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1162 Description: stale data returned from RAID cache
1163 Details : If only a small amount of IO is done to the RAID device before
1164 reading it again it is possible to get stale data from the RAID
1165 cache instead of reading it from disk.
1168 Frequency : always for sles10 kernel
1170 Description: sles10 support
1171 Details : ll_follow_link: compile fixes and using of nd_set_link
1172 under newer kernels.
1175 Frequency : depends on arch, kernel and compiler version, always on sles10
1178 Description: recursive or deep enough symlinks cause stack overflow
1179 Details : getting rid of large stack-allocated variable in
1183 Frequency : depends on hardware
1185 Description: lustre write performance loss in the SLES10 kernel
1186 Details : the performance loss is caused by using of write barriers in the
1187 ext3 code. The SLES10 kernel turns barrier support on by
1188 default. The fix is to undo that change for ldiskfs.
1190 ------------------------------------------------------------------------------
1192 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1194 * Support for kernels:
1195 2.6.9-42.0.3EL (RHEL 4)
1196 2.6.5-7.276 (SLES 9)
1197 2.4.21-47.0.1.EL (RHEL 3)
1198 2.6.12.6 vanilla (kernel.org)
1202 Frequency : quota enabled and large files being deleted
1204 Description: releasing more than 4GB of quota at once hangs OST
1205 Details : If a user deletes more than 4GB of files on a single OST it
1206 will cause the OST to spin in an infinite loop. Release
1207 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1212 Description: statfs data retrieved from /proc may be stale or zero
1213 Details : When reading per-device statfs data from /proc, in the
1214 {kbytes,files}_{total,free,avail} files, it may appear
1215 as zero or be out of date.
1218 Frequency : systems with MD RAID1 external journal devices
1220 Description: lconf's call to blkid is confused by RAID1 journal devices
1221 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1222 of returning all block devices that match the journal UUID.
1225 Frequency : always, for aggregate stripe size over 4GB
1227 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1228 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1229 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1232 Frequency : always if "lfs find" used on a local file/directory
1234 Description: "lfs find" segfaults if used on a local file/directory
1235 Details : The case where a directory component was not specified wasn't
1236 handled correctly. Handle this properly.
1239 Frequency : always on ppc64
1241 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1242 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1243 perform endianness conversion.
1246 Frequency : rarely (truncate to non-zero file size after write under load)
1247 Bugzilla : 10730, 10687
1248 Description: Files padded with zeros to next 4K multiple
1249 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1250 it is possible that files that are truncated to a non-zero size
1251 immediately after being written are filled with zero bytes beyond
1252 the truncated size. No file data is lost.
1254 Severity : enhancement
1256 Description: Allow recovery/failover for liblustre clients.
1257 Details : liblustre clients were unaware of failover configurations until
1260 Severity : enhancement
1262 Description: user file locks should fail when not mounting with flock option
1263 Details : Set up an error-returning stub in ll_file_operations.lock field
1264 to prevent incorrect behaviour when client is mounted without
1265 flock option. Also, set up properly f_op->flock field for
1269 Frequency : always on ia64
1271 Description: "lfs df" loops on printing out MDS statfs information
1272 Details : The obd_ioctl_data was not initialized and in some systems
1273 this caused a failure during the ioctl that did not return
1274 an error. Initialize the struct and return an error on failure.
1277 Frequency : SLES 9 only
1279 Description: Error of copying files with lustre special EAs as root
1280 Details : Client side always return success for setxattr call for lustre
1281 special xattr (currently only "trusted.lov").
1284 Frequency : rarely on clusters with both ia64+i386 clients
1286 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1287 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1288 clients are doing concurrent writes to the same file it is
1289 possible that the ia64 clients may LASSERT because the OST
1290 extent locks are not PAGE_SIZE aligned. Ensure that grown
1291 locks are always aligned on the request boundary.
1294 Frequency : specific use, occasional
1296 Description: Overwriting in use executable truncates on-disk binary image
1297 Details : If one node attempts to overwrite an executable in use by
1298 another node, we now correctly return ETXTBSY instead of
1299 truncating the file.
1301 Severity : enhancement
1303 Description: Async OSC create to avoid the blocking unnecessarily.
1304 Details : If a OST has no remain object, system will block on the creating
1305 when need to create a new object on this OST. Now, ways use
1306 pre-created objects when available, instead of blocking on an
1307 empty osc while others are not empty. If we must block, we block
1308 for the shortest possible period of time.
1313 Description: chmod on Lustre root is propagated to other clients
1314 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1315 invalid by the follow_mount time.
1318 Frequency : liblustre clients only
1320 Description: Race in 'instant cancel' lock handling could lead to such locks
1321 never to be granted in case of SMP MDS
1322 Details : Do not destroy not yet granted but cbpending locks in
1326 Frequency : replay/resend of open
1328 Description: non null lock assetion failure in mds_intent_policy
1329 Details : Trying to replay/resend lockless open requests resulted in
1330 mds_open() returning 0 with no lock. Now it sets a flag if
1331 a lock is going to be returned.
1333 Severity : enhancement
1335 Description: Checksum enhancements
1336 Details : New checksum enhancements allow for resending RPCs that failed
1339 Severity : enhancement
1341 Description: Tunables on number of dirty pages in cacche
1342 Details : Allow to set limit on number of dirty pages cached.
1347 Description: client crash on unmount - lock still has references
1348 Details : In some error handling cases it was possible to leak a lock
1349 reference on a client while accessing a file. This was not
1350 harmful to the client during operation, but would cause the
1351 client to crash when the filesystem is unmounted.
1354 Frequency : specific case, rare
1356 Description: ETXTBSY on mds though file not in use by client
1357 Details : ETXTBSY is no longer incorrectly returned when attempting to
1358 chmod or chown a directory that the user previously tried to
1359 execute or a currently-executing binary.
1362 Frequency : extremely rare except on liblustre-based clients
1364 Description: Lustre space not freed when files are deleted
1365 Details : Clean up open-unlinked files after client eviction. Previously
1366 the unlink was skipped and the files remained as orphans.
1371 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1372 Details : In some cases it was possible to send a blocking callback to
1373 a client doing a glimpse, even though that client didn't get
1374 a lock granted. When the glimpse lock is cancelled on the OST
1375 the freed lock is left on the waiting list and corrupted the list.
1378 Frequency : all core dumps
1380 Description: Broke core dumps to lustre
1381 Details : Negative dentry may be unhashed if parent does not have UPDATE
1382 lock, but some callers, e.g. do_coredump, expect dentry to be
1383 hashed after successful create, hash it in ll_create_it.
1385 ------------------------------------------------------------------------------
1387 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1389 * Support for kernels:
1390 2.6.9-42.0.2.EL (RHEL 4)
1391 2.6.5-7.276 (SLES 9)
1392 2.4.21-40.EL (RHEL 3)
1393 2.6.12.6 vanilla (kernel.org)
1397 Frequency : always on RHEL 3
1399 Description: Number of open files grows over time
1400 Details : The number of open files grows over time, whether or not
1401 Lustre is started. This was due to a filp leak introduced
1402 by one of our kernel patches.
1404 ------------------------------------------------------------------------------
1406 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1408 * Support for kernels:
1409 2.6.9-42.EL (RHEL 4)
1410 2.6.5-7.276 (SLES 9)
1411 2.4.21-40.EL (RHEL 3)
1412 2.6.12.6 vanilla (kernel.org)
1417 Bugzilla : 5719, 9635, 9792, 9684
1418 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1419 Details : If a server is under heavy load and cannot reply to new
1420 connection requests before the client resends the (re)connect,
1421 the connection handling code can behave badly if two service
1422 threads are concurrently handing separate (re)connections from
1423 the same client. Add better locking to the connection handling
1424 code, and ensure that only a single connection will be processed
1425 for a given client UUID, even if the lock is dropped.
1427 Severity : enhancement
1429 Description: add TCP zero-copy support to kernel
1430 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1431 sends if the hardware supports scatter-gather and checksumming.
1432 This allows socklnd to do client-write and server-read more
1433 efficiently and reduce CPU utilization from skbuf copying.
1436 Frequency : only if NFS exporting from client
1438 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1439 Details : If setting attributes on a file created under NFS that had
1440 never been opened it would be possible to oops the client
1441 if the file had no objects.
1444 Frequency : always for liblustre
1446 Description: liblustre client does MDS+OSTs setattr RPC for each write
1447 Details : When doing a write from a liblustre client, the client
1448 incorrectly issued an RPC to the MDS and each OST the file was
1449 striped over in order to update the timestamps. When writing
1450 with small chunks and many clients this could overwhelm the MDS
1451 with RPCs. In all cases it would slow down the write because
1452 these RPCs are unnecessary.
1454 Severity : enhancement
1456 Description: allow number of MDS service threads to be changed at module load
1457 Details : It is now possible to change the number of MDS service threads
1458 running. Adding "options mds mds_num_threads={N}" to the MDS's
1459 /etc/modprobe.conf will set the number of threads for the next
1460 time Lustre is restarted (assuming the "mds" module is also
1461 reloaded at that time). The default number of threads will
1462 stay the same, 32 for most systems.
1467 Description: OST crash if filesystem is unformatted or corrupt
1468 Details : If an OST is started on a device that has never been formatted
1469 or if the filesystem is corrupt and cannot even mount then the
1470 error handling cleanup routines would dereference a NULL pointer.
1475 Description: NULL pointer deref in llap_from_page.
1476 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1477 incorrect) mapping if the page was truncated/reclaimed while it was
1478 searched for. Check for this condition and skip such pages when
1479 doing readahead. Introduce extra check to llap_from_page() to
1480 verify page->mapping->host is non-NULL (so page is not anonymous).
1483 Frequency : Sometimes when using sys_sendfile
1485 Description: "page not covered by a lock" warnings from ll_readpage
1486 Details : sendfile called ll_readpage without right page locks present.
1487 Now we introduced ll_file_sendfile that does necessary locking
1488 around call to generic_file_sendfile() much like we do in
1492 Frequency : with certain MDS communication failures at client mount time
1494 Description: NULL pointer deref after failed client mount
1495 Details : a client connection request may delayed by the network layer
1496 and not be sent until after the PTLRPC layer has timed out the
1497 request. If the client fails the mount immediately it will try
1498 to clean up before the network times out the request. Add a
1499 reference from the request import to the obd device and delay
1500 the cleanup until the network drops the request.
1503 Frequency : occasionally during client (re)connect
1505 Description: assertion failure during client (re)connect
1506 Details : processing a client connection request may be delayed by the
1507 client or server longer than the client connect timeout. This
1508 causes the client to resend the connection request. If the
1509 original connection request is replied in this interval, the
1510 client may trip an assertion failure in ptlrpc_connect_interpret()
1511 which thought it would be the only running connect process.
1514 Frequency : only with obd_echo servers and clients that are rebooted
1516 Description: kernel BUG accessing uninitialized data structure
1517 Details : When running an obd_echo server it did not start the ping_evictor
1518 thread, and when a client was evicted an uninitialized data
1519 structure was accessed. Start the ping_evictor in the RPC
1520 service startup instead of the OBD startup.
1522 Severity : enhancement
1523 Bugzilla : 10193 (patchless)
1524 Description: Remove dependency on various unexported kernel interfaces.
1525 Details : No longer need reparent_to_init, exit_mm, exit_files,
1526 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1529 Frequency : rare (only users of deprecated and unsupported LDAP config)
1531 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1532 Details : LDAP apparently contains 'lustreName' attributes instead of
1533 'name'. A simple remapping of the name is sufficient.
1536 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1538 Description: waiting_locks_callback trips kernel BUG if client is evicted
1539 Details : Running with the dump_on_timeout debug flag turned on makes
1540 it possible that the waiting_locks_callback() can try to dump
1541 the Lustre kernel debug logs from an interrupt handler. Defer
1542 this log dumping to the expired_lock_main() thread.
1544 Severity : enhancement
1546 Description: Support NFS exporting on 2.6 kernels.
1547 Details : Implement non-rawops metadata methods for NFS server to use without
1548 changing NFS server code.
1551 Frequency : very rare (synthetic metadata workload only)
1553 Description: two racing renames might cause an MDS thread to deadlock
1554 Details : Running the "racer" program may cause one MDS thread to rename
1555 a file from being the source of a rename to being the target of
1556 a rename at exactly the same time that another thread is doing
1557 so, and the second thread has already enqueued these locks after
1558 doing a lookup of the target and is trying to relock them in
1559 order. Ensure that we don't try to re-lock the same resource.
1562 Frequency : only very large systems with liblustre clients
1564 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1565 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1566 all OSTs in parallel. This allows the eviction of stale liblustre
1567 clients to proceed much faster than if they were done in series,
1568 and also offers similar improvements for other set_info RPCs.
1573 Description: excessive CPU usage during initial read phase on client
1574 Details : During the initial read phase on a client, it would agressively
1575 retry readahead on the file, consuming too much CPU and impacting
1576 performance (since 1.4.5.8). Improve the readahead algorithm
1577 to avoid this, and also improve some other common cases (read
1578 of small files in particular, where "small" is files smaller than
1579 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1584 Description: MDS crash when receiving packet with unknown intent.
1585 Details : Do not LBUG in unknown intent case, just return -EFAULT
1587 Severity : enhancement
1588 Bugzilla : 9293, 9385
1589 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1590 Details : Do not serialize getattr (non-intent version) and statfs.
1593 Frequency : occasional, when OST network is overloaded/intermittent
1595 Description: client evicted by OST after bulk IO timeout
1596 Details : If a client sends a bulk IO request (read or write) the OST
1597 may evict the client if it is unresposive to its data GET/PUT
1598 request. This is incorrect if the network is overloaded (takes
1599 too long to transfer the RPC data) or dropped the OST GET/PUT
1600 request. There is no need to evict the client at all, since
1601 the pinger and/or lock callbacks will handle this, and the
1602 client can restart the bulk request.
1605 Frequency : Always when mmapping file with no objects
1607 Description: client crashes when mmapping file with no objects
1608 Details : Check that we actually have objects in a file before doing any
1609 operations on objects in ll_vm_open, ll_vm_close and
1615 Description: Request leak when working with deleted CWD
1616 Details : Introduce advanced request refcount tracking for requests
1617 referenced from lustre intent.
1619 Severity : Enhancement
1621 Description: Cache open file handles on client.
1622 Details : MDS now will return special lock along with openhandle, if
1623 requested and client is allowed to hold openhandle, even if unused,
1624 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1625 closing files for every read/write openration.
1627 Severity : Enhancement
1629 Description: Cache open negative dentries on client when possible.
1630 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1631 negative dentries on lock revocation.
1636 Description: Remounting a client read-only wasn't possible with a zconf mount
1637 Details : It wasn't possible to remount a client read-only with llmount.
1639 Severity : enhancement
1640 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1641 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1642 MPICH is included to add Lustre-specific ADIO interfaces.
1643 This is based closely on the UFS ADIO layer and only differs
1644 in file creation, in order to allow the OST striping to be set.
1645 This is user-contributed code and not supported by CFS.
1650 Description: extended inode attributes (immutable, append-only) work improperly
1651 when 2.4 and 2.6 kernels are used on client/server or vice versa
1652 Details : Introduce kernel-independent values for these flags.
1654 Severity : enhancement
1657 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1658 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1659 granularity. Now, Fractional MB granularity is be supported,
1660 this is very useful for low memory system.
1662 Severity : enhancement
1664 Description: Getattr by fid
1665 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1666 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1667 have locks and avoids getattr rpc requests for it on every stat.
1670 Frequency : Always, for filesystems larger than 2TB
1672 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1673 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1674 filesystems larger than 2TB to be mounted with mballoc enabled.
1677 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1679 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1680 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1681 filesystems larger than 2TB is mounted then the kernel will
1682 silently corrupt the start of the filesystem. CONFIG_LBD is
1683 enabled for all CFS-supported kernels, but the possibility of
1684 this happening with a modified kernel config exists.
1686 Severity : enhancement
1688 Description: add client O_DIRECT support for 2.6 kernels
1689 Details : It is now possible to do O_DIRECT reads and writes to files
1690 in the Lustre client mountpoint on 2.6 kernel clients.
1692 Severity : enhancement
1694 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1695 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1696 parallel, not waiting for response from every OST before sending
1697 a rpc to the next OST.
1702 Description: setattr vs write race when updating file timestamps
1703 Details : Client processes that update a file timestamp into the past
1704 right after writing to the file (e.g. tar) it is possible that
1705 the updated file modification time can be reset to the current
1706 time due to a race between processing the setattr and write RPC.
1708 Severity : enhancement
1710 Description: Bring 'lfs find' closer in line with regular Linux find.
1711 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1712 -print0 options and obtains all the needed info through the lustre
1715 Severity : enhancement
1717 Description: support up to 1024 configured devices on one node
1718 Details : change obd_dev array from statically allocated to dynamically
1719 allocated structs as they are first used to reduce memory usage
1724 Description: Flush dirty partially truncated pages during truncate
1725 Details : Immediatelly flush partially truncated pages in filter_setattr,
1726 this way we completely avoid having any pages in page cache on OST
1727 and can retire ugly workarounds during writes to flush such pages.
1732 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1733 Details : protect inode from truncation within vfs_unlink() context
1734 just take a reference before calling vfs_unlink() and release it
1735 when parent's i_sem is free.
1740 Description: last_id value checked outside lock on OST caused LASSERT failure
1741 Details : If there were multiple MDS->OST object precreate requests in
1742 flight, it was possible that the OST's last object id was checked
1743 outside a lock and incorrectly tripped an assertion. Move checks
1744 inside locks, and discard old precreate requests.
1747 Frequency : always, if extents are used on OSTs
1749 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1750 Details : OSTs using the extents format would not zero the high 16 bits of
1751 the index physical block number. This is not a problem for any
1752 OST filesystems smaller than 16TB, and no kernels support ext3
1753 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1754 new/modified files) and can be fixed for existing filesystems
1755 with e2fsprogs-1.39-cfs1.
1760 Description: import connection selection may be incorrect if timer wraps
1761 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1762 import connections to be ignored if the 32-bit jiffies counter
1763 wraps. Use a 64-bit jiffies counter.
1766 Frequency : very large clusters immediately after boot
1768 Description: LNET request buffers exhausted under heavy short-term load
1769 Details : If a large number of client requests are generated on a service
1770 that has previously never seen so many requests it is possible
1771 that the request buffer growth cannot keep up with the spike in
1772 demand. Instead of dropping incoming requests, they are held in
1773 the LND until the RPC service can accept more requests.
1776 Frequency : Sometimes during replay
1778 Description: Assertion failure in ll_local_open after replay.
1779 Details : If replay happened on an open request reply before we were able
1780 to set replay handler, reply will become not swabbed tripping the
1781 assertion in ll_local_open. Now we set the handler right after
1782 recognising of open request
1785 Frequency : very rare
1787 Description: kernel reports "badness in vsnprintf"
1788 Details : Reading from the "recovery_status" /proc file in small chunks
1789 may cause a negative length in lprocfs_obd_rd_recovery_status()
1790 call to vsnprintf() (which is otherwise harmless). Exit early
1791 if there is no more space in the output buffer.
1793 Severity : enhancement
1795 Description: clear OBD RPC statistics by writing to them
1796 Details : It is now possible to clear the OBD RPC statistics by writing
1797 to the "stats" file.
1802 Description: Client mtime is not the same on different clients after utimes
1803 Details : In some cases, the client was using the utimes() syscall on
1804 a file cached on another node. The clients now validate the
1805 ctime from the MDS + OSTs to determine which one is right.
1810 Description: Inability to activate failout mode
1811 Details : lconf script incorrectly assumed that in pythong string's numeric
1812 value is used in comparisons.
1815 Frequency : always with multiple stripes per file
1817 Description: Inefficient object allocation for mutli-stripe files
1818 Details : When selecting which OSTs to stripe files over, for files with
1819 a stripe count that divides evenly into the number of OSTs,
1820 the MDS is always picking the same starting OST for each file.
1821 Return the OST selection heuristic to the original design.
1826 Description: mount failures may take full timeout to return an error
1827 Details : Under some heavy load conditions it is possible that a
1828 failed mount can wait for the full obd_timeout interval,
1829 possibly several minutes, before reporting an error.
1830 Instead return an error as soon as the status is known.
1832 Frequency : quota enabled and large files being deleted
1834 Description: releasing more than 4GB of quota at once hangs OST
1835 Details : If a user deletes more than 4GB of files on a single OST it
1836 will cause the OST to spin in an infinite loop. Release
1837 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1842 Description: statfs data retrieved from /proc may be stale or zero
1843 Details : When reading per-device statfs data from /proc, in the
1844 {kbytes,files}_{total,free,avail} files, it may appear
1845 as zero or be out of date.
1848 Frequency : systems with MD RAID1 external journal devices
1850 Description: lconf's call to blkid is confused by RAID1 journal devices
1851 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1852 of returning all block devices that match the journal UUID.
1855 Frequency : always, for aggregate stripe size over 4GB
1857 Description: assertion fails when trying to use 4GB stripe size
1858 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1859 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1862 Frequency : always on ppc64
1864 Description: the first write on an ext3 filesystem with mballoc got stuck
1865 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1866 perform endianness conversion.
1868 ------------------------------------------------------------------------------
1870 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1872 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1873 INTEROPERATE with older versions automatically. Please read the
1874 user documentation before upgrading any part of a live system.
1875 * WARNING: Lustre networking configuration changes are required with
1876 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1879 * Support for kernels:
1880 2.6.9-22.0.2.EL (RHEL 4)
1881 2.6.5-7.244 (SLES 9)
1882 2.6.12.6 vanilla (kernel.org)
1885 Severity : enhancement
1886 Bugzilla : 7981/8208
1887 Description: Introduced Lustre Networking (LNET)
1888 Details : LNET is new networking infrastructure for Lustre, it includes
1889 a reorganized network configuration mode (see the user
1890 documentation for full details) as well as support for routing
1891 between different network fabrics. Lustre Networking Devices
1892 (LNDs) for the supported network fabrics have also been
1893 created for this new infrastructure.
1895 Severity : enhancement
1896 Description: Introduced Access control lists
1897 Details : clients can set ACLs on files and directories in order to have
1898 more fine-grained permissions than the standard Unix UGO+RWX.
1899 The MDS must be started with the "-o acl" mount option.
1901 Severity : enhancement
1902 Description: Introduced filesystem quotas
1903 Details : Administrators may now establish per-user quotas on the
1906 Severity : enhancement
1908 Description: Configuration change for the XT3
1909 The PTLLND is now used to run Lustre over Portals on the XT3
1910 The configure option(s) --with-cray-portals are no longer used.
1911 Rather --with-portals=<path-to-portals-includes> is used to
1912 enable building on the XT3. In addition to enable XT3 specific
1913 features the option --enable-cray-xt3 must be used.
1918 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1919 Details : A race between allocating a new llcd and re-getting the llcd_lock
1920 allowed another thread to grab newly-allocated llcd.
1922 Severity : enhancement
1924 Description: 2.6 OST async journal commit and locking fix to improve performance
1925 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1926 2.6 kernels are now async as they already were in 2.4 kernels so
1927 that they can commit concurrently with the network bulk transfer.
1928 For block-allocated files the filter allocation semaphore is held
1929 to avoid filesystem fragmentation during allocation. BKL lock
1930 removed for 2.6 xattr operations where it is no longer needed.
1935 Description: lconf incorrectly determined whether two IP networks could talk
1936 Details : In some more complicated routing and multiple-network
1937 configurations, lconf will avoid trying to make a network
1938 connection to a disjoint part of the IP space. It was doing the
1939 math incorrectly for one set of cases.
1944 Description: Fix for potential infinite loop processing records in an llog.
1945 Details : If an llog record is corrupted/zeroed, it is possible to loop
1946 forever in llog_process(). Validate the llog record length
1947 and skip the remainder of the block on error.
1950 Frequency : occasional (liblustre only)
1952 Description: liblustre could not open files whose last component is a symlink
1953 Details : sysio_path_walk() would incorrectly pass the open intent to
1954 intermediate path components.
1957 Frequency : rare (liblustre only with non-standard tuning)
1958 Bugzilla : 7201 (7350)
1959 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1960 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1961 after it has been released, only do this for referenced locks
1963 Severity : enhancement
1965 Description: specifying an (invalid) directory default stripe_size of -1
1966 would reset the directory default striping
1967 Details : stripe_size -1 was used internally to signal directory stripe
1968 removal, now use "all default" to signal dir stripe removal
1969 as a directory striping of "all default" is not useful
1972 Frequency : common for large clusters running liblustre clients
1974 Description: doing an ls when liblustre clients are running is slow
1975 Details : sending a glimpse AST to a liblustre client waits for every AST
1976 to time out, as liblustre clients will not respond. Since they
1977 cannot cache data we refresh the OST lock LVB from disk instead.
1979 Severity : enhancement
1981 Description: doing an ls at the same time as file IO can be slow
1982 Details : enqueue and other "small" requests can be blocked behind many
1983 large IO requests. Create a new OST IO portal for non-IO
1984 requests so they can be processed faster.
1987 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1989 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1990 created file with mode 000
1991 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1992 allow an ftruncate() as a non-root user to a file with mode 000.
1993 Lustre now respects this flag to disable mode checks when
1994 truncating a file owned by the user
1997 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2000 Description: Revoking locks from clients that went dead or catatonic might take
2002 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2003 cancellation of such locks instant on servers without waiting for
2004 any reply from clients. Clients drops these locks when cancel
2005 notification from server is received without replying.
2008 Frequency : liblustre-only, when liblustre client dies or becomes busy
2010 Description: Doing ls on Linux clients can take a long time with active
2012 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2013 granting such locks to it in the first place if possible. Locks
2014 are taken by proxy on the OST during the read or write and
2015 dropped immediately afterward. Add connect flags handling, do
2016 not grant locks to liblustre clients for glimpse ASTs.
2018 Severity : enhancement
2020 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2021 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2022 sequential reads and behaves badly with applications doing random
2023 reads. Improve it by reading ahead at least read region, and
2024 avoiding excessive large RPC for small reads.
2026 Severity : enhancement
2028 Description: Creating more than 1000 files for a single job may cause a load
2029 imbalance on the OSTs if there are also a large number of OSTs.
2030 Details : qos_prep_create() uses an OST index reseed value that is an
2031 even multiple of the number of available OSTs so that if the
2032 reseed happens in the middle of the object allocation it will
2033 still utilize the OSTs as uniformly as possible.
2038 Description: OST or MDS may oops in ping_evictor_main()
2039 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2040 but doesn't restart at beginning of obd_exports_timed list
2043 Severity : enhancement
2045 Description: improve by-nid export eviction on the MDS and OST
2046 Details : allow multiple exports with the same NID to be evicted at one
2047 time without re-searching the exports list.
2050 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2052 Description: MDS may oops in groups_free()
2053 Details : in rare race conditions a newly allocated group_info struct is
2054 freed again, and this can be NULL. The 2.4 compatibility code
2055 for groups_free() checked for a NULL pointer, but 2.6 did not.
2058 Frequency : common for liblustre clients doing little filesystem IO
2059 Bugzilla : 9352, 7313
2060 Description: server may evict liblustre clients accessing contended locks
2061 Details : if a client is granted a lock or receives a completion AST
2062 with a blocking AST already set it would not reply to the AST
2063 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2066 Frequency : lfs setstripe, only systems with more than 160 OSTs
2068 Description: unable to set striping with a starting offset beyond OST 160
2069 Details : llapi_create_file() incorrectly limited the starting stripe
2070 index to the maximum single-file stripe count.
2073 Frequency : LDAP users only
2075 Description: lconf did not handle in-kernel recovery with LDAP properly
2076 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2078 Severity : enhancement
2080 Description: bind OST threads to NUMA nodes to improve performance
2081 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2082 node and do their allocations there to localize memory access
2084 Severity : enhancement
2086 Description: llmount can determine client NID directly from Myrinet (GM)
2087 Details : the client NID code from gmnalnid was moved directly into
2088 llmount, removing the need to use this or specifying the
2089 client NID explicitly when mounting GM clients with zeroconf
2092 Frequency : if client is started with down MDS
2094 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2095 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2096 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2097 ptlrpc_import_delayed_req() because we didn't check if the
2098 request was interrupted, and we also didn't break out of the
2099 event loop if there was no timeout
2104 Description: data loss during non-page-aligned writes to a single file from
2105 both multiple nodes and multiple threads on one node at same time
2106 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2107 inconsistency led to false short-reads, that were cached and later
2108 used by ->prepare_write() to fill in partially written page,
2109 leading to data loss.
2112 Frequency : always, if lconf --abort_recovery used
2114 Description: lconf --abort_recovery fails with 'Operation not supported'
2115 Details : lconf was attempting to abort recovery on the MDT device and not
2118 Severity : enhancement
2120 Description: remove cleanup logs
2121 Details : replace lconf-generated cleanup logs with lustre internal
2122 cleanup routines. Eliminates the need for client-cleanup and
2125 Severity : enhancement
2127 Description: add support for EAs (user and system) on lustre filesystems
2128 Details : it is now possible to store extended attributes in the Lustre
2129 client filesystem, and with the user_xattr mount option it
2130 is possible to allow users to store EAs on their files also
2132 Severity : enhancement
2134 Description: Add possibility (config option) to show minimal available OST free
2136 Details : When compiled with --enable-mindf configure option, statfs(2)
2137 (and so, df) will return least minimal free space available from
2138 all OSTs as amount of free space on FS, instead of summary of
2139 free spaces of all OSTs.
2141 Severity : enhancement
2143 Description: do not expand extent locks acquired on OST-side
2144 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2145 by server: they are not cached anyway.
2148 Frequency : when mmap is used/binaries executed from Lustre
2150 Description: Unmmap pages before throwing them away from read cache.
2151 Details : llap_shrink cache now attempts to unmap pages before discarding
2152 them (if unmapping failed - do not discard). SLES9 kernel has
2153 extra checks that trigger if this unmapping is not done first.
2158 Description: lconf didn't resolve symlinks before checking to see whether a
2159 given mountpoint was already in use
2162 Frequency : when migrating failover services
2163 Bugzilla : 6395, 9514
2164 Description: When migrating a subset of services from a node (e.g. failback
2165 from a failover service node) the remaining services would
2166 time out and evict clients.
2167 Details : lconf --force (implied by --failover) sets the global obd_timeout
2168 to 5 seconds in order to quickly disconnect, but this caused
2169 other RPCs to time out too quickly. Do not change the global
2170 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2172 Severity : enhancement
2173 Frequency : if MDS is started with down OST
2174 Bugzilla : 9439,5706
2175 Description: Allow startup/shutdown of an MDS without depending on the
2176 availability of the OSTs.
2177 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2178 Add appropriate locking and lov-osc refcounts for safe
2179 cleaning. Add osc abort_inflight calls in case the
2180 synchronize never started.
2183 Frequency : occasional (Cray XT3 only)
2185 Description: root not authorized to access files in CRAY_PORTALS environment
2186 Details : The client process capabilities were not honoured on the MDS in
2187 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2188 been accessed by an authorized user then root was able to access
2189 the file on the local client also. The root user capabilities
2190 are now allowed on the MDS, as this environment has secure UID.
2193 Frequency : occasional
2195 Description: ldiskfs "too long searching" message happens too often
2196 Details : A debugging message (otherwise harmless) prints too often on
2197 the OST console. This has been reduced to only happen when
2198 there are fragmentation problems on the filesystem.
2203 Description: Division by zero in statfs when all OSCs are inactive
2204 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2205 lov_statfs divides by value returned by lov_get_stripecnt().
2209 Bugzilla : 9489, 3273
2210 Description: First write from each client to each OST was only 4kB in size,
2211 to initialize client writeback cache, which caused sub-optimal
2212 RPCs and poor layout on disk for the first writen file.
2213 Details : Clients now request an initial cache grant at (re)connect time
2214 and so that they can start streaming writes to the cache right
2215 away and always do full-sized RPCs if there is enough data.
2216 If the OST is rebooted the client also re-establishes its grant
2217 so that client cached writes will be honoured under the grant.
2222 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2223 Details : Now I/O RPCs go to different portal number and (presumably) fast
2224 lock requests (and glimses) and other RPCs get their own service
2225 threads pool that should be able to service those RPCs
2228 Severity : enhancement
2230 Description: Ability to exchange lustre version between client and servers and
2231 issue warnings at client side if client is too old. Also for
2232 liblustre clients there is ability to refuse connection of too old
2234 Details : New 'version' field is added to connect data structure that is
2235 filled with version info. That info is later checked by server and
2239 Frequency : rare, liblustre only.
2240 Bugzilla : 9296, 9581
2241 Description: Two simultaneous writes from liblustre at offset within same page
2242 might proceed at the same time overwriting eachother with stale
2244 Details : I/O lock withing llu_file_prwv was released too early, before data
2245 actually was hitting the wire. Extended lock-holding time until
2246 server acknowledges receiving data.
2249 Frequency : extremely rare. Never observed in practice.
2251 Description: avoid generating lustre_handle cookie of 0.
2252 Details : class_handle_hash() generates handle cookies by incrementing
2253 global counter, and can hit 0 occasionaly (this is unlikely, but
2254 not impossible, because initial value of cookie counter is
2255 selected randonly). Value of 0 is used as a sentinel meaning
2256 "unassigned handle" --- avoid it. Also coalesce two critical
2257 sections in this function into one.
2259 Severity : enhancement
2261 Description: allow liblustre clients to delegate truncate locking to OST
2262 Details : To avoid overhead of locking, liblustre client instructs OST to
2263 take extent lock in ost_punch() on client's behalf. New connection
2264 flag is added to handle backward compatibility.
2266 Severity : enhancement
2267 Bugzilla : 4928, 7341, 9758
2268 Description: allow number of OST service threads to be specified
2269 Details : a module parameter allows the number of OST service threads
2270 to be specified via "options ost ost_num_threads={N}" in the
2271 OSS's /etc/modules.conf or /etc/modprobe.conf.
2275 Bugzilla : 6146, 9635, 9895
2276 Description: servers crash with bad pointer in target_handle_connect()
2277 Details : In rare cases when a client is reconnecting it was possible that
2278 the connection request was the last reference for that export.
2279 We would temporarily drop the export reference and get a new
2280 one, but this may have been the last reference and the export
2281 was just destroyed. Get new reference before dropping old one.
2283 Severity : enhancement
2284 Frequency : if client is started with failover MDS
2286 Description: Allow multiple MDS hostnames in the mount command
2287 Details : Try to read the configuration from all specified MDS
2288 hostnames during a client mount in case the "primary"
2291 Severity : enhancement
2293 Description: Stop sending data to evicted clients as soon as possible.
2294 Details : Check if the client we are about to send or are sending data to
2295 was evicted already. (Check is done every second of waiting,
2296 for which l_wait_event interface was extended to allow checking
2297 of exit condition at specified intervals).
2300 Frequency : rare, normally only when NFS exporting is done from client
2302 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2304 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2307 Frequency : always, if srand() is called before liblustre initialization
2309 Description: Liblustre uses system PRNG disturbing its usage by user application
2310 Details : Introduce internal to lustre fast and high-quality PRNG for
2311 lustre usage and make liblustre and some other places in generic
2312 lustre code to use it.
2314 Severity : enhancement
2315 Bugzilla : 9477, 9557, 9870
2316 Description: Verify that the MDS configuration logs are updated when xml is
2317 Details : Check if the .xml configuration logs are newer than the config
2318 logs stored on the MDS and report an error if this is the case.
2319 Request --write-conf, or allow starting with --old_conf.
2321 Severity : enhancement
2323 Description: Handle symlinks in the path when checking if Lustre is mounted.
2324 Details : Resolve intermediate symlinks when checking if a client has
2325 mounted a filesystem to avoid duplicate client mounts.
2330 Description: lconf can hit an error exception but still return success.
2331 Details : The lconf command catches the Command error exception at the top
2332 level script context and will exit with the associated exit
2333 status, but doesn't ensure that this exit status is non-zero.
2338 Description: failure of ptlrpc thread startup can cause oops
2339 Details : Starting a ptlrpc service thread can fail if there are a large
2340 number of threads or the server memory is very fragmented.
2341 Handle this without oopsing.
2344 Frequency : always, only if liblustre and non-default acceptor port was used
2346 Description: liblustre cannot connect to servers with non-default acceptor port
2347 Details : tcpnal_set_default_params() was not called and was therefore
2348 ignoring the environment varaible TCPNAL_PORT, as well as other
2349 TCPNAL_ environment variables
2354 Description: two objects could be created on the same OST for a single file
2355 Details : If an OST is down, in some cases it was possible to create two
2356 objects on a single OST for a single file. No problems other
2357 than potential performance impact and spurious error messages.
2361 Bugzilla : 5681, 9562
2362 Description: Client may oops in ll_unhash_aliases
2363 Details : Client dcache may become inconsistent in race condition.
2364 In some cases "getcwd" can fail if the current directory is
2370 Description: Inode refcounting problems in NFS export code
2371 Details : link_raw functions used to call d_instantiate without obtaining
2372 extra inode reference first.
2376 Bugzilla : 9942, 9903
2377 Description: Referencing freed requests leading to crash, memleaks with NFS.
2378 Details : We used to require that call to ll_revalidate_it was always
2379 followed by ll_lookup_it. Also with revalidate_special() it is
2380 possible to call ll_revalidate_it() twice for the same dentry
2381 even if first occurence returned success. This fix changes semantic
2382 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2383 reference on a request referred from the intent.
2384 ll_intent_release() then releases such a request.
2387 Frequency : rare, normally benchmark loads only
2389 Description: unlinked inodes were kept in memory on the client
2390 Details : If a client is repeatedly creating and unlinking files it
2391 can accumulate a lot of stale inodes in the inode slab cache.
2392 If there is no other client load running this can cause the
2393 client node to run out of memory. Instead flush old inodes
2394 from client cache that have the same inode number as a new inode.
2397 Frequency : SLES9 2.6.5 kernel and long filenames only
2398 Bugzilla : 9969, 10379
2399 Description: utime reports stale NFS file handle
2400 Details : SLES9 uses out-of-dentry names in some cases, which confused
2401 the lustre dentry revalidation. Change it to always use the
2405 Frequency : rare, unless heavy write-truncate concurrency is continuous
2406 Bugzilla : 4180, 6984, 7171, 9963, 9331
2407 Description: OST becomes very slow and/or deadlocked during object unlink
2408 Details : filter_destroy() was holding onto the parent directory lock
2409 while truncating+unlinking objects. For very large objects this
2410 may block other threads for a long time and slow overall OST
2411 responsiveness. It may also be possible to get a lock ordering
2412 deadlock in this case, or run out of journal credits because of
2413 the combined truncate+unlink. Solution is to do object truncate
2414 first in one transaction without parent lock, and then do the
2415 final unlink in a new transaction with the parent lock. This
2416 reduces the lock hold time dramatically.
2419 Frequency : rare, 2.4 kernels only
2421 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2422 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2423 before unmounting their filesystem in order to avoid deadlock.
2424 The kernel_locked() function in 2.4 kernels only checks whether
2425 the kernel lock is held, not whether it is this process that is
2426 holding it as 2.6 kernels do.
2431 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2432 Details : The client ptlrpc code may be trying to reconnect to a down
2433 server before a previous connection attempt has timed out.
2434 Increase the reconnect interval to be longer than the connection
2435 timeout interval to avoid sending duplicate connections to
2439 Frequency : echo_client brw_test command
2441 Description: fix echo_client to work with OST preallocated code
2442 Details : OST preallocation code (5137) didn't take echo_client IO path
2443 into account: echo_client calls filter methods outside of any
2444 OST thread and, hence, there is no per-thread preallocated
2445 pages and buffers to use. Solution: hijack pga pages for IO. As
2446 a byproduct, this avoids unnecessary data copying.
2450 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2451 Description: Client can oops in mdc_commit_close() after open replay
2452 Details : It was possible for the MDS to return an open request with no
2453 transaction number in mds_finish_transno() if the client was
2454 evicted, but without actually returning an error. Clients
2455 would later try to replay that open and may trip an assertion
2456 Simplify the client close codepath, and always return an error
2457 from the MDS in case the open is not successful.
2460 Frequency : rare, 2.6 OSTs only
2462 Description: OST may deadlock under high load on fragmented files
2463 Details : If there was a heavy load and highly-fragmented OST filesystems
2464 it was possible to have all the OST threads deadlock waiting on
2465 allocation of biovecs, because the biovecs were not released
2466 until the entire RPC IO was completed. Instead, release biovecs
2467 as soon as they are complete to ensure forward IO progress.
2469 Severity : enhancement
2471 Description: Support for specifying external journal device at mount
2472 Details : If an OST or MDS device is formatted with an external journal
2473 device, this device major/minor is stored in the ext3 superblock
2474 and may not be valid for failover. Allow detecting and
2475 specifying the external journal at mount time.
2480 Description: Mounting an MDS with pending unlinked files may cause oops
2481 Details : target_finish_recovery() calls mds_postrecov() which returned
2482 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2483 considers this an error and immediately begins cleaning up the
2484 lov, just after starting the mds_lov process
2486 Severity : enhancement
2488 Description: Implement 'lfs df' to report actual free space on per-OST basis
2489 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2490 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2491 usage of objects; '-h' to report in human readable format.
2493 ------------------------------------------------------------------------------
2495 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2502 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2503 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2504 at startup during orphan unlinking. Instead, skip buddy update
2505 before setup, as it will be regenerated after recovery is complete.
2510 Description: If an OST is inactive, its locks might reference stale inodes.
2511 Details : lov_change_cbdata() must iterate over all namespaces, even if
2512 they are inactive to clear inode references from the lock.
2514 Severity : enhancement
2515 Frequency : occasional, if non-standard max_dirty_mb used
2517 Description: Client will block write RPCs if not enough grant
2518 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2519 then the client will block writes while waiting for another RPC
2520 to complete instead of consuming its dirty limit. With change
2521 we get improved performance when max_dirty_mb is small.
2523 Severity : enhancement
2524 Bugzilla : 3389, 6253
2525 Description: Add support for supplementary groups on the MDS.
2526 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2527 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2528 lookups for user supplementary groups into a cache.
2532 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2533 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2534 for liblustre O_EXCL is enforced by MDS.
2538 Description: readdir never returns NULL in liblustre.
2539 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2540 dentry in struct dirent.
2544 Frequency : liblustre only, depends on application IO pattern
2545 Description: liblustre clients evicted if not contacting servers
2546 Details : Don't put liblustre clients into the ping_evictor list, so
2547 they will not be evicted by the pinger ever.
2549 Severity : enhancement
2551 Description: Add ability to evict clients by NID from MDS.
2552 Details : By echoing "nid:$NID" string into
2553 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2554 $NID would be instantly evicted from this MDS and from all active
2555 OSTs connected to it.
2559 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2560 Details : lookup_it_finish() used to query file size from OSTs that was not
2565 Description: service threads change working directory to that of init
2566 Details : Starting lustre service threads may pin the working directory
2567 of the parent thread, making that filesystem busy. Threads
2568 now change to the working directory of init to avoid this.
2572 Frequency : during shutdown only
2573 Description: shutdown with a failed MDS or OST can cause unmount to hang
2574 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2579 Frequency : 2.6 only
2580 Description: chmod/chown may include an extra supplementary group
2581 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2582 supplementary group and if none is specified this is used.
2585 Bugzilla : 5479 (6816)
2587 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2588 Details : If lookup is in progress on a file that is unlinked we might try
2589 to revalidate the inode and fail in revalidate after lookup is
2590 complete and ll_file_open() enqueues the open again but
2591 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2594 Frequency : always, if lconf --abort_recovery used
2596 Description: lconf --abort_recovery fails with 'Operation not supported'
2597 Details : lconf was attempting to abort recovery on the MDT device and not
2600 ------------------------------------------------------------------------------
2602 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2607 Frequency : rare (only unsupported configurations with a node running as an
2609 Bugzilla : 6514, 5137
2610 Description: Mounting a Lustre file system on a node running as an OST could
2612 Details : OSTs now preallocates memory needed to write out data at
2613 startup, instead of when needed, to avoid having to
2614 allocate memory in possibly low memory situations.
2615 Specifically, if the file system is mounted on on OST,
2616 memory pressure could force it to try to write out data,
2617 which it needed to allocate memory to do. Due to the low
2618 memory, it would be unable to do so and the node would
2619 become unresponsive.
2621 Severity : enhancement
2623 Description: Addition of lconf --service command line option
2624 Details : lconf now accepts a '--service <arg>' option, which is
2625 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2627 Severity : enhancement
2629 Description: Failover mode is now the default for OSTs.
2630 Details : By default, OSTs will now run in failover mode. To return to
2631 the old behaviour, add '--failout' to the lmc line for OSTs.
2633 Severity : enhancement
2635 Description: Health checks are now provided for MDS and OSTs
2636 Details : Additional detailed health check information on MSD and OSTs
2637 is now provided through the procfs health_check value.
2640 Frequency : occasional, depends on IO load
2642 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2643 after numerous create/delete cycles
2644 Details : The ext3 inode allocation policy would not allocate new inodes
2645 very well on the OSTs because there are no new directories
2646 being created. Instead we look for groups with free space if
2647 the parent directories are nearly full.
2652 Description: Network or server problems during mount may cause partially
2653 mounted clients instead of returning an error.
2654 Details : The config llog parsing code may overwrite the error return
2655 code during mount error handling, returning success instead
2661 Description: MDS can fail to allocate large reply buffers
2662 Details : After long uptimes the MDS can fail to allocate large reply
2663 buffers (e.g. zconf client mount config records) due to memory
2664 fragmentation or consumption by the buffer cache. Preallocate
2665 some large reply buffers so that these replies can be sent even
2666 under memory pressure.
2670 Frequency : rare (liblustre)
2671 Description: fsx running with liblustre complained that using truncate() to
2672 extend the file doesn't work. This patch corrects that issue.
2673 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2674 ATTR_SIZE and lsm use in llu_setattr_raw.
2678 Frequency : rare, only 2.6 kernels
2679 Description: Unusual file access patterns on the MDS may result in inode
2680 data being lost in very rare circumstances.
2681 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2682 code in the 2.6 kernel caused the inode and/or EA data not to
2683 be read from disk, causing single-file corruption.
2687 Frequency : rare, only 2.6 filesystems using extents
2688 Description: Heavy concurrent write and delete load may cause data corruption.
2689 Details : It was possible under high-load situations to have an extent
2690 metadata block in the block device cache from a just-unlinked
2691 file overwrite a newly-allocated data block. We now unmap any
2692 metadata buffers that alias just-allocated data blocks.
2696 Frequency : filesystems with default stripe_count larger than 77
2697 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2698 Details : lconf specifies an inode size of 4096 bytes when the default
2699 stripe_count is larger than 77. This conflicts with the default
2700 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2701 this case to avoid pinning too much memory for large EAs.
2703 ------------------------------------------------------------------------------
2705 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2710 Frequency : rare (extremely heavy IO load with hundreds of clients)
2712 Description: Client is evicted, gets IO error writing to file
2713 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2714 caused clients to be evicted for AST timeouts. The fixes in
2715 bug 5192 mean we no longer need to have such short AST timeouts
2716 so ldlm_timeout has been increased.
2719 Frequency : occasional during --force or --failover shutdown under load
2720 Bugzilla : 5949, 4834
2721 Description: Server oops/LBUG if stopped with --force or --failover under load
2722 Details : a collection of import/export refcount and cleanup ordering
2723 issues fixed for safer force cleanup
2726 Frequency : only filesystems larger than 120 OSTs
2727 Bugzilla : 5990, 6223
2728 Description: lfs getstripe would oops on a very large filesystem
2729 Details : lov_getconfig used kfree on vmalloc'd memory
2732 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2733 Bugzilla : 6242, 6243
2734 Description: reading from files that had been truncated to a non-zero size
2735 but never opened returned no data
2736 Details : ll_file_read() reads zeros from no-object files to EOF
2741 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2742 Details : There's little harm in aborting MDS/OSS recovery and letting it
2743 try again, so I removed the LASSERT and return an error instead.
2745 Severity : enhancement
2747 Description: New debugging infrastructure for tracking down data corruption
2748 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2749 (b) cover more of the client-side code path, and (c) try to narrow
2750 down where problems occurred
2754 Bugzilla : 3819, 4364, 4397, 6313
2755 Description: Racing close and eviction MDS could cause assertion in mds_close
2756 Details : It was possible to get multiple mfd references during close and
2757 client eviction, leading to one thread referencing a freed mfd.
2759 Severity: : enhancement
2760 Bugzilla : 3262, 6359
2761 Description: Attempts to reconnect to servers are now more aggressive.
2762 Details : This builds on the enhanced upcall-less recovery that was added
2763 in 1.4.2. When trying to reconnect to servers, clients will
2764 now try each server in the failover group every 10 seconds. By
2765 default, clients would previously try one server every 25 seconds.
2770 Description: After recovery, certain operations trigger a failed
2771 assertion on a client.
2772 Details : Failing over an mds, using lconf -d --failover, while a
2773 client was doing a readdir() call would cause the client to
2774 LBUG after recovery completed and the readdir() was resent.
2776 Severity : enhancement
2778 Description: Default groups are now added by lconf
2779 Details : You can now run lconf --group <servicename> without having to
2780 manually add groups with lmc.
2783 Frequency : occasional
2785 Description: Nodes with an elan id of 0 trigger a failed assertion
2788 Frequency : always when accessing e.g. tty/console device nodes
2790 Description: tty and some other devices nodes cannot be used on lustre
2791 Details : file's private_data field is used by device data and lustre
2792 values in there got lost. New field was added to struct file to
2793 store fs-specific private data.
2796 Frequency : when exporting Lustre via NFS
2798 Description: NFSD failed occasionally when looking up a path component
2799 Details : NFSD is looking up ".." which was broken in ext3 directories
2800 that had grown large enough to become hashed.
2803 Frequency : Clusters with multiple interfaces not on the same subnet
2805 Description: Nodes will repeatedly try to reconnect to an interface which it
2806 cannot reach and report an error to the log.
2807 Details : Extra peer list entries will be created by lconf with some peers
2808 unreachable. lconf now validates the peer before adding it.
2811 Frequency : Only if a default stripe is set on the filesystem root.
2813 Description: Setting a default stripe on the filesystem root prevented the
2814 filesystem from being remounted.
2815 Details : The client was sending extra request flags in the root getattr
2816 request and did not allocate a reply buffer for the dir EA.
2819 Frequency : occasional, higher if lots of files are accessed by one client
2820 Bugzilla : 6159, 6097
2821 Description: Client trips assertion regarding lsm mismatch/magic
2822 Details : While revalidating inodes the VFS looks up inodes with ifind()
2823 and in rare cases can find an inode that is being freed.
2824 The ll_test_inode() code will free the lsm during ifind()
2825 when it finds an existing inode and then the VFS later attaches
2826 this free lsm to a new inode.
2830 Bugzilla : 6422, 7030
2831 Description: MDS deadlock between mkdir and client eviction
2832 Details : Creating a new file via mkdir or mknod (starting a transaction
2833 and getting the ns lock) can deadlock with client eviction
2834 (gets ns lock and trying to finish a synchronous transaction).
2837 Frequency : occasional
2838 Description: While starting a server, the fsfilt_ext3 module could not be
2840 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2841 kernels. Previously, lconf would still use the ext3 name
2842 when trying to load modules. Now, it will correctly use
2843 ext3 on 2.4 and ldiskfs on 2.6.
2845 Severity : enhancement
2846 Description: The default stripe count has been changed to 1
2847 Details : The interpretation of the default stripe count (0, to lfs
2848 or lmc) has been changed to mean striping across a single
2849 OST, rather than all available. For general usage we have
2850 found a stripe count of 1 or 2 works best.
2852 Severity : enhancement
2853 Description: Add support for compiling against Cray portals.
2854 Details : Conditional compiling for some areas that are different
2858 Frequency : occasional
2859 Bugzilla : 6409, 6834
2860 Description: Creating files with an explicit stripe count may lead to
2861 a failed assertion on the MDS
2862 Details : If some OSTs are full or unavailable, creating files may
2863 trigger a failed assertion on the MDS. Now, Lustre will
2864 try to use other servers or return an error to the
2868 Frequency : occasional
2870 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2871 caused lock timeout during readahead (since 1.4.2).
2872 Details : Processes doing readahead might match a lock that hasn't been
2873 granted yet if there are overlapping and conflicting lock
2874 requests. The readahead process waits on ungranted lock
2875 (original lock is CBPENDING), while OST waits for that process
2876 to cancel CBPENDING read lock and eventually evicts client.
2878 Severity : enhancement
2880 Description: Initial enabling of flock support for clients
2881 Details : Implements fcntl advisory locking and file status functions.
2882 This feature is provided as an optional mount flag (default
2883 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2884 locking are implemented yet, and those that are are not guaranteed
2885 to be completely correct in production environments.
2886 mount -t lustre -o [flock|noflock] ...
2889 Frequency : occasional
2891 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2892 assertion in the ext3 JBD (journaling) layer.
2893 Details : The b_committed_data struct is protected by the big kernel lock
2894 in 2.4 kernels, serializing journal_commit_transaction() and
2895 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2896 access to this struct. In 2.6 kernels there is finer grained
2897 locking to improve SMP performance of the JBD layer.
2901 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2906 Description: merge and cleanup kernel patches.
2909 -----------------------------------------------------------------------------
2911 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2913 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2914 versions of Lustre. Please update all servers and clients to
2915 version 1.4.2 or later at the same time. You must also run
2916 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2917 to update the configuration logs.
2919 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2920 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2921 - swab llog records of type '0' so we get proper header size/idx (5861)
2922 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2923 - fix rename of one directory over another leaking an inode (5953)
2924 - avoid SetPageDirty on 2.6 (5981)
2925 - don't re-add just-being-destroyed locks to the waiting list (5653)
2926 - when creating new directories, inherit the parent's custom
2927 striping settings if present parent (3048)
2928 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2929 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2930 - don't hold client locks on temporary worklist from l_lru (5666)
2931 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2932 - automatically evict dead clients (5921)
2933 - Update file size properly in create+truncate+fstat case (6196)
2934 - Do not unhash mountpoint dentries, do not allow removal of
2936 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2937 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2938 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2939 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2940 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2941 - Correctly handle reads of files with no objects (6243)
2942 - lctl recover will also mark a device active if deactivate used (5933)
2944 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2945 - allow --write-conf on an MDS with different nettype than client (5619)
2946 - don't write config llogs to MDS for mounts not from that MDS (5617)
2947 - lconf should create multiple TCP connections from a client (5201)
2948 - init scripts are now turned off by default; run chkconfig --on
2949 lustre and chkconfig --on lustrefs to use them
2950 - upcalls are no longer needed for clients to recover to failover
2952 - add --abort-recovery option to lconf to abort recovery on device
2954 - add support for an arbitrary number of OSTs (3026)
2955 - Quota support protocol changes.
2956 - forward compatibility changes to wire structs (6007)
2957 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2958 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2959 - improved "lustre status" script
2960 - initialize blocksize for non-regular files (6062)
2961 - added --disable-server and --disable-client configure options (5782)
2962 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2963 - Vanilla 2.4.29 support
2964 - increase maximum number of obd devices to 520 (6242)
2965 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2966 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2967 SLES 9 2.6.5 kernels
2968 - limit stripes per file to 160 (the maximum EA size) (6093)
2970 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2973 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2974 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2975 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2976 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2977 - small ext3 extents cleanups and fixes (5733)
2978 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2979 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2980 - avoid needless client->OST connect, fix handle mismatch (5317)
2981 - fix DLM error path that led to out-of-sync client, long delays (5779)
2982 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2983 - fix several locking issues related to i_size (5492,5624,5654,5672)
2984 - don't move pending lock onto export if it is already evicted (5683)
2985 - fix kernel oops when creating .foo in unlinked directory (5548)
2986 - fix deadlock in obdfilter statistics vs. object create (5811)
2987 - use time_{before,after} to avoid timer jiffies wrap (5882)
2988 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2989 - Do not leak request if server was not able to process it (5154)
2990 - If mds_open unable to find parent dir, make that negative lookup(5154)
2991 - don't create new directories with extent-mapping (5909, 5936)
2993 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2994 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2995 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2996 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2997 - support for RHEL4 kernel on i686 (5773)
2998 - provide error messages when incompatible logs are encountered (5898)
3000 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3001 * version 1.4.0.10 (1.4.1 release candidate 1)
3003 - don't keep a lock reference when lock is not granted (4238)
3004 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3005 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3006 - drop import inflight refcount on signal_completed_replay error (5255)
3007 - unlock page after async write error during send (3677)
3008 - handle missing objects in filter_preprw_read properly (5265)
3009 - no transno return for symlink open, don't save no-trasno open (3440)
3010 - don't try to complete elan receive that already failed (4012)
3011 - free RPC server reply state on error (5406)
3012 - clean up thread from ptlrpc_start_thread() on error (5160)
3013 - readahead could read extra page into cache that wasn't ejected (5388)
3014 - prevent races in class_attach/setup/cleanup/detach (5260)
3015 - don't dereference de->d_inode after l_dput of de (5458)
3016 - use "int" for stripe value returned from lock_to_stripe (5544)
3017 - mballoc allocation and error-checking fixes in 2.6 (5504)
3018 - block device patches to fix I/O request sizes in 2.6 (5482)
3019 - look up hostnames for IB nals (5602)
3020 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3021 - don't start multiple acceptors for the same port (5277)
3022 - fix incorrect LASSERT in mds_getattr_name (5635)
3023 - export a proc file for general "ping" checking (5628)
3024 - fix "lfs check" to not block when the MDS is down (5628)
3026 - service request history (4965)
3027 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3028 - create an "evict_client" /proc entry on OSTs, like the MDS has
3029 - fix mount usage message, return errors per mount(8) (5168)
3030 - change grep [] to grep "[]" in tests so they work in more UMLs
3031 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3032 - remove ancient LOV_MAGIC_V0 EA support (5047)
3033 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3034 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3035 - allow readahead allocations to fail when low on memory (5383)
3036 - mmap locking landed again, after considerable improvement (2828)
3037 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3039 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3042 - send OST transaction number in read/write reply to free req (4966)
3043 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3044 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3045 - the watchdog thread now runs as interruptible (5246)
3046 - flock/lockf fixes (but it's still disabled, pending 5135)
3047 - don't use EXT3 constants in llite code (5094)
3048 - memory shortage at startup could cause assertion (5176)
3050 - reorganization of lov code
3051 - single portals codebase
3053 - add extents/mballoc support (5025)
3054 - direct I/O reads in the obdfilter (4048)
3055 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3057 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3060 - send OST transaction number in read/write reply to free req (4966)
3061 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3062 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3063 - the watchdog thread now runs as interruptible (5246)
3064 - handle missing objects in filter_preprw_read properly (5265)
3065 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3066 - drop import inflight refcount on signal_completed_replay error (5255)
3067 - unlock page after async write error during send (3677)
3068 - return original error code on reconstructed replies (3761)
3069 - no transno return for symlink open, don't save no-trasno open (3440)
3071 - add pid to ldlm debugging output (4922)
3072 - bump the watchdog timeouts -- we can't handle 30sec yet
3073 - extra debugging for orphan dentry/inode bug (5259)
3075 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3078 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3079 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3080 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3081 - fix NULL dereference in /proc/sys/portals/routes (4827)
3082 - allow failed mdc_close() operations to be interrupted (4561)
3083 - stop precreate on OST before MDS would time out on it (4778)
3084 - don't send partial-page writes before EOF from client (4410)
3085 - discard client grant for sub-page writes on large-page clients (4520)
3086 - don't free dentries not owned by NFS code, check generation (4806)
3087 - fix lsm leak if mds_create_objects() fails (4801)
3088 - limit debug_daemon file size, always print CERROR messages (4789)
3089 - use transno after validating reply (3892)
3090 - process timed out requests if import state changes (3754)
3091 - update mtime on OST during writes, return in glimpse (4829)
3092 - add mkfsoptions to LDAP (4679)
3093 - use ->max_readahead method instead of zapping global ra (5039)
3094 - don't interrupt __l_wait_event() during strace
3096 - add software watchdogs to catch hung threads quickly (4941)
3097 - make lustrefs init script start after nfs is mounted
3098 - fix CWARN/ERROR duplication (4930)
3099 - return async write errors to application if possible (2248)
3100 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3101 - print NAL number in %x format (4645)
3102 - update barely-supported suse-2.4.21-171 series (4842)
3103 - support for sles 9 %post scripts
3104 - support for building 2.6 kernel-source packages
3105 - support for sles km_* packages
3107 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3110 - ignore -ENOENT errors in osc_destroy (3639)
3111 - notify osc create thread that OSC is being cleaned up (4600)
3112 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3113 - reconstruct ost_handle() like mds_handle() (4657)
3114 - create a new thread to do import eviction to avoid deadlock (3969)
3115 - let lconf resolve symlinked-to devices (4629)
3116 - don't unlink "objects" from directory with default EA (4554)
3117 - hold socknal file ref over connect in case target is down (4394)
3118 - allow more than 32000 subdirectories in a single directory (3244)
3119 - fix blocks count for O_DIRECT writes (3751)
3120 - OST returns ENOSPC from object create when no space left (4539)
3121 - don't send truncate RPC if file size isn't changing (4410)
3122 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3123 - bind to privileged port in socknal and tcpnal (3689)
3125 - rate limit CERROR/CWARN console message to avoid overload (4519)
3126 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3127 - basic mmap support (3918)
3128 - kernel patch series update from b1_4 (4711)
3130 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3133 - avoid crash during MDS cleanup with OST shut down (2775)
3134 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3135 - don't use bad inodes on the MDS (3744)
3136 - dynamic object preallocation to improve recovery speed (4236)
3137 - don't hold spinlock over lock dumping or change debug flags (4401)
3138 - don't zero obd_dev when it is force cleaned (3651)
3139 - print grants to console if they go negative (4431)
3140 - "lctl deactivate" will stop automatic recovery attempts (3406)
3141 - look for existing locks in ldlm_handle_enqueue() (3764)
3142 - don't resolve lock handle twice in recovery avoiding race (4401)
3143 - revalidate should check working dir is a directory (4134)
3145 - don't always mark "slow" obdfilter messages as errors (4418)
3147 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3150 - don't close LustreDB during write_conf until it is done (3860)
3151 - fix typo in lconf for_each_profile (3821)
3152 - allow dumping logs from multiple threads at one time (3820)
3153 - don't allow multiple threads in OSC recovery (3812)
3154 - fix debug_size parameters (3864)
3155 - fix mds_postrecov to initialize import for llog ctxt (3121)
3156 - replace config semaphore with spinlock (3306)
3157 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3158 - don't allow enqueue to complete on a destroyed export (3822)
3159 - down write_lock before checking llog header bitmap (3825)
3160 - recover from lock replay timeout (3764)
3161 - up llog sem before sending rpc (3652)
3162 - reduce ns lock hold times when setting kms (3267)
3163 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3164 - fix NULL deref and obd_dev leak on setup error (3312)
3165 - replace some LBUG about llog ops with error handling (3841)
3166 - don't match INVALID dentries from d_lookup and spin (3784)
3167 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3168 - fix invalid assertion in ptlrpc_set_wait (3880)
3170 - add libwrap support for the TCP acceptor (3996)
3171 - add /proc/sys/portals/routes for non-root route listing (3994)
3172 - allow setting MDS UUID in .xml (2580)
3173 - print the stack of a process that LBUGs (4228)
3175 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3178 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3179 - make sure to unset replay flag from failed open requests (3440)
3180 - if default stripe count is 0, use OST count for inode size (3636)
3181 - update parent mtime/ctime on client for create/unlink (2611)
3182 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3183 - free recovery state on server during a forced cleanup (3571)
3184 - unregister_reply for resent reqs (3063)
3185 - loop back devices mounting and status check on 2.6 (3563)
3186 - fix resource-creation race that can provoke i_size == 0 (3513)
3187 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3188 - more debugging for page-accounting assertion (3746)
3189 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3190 - avoid deadlock after precreation failure (3758)
3191 - fix race and lock order deadlock in orphan handling (3450, 3750)
3192 - add validity checks when grabbing inodes from l_ast_data (3599)
3194 - add /proc/.../recovery_status to obdfilter (3428)
3195 - lightweight CDEBUG infrastructure, debug daemon (3668)
3196 - change default OSC RPC parameters to be better on small clusters
3197 - turn off OST read cache for files smaller than 32MB
3198 - install man pages and include them in rpms (3100)
3199 - add new init script for (un)mounting lustre filesystems (2593)
3200 - run chkconfig in %post for init scripts (3701)
3201 - drop scimac NAL (unmaintained)
3203 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3206 - clean kiobufs before and after use (3485)
3207 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3208 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3209 - change init script to fail more clearly if not run as root (1528)
3210 - allow clients to reconnect during replay (1742)
3211 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3212 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3213 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3214 - fix cond_resched (3554)
3215 - only evict unfinished clients after recovery (3515)
3216 - allow bulk resend, prevent data loss (3570)
3217 - dynamic ptlrpc request buffer allocation (2102)
3218 - don't allow unlinking open directory if it isn't empty (2904)
3219 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3220 - remove extraneous obd dereference causing LASSERT failure (3334)
3221 - don't use get_cycles() when creating temp. files on the mds (3156)
3222 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3223 - handle EEXIST for set-stripe, set proper directory name (3336)
3225 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3226 - fix ksocknal_fmb_callback() error messages (2918)
3228 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3231 - don't copy lvb into (possibly NULL) reply on error (2983)
3232 - don't deref dentry after dput, don't free lvb on error (2922)
3233 - use the kms to determine writeback rpc length (2947)
3234 - increment oti_logcookies when osc is inactive (2948)
3235 - update client's i_blocks count via lvb messages (2543)
3236 - handle intent open/close of special files properly (1557)
3237 - mount MDS with errors=remount-ro, like obdfilter (2009)
3238 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3239 - don't use cancelling-locks' kms values (2947)
3240 - use highest lock extent for kms, not last one (2925)
3241 - don't dereference ERR_PTR() dentry in error handling path (3107)
3242 - fix thread race in portals_debug_dumplog() (3122)
3243 - create lprocfs device entries at setup instead of at attach (1519)
3244 - common AST error handler, don't evict client on completion race (3145)
3245 - zero nameidata in detach_mnt in 2.6 (3118)
3246 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3247 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3248 - initialize RPC timeout timer earlier for 2.6 (3219)
3249 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3250 - print nal/nid for unknown nid (3258)
3251 - additional checks for oscc recovery before doing precreate (3284)
3252 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3253 - don't crash in mdc_close for bad permissions on open (3285)
3254 - zero i_rdev for non-device files (3147)
3255 - clear page->private before handing to FS, better assertion (3119)
3256 - tune the read pipeline (3236)
3257 - fix incorrect decref of invalidated dentry (2350)
3258 - provide read-ahead stats and refine rpc in flight stats (3328)
3259 - don't hold journal transaction open across create RPC (3313)
3260 - update atime on MDS at close time (3265)
3261 - close LDAP connection when recovering to avoid server load (3315)
3262 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3263 - don't leak open file on MDS after open resend (3325)
3264 - serialize filter_precreate and filter_destroy_precreated (3329)
3265 - loop device shouldn't call sync_dev() for nul device (3092)
3266 - clear page cache after eviction (2766)
3267 - resynchronize MDS->OST in background (2824)
3268 - refuse to mount the same filesystem twice on same mountpoint (3394)
3269 - allow llmount to create routes for mounting behind routers (3320)
3270 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3271 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3272 - fix rare problem with rename on htree directories (3417)
3274 - allow default OST striping configuration per directory (1414)
3275 - fix compilation for qswnal for 2.6 kernels (3125)
3276 - increase maximum number of MDS request buffers for large systems
3277 - change liblustreapi to be useful for external progs like lfsck (3098)
3278 - increase local configuration timeout for slow disks (3353)
3279 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3281 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3284 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3285 - don't overwrite extent policy data in reply if lock was blocked (2901)
3286 - drop filter export grants atomically with removal from device (2663)
3287 - del obd_self_export from work_list in class_disconnect_exports (2908)
3288 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3289 - swab reply message in mdc_close, other PPC fixes (2464)
3290 - fix destroying of named logs (2325)
3291 - overwrite old logs when running lconf --write_conf (2264)
3292 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3293 - fix race in target_handle_connect (2898)
3294 - mds_reint_create() should take same inode create lock (2926)
3295 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3296 - don't close files for self_export to avoid uninitialized obd (2936)
3297 - allow MDS with the same name as client node (2939)
3298 - hold dentry reference for closed log files for unlink (2325)
3299 - reserve space for all logs during transactions (2059)
3300 - don't evict page beyond end of stripe extent (2925)
3301 - don't oops on a deleted current working directory (2399)
3302 - handle hard links to targets without a parent properly (2517)
3303 - don't dereference NULL lock when racing during eviction (2867)
3304 - don't grow lock extents when lots of conflicting locks (2919)
3306 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3309 - account for cache space usage on clients to avoid data loss (974)
3310 - lfsck support in lustre kernel code (2349)
3311 - reduce journal credits needed for BRW writes (2370)
3312 - orphan handling to avoid losing space on client/server crashes
3313 - ptlrpcd can be blocked, stopping ALL progress (2477)
3314 - use lock value blocks to assist in proper KMS, faster stat (1021)
3315 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3316 - recovery for initial connections (2355)
3317 - fixes for mds_cleanup_orphans (1934)
3318 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3319 - block all file creations until orphan recovery completes (1901)
3320 - client remove rq_connection from request struct (2423)
3321 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3322 - recovery timer race (2670)
3323 - mdc_close recovey bug (2532)
3324 - ptlrpc cleanup bug (2710)
3325 - mds timeout on local locks (2588)
3326 - namespace lock held during RPCs (2431)
3327 - handle interrupted sync write properly (2503)
3328 - don't try to handle a message that hasn't been replied to (2699)
3329 - client assert failure during cleanup after abort recovery (2701)
3330 - leak mdc device after failed mount (2712)
3331 - ptlrpc_check_set allows timedout requests to complete (2714)
3332 - wait for inflight reqs when ptlrpcd finishes (2710)
3333 - make sure unregistered services are removed from the srv_list
3334 - reset bulk XID's when resending them (caught by 1138 test)
3335 - unregister_bulk after timeout
3336 - fix lconf error (2694)
3337 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3338 - readahead locks pages, leaves pending causing memory pressure (2673)
3339 - increase OST request buffers to 4096 on large machines (2729)
3340 - fix up permission of existing directories in simple_mkdir (2661)
3341 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3342 - don't assign transno to errored transactions (2742)
3343 - don't delete objects on OST if given a bogus objid from MDS (2751)
3344 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3345 - if rq_no_resend, then timeout request after recovery (2432)
3346 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3347 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3348 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3349 - MDS recovery shouldn't delete objects that it has given out (2730)
3350 - if enqueue arrives after completion, don't clobber LVB (2819)
3351 - don't unlock pages twice when trigger_group_io returns error (2814)
3352 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3353 - don't write pages to disk if there was an error (1450)
3354 - don't ping imports that have recovery disabled (2676)
3355 - take buffered bytes into account when balancing socknal conn (2817)
3356 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3357 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3358 - remove little-endian swabbing of llog records (1987)
3359 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3360 - retry reposting request buffers if they fail (1191)
3361 - grow extent at grant time to avoid granting a revoked lock (2809)
3362 - lock revoke doesn't evict page if covered by a second lock (2765)
3363 - disable VM readahead to avoid reading outside lock extents (2805)
3365 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3366 - updated kernel patches for hp-2.4.20 kernel (2681)
3368 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3371 - fix truncated write corruption (2366)
3372 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3374 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3375 - ensure len(uuid) < 37 in lmc (1171)
3376 - fix ia64 OOPS in llog_test (2255)
3377 - zero end of page at obdfilter for partial page writes (2648)
3378 - don't leave stale dentries around after renames (bug 2428)
3379 - fix timeouts when evicting a client with a single lock held (2642)
3380 - set deadline for the initial HELLO message to drain (2634)
3381 - print out dotted-quad IP addresses in the socknal (2302)
3383 - additional debugging for MDS client eviction problem (2443)
3384 - fix mkfsoptions support for osts (2603, 2604)
3386 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3389 - add series for the vanilla 2.6.0 kernel
3390 - add series for the vanilla 2.4.24 kernel
3391 - add series for a cray x86/64 UL kernel drop
3392 - fix xattr patches for the vanilla 2.4.19 series
3394 - generate true UUIDs in lmc (1171)
3395 - have portals stack dumping break in UML (2466)
3396 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3397 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3398 - wake callback waiting threads on client eviction (2460)
3399 - Add --ptldebug and --subsystem to lmc (1719)
3400 - update assertion to allow safe interrupt allocation
3401 - set rq_no_resend for cancel requests (2432)
3402 - recalculate ptlrpcd timeout after resend (2494)
3403 - call vfs_rmdir when removing pending directories (2368)
3404 - fix renaming a file to itself (2429)
3405 - lmc creates a default one-stripe lov (2454)
3406 - expand procfs space to handle large clusters (2326)
3407 - increase UML stack to avoid overflow
3408 - update lconf's list of debug and subsystem masks
3409 - fix lfs find --obd (2510)
3410 - /proc tunable for disabling filter read caching (2591)
3411 - stop rpm packages from altering slapd.conf (2301)
3412 - disable nagle in the socknal under 0conf (2578)
3413 - choose mds inode size based on stripe count (2572)
3414 - fix kernel-source rpm problems (2516)
3416 - add --disable-doc to avoid pdf generation (2421)
3417 - update documentation, tests, type-os, comments
3418 - avoid format warnings on ia64
3419 - remove the TOE NAL
3420 - tiny code cleanups by removing unused fields
3422 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3425 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3426 - remove the most blatant lies from BUILDING file (2371)
3427 - change default debug level to reasonable production setting
3428 - reduce client side cache size to reduce cache flush time
3429 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3430 - make TCP zerocopy and pinger support enabled by default (2476)
3431 - sync writes completed after process exits caused crashes (2319)
3432 - maintain correct mount count on the MDS (2356)
3433 - backout 1557, because 2316 wasn't really fixed
3434 - better file I/O statistics gathering in /proc
3435 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3436 - another kernel patch to fix zero-copy TCP function export
3437 - don't take duplicate lock when processing re-sent getattr (2420)
3438 - lctl uses obd_self_export instead of creating new conn (2353)
3439 - MDS/OST recovery case which requires object creation asserted (2425)
3440 - move lfs from /usr/sbin to /usr/bin in packages
3441 - fix race between mds_client_add and mds_client_free (2417)
3442 - use kmalloc instead of slabs in portals (2430)
3443 - don't create duplicate records when a failover MDS is present (2442)
3444 - remove unnecessary mount age check (2332)
3445 - don't remove directory inodes from locks prematurely (2451)
3446 - don't break if MDS service name is the same as hostname (2103)
3447 - fix races in client write RPC generation when cache full (2482)
3449 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3452 - remove now-unused request->rq_obd (278)
3453 - if an allocation fails, print out how much memory we've used (1933)
3454 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3455 - add the "configurable stack size" patch to most series files (1256)
3456 - ability to write large log records, for 100+ OST configs (2306)
3457 - fix NULL deref when filter_prep fails (2314)
3458 - fix operator precedence error in filter_sync
3459 - dynamic allocation of socknal TX descriptors (2315)
3460 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3461 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3462 - ensure that CWARN messages in Portals always get to the syslog
3463 - __init/__exit are not for prototype decls (ldlm_init/exit)
3464 - x86-64 compile warning fixes
3465 - fix gateway LMC keyword conflict (2318)
3466 - fix MDS lock inversions in getattr/reint paths (1844)
3467 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3468 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3469 - DLM race condition prevented some lock evictions (2328)
3470 - ENOMEM detection and retry on socknal sends (2230)
3471 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3472 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3473 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3474 - better error messages when a client is rejected during recovery (1505)
3475 - avoid cancelling locks which were never granted, after failure (2330)
3476 - fix i_sem/journal inversion in mds_client_add (2333)
3477 - fix truncate/getattr lock cycle deadlock (2334)
3478 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3479 - fix two rare exit paths which could leak an l_lock() ref (2321)
3480 - fencepost error in MDS/OST orphan recovery (2226)
3481 - make log record alignment 8 bytes (1988)
3482 - lstripe now fails when requested offset > ost_count (2237)
3483 - ensure that all kernel series have a complete list.h (1607)
3484 - fix crashes in special-file operations (2316)
3485 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3487 - allow configurable automake binary, for testing new versions
3488 - small update to the lfs documentation
3490 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3492 * fix negative export reference count in fsfilt_sync (2312)
3494 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3495 * release candidate 0.9.1
3497 - orphans are moved into the PENDING directory for possible recovery
3498 - replayed opens now open by fid for orphan/rename safety (1042)
3499 - last close of an orphan inode generates a transno (683)
3500 - chdir() and mount() now pin the directory entry (1020)
3501 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3502 - discard very old requests without processing them (1502)
3503 - remove some common, well-understood CERRORs (1505)
3504 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3505 - clear "grant" flags in OST replies until OST grant code lands (1644)
3506 - fix read performance by not clobbering i_blksize on client (1598)
3507 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3508 - make LRU size a /proc tunable, clears locks when reduced (707)
3509 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3510 - server locks take references on exports now (1558)
3511 - build fixes for 2.4.20-rh trees (1663)
3512 - return an error from lov_create if all OSCs are inactive (1751)
3513 - fix import levels when a reconnect happens without a timeout (1597)
3514 - exit early from mds_open if we get a lookup error (1749)
3515 - partial page read at EOF wouldn't wait for disk before sending (1642)
3516 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3517 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3518 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3519 - write/truncate lock inversion (1639)
3520 - Don't auto-load obdclass, portals modules during cleanup (1495)
3521 - fix timestamps from jumping to "now" (1763)
3522 - extra journal assertions (1648)
3523 - add an extra multiunlink test (1771)
3524 - fix read_record/write_record API (1776)
3525 - fix leak of offset_extent, possible incorrect i_size later (1772)
3526 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3527 - Debugging for the kqswnal_get_idle_tx problems (1820)
3528 - Allow recovery to be attempted multiple times (1536)
3529 - Write out MDS last_rcvd file after it is first created (1600)
3530 - Fix tx_descriptor leak in failed transmit situations (1827)
3531 - ext3 journaling fixes for assertion failure after IO error (1871)
3532 - class_export_put() on freed export after completion AST error (1896)
3533 - Fix revalidate looping in VFS (1322)
3534 - Don't access a freed export during MDS_REINT timeout (1521)
3535 - Add open-unlink recovery support on the MDS (1673,1764)
3536 - Return an error if no MDS data was read from last_rcvd (1946)
3537 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3538 - Don't setup a disk device that doesn't match exported UUID (317)
3539 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3540 - avoid committing NULL handle in force close
3541 - local.sh is now a one-stripe LOV configuration
3542 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3543 - fix ext3 htree duplicate directory entry corruption (1516)
3544 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3545 - update to new LOV EA format (2097)
3546 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3548 2003-06-15 Phil Schwan <phil@clusterfs.com>
3551 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3552 - per-import/export recovery handling (958, 931, 959)
3553 - multiple last-rcvd slots, for serving multiple FSes (949)
3554 - connections are again shared between multiple imp/exports (963, 964)
3555 - "umount -f" would hang if any requests needed to be sent (393, 978)
3556 - avoid pinning large req buffer by copying for queued messages (989)
3557 - add "uuid" to "lctl device" command to help upcalls (991)
3558 - "open" RPCs with transnos would confuse recovery counters (1037)
3559 - do proper endian conversion of all wire messages (288, 340, 891)
3560 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3561 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3562 - fix ptlrpc_connection leak in target_handle_connect (1174)
3563 - fix import refcounting bug in OST and MDS cleanup (1134)
3564 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3565 - fix very unlikely obd_types race condition (501)
3566 - remove osc_open hack for echo_client (1187)
3567 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3568 - a failure in read_inode2 leads to deadlock (1139)
3569 - cancel ack-locks as soon as transaction is committed (1072)
3570 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3571 - make sure to commitrw after any preprw to avoid deadlock (1162)
3572 - failing to execute a file in a lustre FS would lock inode (1203)
3573 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3574 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3575 - obd_uuid2tgt would walk off the end of the list (1255)
3576 - on IA64 the osc would give portals incorrect bulk size (1258)
3577 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3578 - fix lock inversion caused by new llite matching code (1282)
3579 - limit the number of dirty pages on a client to 10MB (1286)
3580 - timed out locks were not being corrected cancelled (1289)
3581 - fix O_DIRECT above 4GB on IA-32 (1292)
3582 * major user-visible changes
3583 - fail out/fail over policy now controlled by the upcall (993)
3585 - add OBD_PING to check server availability and failure (954)
3586 - lustre messages are now sent in sending host order (288, 340, 891)
3587 - add eadatalen to MDS getattr reply (340)
3588 - OST read replies may contain second buffer, with per-page status (593)
3590 2003-03-11 Phil Schwan <phil@clusterfs.com>
3593 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3594 - failed open()s could cause deadlock; fixed (867, 869)
3595 - stop cancelling OST locks when files are closed (481)
3596 - overlapping XID spaces caused network corruption (851, 853)
3597 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3598 - setattr_raw added, to do single-RPC, server-side setattrs
3599 - lmc/lconf syntax change for OST UUIDs
3600 - fix crashy race condition between ptlrpc_free_req and osc_close
3601 - don't use request in mdc_enqueue if we hit a timeout (889)
3602 - don't set the inode i_size for regular files from the MDS (896)
3603 - handle out of order completion AST (842)
3604 - don't LBUG if a lock request times out after receiving AST (913)
3605 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3606 - if a bad lock AST arrives, send an error instead of dropping entirely
3607 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3608 - fix leak in bulk IO when only partially completed (899, 900, 926)
3609 - fix O_DIRECT for ia64 (55)
3610 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3611 - C-z after timeout could hang a process forever; fixed (977)
3613 - client-side I/O cache (678, 924, 929, 941, 970)
3615 - READPAGE and SETATTRs which don't take server-side locks get
3618 2003-02-11 Phil Schwan <phil@clusterfs.com>
3621 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3622 - Fix fsfilt_extN_readpage() to read a full page of directory
3623 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3624 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3625 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3626 - Most intent operations were removed, in favour of a new RPC mode
3627 that does a single RPC to the server and bypasses most of the VFS
3628 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3629 - Aggressively cancel local locks on DLM servers
3630 - mds_reint_unlink sends EA to the client if it's the last nlink.
3631 client uses that EA to unlink OST objects.
3632 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3633 - recursive symlinks were fixed (439)
3634 - fixed NULL deref in DEBUG_REQ
3635 - filter_update_lastobjid no longer calls sync, which annoyed extN
3636 - fixed multi-client small-writes to a single file problem (445)
3637 - fixed mtime updates during file writes (607)
3638 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3639 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3640 - fixed Linux OST size reporting problem (444, 656)
3641 - OST now updates object mtime with writes or setattr (607, 619)
3642 - client verifies file size before zeroing page past EOF (445)
3643 - OST now writes last allocated objid to disk with allocation (108)
3644 - LOV on echo now works (409)
3646 - mds_reint_unlink sends a new buffer, with the EA included. this
3647 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3648 set if a regular file was being unlinked, and it was the last link
3649 - use PtlGet from the target for bulk writes (315)
3650 - OST now updates object mtime with writes or setattr (607, 619)
3651 - LDLM now has a grant-time callback to revalidate locked items, if
3653 - Many MDS operations were reorganized to combat race conditions
3655 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3656 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3658 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3661 - Fully reactivate OST imports after reconnection (512, others)
3662 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3663 - More graceful error handling for truncating on dead OST (515)
3664 - Don't error out unless we're actually accessing dead stripes (474)
3665 - Fix garbage sizes when stripes are missing (410)
3666 - LRU counters were broken, causing constant lock purge (433, 432)
3667 - garbage on read from stripes with failed OSTs (441)
3668 - mark OSCs as active before reconnecting during recovery (438)
3669 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3670 - lfind did not preserve OST order in output (443)
3671 - symlinks cause hung clients, incorrect data (439)
3672 - stop dereferencing request after dropping refcount (457)
3673 - don't LASSERT(spin_is_locked) on non-SMP (455)
3674 - fixes for many rename() bugs
3675 - fstat didn't correctly synchronize attributes (399)
3676 - server must handle lock cancellation during blocking AST prep (487)
3677 - bulk descriptors were free()d too soon (511)
3678 - fix paths in lconf, which would load incorrect modules (451, 507)
3679 - fix confusing lconf 'host not found' error message (386)
3680 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3681 - fix race condition in mdc_blocking_ast() for inode access (526)
3682 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3683 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3684 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3685 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3686 - fix lstripe in several ways (kernel side) (527)
3687 - fix request leak in ldlm_cli_enqueue (262)
3688 - incorrect OSC was marked inactive after OST failure
3689 - call mds_fs_cleanup before unmounting filesystem (524)
3690 - fix races between taking ns_lock and ldlm_lock_change_resource
3691 - fix races updating LOV export open file list
3692 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3693 - fix recovery NULL deref in ldlm_cli_cancel_unused
3694 - fix some DLM races by using new hash table for lock handles (419)
3695 - permit the client to specify desired inodes, at replay
3696 - duplicate requests when we queue them for replay reintegration
3697 - fix last_rcvd offset calculation
3698 - sync after each recovered transaction, so we always make progress
3699 - never, not always, ERESTART requests without transnos
3700 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3702 - skip replay if the MDS says that the client is already connected
3703 - don't check for a recovery-enabled export to match lctl's UUID
3704 - don't INC_USE_COUNT for phantom exports
3705 - don't crash when cleaning up phantom exports (567)
3706 - don't double-finish or set replay data for errored mdc_open requests
3707 - abort requests when they time out, so we don't get old replies
3708 - send/receive replies for AST messages again
3709 - if the client says that it doesn't have the lock, cancel it on the
3711 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3712 mark it as destroyed, it will all work out when decref is called
3713 - fix module use counts (22, 581)
3715 - ASTs now expect a reply (server cancels lock on error reply)
3717 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3720 - fix many simultaneous client startup (392)
3721 - fix dentry->d_it clobbering
3722 - credentials weren't being shipped for readdir/getattr operations
3723 - remove invalid assertions triggered during some concurrent MD
3725 - proper Lustre versions added (336, 389)
3726 - fix memory leak for create error case (398)
3727 - fix LOV locking bug that would get cli/srv out of sync
3728 - fix echo client over LOV (409)
3729 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3730 - fix double-O_EXCL intent crash (424)
3731 - avoid sending multiple lock CANCELs (352)
3733 - MDS can do multi-client recovery (modulo bugs in new code)
3735 - many updates, edits, cleanups
3737 2002-11-18 Phil Schwan <phil@clusterfs.com>
3740 - fix null d_it dereference (346)
3741 - fix full OST/dbench hang (333)
3742 - fix permission problem with file removal (286)
3743 - fix removal of OSCs from LOV when they fail
3744 - fix NULL deref during bulk timeout (214)
3745 - fix problems related to multiple filesystems on one MDS (241)
3746 - fixed serious subtle metadata locking bugs
3747 - free locks on clients when inodes are removed due to memory
3749 - fix inode pointer in lock data (285)
3750 - partial support for multiple MDS on a single host (241)
3751 - data locks weren't cancelled at clear_inode time (290, 311)
3752 - intent locks could lead to unbounded lock growth (205)
3753 - added a maximum lock count, an LRU list, and a flusher
3754 - fix multiple rename (365)
3755 - properly abstracted the echo client
3756 - OSC locked 1 byte too many; fixed
3757 - rewrote brw callback code:
3758 - fixed recovery bugs related to LOVs (306)
3759 - fixed too-many-pages-in-one-write crash (191)
3760 - fixed (again) crash in sync_io_timeout (214)
3761 - probably fixed callback-related race (385)
3763 - Add capability to MDS protocol
3764 - LDLM cancellations and callbacks on different portals
3766 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3769 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3770 - defer open object destruction to close time (601981, 138)
3771 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3772 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3773 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3774 - don't call obd_setattr in ll_file_release for destroyed objects
3775 * protocol change to lustre_msg: move |version| and add |flags|
3776 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3777 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3778 * added replay of create, unlink, link and rename operations during
3779 MDS failover; recovery should be much more robust now
3780 * remove failed OSCs from LOVs (only lov_create uses this so far)
3781 * the lustre-HOWTO was brought (more) up to date (582544)
3783 2002-10-23 Phil Schwan <phil@clusterfs.com>
3786 - in-use dentries weren't being reused properly (617851)
3787 - prevent multiple LDLM setup (599178)
3788 - fix LOV size calculations for truncate (617853)
3789 - fix client handling of MDS intent errors (POSIX)
3790 - fix permission bug in lovstripe.c test (624321)
3791 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3792 - truncate past end of file could corrupt data
3793 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3794 - a race in recovery could return ETIMEDOUT to apps (623947)
3795 - building outside the source directory was fixed
3796 * the lustre-HOWTO was brought (more) up to date (582544)
3797 * major progress was made on recovery functionality
3799 2002-10-10 Phil Schwan <phil@clusterfs.com>
3802 - recovery deadlock fix
3803 - rm -rf causes LBUG fix (617817)
3804 - file open by multiple tasks fix (618962)
3805 - directory permissions bugs (602707 and 620007)
3806 - journal_stop fixed with locking (611313)
3807 - O_APPEND failures resolved (618273, perhaps 614459)
3808 - lconf PATH fix (619770)
3809 - IA64 build fix (621450)
3810 - RPC buffer sizes scale with amount of memory
3812 2002-10-01 Phil Schwan <phil@clusterfs.com>
3815 - locks would be cancelled without throwing away data pages,
3816 resulting in inconsistent data (605627)
3817 - inode attributes were not always being refreshed (605627, 612449)
3818 - lconf now continues to cleanup after lctl reports an error
3819 - MDS now enforces user permissions (602707)
3820 - lprocfs cleanup fixed, but not yet enabled (614157)
3821 - fixed infinite server hang, should a client not respond to an AST
3822 - avoid going into recovery if user calls readlink() with a buffer
3823 that's too small (613941)
3824 - AST RPCs no longer require replies (614867) -- this may be changed
3825 - don't crash server if client sends an IOV that's too big (611336)
3826 - fixed lock conversion deadlock (611892)
3827 - fixed the following of symlinks (614622)
3828 * recovery: the server can remove locks from a client that dies, other
3829 clients can make progress
3830 * more extN patch fixes
3831 * compile-time configurable ptlrpc buffer allocations
3833 - collaborative read cache document
3834 - Lustre Lite Performance CDR document-in-progress
3836 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3839 - fix typo in patch-2.4.18
3841 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3844 - clear ptlrpc request each time in handle_incoming_request()
3845 - unlink of files now destroys the object on the OST
3847 2002-09-19 Peter Braam <braam@clusterfs.com>
3849 * add hard link support
3850 * change obdfile creation method
3851 * kernel patch changed
3853 2002-09-19 Peter Braam <braam@clusterfs.com>
3856 - stack overflow bug in extN fixed
3858 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3860 * documentation updates
3861 - add man pages for config tools
3862 - update tests/README to describe testing with new config tools
3863 - finish metadata API descriptions
3864 * bug fixes and cleanups
3865 - statfs workaround for 16TB limit
3866 - LOV stripe allocation improved, can stripe on subset of OSTs
3867 - LOV file size/IO offset was wrong for files > 4GB in size
3868 - object EA data was being dropped, caused files to be unreadable
3869 - memory overflow with non-LOV OST caused memory corruption
3870 - fixed regression tests to work with new config tools, obdfilter
3871 - fixed bug when directory size became larger than 1 block
3872 - fixed bug (for single client case) when PWD was deleted
3873 - invalidate local directory pages when doing intent-based ops
3874 - avoid LDLM oops when lock callback contained bad data
3876 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3878 * documentation updates
3879 * bug fixes and cleanups
3880 - configuration tools
3883 - 64-bit compile warnings
3884 - 64-bit internal statfs data
3886 * test_brw on persistent OST devices
3888 * lprocfs (disabled)
3890 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3892 * documentation updates
3893 * bug fixes and cleanups
3894 * configuration tools
3896 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3899 * many small fixes to 0_5_4
3900 * io/network handling
3901 * thinkos in MDS operations
3903 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3906 * crucial basic fixes to 0.5.3
3907 * IOR, Iozone work over Elan
3910 2002-08-07 Phil Schwan <phil@clusterfs.com>
3911 * version 0_5_3, our first alpha
3912 * we use the new Portals iovs
3913 * documentation updates
3914 * bug fixes and cleanups
3915 * small changes in the DLM wire protocol
3917 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3918 * version 0_5_1 with some initial stability,
3919 * locking on MD and file I/O.
3920 * documentation updates
3921 * several bug fixes since 0.5.0
3922 * small changes in wire protocol
3924 2002-07-18 Phil Schwan <phil@clusterfs.com>
3926 * delivered as Lustre Light Alpha
3927 * fixed a crash after handling invalid MDS requests
3928 * fixed directory pages for architectures with non-4k pages sizes
3930 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3931 * release version v0_4_4
3932 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3934 * Fixes a number of interruption problems with OST operations.
3935 * Update documentation for portals header changes
3936 * Move all wire protocol structs/defines to lustre_idl.h
3937 * Fixes symlink length bug.
3938 * Add tcpdump to repository.
3940 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3941 * release version v0_4_3
3942 * Fixes statfs for inodes on extN.
3943 * Fixes bug in runtests which would delete /etc/hosts.
3944 * Use 64-bit object IDs wherever possible (not into VFS though)
3945 Remove ost_get_info, which is unused by lustre, and out of date.
3947 2002-07-03 Peter Braam <braam@clusterfs.com>
3948 * release version v0_4_2 Fixes a lookup error (type not passed)
3949 * move forward to head of Portals
3950 * move forward to latest Lustre kernel
3952 2002-06-25 Peter Braam <braam@clusterfs.com>
3953 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob