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.
206 Description: Client eviction due to failover config
207 Details : after a connection loss, the lustre client should attempt to
208 reconnect to the last active server first before trying the
209 other potential connections.
213 Description: when mds and osts use different quota unit(32bit and 64bit),
214 quota will be released repeatly.
215 Details : void sending multiple quota reqs to mds, which will keep the status
217 --------------------------------------------------------------------------------
219 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
221 * Support for kernels:
222 2.6.5-7.283 (SLES 9),
223 2.6.9-55.EL (RHEL 4),
224 2.6.16.46-0.14 (SLES 10),
225 2.6.18.8 vanilla (kernel.org)
226 * Client support for unpatched kernels:
227 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
228 2.6.16 - 2.6.22 vanilla (kernel.org)
229 * Due to recently discovered recovery problems, we do not recommend
230 using patchless RHEL 4 clients with this or any earlier release.
231 * Recommended e2fsprogs version: 1.39.cfs8
232 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
233 * Starting with this release, the ldiskfs backing filesystem required
234 by Lustre is now in its own package, lustre-ldiskfs. This package
235 should be installed. It is versioned separately from Lustre and
236 may be released separately in future.
240 Description: block reactivating mgc import until all deactivates complete
241 Details : Fix race when failing back MDT/MGS to itself (testing)
243 Severity : enhancement
245 Description: add optional extra BUILD_VERSION info
246 Details : add a new environment variable (namely LUSTRE_VERS) which allows
247 to override the lustre version.
250 Frequency : 2.6.18 servers only
252 Description: ll_kern_mount() doesn't release the module reference
253 Details : The ldiskfs module reference count never drops down to 0
254 because ll_kern_mount() doesn't release the module reference.
259 Description: server LBUG when using old ost_num_threads parameter
260 Details : Accept the old ost_num_threads parameter but warn that it
261 is deprecated, and fix an off-by-one error that caused an LBUG.
266 Description: Transient SCSI error results in persistent IO issue
267 Details : iobuf->dr_error is not reinitialized to 0 between two
271 Frequency : sometimes when underlying device returns I/O errors
273 Description: OSTs not going read-only during write failures
274 Details : OSTs are not remounted read-only when the journal commit threads
275 get I/O errors because fsfilt_ext3 calls journal_start/stop()
276 instead of the ext3 wrappers.
280 Description: poor connect scaling with increasing client count
281 Details : Don't run filter_grant_sanity_check for more than 100 exports
282 to improve scaling for large numbers of clients.
285 Frequency : SLES10 only
287 Description: sanity-quota.sh quotacheck failed: rc = -22
288 Details : Quotas cannot be enabled on SLES10.
291 Frequency : liblustre clients only
293 Description: getdirentries does not give error when run on compute nodes
294 Details : getdirentries does not fail when the size specified as an argument
295 is too small to contain at least one entry
297 Severity : enhancement
299 Description: Add LNET router traceability for debug purposes
300 Details : If a checksum failure occurs with a router as part of the
301 IO path, the NID of the last router that forwarded the bulk data
302 is printed so it can be identified.
307 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
308 Details : Due to a race condition, liblustre clients were occasionally
311 Severity : enhancement
313 Description: lfs setstripe use optional parameters instead of postional
316 Severity : enhancement
318 Description: Nanosecond timestamp support for ldiskfs
319 Details : The on-disk ldiskfs filesystem has added support for nanosecond
320 resolution timestamps. There is not yet support for this at
321 the Lustre filesystem level.
324 Frequency : during server recovery
326 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
327 Details : request with rq_no_resend flag not awake l_wait_event if they get a
331 Frequency : nfs export on patchless client
333 Description: connectathon hang when test nfs export over patchless client
334 Details : Disconnected dentry cannot be found with lookup, so we do not need
335 to unhash it or make it invalid
338 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
339 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
340 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
345 Description: open req refcounting wrong on reconnect
346 Details : If reconnect happened between getting open reply from server and
347 call to mdc_set_replay_data in ll_file_open, we will schedule
348 replay for unreferenced request that we are about to free.
349 Subsequent close will crash in variety of ways.
350 Check that request is still eligible for replay in
351 mdc_set_replay_data().
356 Description: disable writes to filesystem when reading health_check file
357 Details : the default for reading the health_check proc file has changed
358 to NOT do a journal transaction and write to disk, because this
359 can cause reads of the /proc file to hang and block HA state
360 checking on a healthy but otherwise heavily loaded system. It
361 is possible to return to the previous behaviour during configure
362 with --enable-health-write.
364 Severity : enhancement
366 Description: 64-bit inode version
367 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
368 to the inode. This will be required for version-based recovery.
373 Description: MDS fails to start if a duplicate client export is detected
374 Details : in some rare cases it was possible for a client to connect to
375 an MDS multiple times. Upon recovery the MDS would detect this
376 and fail during startup. Handle this more gracefully.
378 Severity : enhancement
380 Description: Add -o localflock option to simulate old noflock
382 Details : This will achieve local-only flock/fcntl locks
388 Description: log_commit_thread vs filter_destroy race leads to crash
389 Details : Take import reference before releasing llog record semaphore
394 Description: Wrong request locking in request set processing
395 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
396 imp_delayed_list, in this place should be used imp_lock.
399 Frequency : when reconnection
401 Description: Grant Leak when osc reconnect to OST
402 Details : When osc reconnect ost, OST(filter) should check whether it
403 should grant more space to client by comparing fed_grant and
404 cl_avail_grant, and return the granted space to client instead
405 of "new granted" space, because client will call osc_init_grant
406 to update the client grant space info.
409 Frequency : when client reconnect to OST
411 Description: Grant Leak when osc do resend and replay bulk write
412 Details : When osc reconnect to OST, OST(filter)should clear grant info of
413 bulk write request, because the grant info while be sync between
414 OSC and OST when reconnect, and we should ignore the grant info
415 these of resend/replay write req.
420 Description: Grant space more than avaiable left space sometimes.
421 Details : When then OST is about to be full, if two bulk writing from
422 different clients came to OST. Accord the avaliable space of the
423 OST, the first req should be permitted, and the second one
424 should be denied by ENOSPC. But if the seconde arrived before
425 the first one is commited. The OST might wrongly permit second
426 writing, which will cause grant space > avaiable space.
429 Frequency : when client is evicted
431 Description: Grant might be wrongly erased when osc is evicted by OST
432 Details : when the import is evicted by server, it will fork another
433 thread ptlrpc_invalidate_import_thread to invalidate the
434 import, where the grant will be set to 0. While the original
435 thread will update the grant it got when connecting. So if
436 the former happened latter, the grant will be wrongly errased
437 because of this race.
442 Description: Checking Stale with correct fid
443 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
444 to check whether it is stale, because sometimes, we want the
445 enqueue happened anyway, and op_data.fid2 will not be initialized.
447 Severity : enhancement
449 Description: update patchless client
450 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
453 Frequency : only with 2.4 kernel
455 Description: random memory corruption
456 Details : size of struct ll_inode_info is to big for union inode.u and this
457 can be cause of random memory corruption.
462 Description: Memory leak in recovery
463 Details : Lov_mds_md was not free in an error handler in mds_create_object.
464 It should also check obd_fail before fsfilt_start, otherwise if
465 fsfilt_start return -EROFS,(failover mds during mds recovery).
466 then the req will return with repmsg->transno = 0 and rc = EROFS.
467 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
468 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
469 be freed no matter whether fsfilt_commit success or not.
472 Frequency : only with huge count clients
474 Description: Prevents from taking the superblock lock in llap_from_page for
476 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
477 for llap_from_page call in ll_removepage prevents from taking the
478 superblock lock for a soon died page.
483 Description: Not check open intent error before release open handle
484 Details : in some rare cases, the open intent error is not checked before
485 release open handle, which may cause
486 ASSERTION(open_req->rq_transno != 0), because it tries to release
487 the failed open handle.
492 Description: Set cat log bitmap only after create log success.
493 Details : in some rare cases, the cat log bitmap is set too early. and it
494 should be set only after create log success.
498 Description: Accessing a block bevice can re-enable I/O when Lustre is
499 tearing down a device.
500 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
505 Description: service threads may hog cpus when there are a lot of requests
507 Details : Insert cond_resched to give other threads a chance to use some of
513 Description: the cat log was not initialized in recovery
514 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
515 unlink log on mds will not be initialized until mds post
516 recovery. And also in mds post recovery, the unlink log will
517 initialization will be done asynchronausly, so there will be race
518 between add unlink log and unlink log initialization.
522 Description: brw_stats were being printed incorrectly
523 Details : brw_stats were being printed as log2 but all of them were not
524 recorded as log2. Also remove some code duplication arising from
525 filter_tally_{read,write}.
529 Frequency : rare, only in recovery.
530 Description: ASSERTION(req->rq_type != LI_POISON) failed
531 Details : imp_lock should be held while iterating over imp_sending_list for
532 prevent destroy request after get timeout in ptlrpc_queue_wait.
536 Description: replay-single.sh test 52 fails
537 Details : A lock's skiplist need to be cleanup when it being unlinked
538 from its resource list.
542 Description: Short directio read returns full requested size rather than
544 Details : Direct I/O operations should return actual amount of bytes
545 transferred rather than requested size.
547 Severity : enhancement
549 Description: metadata RPC reduction (e.g. for rm performance)
550 Details : decrease the amount of synchronous RPC between clients and servers
551 by canceling conflicing lock before the operation on the client side
552 and packing thier handles into the main operation RPC to server.
554 Severity : enhancement
556 Description: Async OSC create to avoid the blocking unnecessarily.
557 Details : If a OST has no remain object, system will block on the creating
558 when need to create a new object on this OST. Now, ways use
559 pre-created objects when available, instead of blocking on an
560 empty osc while others are not empty. If we must block, we block
561 for the shortest possible period of time.
563 --------------------------------------------------------------------------------
565 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
570 Frequency : on some architectures
572 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
573 Details : Uninitialized flags sometimes cause configuration commands to
577 Frequency : patchless clients only
579 Description: missing __iget() symbol export
580 Details : The __iget() symbol export is missing. To avoid the need for
581 this on patchless clients the deathrow inode reaper is turned
582 off, and we depend on the VM to clean up old inodes. This
583 dependency was during via the fix for bug 12181.
585 --------------------------------------------------------------------------------
587 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
589 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
590 INTEROPERATE with older versions automatically. In many cases a
591 special upgrade step is needed. Please read the
592 user documentation before upgrading any part of a live system.
593 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
594 version will not interoperate with 1.6 betas before beta5 (1.5.95).
595 * WARNING: Lustre configuration and startup changes are required with
596 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
602 Severity : enhancement
604 Description: MountConf
605 Details : Lustre configuration is now managed via mkfs and mount
606 commands instead of lmc and lconf. New obd types (MGS, MGC)
607 are added for dynamic configuration management. See
608 https://mail.clusterfs.com/wikis/lustre/MountConf for
611 Severity : enhancement
613 Description: dynamic OST addition
614 Details : OSTs can now be added to a live filesystem
616 Severity : enhancement
618 Description: startup order invariance
619 Details : MDTs and OSTs can be started in any order. Clients only
620 require the MDT to complete startup.
622 Severity : enhancement
624 Description: parallel, asynchronous orphan cleanup
625 Details : orphan cleanup is now performed in separate threads for each
626 OST, allowing parallel non-blocking operation.
628 Severity : enhancement
630 Description: optimized stripe assignment
631 Details : stripe assignments are now made based on ost space available,
632 ost previous usage, and OSS previous usage, in order to try
633 to optimize storage space and networking resources.
635 Severity : enhancement
637 Description: Permanently set tunables
638 Details : All writable /proc/fs/lustre tunables can now be permanently
639 set on a per-server basis, at mkfs time or on a live
642 Severity : enhancement
644 Description: Lustre message v2
645 Details : Add lustre message format v2.
647 Severity : enhancement
649 Description: client OST exclusion list
650 Details : Clients can be started with a list of OSTs that should be
651 declared "inactive" for known non-responsive OSTs.
655 Description: SPEC SFS validation failure on NFS v2 over lustre.
656 Details : Changes the blocksize for regular files to be 2x RPC size,
657 and not depend on stripe size.
659 Severity : enhancement
661 Description: Multiple MD RPCs in flight.
662 Details : Further unserialise some read-only MDS RPCs - learn about intents.
663 To avoid overly-overloading MDS, introduce a limit on number of
664 MDS RPCs in flight for a single client and add /proc controls
665 to adjust this limit.
667 Severity : enhancement
669 Description: client read/write statistics
670 Details : Add client read/write call usage stats for performance
671 analysis of user processes.
672 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
673 file access. extents_stats shows chunk size distribution.
674 extents_stats_per_process show chunk size distribution per
677 Severity : enhancement
679 Description: mds statistics
680 Details : Add detailed mds operations statistics in
681 /proc/fs/lustre/mds/*/stats.
685 Description: Failure of copying files with lustre special EAs.
686 Details : Client side always return success for setxattr call for lustre
687 special xattr (currently only "trusted.lov").
692 Description: Refcount LNET uuids
693 Details : The global LNET uuid list grew linearly with every startup;
694 refcount repeated list entries instead of always adding to
697 Severity : enhancement
699 Description: Dynamic service threads
700 Details : Within a small range, start extra service threads
701 automatically when the request queue builds up.
704 Frequency : mixed-endian client/server environments
706 Description: mixed-endian crashes
707 Details : The new msg_v2 system had some failures in mixed-endian
710 Severity : enhancement
712 Description: Easy OST removal
713 Details : OSTs can be permanently deactivated with e.g. 'lctl
714 conf_param lustre-OST0001.osc.active=0'
716 Severity : enhancement
718 Description: MGS proc entries
719 Details : Added basic proc entries for the MGS showing what filesystems
722 Severity : enhancement
724 Description: provide MGS failover
725 Details : Added config lock reacquisition after MGS server failover.
727 Severity : enhancement
729 Description: add Linux 2.4 support
730 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
734 Description: a large application tries to do I/O to the same resource and dies
736 Details : Check the req->rq_arrival time after the call to
737 ost_brw_lock_get(), but before we do anything about
738 processing it & sending the BULK transfer request. This
739 should help move old stale pending locks off the queue as
740 quickly as obd_timeout.
743 Frequency : when an incorrect nid is specified during startup
745 Description: ptlrpc connect to non-existant node causes kernel crash
746 Details : LNET can't be re-entered from an event callback, which
747 happened when we expire a message after the export has been
748 cleaned up. Instead, hand the zombie cleanup off to another
751 Severity : enhancement
753 Description: plain/inodebits lock performance improvement
754 Details : Grouping plain/inodebits in granted list by their request modes
755 and bits policy, thus improving the performance of search through
759 Frequency : only if OST filesystem is corrupted
761 Description: client incorrectly hits assertion in ptlrpc_replay_req()
762 Details : for a short time RPCs with bulk IO are in the replay list,
763 but replay of bulk IOs is unimplemented. If the OST filesystem
764 is corrupted due to disk cache incoherency and then replay is
765 started it is possible to trip an assertion. Avoid putting
766 committed RPCs into the replay list at all to avoid this issue.
769 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
771 Description: System hang on startup
772 Details : This bug allowed the liblustre (e.g. catamount) client to
773 return to the app before handling all startup RPCs. This
774 could leave the node unresponsive to lustre network traffic
775 and manifested as a server ptllnd timeout.
777 Severity : enhancement
779 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
780 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
781 boolean to control whether to print peer debug info when a
782 client's RPC times out.
785 Frequency : only for kernels with patches from Lustre below 1.4.3
787 Description: Remove old rdonly API
788 Details : Remove old rdonly API which unused from at least lustre 1.4.3
791 Frequency : only for devices with external journals
793 Description: Set external device read-only also
794 Details : During a commanded failover stop, we set the disk device
795 read-only while the server shuts down. We now also set any
796 external journal device read-only at the same time.
799 Frequency : when upgrading from 1.4 while trying to change parameters
801 Description: The wrong (new) MDC name was used when setting parameters for
802 upgraded MDT's. Also allows changing of OSC (and MDC)
803 parameters if --writeconf is specified at tunefs upgrade time.
806 Frequency : when setting specific ost indicies
808 Description: QOS code breaks on skipped indicies
809 Details : Add checks for missing OST indicies in the QOS code, so OSTs
810 created with --index need not be sequential.
812 Severity : enhancement
814 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
815 Details : The uninit_groups feature works in conjunction with the kernel
816 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
817 up the pass1 processing of e2fsck. This is a read-only feature
818 in ldiskfs2 only, so older kernels and current ldiskfs cannot
819 mount filesystems that have had this feature enabled.
821 Severity : enhancement
823 Description: Improve multi-block allocation algorithm to avoid fragmentation
824 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
825 allocation locality and avoid filesystem fragmentation.
827 ------------------------------------------------------------------------------
829 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
831 * Support for kernels:
832 2.4.21-47.0.1.EL (RHEL 3)
834 2.6.9-42.0.10.EL (RHEL 4)
835 2.6.12.6 vanilla (kernel.org)
836 2.6.16.27-0.9 (SLES 10)
837 * Recommended e2fsprogs version: 1.39.cfs5
839 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
843 Frequency : occasional, depends on client load and configuration
844 Bugzilla : 12181, 12203
845 Description: data loss for recently-modified files
847 Details : In some cases it is possible that recently written or created
848 files may not be written to disk in a timely manner (this should
849 normally be within 30s unless client IO load is very high).
850 The problem appears as zero-length files or files that are a
851 multiple of 1MB in size after a client crash or client eviction
852 that are missing data at the end of the file.
854 This problem is more likely to be hit on clients where files are
855 repeatedly created and unlinked in the same directory, clients
856 have a large amount of RAM, have many CPUs, the filesystem has
857 many OSTs, the clients are rebooted frequently, and/or the files
858 are not accessed by other nodes after being written.
860 The presence of the problem can be detected by looking at
861 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
862 smaller than the second (nr_unused) then dirty files will not
863 be flushed automatically to disk. "sync; sleep 10" should be
864 run several times on the node before unmounting it to update
865 Lustre (this is also safe to run on nodes without this problem).
867 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
868 that can cause this same problem, so customers using that kernel
869 also need to update the kernel in addition to Lustre. In order
870 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
872 It is normal that files written just before a client crash (less
873 than 30s) may not yet have been flushed to disk, even for local
877 Frequency : frequent on thin XT3 nodes
879 Description: UUID collision on thin XT3 Linux nodes
880 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
881 randomly, since we relied on an insufficiently-seeded PRNG.
886 Description: OSS hangs after "All ost request buffers busy"
887 Details : A deadlock between quota and journal operations caused OSS
888 hangs after printing "All ost request buffers busy."
891 Frequency : always on liblustre builds
893 Description: Cleanup compiler warnings on liblustre
896 Frequency : always on liblustre builds on XT3
898 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
902 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
903 > 32000 subdirectories
904 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
905 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
906 subdirectory count crosses 32000. This will aid e2fsck to
907 correctly handle more than 32000 subdirectories.
910 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
912 Description: System hang on startup
913 Details : This bug allowed the liblustre (e.g. catamount) client to
914 return to the app before handling all startup RPCs. This
915 could leave the node unresponsive to lustre network traffic
916 and manifested as a server ptllnd timeout.
918 Severity : enhancement
920 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
921 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
922 boolean to control whether to print peer debug info when a
923 client's RPC times out.
928 Description: make O_SYNC working on 2.6 kernels
929 Details : 2.6 kernels use different method for mark pages for write,
930 so need add a code to lustre for O_SYNC work.
935 Description: Failure to close file and release space on NFS
936 Details : Put inode details into lock acquired in ll_intent_file_open.
937 Use mdc_intent_lock in ll_intent_open to properly
938 detect all kind of errors unhandled by mdc_enqueue.
943 Description: proc file read during shutdown sometimes raced obd removal,
945 Details : Add lock to prevent obd access after proc file removal.
948 Frequency : Only for files larger than 4GB on 32-bit clients.
950 Description: improperly doing page alignment of locks
951 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
952 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
957 Description: RPCs being resent when they shouldn't be
958 Details : Some RPCs that should not be resent are being resent. This
959 can cause inconsistencies in the RPC state machine. Do not
960 resend such requests.
963 Frequency : rare, only with NFS export
965 Description: Crash on NFS re-export node
966 Details : under very unusual load conditions an assertion is hit in
967 ll_intent_file_open()
970 Frequency : only if OST filesystem is corrupted
972 Description: client incorrectly hits assertion in ptlrpc_replay_req()
973 Details : for a short time RPCs with bulk IO are in the replay list,
974 but replay of bulk IOs is unimplemented. If the OST filesystem
975 is corrupted due to disk cache incoherency and then replay is
976 started it is possible to trip an assertion. Avoid putting
977 committed RPCs into the replay list at all to avoid this issue.
982 Description: large O_DIRECT requests fail under memory pressure/fragmentation
983 Details : Large single O_DIRECT read and write calls can fail to allocate
984 a sufficiently large buffer to process the request. In case of
985 allocation failure the allocation is retried with a smaller
986 buffer and broken into smaller requests.
988 Severity : enhancement
990 Description: Add -o localflock option to simulate old noflock behaviour.
991 Details : This will achieve local-only flock/fcntl locks coherentness.
996 Description: versioning check is incomplete
997 Details : Checking the version difference of client vs. server, report
998 error if the gap is too big.
1003 Description: add support PG_writeback bit
1004 Details : add support for PG_writeback bit for Lustre, for more carefull
1005 work with page cache in 2.6 kernel. This also fix some deadlocks
1006 and remove hack for work O_SYNC with 2.6 kernel.
1008 Severity : enhancement
1010 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1011 Details : The uninit_groups feature works in conjunction with the kernel
1012 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1013 up the pass1 processing of e2fsck. This is a read-only feature
1014 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1015 mount filesystems that have had this feature enabled.
1017 Severity : enhancement
1019 Description: Improve multi-block allocation algorithm to avoid fragmentation
1020 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1021 allocation locality and avoid filesystem fragmentation.
1023 ------------------------------------------------------------------------------
1025 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1027 * Support for kernels:
1028 2.6.9-42.0.3.EL (RHEL 4)
1029 2.6.5-7.276 (SLES 9)
1030 2.4.21-47.0.1.EL (RHEL 3)
1031 2.6.12.6 vanilla (kernel.org)
1032 2.6.16.21-0.8 (SLES10)
1033 * Recommended e2fsprogs version: 1.39.cfs2-0
1035 * The backwards-compatible /proc/sys/portals symlink has been removed
1036 in this release. Before upgrading, please ensure that you change
1037 any configuration scripts or /etc/sysctl.conf files that access
1038 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1039 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1040 in advance of the upgrade on any system running Lustre 1.4.6 or
1041 newer, since /proc/sys/lnet was added in that version.
1042 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1046 Frequency : only when quota is used
1048 Description: avoid scanning export list for quota master
1049 Details : Change the algorithms to avoid scanning export list in order
1050 to improve the efficiency.
1053 Frequency : MDS failover only, very rarely
1055 Description: "went back in time" messages on mds failover
1056 Details : The greatest transno may be lost when the current operation
1057 finishes with an error (transno==0) and the client's last_rcvd
1058 record is over-written. Save the greatest transno in the
1059 mds_last_transno for this case.
1062 Frequency : always for specific kernels and striping counts
1064 Description: client may get "Matching packet too big" without ACL support
1065 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1066 when trying to access files in certain configurations. The
1067 clients should in fact be denied when mounting because they do
1068 not understand ACLs.
1071 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1073 Description: many clients connecting with IO in progress causes connect timeouts
1074 Details : Avoid synchronous journal commits to avoid delays caused by many
1075 clients connecting/disconnecting when bulk IO is in progress.
1076 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1077 OST_IO_PORTAL to avoid delays behind potentially many pending
1081 Frequency : occasionally with multiple writers to a single file
1083 Description: shared writes to file may result in wrong size reported by stat()
1084 Details : Allow growing of kms when extent lock is cancelled
1087 Frequency : always with random mmap IO to multi-striped file
1089 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1090 Details : Only if the hole is at the end of OST object so that kms is too
1091 small. Fix is to increase kms accordingly in ll_nopage.
1094 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1096 Description: writes to a missing object would leak memory on the OST
1097 Details : If there is an inconsistency between the MDS and OST filesystems,
1098 such that the MDS references an object that doesn't exist, writes
1099 to that object will leak memory due to incorrect cleanup in the
1100 error handling path, eventually running out of memory on the OST.
1105 Description: Creating too long symlink causes lustre errors
1106 Details : Check symlink and name lengths before sending requests to MDS.
1109 Frequency : only if flock is enabled (not on by default)
1111 Description: posix locks not released on fd closure on 2.6.9+
1112 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1113 kernels, this caused such locks not to be released on fd close and
1114 then assertions on fs unmount about still used locks.
1117 Frequency : MDS failover only, very rarely
1119 Description: clients may get ASSERTION(granted_lock != NULL)
1120 Details : When request was taking a long time, and a client was resending
1121 a getattr by name lock request. The were multiple lock requests
1122 with the same client lock handle and
1123 mds_getattr_name->fixup_handle_for_resent_request found one of the
1124 lock handles but later failed with ASSERTION(granted_lock != NULL).
1129 Description: handle->h_buffer_credits > 0, assertion failure
1130 Details : h_buffer_credits is zero after truncate, causing assertion
1131 failure. This patch extends the transaction or creates a new
1135 Frequency : NFS re-export or patchless client
1136 Bugzilla : 11179, 10796
1137 Description: Crash on NFS re-export node (__d_move)
1138 Details : We do not want to hash the dentry if we don't have a lock.
1139 But if this dentry is later used in d_move, we'd hit uninitialised
1140 list head d_hash, so we just do this to init d_hash field but
1141 leave dentry unhashed.
1144 Frequency : NFS re-export or patchless client
1146 Description: NFS exports has problem with symbolic link
1147 Details : lustre client didn't properly install dentry when re-exported
1148 to NFS or running patchless client.
1151 Frequency : NFS re-export or patchless client
1153 Description: Various nfs/patchless fixes.
1154 Details : fixes reuse disconected alias for lookup process - this fixes
1155 warning "find_exported_dentry: npd != pd",
1156 fix permission error with open files at nfs.
1157 fix apply umask when do revalidate.
1160 Frequency : occasional
1162 Description: Crash on NFS re-export node
1163 Details : calling clear_page() on the wrong pointer triggered oops in
1164 generic_mapping_read().
1167 Frequency : rarely, using O_DIRECT IO
1169 Description: unaligned directio crashes client with LASSERT
1170 Details : check for unaligned buffers before trying any requests.
1173 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1175 Description: stale data returned from RAID cache
1176 Details : If only a small amount of IO is done to the RAID device before
1177 reading it again it is possible to get stale data from the RAID
1178 cache instead of reading it from disk.
1181 Frequency : always for sles10 kernel
1183 Description: sles10 support
1184 Details : ll_follow_link: compile fixes and using of nd_set_link
1185 under newer kernels.
1188 Frequency : depends on arch, kernel and compiler version, always on sles10
1191 Description: recursive or deep enough symlinks cause stack overflow
1192 Details : getting rid of large stack-allocated variable in
1196 Frequency : depends on hardware
1198 Description: lustre write performance loss in the SLES10 kernel
1199 Details : the performance loss is caused by using of write barriers in the
1200 ext3 code. The SLES10 kernel turns barrier support on by
1201 default. The fix is to undo that change for ldiskfs.
1203 ------------------------------------------------------------------------------
1205 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1207 * Support for kernels:
1208 2.6.9-42.0.3EL (RHEL 4)
1209 2.6.5-7.276 (SLES 9)
1210 2.4.21-47.0.1.EL (RHEL 3)
1211 2.6.12.6 vanilla (kernel.org)
1215 Frequency : quota enabled and large files being deleted
1217 Description: releasing more than 4GB of quota at once hangs OST
1218 Details : If a user deletes more than 4GB of files on a single OST it
1219 will cause the OST to spin in an infinite loop. Release
1220 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1225 Description: statfs data retrieved from /proc may be stale or zero
1226 Details : When reading per-device statfs data from /proc, in the
1227 {kbytes,files}_{total,free,avail} files, it may appear
1228 as zero or be out of date.
1231 Frequency : systems with MD RAID1 external journal devices
1233 Description: lconf's call to blkid is confused by RAID1 journal devices
1234 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1235 of returning all block devices that match the journal UUID.
1238 Frequency : always, for aggregate stripe size over 4GB
1240 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1241 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1242 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1245 Frequency : always if "lfs find" used on a local file/directory
1247 Description: "lfs find" segfaults if used on a local file/directory
1248 Details : The case where a directory component was not specified wasn't
1249 handled correctly. Handle this properly.
1252 Frequency : always on ppc64
1254 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1255 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1256 perform endianness conversion.
1259 Frequency : rarely (truncate to non-zero file size after write under load)
1260 Bugzilla : 10730, 10687
1261 Description: Files padded with zeros to next 4K multiple
1262 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1263 it is possible that files that are truncated to a non-zero size
1264 immediately after being written are filled with zero bytes beyond
1265 the truncated size. No file data is lost.
1267 Severity : enhancement
1269 Description: Allow recovery/failover for liblustre clients.
1270 Details : liblustre clients were unaware of failover configurations until
1273 Severity : enhancement
1275 Description: user file locks should fail when not mounting with flock option
1276 Details : Set up an error-returning stub in ll_file_operations.lock field
1277 to prevent incorrect behaviour when client is mounted without
1278 flock option. Also, set up properly f_op->flock field for
1282 Frequency : always on ia64
1284 Description: "lfs df" loops on printing out MDS statfs information
1285 Details : The obd_ioctl_data was not initialized and in some systems
1286 this caused a failure during the ioctl that did not return
1287 an error. Initialize the struct and return an error on failure.
1290 Frequency : SLES 9 only
1292 Description: Error of copying files with lustre special EAs as root
1293 Details : Client side always return success for setxattr call for lustre
1294 special xattr (currently only "trusted.lov").
1297 Frequency : rarely on clusters with both ia64+i386 clients
1299 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1300 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1301 clients are doing concurrent writes to the same file it is
1302 possible that the ia64 clients may LASSERT because the OST
1303 extent locks are not PAGE_SIZE aligned. Ensure that grown
1304 locks are always aligned on the request boundary.
1307 Frequency : specific use, occasional
1309 Description: Overwriting in use executable truncates on-disk binary image
1310 Details : If one node attempts to overwrite an executable in use by
1311 another node, we now correctly return ETXTBSY instead of
1312 truncating the file.
1314 Severity : enhancement
1316 Description: Async OSC create to avoid the blocking unnecessarily.
1317 Details : If a OST has no remain object, system will block on the creating
1318 when need to create a new object on this OST. Now, ways use
1319 pre-created objects when available, instead of blocking on an
1320 empty osc while others are not empty. If we must block, we block
1321 for the shortest possible period of time.
1326 Description: chmod on Lustre root is propagated to other clients
1327 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1328 invalid by the follow_mount time.
1331 Frequency : liblustre clients only
1333 Description: Race in 'instant cancel' lock handling could lead to such locks
1334 never to be granted in case of SMP MDS
1335 Details : Do not destroy not yet granted but cbpending locks in
1339 Frequency : replay/resend of open
1341 Description: non null lock assetion failure in mds_intent_policy
1342 Details : Trying to replay/resend lockless open requests resulted in
1343 mds_open() returning 0 with no lock. Now it sets a flag if
1344 a lock is going to be returned.
1346 Severity : enhancement
1348 Description: Checksum enhancements
1349 Details : New checksum enhancements allow for resending RPCs that failed
1352 Severity : enhancement
1354 Description: Tunables on number of dirty pages in cacche
1355 Details : Allow to set limit on number of dirty pages cached.
1360 Description: client crash on unmount - lock still has references
1361 Details : In some error handling cases it was possible to leak a lock
1362 reference on a client while accessing a file. This was not
1363 harmful to the client during operation, but would cause the
1364 client to crash when the filesystem is unmounted.
1367 Frequency : specific case, rare
1369 Description: ETXTBSY on mds though file not in use by client
1370 Details : ETXTBSY is no longer incorrectly returned when attempting to
1371 chmod or chown a directory that the user previously tried to
1372 execute or a currently-executing binary.
1375 Frequency : extremely rare except on liblustre-based clients
1377 Description: Lustre space not freed when files are deleted
1378 Details : Clean up open-unlinked files after client eviction. Previously
1379 the unlink was skipped and the files remained as orphans.
1384 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1385 Details : In some cases it was possible to send a blocking callback to
1386 a client doing a glimpse, even though that client didn't get
1387 a lock granted. When the glimpse lock is cancelled on the OST
1388 the freed lock is left on the waiting list and corrupted the list.
1391 Frequency : all core dumps
1393 Description: Broke core dumps to lustre
1394 Details : Negative dentry may be unhashed if parent does not have UPDATE
1395 lock, but some callers, e.g. do_coredump, expect dentry to be
1396 hashed after successful create, hash it in ll_create_it.
1398 ------------------------------------------------------------------------------
1400 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1402 * Support for kernels:
1403 2.6.9-42.0.2.EL (RHEL 4)
1404 2.6.5-7.276 (SLES 9)
1405 2.4.21-40.EL (RHEL 3)
1406 2.6.12.6 vanilla (kernel.org)
1410 Frequency : always on RHEL 3
1412 Description: Number of open files grows over time
1413 Details : The number of open files grows over time, whether or not
1414 Lustre is started. This was due to a filp leak introduced
1415 by one of our kernel patches.
1417 ------------------------------------------------------------------------------
1419 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1421 * Support for kernels:
1422 2.6.9-42.EL (RHEL 4)
1423 2.6.5-7.276 (SLES 9)
1424 2.4.21-40.EL (RHEL 3)
1425 2.6.12.6 vanilla (kernel.org)
1430 Bugzilla : 5719, 9635, 9792, 9684
1431 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1432 Details : If a server is under heavy load and cannot reply to new
1433 connection requests before the client resends the (re)connect,
1434 the connection handling code can behave badly if two service
1435 threads are concurrently handing separate (re)connections from
1436 the same client. Add better locking to the connection handling
1437 code, and ensure that only a single connection will be processed
1438 for a given client UUID, even if the lock is dropped.
1440 Severity : enhancement
1442 Description: add TCP zero-copy support to kernel
1443 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1444 sends if the hardware supports scatter-gather and checksumming.
1445 This allows socklnd to do client-write and server-read more
1446 efficiently and reduce CPU utilization from skbuf copying.
1449 Frequency : only if NFS exporting from client
1451 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1452 Details : If setting attributes on a file created under NFS that had
1453 never been opened it would be possible to oops the client
1454 if the file had no objects.
1457 Frequency : always for liblustre
1459 Description: liblustre client does MDS+OSTs setattr RPC for each write
1460 Details : When doing a write from a liblustre client, the client
1461 incorrectly issued an RPC to the MDS and each OST the file was
1462 striped over in order to update the timestamps. When writing
1463 with small chunks and many clients this could overwhelm the MDS
1464 with RPCs. In all cases it would slow down the write because
1465 these RPCs are unnecessary.
1467 Severity : enhancement
1469 Description: allow number of MDS service threads to be changed at module load
1470 Details : It is now possible to change the number of MDS service threads
1471 running. Adding "options mds mds_num_threads={N}" to the MDS's
1472 /etc/modprobe.conf will set the number of threads for the next
1473 time Lustre is restarted (assuming the "mds" module is also
1474 reloaded at that time). The default number of threads will
1475 stay the same, 32 for most systems.
1480 Description: OST crash if filesystem is unformatted or corrupt
1481 Details : If an OST is started on a device that has never been formatted
1482 or if the filesystem is corrupt and cannot even mount then the
1483 error handling cleanup routines would dereference a NULL pointer.
1488 Description: NULL pointer deref in llap_from_page.
1489 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1490 incorrect) mapping if the page was truncated/reclaimed while it was
1491 searched for. Check for this condition and skip such pages when
1492 doing readahead. Introduce extra check to llap_from_page() to
1493 verify page->mapping->host is non-NULL (so page is not anonymous).
1496 Frequency : Sometimes when using sys_sendfile
1498 Description: "page not covered by a lock" warnings from ll_readpage
1499 Details : sendfile called ll_readpage without right page locks present.
1500 Now we introduced ll_file_sendfile that does necessary locking
1501 around call to generic_file_sendfile() much like we do in
1505 Frequency : with certain MDS communication failures at client mount time
1507 Description: NULL pointer deref after failed client mount
1508 Details : a client connection request may delayed by the network layer
1509 and not be sent until after the PTLRPC layer has timed out the
1510 request. If the client fails the mount immediately it will try
1511 to clean up before the network times out the request. Add a
1512 reference from the request import to the obd device and delay
1513 the cleanup until the network drops the request.
1516 Frequency : occasionally during client (re)connect
1518 Description: assertion failure during client (re)connect
1519 Details : processing a client connection request may be delayed by the
1520 client or server longer than the client connect timeout. This
1521 causes the client to resend the connection request. If the
1522 original connection request is replied in this interval, the
1523 client may trip an assertion failure in ptlrpc_connect_interpret()
1524 which thought it would be the only running connect process.
1527 Frequency : only with obd_echo servers and clients that are rebooted
1529 Description: kernel BUG accessing uninitialized data structure
1530 Details : When running an obd_echo server it did not start the ping_evictor
1531 thread, and when a client was evicted an uninitialized data
1532 structure was accessed. Start the ping_evictor in the RPC
1533 service startup instead of the OBD startup.
1535 Severity : enhancement
1536 Bugzilla : 10193 (patchless)
1537 Description: Remove dependency on various unexported kernel interfaces.
1538 Details : No longer need reparent_to_init, exit_mm, exit_files,
1539 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1542 Frequency : rare (only users of deprecated and unsupported LDAP config)
1544 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1545 Details : LDAP apparently contains 'lustreName' attributes instead of
1546 'name'. A simple remapping of the name is sufficient.
1549 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1551 Description: waiting_locks_callback trips kernel BUG if client is evicted
1552 Details : Running with the dump_on_timeout debug flag turned on makes
1553 it possible that the waiting_locks_callback() can try to dump
1554 the Lustre kernel debug logs from an interrupt handler. Defer
1555 this log dumping to the expired_lock_main() thread.
1557 Severity : enhancement
1559 Description: Support NFS exporting on 2.6 kernels.
1560 Details : Implement non-rawops metadata methods for NFS server to use without
1561 changing NFS server code.
1564 Frequency : very rare (synthetic metadata workload only)
1566 Description: two racing renames might cause an MDS thread to deadlock
1567 Details : Running the "racer" program may cause one MDS thread to rename
1568 a file from being the source of a rename to being the target of
1569 a rename at exactly the same time that another thread is doing
1570 so, and the second thread has already enqueued these locks after
1571 doing a lookup of the target and is trying to relock them in
1572 order. Ensure that we don't try to re-lock the same resource.
1575 Frequency : only very large systems with liblustre clients
1577 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1578 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1579 all OSTs in parallel. This allows the eviction of stale liblustre
1580 clients to proceed much faster than if they were done in series,
1581 and also offers similar improvements for other set_info RPCs.
1586 Description: excessive CPU usage during initial read phase on client
1587 Details : During the initial read phase on a client, it would agressively
1588 retry readahead on the file, consuming too much CPU and impacting
1589 performance (since 1.4.5.8). Improve the readahead algorithm
1590 to avoid this, and also improve some other common cases (read
1591 of small files in particular, where "small" is files smaller than
1592 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1597 Description: MDS crash when receiving packet with unknown intent.
1598 Details : Do not LBUG in unknown intent case, just return -EFAULT
1600 Severity : enhancement
1601 Bugzilla : 9293, 9385
1602 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1603 Details : Do not serialize getattr (non-intent version) and statfs.
1606 Frequency : occasional, when OST network is overloaded/intermittent
1608 Description: client evicted by OST after bulk IO timeout
1609 Details : If a client sends a bulk IO request (read or write) the OST
1610 may evict the client if it is unresposive to its data GET/PUT
1611 request. This is incorrect if the network is overloaded (takes
1612 too long to transfer the RPC data) or dropped the OST GET/PUT
1613 request. There is no need to evict the client at all, since
1614 the pinger and/or lock callbacks will handle this, and the
1615 client can restart the bulk request.
1618 Frequency : Always when mmapping file with no objects
1620 Description: client crashes when mmapping file with no objects
1621 Details : Check that we actually have objects in a file before doing any
1622 operations on objects in ll_vm_open, ll_vm_close and
1628 Description: Request leak when working with deleted CWD
1629 Details : Introduce advanced request refcount tracking for requests
1630 referenced from lustre intent.
1632 Severity : Enhancement
1634 Description: Cache open file handles on client.
1635 Details : MDS now will return special lock along with openhandle, if
1636 requested and client is allowed to hold openhandle, even if unused,
1637 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1638 closing files for every read/write openration.
1640 Severity : Enhancement
1642 Description: Cache open negative dentries on client when possible.
1643 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1644 negative dentries on lock revocation.
1649 Description: Remounting a client read-only wasn't possible with a zconf mount
1650 Details : It wasn't possible to remount a client read-only with llmount.
1652 Severity : enhancement
1653 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1654 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1655 MPICH is included to add Lustre-specific ADIO interfaces.
1656 This is based closely on the UFS ADIO layer and only differs
1657 in file creation, in order to allow the OST striping to be set.
1658 This is user-contributed code and not supported by CFS.
1663 Description: extended inode attributes (immutable, append-only) work improperly
1664 when 2.4 and 2.6 kernels are used on client/server or vice versa
1665 Details : Introduce kernel-independent values for these flags.
1667 Severity : enhancement
1670 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1671 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1672 granularity. Now, Fractional MB granularity is be supported,
1673 this is very useful for low memory system.
1675 Severity : enhancement
1677 Description: Getattr by fid
1678 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1679 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1680 have locks and avoids getattr rpc requests for it on every stat.
1683 Frequency : Always, for filesystems larger than 2TB
1685 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1686 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1687 filesystems larger than 2TB to be mounted with mballoc enabled.
1690 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1692 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1693 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1694 filesystems larger than 2TB is mounted then the kernel will
1695 silently corrupt the start of the filesystem. CONFIG_LBD is
1696 enabled for all CFS-supported kernels, but the possibility of
1697 this happening with a modified kernel config exists.
1699 Severity : enhancement
1701 Description: add client O_DIRECT support for 2.6 kernels
1702 Details : It is now possible to do O_DIRECT reads and writes to files
1703 in the Lustre client mountpoint on 2.6 kernel clients.
1705 Severity : enhancement
1707 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1708 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1709 parallel, not waiting for response from every OST before sending
1710 a rpc to the next OST.
1715 Description: setattr vs write race when updating file timestamps
1716 Details : Client processes that update a file timestamp into the past
1717 right after writing to the file (e.g. tar) it is possible that
1718 the updated file modification time can be reset to the current
1719 time due to a race between processing the setattr and write RPC.
1721 Severity : enhancement
1723 Description: Bring 'lfs find' closer in line with regular Linux find.
1724 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1725 -print0 options and obtains all the needed info through the lustre
1728 Severity : enhancement
1730 Description: support up to 1024 configured devices on one node
1731 Details : change obd_dev array from statically allocated to dynamically
1732 allocated structs as they are first used to reduce memory usage
1737 Description: Flush dirty partially truncated pages during truncate
1738 Details : Immediatelly flush partially truncated pages in filter_setattr,
1739 this way we completely avoid having any pages in page cache on OST
1740 and can retire ugly workarounds during writes to flush such pages.
1745 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1746 Details : protect inode from truncation within vfs_unlink() context
1747 just take a reference before calling vfs_unlink() and release it
1748 when parent's i_sem is free.
1753 Description: last_id value checked outside lock on OST caused LASSERT failure
1754 Details : If there were multiple MDS->OST object precreate requests in
1755 flight, it was possible that the OST's last object id was checked
1756 outside a lock and incorrectly tripped an assertion. Move checks
1757 inside locks, and discard old precreate requests.
1760 Frequency : always, if extents are used on OSTs
1762 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1763 Details : OSTs using the extents format would not zero the high 16 bits of
1764 the index physical block number. This is not a problem for any
1765 OST filesystems smaller than 16TB, and no kernels support ext3
1766 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1767 new/modified files) and can be fixed for existing filesystems
1768 with e2fsprogs-1.39-cfs1.
1773 Description: import connection selection may be incorrect if timer wraps
1774 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1775 import connections to be ignored if the 32-bit jiffies counter
1776 wraps. Use a 64-bit jiffies counter.
1779 Frequency : very large clusters immediately after boot
1781 Description: LNET request buffers exhausted under heavy short-term load
1782 Details : If a large number of client requests are generated on a service
1783 that has previously never seen so many requests it is possible
1784 that the request buffer growth cannot keep up with the spike in
1785 demand. Instead of dropping incoming requests, they are held in
1786 the LND until the RPC service can accept more requests.
1789 Frequency : Sometimes during replay
1791 Description: Assertion failure in ll_local_open after replay.
1792 Details : If replay happened on an open request reply before we were able
1793 to set replay handler, reply will become not swabbed tripping the
1794 assertion in ll_local_open. Now we set the handler right after
1795 recognising of open request
1798 Frequency : very rare
1800 Description: kernel reports "badness in vsnprintf"
1801 Details : Reading from the "recovery_status" /proc file in small chunks
1802 may cause a negative length in lprocfs_obd_rd_recovery_status()
1803 call to vsnprintf() (which is otherwise harmless). Exit early
1804 if there is no more space in the output buffer.
1806 Severity : enhancement
1808 Description: clear OBD RPC statistics by writing to them
1809 Details : It is now possible to clear the OBD RPC statistics by writing
1810 to the "stats" file.
1815 Description: Client mtime is not the same on different clients after utimes
1816 Details : In some cases, the client was using the utimes() syscall on
1817 a file cached on another node. The clients now validate the
1818 ctime from the MDS + OSTs to determine which one is right.
1823 Description: Inability to activate failout mode
1824 Details : lconf script incorrectly assumed that in pythong string's numeric
1825 value is used in comparisons.
1828 Frequency : always with multiple stripes per file
1830 Description: Inefficient object allocation for mutli-stripe files
1831 Details : When selecting which OSTs to stripe files over, for files with
1832 a stripe count that divides evenly into the number of OSTs,
1833 the MDS is always picking the same starting OST for each file.
1834 Return the OST selection heuristic to the original design.
1839 Description: mount failures may take full timeout to return an error
1840 Details : Under some heavy load conditions it is possible that a
1841 failed mount can wait for the full obd_timeout interval,
1842 possibly several minutes, before reporting an error.
1843 Instead return an error as soon as the status is known.
1845 Frequency : quota enabled and large files being deleted
1847 Description: releasing more than 4GB of quota at once hangs OST
1848 Details : If a user deletes more than 4GB of files on a single OST it
1849 will cause the OST to spin in an infinite loop. Release
1850 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1855 Description: statfs data retrieved from /proc may be stale or zero
1856 Details : When reading per-device statfs data from /proc, in the
1857 {kbytes,files}_{total,free,avail} files, it may appear
1858 as zero or be out of date.
1861 Frequency : systems with MD RAID1 external journal devices
1863 Description: lconf's call to blkid is confused by RAID1 journal devices
1864 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1865 of returning all block devices that match the journal UUID.
1868 Frequency : always, for aggregate stripe size over 4GB
1870 Description: assertion fails when trying to use 4GB stripe size
1871 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1872 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1875 Frequency : always on ppc64
1877 Description: the first write on an ext3 filesystem with mballoc got stuck
1878 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1879 perform endianness conversion.
1881 ------------------------------------------------------------------------------
1883 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1885 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1886 INTEROPERATE with older versions automatically. Please read the
1887 user documentation before upgrading any part of a live system.
1888 * WARNING: Lustre networking configuration changes are required with
1889 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1892 * Support for kernels:
1893 2.6.9-22.0.2.EL (RHEL 4)
1894 2.6.5-7.244 (SLES 9)
1895 2.6.12.6 vanilla (kernel.org)
1898 Severity : enhancement
1899 Bugzilla : 7981/8208
1900 Description: Introduced Lustre Networking (LNET)
1901 Details : LNET is new networking infrastructure for Lustre, it includes
1902 a reorganized network configuration mode (see the user
1903 documentation for full details) as well as support for routing
1904 between different network fabrics. Lustre Networking Devices
1905 (LNDs) for the supported network fabrics have also been
1906 created for this new infrastructure.
1908 Severity : enhancement
1909 Description: Introduced Access control lists
1910 Details : clients can set ACLs on files and directories in order to have
1911 more fine-grained permissions than the standard Unix UGO+RWX.
1912 The MDS must be started with the "-o acl" mount option.
1914 Severity : enhancement
1915 Description: Introduced filesystem quotas
1916 Details : Administrators may now establish per-user quotas on the
1919 Severity : enhancement
1921 Description: Configuration change for the XT3
1922 The PTLLND is now used to run Lustre over Portals on the XT3
1923 The configure option(s) --with-cray-portals are no longer used.
1924 Rather --with-portals=<path-to-portals-includes> is used to
1925 enable building on the XT3. In addition to enable XT3 specific
1926 features the option --enable-cray-xt3 must be used.
1931 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1932 Details : A race between allocating a new llcd and re-getting the llcd_lock
1933 allowed another thread to grab newly-allocated llcd.
1935 Severity : enhancement
1937 Description: 2.6 OST async journal commit and locking fix to improve performance
1938 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1939 2.6 kernels are now async as they already were in 2.4 kernels so
1940 that they can commit concurrently with the network bulk transfer.
1941 For block-allocated files the filter allocation semaphore is held
1942 to avoid filesystem fragmentation during allocation. BKL lock
1943 removed for 2.6 xattr operations where it is no longer needed.
1948 Description: lconf incorrectly determined whether two IP networks could talk
1949 Details : In some more complicated routing and multiple-network
1950 configurations, lconf will avoid trying to make a network
1951 connection to a disjoint part of the IP space. It was doing the
1952 math incorrectly for one set of cases.
1957 Description: Fix for potential infinite loop processing records in an llog.
1958 Details : If an llog record is corrupted/zeroed, it is possible to loop
1959 forever in llog_process(). Validate the llog record length
1960 and skip the remainder of the block on error.
1963 Frequency : occasional (liblustre only)
1965 Description: liblustre could not open files whose last component is a symlink
1966 Details : sysio_path_walk() would incorrectly pass the open intent to
1967 intermediate path components.
1970 Frequency : rare (liblustre only with non-standard tuning)
1971 Bugzilla : 7201 (7350)
1972 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1973 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1974 after it has been released, only do this for referenced locks
1976 Severity : enhancement
1978 Description: specifying an (invalid) directory default stripe_size of -1
1979 would reset the directory default striping
1980 Details : stripe_size -1 was used internally to signal directory stripe
1981 removal, now use "all default" to signal dir stripe removal
1982 as a directory striping of "all default" is not useful
1985 Frequency : common for large clusters running liblustre clients
1987 Description: doing an ls when liblustre clients are running is slow
1988 Details : sending a glimpse AST to a liblustre client waits for every AST
1989 to time out, as liblustre clients will not respond. Since they
1990 cannot cache data we refresh the OST lock LVB from disk instead.
1992 Severity : enhancement
1994 Description: doing an ls at the same time as file IO can be slow
1995 Details : enqueue and other "small" requests can be blocked behind many
1996 large IO requests. Create a new OST IO portal for non-IO
1997 requests so they can be processed faster.
2000 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2002 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2003 created file with mode 000
2004 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2005 allow an ftruncate() as a non-root user to a file with mode 000.
2006 Lustre now respects this flag to disable mode checks when
2007 truncating a file owned by the user
2010 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2013 Description: Revoking locks from clients that went dead or catatonic might take
2015 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2016 cancellation of such locks instant on servers without waiting for
2017 any reply from clients. Clients drops these locks when cancel
2018 notification from server is received without replying.
2021 Frequency : liblustre-only, when liblustre client dies or becomes busy
2023 Description: Doing ls on Linux clients can take a long time with active
2025 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2026 granting such locks to it in the first place if possible. Locks
2027 are taken by proxy on the OST during the read or write and
2028 dropped immediately afterward. Add connect flags handling, do
2029 not grant locks to liblustre clients for glimpse ASTs.
2031 Severity : enhancement
2033 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2034 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2035 sequential reads and behaves badly with applications doing random
2036 reads. Improve it by reading ahead at least read region, and
2037 avoiding excessive large RPC for small reads.
2039 Severity : enhancement
2041 Description: Creating more than 1000 files for a single job may cause a load
2042 imbalance on the OSTs if there are also a large number of OSTs.
2043 Details : qos_prep_create() uses an OST index reseed value that is an
2044 even multiple of the number of available OSTs so that if the
2045 reseed happens in the middle of the object allocation it will
2046 still utilize the OSTs as uniformly as possible.
2051 Description: OST or MDS may oops in ping_evictor_main()
2052 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2053 but doesn't restart at beginning of obd_exports_timed list
2056 Severity : enhancement
2058 Description: improve by-nid export eviction on the MDS and OST
2059 Details : allow multiple exports with the same NID to be evicted at one
2060 time without re-searching the exports list.
2063 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2065 Description: MDS may oops in groups_free()
2066 Details : in rare race conditions a newly allocated group_info struct is
2067 freed again, and this can be NULL. The 2.4 compatibility code
2068 for groups_free() checked for a NULL pointer, but 2.6 did not.
2071 Frequency : common for liblustre clients doing little filesystem IO
2072 Bugzilla : 9352, 7313
2073 Description: server may evict liblustre clients accessing contended locks
2074 Details : if a client is granted a lock or receives a completion AST
2075 with a blocking AST already set it would not reply to the AST
2076 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2079 Frequency : lfs setstripe, only systems with more than 160 OSTs
2081 Description: unable to set striping with a starting offset beyond OST 160
2082 Details : llapi_create_file() incorrectly limited the starting stripe
2083 index to the maximum single-file stripe count.
2086 Frequency : LDAP users only
2088 Description: lconf did not handle in-kernel recovery with LDAP properly
2089 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2091 Severity : enhancement
2093 Description: bind OST threads to NUMA nodes to improve performance
2094 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2095 node and do their allocations there to localize memory access
2097 Severity : enhancement
2099 Description: llmount can determine client NID directly from Myrinet (GM)
2100 Details : the client NID code from gmnalnid was moved directly into
2101 llmount, removing the need to use this or specifying the
2102 client NID explicitly when mounting GM clients with zeroconf
2105 Frequency : if client is started with down MDS
2107 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2108 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2109 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2110 ptlrpc_import_delayed_req() because we didn't check if the
2111 request was interrupted, and we also didn't break out of the
2112 event loop if there was no timeout
2117 Description: data loss during non-page-aligned writes to a single file from
2118 both multiple nodes and multiple threads on one node at same time
2119 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2120 inconsistency led to false short-reads, that were cached and later
2121 used by ->prepare_write() to fill in partially written page,
2122 leading to data loss.
2125 Frequency : always, if lconf --abort_recovery used
2127 Description: lconf --abort_recovery fails with 'Operation not supported'
2128 Details : lconf was attempting to abort recovery on the MDT device and not
2131 Severity : enhancement
2133 Description: remove cleanup logs
2134 Details : replace lconf-generated cleanup logs with lustre internal
2135 cleanup routines. Eliminates the need for client-cleanup and
2138 Severity : enhancement
2140 Description: add support for EAs (user and system) on lustre filesystems
2141 Details : it is now possible to store extended attributes in the Lustre
2142 client filesystem, and with the user_xattr mount option it
2143 is possible to allow users to store EAs on their files also
2145 Severity : enhancement
2147 Description: Add possibility (config option) to show minimal available OST free
2149 Details : When compiled with --enable-mindf configure option, statfs(2)
2150 (and so, df) will return least minimal free space available from
2151 all OSTs as amount of free space on FS, instead of summary of
2152 free spaces of all OSTs.
2154 Severity : enhancement
2156 Description: do not expand extent locks acquired on OST-side
2157 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2158 by server: they are not cached anyway.
2161 Frequency : when mmap is used/binaries executed from Lustre
2163 Description: Unmmap pages before throwing them away from read cache.
2164 Details : llap_shrink cache now attempts to unmap pages before discarding
2165 them (if unmapping failed - do not discard). SLES9 kernel has
2166 extra checks that trigger if this unmapping is not done first.
2171 Description: lconf didn't resolve symlinks before checking to see whether a
2172 given mountpoint was already in use
2175 Frequency : when migrating failover services
2176 Bugzilla : 6395, 9514
2177 Description: When migrating a subset of services from a node (e.g. failback
2178 from a failover service node) the remaining services would
2179 time out and evict clients.
2180 Details : lconf --force (implied by --failover) sets the global obd_timeout
2181 to 5 seconds in order to quickly disconnect, but this caused
2182 other RPCs to time out too quickly. Do not change the global
2183 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2185 Severity : enhancement
2186 Frequency : if MDS is started with down OST
2187 Bugzilla : 9439,5706
2188 Description: Allow startup/shutdown of an MDS without depending on the
2189 availability of the OSTs.
2190 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2191 Add appropriate locking and lov-osc refcounts for safe
2192 cleaning. Add osc abort_inflight calls in case the
2193 synchronize never started.
2196 Frequency : occasional (Cray XT3 only)
2198 Description: root not authorized to access files in CRAY_PORTALS environment
2199 Details : The client process capabilities were not honoured on the MDS in
2200 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2201 been accessed by an authorized user then root was able to access
2202 the file on the local client also. The root user capabilities
2203 are now allowed on the MDS, as this environment has secure UID.
2206 Frequency : occasional
2208 Description: ldiskfs "too long searching" message happens too often
2209 Details : A debugging message (otherwise harmless) prints too often on
2210 the OST console. This has been reduced to only happen when
2211 there are fragmentation problems on the filesystem.
2216 Description: Division by zero in statfs when all OSCs are inactive
2217 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2218 lov_statfs divides by value returned by lov_get_stripecnt().
2222 Bugzilla : 9489, 3273
2223 Description: First write from each client to each OST was only 4kB in size,
2224 to initialize client writeback cache, which caused sub-optimal
2225 RPCs and poor layout on disk for the first writen file.
2226 Details : Clients now request an initial cache grant at (re)connect time
2227 and so that they can start streaming writes to the cache right
2228 away and always do full-sized RPCs if there is enough data.
2229 If the OST is rebooted the client also re-establishes its grant
2230 so that client cached writes will be honoured under the grant.
2235 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2236 Details : Now I/O RPCs go to different portal number and (presumably) fast
2237 lock requests (and glimses) and other RPCs get their own service
2238 threads pool that should be able to service those RPCs
2241 Severity : enhancement
2243 Description: Ability to exchange lustre version between client and servers and
2244 issue warnings at client side if client is too old. Also for
2245 liblustre clients there is ability to refuse connection of too old
2247 Details : New 'version' field is added to connect data structure that is
2248 filled with version info. That info is later checked by server and
2252 Frequency : rare, liblustre only.
2253 Bugzilla : 9296, 9581
2254 Description: Two simultaneous writes from liblustre at offset within same page
2255 might proceed at the same time overwriting eachother with stale
2257 Details : I/O lock withing llu_file_prwv was released too early, before data
2258 actually was hitting the wire. Extended lock-holding time until
2259 server acknowledges receiving data.
2262 Frequency : extremely rare. Never observed in practice.
2264 Description: avoid generating lustre_handle cookie of 0.
2265 Details : class_handle_hash() generates handle cookies by incrementing
2266 global counter, and can hit 0 occasionaly (this is unlikely, but
2267 not impossible, because initial value of cookie counter is
2268 selected randonly). Value of 0 is used as a sentinel meaning
2269 "unassigned handle" --- avoid it. Also coalesce two critical
2270 sections in this function into one.
2272 Severity : enhancement
2274 Description: allow liblustre clients to delegate truncate locking to OST
2275 Details : To avoid overhead of locking, liblustre client instructs OST to
2276 take extent lock in ost_punch() on client's behalf. New connection
2277 flag is added to handle backward compatibility.
2279 Severity : enhancement
2280 Bugzilla : 4928, 7341, 9758
2281 Description: allow number of OST service threads to be specified
2282 Details : a module parameter allows the number of OST service threads
2283 to be specified via "options ost ost_num_threads={N}" in the
2284 OSS's /etc/modules.conf or /etc/modprobe.conf.
2288 Bugzilla : 6146, 9635, 9895
2289 Description: servers crash with bad pointer in target_handle_connect()
2290 Details : In rare cases when a client is reconnecting it was possible that
2291 the connection request was the last reference for that export.
2292 We would temporarily drop the export reference and get a new
2293 one, but this may have been the last reference and the export
2294 was just destroyed. Get new reference before dropping old one.
2296 Severity : enhancement
2297 Frequency : if client is started with failover MDS
2299 Description: Allow multiple MDS hostnames in the mount command
2300 Details : Try to read the configuration from all specified MDS
2301 hostnames during a client mount in case the "primary"
2304 Severity : enhancement
2306 Description: Stop sending data to evicted clients as soon as possible.
2307 Details : Check if the client we are about to send or are sending data to
2308 was evicted already. (Check is done every second of waiting,
2309 for which l_wait_event interface was extended to allow checking
2310 of exit condition at specified intervals).
2313 Frequency : rare, normally only when NFS exporting is done from client
2315 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2317 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2320 Frequency : always, if srand() is called before liblustre initialization
2322 Description: Liblustre uses system PRNG disturbing its usage by user application
2323 Details : Introduce internal to lustre fast and high-quality PRNG for
2324 lustre usage and make liblustre and some other places in generic
2325 lustre code to use it.
2327 Severity : enhancement
2328 Bugzilla : 9477, 9557, 9870
2329 Description: Verify that the MDS configuration logs are updated when xml is
2330 Details : Check if the .xml configuration logs are newer than the config
2331 logs stored on the MDS and report an error if this is the case.
2332 Request --write-conf, or allow starting with --old_conf.
2334 Severity : enhancement
2336 Description: Handle symlinks in the path when checking if Lustre is mounted.
2337 Details : Resolve intermediate symlinks when checking if a client has
2338 mounted a filesystem to avoid duplicate client mounts.
2343 Description: lconf can hit an error exception but still return success.
2344 Details : The lconf command catches the Command error exception at the top
2345 level script context and will exit with the associated exit
2346 status, but doesn't ensure that this exit status is non-zero.
2351 Description: failure of ptlrpc thread startup can cause oops
2352 Details : Starting a ptlrpc service thread can fail if there are a large
2353 number of threads or the server memory is very fragmented.
2354 Handle this without oopsing.
2357 Frequency : always, only if liblustre and non-default acceptor port was used
2359 Description: liblustre cannot connect to servers with non-default acceptor port
2360 Details : tcpnal_set_default_params() was not called and was therefore
2361 ignoring the environment varaible TCPNAL_PORT, as well as other
2362 TCPNAL_ environment variables
2367 Description: two objects could be created on the same OST for a single file
2368 Details : If an OST is down, in some cases it was possible to create two
2369 objects on a single OST for a single file. No problems other
2370 than potential performance impact and spurious error messages.
2374 Bugzilla : 5681, 9562
2375 Description: Client may oops in ll_unhash_aliases
2376 Details : Client dcache may become inconsistent in race condition.
2377 In some cases "getcwd" can fail if the current directory is
2383 Description: Inode refcounting problems in NFS export code
2384 Details : link_raw functions used to call d_instantiate without obtaining
2385 extra inode reference first.
2389 Bugzilla : 9942, 9903
2390 Description: Referencing freed requests leading to crash, memleaks with NFS.
2391 Details : We used to require that call to ll_revalidate_it was always
2392 followed by ll_lookup_it. Also with revalidate_special() it is
2393 possible to call ll_revalidate_it() twice for the same dentry
2394 even if first occurence returned success. This fix changes semantic
2395 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2396 reference on a request referred from the intent.
2397 ll_intent_release() then releases such a request.
2400 Frequency : rare, normally benchmark loads only
2402 Description: unlinked inodes were kept in memory on the client
2403 Details : If a client is repeatedly creating and unlinking files it
2404 can accumulate a lot of stale inodes in the inode slab cache.
2405 If there is no other client load running this can cause the
2406 client node to run out of memory. Instead flush old inodes
2407 from client cache that have the same inode number as a new inode.
2410 Frequency : SLES9 2.6.5 kernel and long filenames only
2411 Bugzilla : 9969, 10379
2412 Description: utime reports stale NFS file handle
2413 Details : SLES9 uses out-of-dentry names in some cases, which confused
2414 the lustre dentry revalidation. Change it to always use the
2418 Frequency : rare, unless heavy write-truncate concurrency is continuous
2419 Bugzilla : 4180, 6984, 7171, 9963, 9331
2420 Description: OST becomes very slow and/or deadlocked during object unlink
2421 Details : filter_destroy() was holding onto the parent directory lock
2422 while truncating+unlinking objects. For very large objects this
2423 may block other threads for a long time and slow overall OST
2424 responsiveness. It may also be possible to get a lock ordering
2425 deadlock in this case, or run out of journal credits because of
2426 the combined truncate+unlink. Solution is to do object truncate
2427 first in one transaction without parent lock, and then do the
2428 final unlink in a new transaction with the parent lock. This
2429 reduces the lock hold time dramatically.
2432 Frequency : rare, 2.4 kernels only
2434 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2435 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2436 before unmounting their filesystem in order to avoid deadlock.
2437 The kernel_locked() function in 2.4 kernels only checks whether
2438 the kernel lock is held, not whether it is this process that is
2439 holding it as 2.6 kernels do.
2444 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2445 Details : The client ptlrpc code may be trying to reconnect to a down
2446 server before a previous connection attempt has timed out.
2447 Increase the reconnect interval to be longer than the connection
2448 timeout interval to avoid sending duplicate connections to
2452 Frequency : echo_client brw_test command
2454 Description: fix echo_client to work with OST preallocated code
2455 Details : OST preallocation code (5137) didn't take echo_client IO path
2456 into account: echo_client calls filter methods outside of any
2457 OST thread and, hence, there is no per-thread preallocated
2458 pages and buffers to use. Solution: hijack pga pages for IO. As
2459 a byproduct, this avoids unnecessary data copying.
2463 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2464 Description: Client can oops in mdc_commit_close() after open replay
2465 Details : It was possible for the MDS to return an open request with no
2466 transaction number in mds_finish_transno() if the client was
2467 evicted, but without actually returning an error. Clients
2468 would later try to replay that open and may trip an assertion
2469 Simplify the client close codepath, and always return an error
2470 from the MDS in case the open is not successful.
2473 Frequency : rare, 2.6 OSTs only
2475 Description: OST may deadlock under high load on fragmented files
2476 Details : If there was a heavy load and highly-fragmented OST filesystems
2477 it was possible to have all the OST threads deadlock waiting on
2478 allocation of biovecs, because the biovecs were not released
2479 until the entire RPC IO was completed. Instead, release biovecs
2480 as soon as they are complete to ensure forward IO progress.
2482 Severity : enhancement
2484 Description: Support for specifying external journal device at mount
2485 Details : If an OST or MDS device is formatted with an external journal
2486 device, this device major/minor is stored in the ext3 superblock
2487 and may not be valid for failover. Allow detecting and
2488 specifying the external journal at mount time.
2493 Description: Mounting an MDS with pending unlinked files may cause oops
2494 Details : target_finish_recovery() calls mds_postrecov() which returned
2495 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2496 considers this an error and immediately begins cleaning up the
2497 lov, just after starting the mds_lov process
2499 Severity : enhancement
2501 Description: Implement 'lfs df' to report actual free space on per-OST basis
2502 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2503 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2504 usage of objects; '-h' to report in human readable format.
2506 ------------------------------------------------------------------------------
2508 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2515 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2516 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2517 at startup during orphan unlinking. Instead, skip buddy update
2518 before setup, as it will be regenerated after recovery is complete.
2523 Description: If an OST is inactive, its locks might reference stale inodes.
2524 Details : lov_change_cbdata() must iterate over all namespaces, even if
2525 they are inactive to clear inode references from the lock.
2527 Severity : enhancement
2528 Frequency : occasional, if non-standard max_dirty_mb used
2530 Description: Client will block write RPCs if not enough grant
2531 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2532 then the client will block writes while waiting for another RPC
2533 to complete instead of consuming its dirty limit. With change
2534 we get improved performance when max_dirty_mb is small.
2536 Severity : enhancement
2537 Bugzilla : 3389, 6253
2538 Description: Add support for supplementary groups on the MDS.
2539 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2540 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2541 lookups for user supplementary groups into a cache.
2545 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2546 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2547 for liblustre O_EXCL is enforced by MDS.
2551 Description: readdir never returns NULL in liblustre.
2552 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2553 dentry in struct dirent.
2557 Frequency : liblustre only, depends on application IO pattern
2558 Description: liblustre clients evicted if not contacting servers
2559 Details : Don't put liblustre clients into the ping_evictor list, so
2560 they will not be evicted by the pinger ever.
2562 Severity : enhancement
2564 Description: Add ability to evict clients by NID from MDS.
2565 Details : By echoing "nid:$NID" string into
2566 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2567 $NID would be instantly evicted from this MDS and from all active
2568 OSTs connected to it.
2572 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2573 Details : lookup_it_finish() used to query file size from OSTs that was not
2578 Description: service threads change working directory to that of init
2579 Details : Starting lustre service threads may pin the working directory
2580 of the parent thread, making that filesystem busy. Threads
2581 now change to the working directory of init to avoid this.
2585 Frequency : during shutdown only
2586 Description: shutdown with a failed MDS or OST can cause unmount to hang
2587 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2592 Frequency : 2.6 only
2593 Description: chmod/chown may include an extra supplementary group
2594 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2595 supplementary group and if none is specified this is used.
2598 Bugzilla : 5479 (6816)
2600 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2601 Details : If lookup is in progress on a file that is unlinked we might try
2602 to revalidate the inode and fail in revalidate after lookup is
2603 complete and ll_file_open() enqueues the open again but
2604 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2607 Frequency : always, if lconf --abort_recovery used
2609 Description: lconf --abort_recovery fails with 'Operation not supported'
2610 Details : lconf was attempting to abort recovery on the MDT device and not
2613 ------------------------------------------------------------------------------
2615 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2620 Frequency : rare (only unsupported configurations with a node running as an
2622 Bugzilla : 6514, 5137
2623 Description: Mounting a Lustre file system on a node running as an OST could
2625 Details : OSTs now preallocates memory needed to write out data at
2626 startup, instead of when needed, to avoid having to
2627 allocate memory in possibly low memory situations.
2628 Specifically, if the file system is mounted on on OST,
2629 memory pressure could force it to try to write out data,
2630 which it needed to allocate memory to do. Due to the low
2631 memory, it would be unable to do so and the node would
2632 become unresponsive.
2634 Severity : enhancement
2636 Description: Addition of lconf --service command line option
2637 Details : lconf now accepts a '--service <arg>' option, which is
2638 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2640 Severity : enhancement
2642 Description: Failover mode is now the default for OSTs.
2643 Details : By default, OSTs will now run in failover mode. To return to
2644 the old behaviour, add '--failout' to the lmc line for OSTs.
2646 Severity : enhancement
2648 Description: Health checks are now provided for MDS and OSTs
2649 Details : Additional detailed health check information on MSD and OSTs
2650 is now provided through the procfs health_check value.
2653 Frequency : occasional, depends on IO load
2655 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2656 after numerous create/delete cycles
2657 Details : The ext3 inode allocation policy would not allocate new inodes
2658 very well on the OSTs because there are no new directories
2659 being created. Instead we look for groups with free space if
2660 the parent directories are nearly full.
2665 Description: Network or server problems during mount may cause partially
2666 mounted clients instead of returning an error.
2667 Details : The config llog parsing code may overwrite the error return
2668 code during mount error handling, returning success instead
2674 Description: MDS can fail to allocate large reply buffers
2675 Details : After long uptimes the MDS can fail to allocate large reply
2676 buffers (e.g. zconf client mount config records) due to memory
2677 fragmentation or consumption by the buffer cache. Preallocate
2678 some large reply buffers so that these replies can be sent even
2679 under memory pressure.
2683 Frequency : rare (liblustre)
2684 Description: fsx running with liblustre complained that using truncate() to
2685 extend the file doesn't work. This patch corrects that issue.
2686 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2687 ATTR_SIZE and lsm use in llu_setattr_raw.
2691 Frequency : rare, only 2.6 kernels
2692 Description: Unusual file access patterns on the MDS may result in inode
2693 data being lost in very rare circumstances.
2694 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2695 code in the 2.6 kernel caused the inode and/or EA data not to
2696 be read from disk, causing single-file corruption.
2700 Frequency : rare, only 2.6 filesystems using extents
2701 Description: Heavy concurrent write and delete load may cause data corruption.
2702 Details : It was possible under high-load situations to have an extent
2703 metadata block in the block device cache from a just-unlinked
2704 file overwrite a newly-allocated data block. We now unmap any
2705 metadata buffers that alias just-allocated data blocks.
2709 Frequency : filesystems with default stripe_count larger than 77
2710 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2711 Details : lconf specifies an inode size of 4096 bytes when the default
2712 stripe_count is larger than 77. This conflicts with the default
2713 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2714 this case to avoid pinning too much memory for large EAs.
2716 ------------------------------------------------------------------------------
2718 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2723 Frequency : rare (extremely heavy IO load with hundreds of clients)
2725 Description: Client is evicted, gets IO error writing to file
2726 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2727 caused clients to be evicted for AST timeouts. The fixes in
2728 bug 5192 mean we no longer need to have such short AST timeouts
2729 so ldlm_timeout has been increased.
2732 Frequency : occasional during --force or --failover shutdown under load
2733 Bugzilla : 5949, 4834
2734 Description: Server oops/LBUG if stopped with --force or --failover under load
2735 Details : a collection of import/export refcount and cleanup ordering
2736 issues fixed for safer force cleanup
2739 Frequency : only filesystems larger than 120 OSTs
2740 Bugzilla : 5990, 6223
2741 Description: lfs getstripe would oops on a very large filesystem
2742 Details : lov_getconfig used kfree on vmalloc'd memory
2745 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2746 Bugzilla : 6242, 6243
2747 Description: reading from files that had been truncated to a non-zero size
2748 but never opened returned no data
2749 Details : ll_file_read() reads zeros from no-object files to EOF
2754 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2755 Details : There's little harm in aborting MDS/OSS recovery and letting it
2756 try again, so I removed the LASSERT and return an error instead.
2758 Severity : enhancement
2760 Description: New debugging infrastructure for tracking down data corruption
2761 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2762 (b) cover more of the client-side code path, and (c) try to narrow
2763 down where problems occurred
2767 Bugzilla : 3819, 4364, 4397, 6313
2768 Description: Racing close and eviction MDS could cause assertion in mds_close
2769 Details : It was possible to get multiple mfd references during close and
2770 client eviction, leading to one thread referencing a freed mfd.
2772 Severity: : enhancement
2773 Bugzilla : 3262, 6359
2774 Description: Attempts to reconnect to servers are now more aggressive.
2775 Details : This builds on the enhanced upcall-less recovery that was added
2776 in 1.4.2. When trying to reconnect to servers, clients will
2777 now try each server in the failover group every 10 seconds. By
2778 default, clients would previously try one server every 25 seconds.
2783 Description: After recovery, certain operations trigger a failed
2784 assertion on a client.
2785 Details : Failing over an mds, using lconf -d --failover, while a
2786 client was doing a readdir() call would cause the client to
2787 LBUG after recovery completed and the readdir() was resent.
2789 Severity : enhancement
2791 Description: Default groups are now added by lconf
2792 Details : You can now run lconf --group <servicename> without having to
2793 manually add groups with lmc.
2796 Frequency : occasional
2798 Description: Nodes with an elan id of 0 trigger a failed assertion
2801 Frequency : always when accessing e.g. tty/console device nodes
2803 Description: tty and some other devices nodes cannot be used on lustre
2804 Details : file's private_data field is used by device data and lustre
2805 values in there got lost. New field was added to struct file to
2806 store fs-specific private data.
2809 Frequency : when exporting Lustre via NFS
2811 Description: NFSD failed occasionally when looking up a path component
2812 Details : NFSD is looking up ".." which was broken in ext3 directories
2813 that had grown large enough to become hashed.
2816 Frequency : Clusters with multiple interfaces not on the same subnet
2818 Description: Nodes will repeatedly try to reconnect to an interface which it
2819 cannot reach and report an error to the log.
2820 Details : Extra peer list entries will be created by lconf with some peers
2821 unreachable. lconf now validates the peer before adding it.
2824 Frequency : Only if a default stripe is set on the filesystem root.
2826 Description: Setting a default stripe on the filesystem root prevented the
2827 filesystem from being remounted.
2828 Details : The client was sending extra request flags in the root getattr
2829 request and did not allocate a reply buffer for the dir EA.
2832 Frequency : occasional, higher if lots of files are accessed by one client
2833 Bugzilla : 6159, 6097
2834 Description: Client trips assertion regarding lsm mismatch/magic
2835 Details : While revalidating inodes the VFS looks up inodes with ifind()
2836 and in rare cases can find an inode that is being freed.
2837 The ll_test_inode() code will free the lsm during ifind()
2838 when it finds an existing inode and then the VFS later attaches
2839 this free lsm to a new inode.
2843 Bugzilla : 6422, 7030
2844 Description: MDS deadlock between mkdir and client eviction
2845 Details : Creating a new file via mkdir or mknod (starting a transaction
2846 and getting the ns lock) can deadlock with client eviction
2847 (gets ns lock and trying to finish a synchronous transaction).
2850 Frequency : occasional
2851 Description: While starting a server, the fsfilt_ext3 module could not be
2853 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2854 kernels. Previously, lconf would still use the ext3 name
2855 when trying to load modules. Now, it will correctly use
2856 ext3 on 2.4 and ldiskfs on 2.6.
2858 Severity : enhancement
2859 Description: The default stripe count has been changed to 1
2860 Details : The interpretation of the default stripe count (0, to lfs
2861 or lmc) has been changed to mean striping across a single
2862 OST, rather than all available. For general usage we have
2863 found a stripe count of 1 or 2 works best.
2865 Severity : enhancement
2866 Description: Add support for compiling against Cray portals.
2867 Details : Conditional compiling for some areas that are different
2871 Frequency : occasional
2872 Bugzilla : 6409, 6834
2873 Description: Creating files with an explicit stripe count may lead to
2874 a failed assertion on the MDS
2875 Details : If some OSTs are full or unavailable, creating files may
2876 trigger a failed assertion on the MDS. Now, Lustre will
2877 try to use other servers or return an error to the
2881 Frequency : occasional
2883 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2884 caused lock timeout during readahead (since 1.4.2).
2885 Details : Processes doing readahead might match a lock that hasn't been
2886 granted yet if there are overlapping and conflicting lock
2887 requests. The readahead process waits on ungranted lock
2888 (original lock is CBPENDING), while OST waits for that process
2889 to cancel CBPENDING read lock and eventually evicts client.
2891 Severity : enhancement
2893 Description: Initial enabling of flock support for clients
2894 Details : Implements fcntl advisory locking and file status functions.
2895 This feature is provided as an optional mount flag (default
2896 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2897 locking are implemented yet, and those that are are not guaranteed
2898 to be completely correct in production environments.
2899 mount -t lustre -o [flock|noflock] ...
2902 Frequency : occasional
2904 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2905 assertion in the ext3 JBD (journaling) layer.
2906 Details : The b_committed_data struct is protected by the big kernel lock
2907 in 2.4 kernels, serializing journal_commit_transaction() and
2908 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2909 access to this struct. In 2.6 kernels there is finer grained
2910 locking to improve SMP performance of the JBD layer.
2914 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2919 Description: merge and cleanup kernel patches.
2922 -----------------------------------------------------------------------------
2924 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2926 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2927 versions of Lustre. Please update all servers and clients to
2928 version 1.4.2 or later at the same time. You must also run
2929 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2930 to update the configuration logs.
2932 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2933 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2934 - swab llog records of type '0' so we get proper header size/idx (5861)
2935 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2936 - fix rename of one directory over another leaking an inode (5953)
2937 - avoid SetPageDirty on 2.6 (5981)
2938 - don't re-add just-being-destroyed locks to the waiting list (5653)
2939 - when creating new directories, inherit the parent's custom
2940 striping settings if present parent (3048)
2941 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2942 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2943 - don't hold client locks on temporary worklist from l_lru (5666)
2944 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2945 - automatically evict dead clients (5921)
2946 - Update file size properly in create+truncate+fstat case (6196)
2947 - Do not unhash mountpoint dentries, do not allow removal of
2949 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2950 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2951 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2952 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2953 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2954 - Correctly handle reads of files with no objects (6243)
2955 - lctl recover will also mark a device active if deactivate used (5933)
2957 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2958 - allow --write-conf on an MDS with different nettype than client (5619)
2959 - don't write config llogs to MDS for mounts not from that MDS (5617)
2960 - lconf should create multiple TCP connections from a client (5201)
2961 - init scripts are now turned off by default; run chkconfig --on
2962 lustre and chkconfig --on lustrefs to use them
2963 - upcalls are no longer needed for clients to recover to failover
2965 - add --abort-recovery option to lconf to abort recovery on device
2967 - add support for an arbitrary number of OSTs (3026)
2968 - Quota support protocol changes.
2969 - forward compatibility changes to wire structs (6007)
2970 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2971 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2972 - improved "lustre status" script
2973 - initialize blocksize for non-regular files (6062)
2974 - added --disable-server and --disable-client configure options (5782)
2975 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2976 - Vanilla 2.4.29 support
2977 - increase maximum number of obd devices to 520 (6242)
2978 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2979 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2980 SLES 9 2.6.5 kernels
2981 - limit stripes per file to 160 (the maximum EA size) (6093)
2983 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2986 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2987 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2988 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2989 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2990 - small ext3 extents cleanups and fixes (5733)
2991 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2992 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2993 - avoid needless client->OST connect, fix handle mismatch (5317)
2994 - fix DLM error path that led to out-of-sync client, long delays (5779)
2995 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2996 - fix several locking issues related to i_size (5492,5624,5654,5672)
2997 - don't move pending lock onto export if it is already evicted (5683)
2998 - fix kernel oops when creating .foo in unlinked directory (5548)
2999 - fix deadlock in obdfilter statistics vs. object create (5811)
3000 - use time_{before,after} to avoid timer jiffies wrap (5882)
3001 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3002 - Do not leak request if server was not able to process it (5154)
3003 - If mds_open unable to find parent dir, make that negative lookup(5154)
3004 - don't create new directories with extent-mapping (5909, 5936)
3006 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3007 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3008 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3009 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3010 - support for RHEL4 kernel on i686 (5773)
3011 - provide error messages when incompatible logs are encountered (5898)
3013 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3014 * version 1.4.0.10 (1.4.1 release candidate 1)
3016 - don't keep a lock reference when lock is not granted (4238)
3017 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3018 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3019 - drop import inflight refcount on signal_completed_replay error (5255)
3020 - unlock page after async write error during send (3677)
3021 - handle missing objects in filter_preprw_read properly (5265)
3022 - no transno return for symlink open, don't save no-trasno open (3440)
3023 - don't try to complete elan receive that already failed (4012)
3024 - free RPC server reply state on error (5406)
3025 - clean up thread from ptlrpc_start_thread() on error (5160)
3026 - readahead could read extra page into cache that wasn't ejected (5388)
3027 - prevent races in class_attach/setup/cleanup/detach (5260)
3028 - don't dereference de->d_inode after l_dput of de (5458)
3029 - use "int" for stripe value returned from lock_to_stripe (5544)
3030 - mballoc allocation and error-checking fixes in 2.6 (5504)
3031 - block device patches to fix I/O request sizes in 2.6 (5482)
3032 - look up hostnames for IB nals (5602)
3033 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3034 - don't start multiple acceptors for the same port (5277)
3035 - fix incorrect LASSERT in mds_getattr_name (5635)
3036 - export a proc file for general "ping" checking (5628)
3037 - fix "lfs check" to not block when the MDS is down (5628)
3039 - service request history (4965)
3040 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3041 - create an "evict_client" /proc entry on OSTs, like the MDS has
3042 - fix mount usage message, return errors per mount(8) (5168)
3043 - change grep [] to grep "[]" in tests so they work in more UMLs
3044 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3045 - remove ancient LOV_MAGIC_V0 EA support (5047)
3046 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3047 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3048 - allow readahead allocations to fail when low on memory (5383)
3049 - mmap locking landed again, after considerable improvement (2828)
3050 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3052 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3055 - send OST transaction number in read/write reply to free req (4966)
3056 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3057 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3058 - the watchdog thread now runs as interruptible (5246)
3059 - flock/lockf fixes (but it's still disabled, pending 5135)
3060 - don't use EXT3 constants in llite code (5094)
3061 - memory shortage at startup could cause assertion (5176)
3063 - reorganization of lov code
3064 - single portals codebase
3066 - add extents/mballoc support (5025)
3067 - direct I/O reads in the obdfilter (4048)
3068 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3070 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3073 - send OST transaction number in read/write reply to free req (4966)
3074 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3075 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3076 - the watchdog thread now runs as interruptible (5246)
3077 - handle missing objects in filter_preprw_read properly (5265)
3078 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3079 - drop import inflight refcount on signal_completed_replay error (5255)
3080 - unlock page after async write error during send (3677)
3081 - return original error code on reconstructed replies (3761)
3082 - no transno return for symlink open, don't save no-trasno open (3440)
3084 - add pid to ldlm debugging output (4922)
3085 - bump the watchdog timeouts -- we can't handle 30sec yet
3086 - extra debugging for orphan dentry/inode bug (5259)
3088 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3091 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3092 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3093 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3094 - fix NULL dereference in /proc/sys/portals/routes (4827)
3095 - allow failed mdc_close() operations to be interrupted (4561)
3096 - stop precreate on OST before MDS would time out on it (4778)
3097 - don't send partial-page writes before EOF from client (4410)
3098 - discard client grant for sub-page writes on large-page clients (4520)
3099 - don't free dentries not owned by NFS code, check generation (4806)
3100 - fix lsm leak if mds_create_objects() fails (4801)
3101 - limit debug_daemon file size, always print CERROR messages (4789)
3102 - use transno after validating reply (3892)
3103 - process timed out requests if import state changes (3754)
3104 - update mtime on OST during writes, return in glimpse (4829)
3105 - add mkfsoptions to LDAP (4679)
3106 - use ->max_readahead method instead of zapping global ra (5039)
3107 - don't interrupt __l_wait_event() during strace
3109 - add software watchdogs to catch hung threads quickly (4941)
3110 - make lustrefs init script start after nfs is mounted
3111 - fix CWARN/ERROR duplication (4930)
3112 - return async write errors to application if possible (2248)
3113 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3114 - print NAL number in %x format (4645)
3115 - update barely-supported suse-2.4.21-171 series (4842)
3116 - support for sles 9 %post scripts
3117 - support for building 2.6 kernel-source packages
3118 - support for sles km_* packages
3120 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3123 - ignore -ENOENT errors in osc_destroy (3639)
3124 - notify osc create thread that OSC is being cleaned up (4600)
3125 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3126 - reconstruct ost_handle() like mds_handle() (4657)
3127 - create a new thread to do import eviction to avoid deadlock (3969)
3128 - let lconf resolve symlinked-to devices (4629)
3129 - don't unlink "objects" from directory with default EA (4554)
3130 - hold socknal file ref over connect in case target is down (4394)
3131 - allow more than 32000 subdirectories in a single directory (3244)
3132 - fix blocks count for O_DIRECT writes (3751)
3133 - OST returns ENOSPC from object create when no space left (4539)
3134 - don't send truncate RPC if file size isn't changing (4410)
3135 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3136 - bind to privileged port in socknal and tcpnal (3689)
3138 - rate limit CERROR/CWARN console message to avoid overload (4519)
3139 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3140 - basic mmap support (3918)
3141 - kernel patch series update from b1_4 (4711)
3143 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3146 - avoid crash during MDS cleanup with OST shut down (2775)
3147 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3148 - don't use bad inodes on the MDS (3744)
3149 - dynamic object preallocation to improve recovery speed (4236)
3150 - don't hold spinlock over lock dumping or change debug flags (4401)
3151 - don't zero obd_dev when it is force cleaned (3651)
3152 - print grants to console if they go negative (4431)
3153 - "lctl deactivate" will stop automatic recovery attempts (3406)
3154 - look for existing locks in ldlm_handle_enqueue() (3764)
3155 - don't resolve lock handle twice in recovery avoiding race (4401)
3156 - revalidate should check working dir is a directory (4134)
3158 - don't always mark "slow" obdfilter messages as errors (4418)
3160 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3163 - don't close LustreDB during write_conf until it is done (3860)
3164 - fix typo in lconf for_each_profile (3821)
3165 - allow dumping logs from multiple threads at one time (3820)
3166 - don't allow multiple threads in OSC recovery (3812)
3167 - fix debug_size parameters (3864)
3168 - fix mds_postrecov to initialize import for llog ctxt (3121)
3169 - replace config semaphore with spinlock (3306)
3170 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3171 - don't allow enqueue to complete on a destroyed export (3822)
3172 - down write_lock before checking llog header bitmap (3825)
3173 - recover from lock replay timeout (3764)
3174 - up llog sem before sending rpc (3652)
3175 - reduce ns lock hold times when setting kms (3267)
3176 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3177 - fix NULL deref and obd_dev leak on setup error (3312)
3178 - replace some LBUG about llog ops with error handling (3841)
3179 - don't match INVALID dentries from d_lookup and spin (3784)
3180 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3181 - fix invalid assertion in ptlrpc_set_wait (3880)
3183 - add libwrap support for the TCP acceptor (3996)
3184 - add /proc/sys/portals/routes for non-root route listing (3994)
3185 - allow setting MDS UUID in .xml (2580)
3186 - print the stack of a process that LBUGs (4228)
3188 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3191 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3192 - make sure to unset replay flag from failed open requests (3440)
3193 - if default stripe count is 0, use OST count for inode size (3636)
3194 - update parent mtime/ctime on client for create/unlink (2611)
3195 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3196 - free recovery state on server during a forced cleanup (3571)
3197 - unregister_reply for resent reqs (3063)
3198 - loop back devices mounting and status check on 2.6 (3563)
3199 - fix resource-creation race that can provoke i_size == 0 (3513)
3200 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3201 - more debugging for page-accounting assertion (3746)
3202 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3203 - avoid deadlock after precreation failure (3758)
3204 - fix race and lock order deadlock in orphan handling (3450, 3750)
3205 - add validity checks when grabbing inodes from l_ast_data (3599)
3207 - add /proc/.../recovery_status to obdfilter (3428)
3208 - lightweight CDEBUG infrastructure, debug daemon (3668)
3209 - change default OSC RPC parameters to be better on small clusters
3210 - turn off OST read cache for files smaller than 32MB
3211 - install man pages and include them in rpms (3100)
3212 - add new init script for (un)mounting lustre filesystems (2593)
3213 - run chkconfig in %post for init scripts (3701)
3214 - drop scimac NAL (unmaintained)
3216 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3219 - clean kiobufs before and after use (3485)
3220 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3221 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3222 - change init script to fail more clearly if not run as root (1528)
3223 - allow clients to reconnect during replay (1742)
3224 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3225 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3226 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3227 - fix cond_resched (3554)
3228 - only evict unfinished clients after recovery (3515)
3229 - allow bulk resend, prevent data loss (3570)
3230 - dynamic ptlrpc request buffer allocation (2102)
3231 - don't allow unlinking open directory if it isn't empty (2904)
3232 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3233 - remove extraneous obd dereference causing LASSERT failure (3334)
3234 - don't use get_cycles() when creating temp. files on the mds (3156)
3235 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3236 - handle EEXIST for set-stripe, set proper directory name (3336)
3238 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3239 - fix ksocknal_fmb_callback() error messages (2918)
3241 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3244 - don't copy lvb into (possibly NULL) reply on error (2983)
3245 - don't deref dentry after dput, don't free lvb on error (2922)
3246 - use the kms to determine writeback rpc length (2947)
3247 - increment oti_logcookies when osc is inactive (2948)
3248 - update client's i_blocks count via lvb messages (2543)
3249 - handle intent open/close of special files properly (1557)
3250 - mount MDS with errors=remount-ro, like obdfilter (2009)
3251 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3252 - don't use cancelling-locks' kms values (2947)
3253 - use highest lock extent for kms, not last one (2925)
3254 - don't dereference ERR_PTR() dentry in error handling path (3107)
3255 - fix thread race in portals_debug_dumplog() (3122)
3256 - create lprocfs device entries at setup instead of at attach (1519)
3257 - common AST error handler, don't evict client on completion race (3145)
3258 - zero nameidata in detach_mnt in 2.6 (3118)
3259 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3260 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3261 - initialize RPC timeout timer earlier for 2.6 (3219)
3262 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3263 - print nal/nid for unknown nid (3258)
3264 - additional checks for oscc recovery before doing precreate (3284)
3265 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3266 - don't crash in mdc_close for bad permissions on open (3285)
3267 - zero i_rdev for non-device files (3147)
3268 - clear page->private before handing to FS, better assertion (3119)
3269 - tune the read pipeline (3236)
3270 - fix incorrect decref of invalidated dentry (2350)
3271 - provide read-ahead stats and refine rpc in flight stats (3328)
3272 - don't hold journal transaction open across create RPC (3313)
3273 - update atime on MDS at close time (3265)
3274 - close LDAP connection when recovering to avoid server load (3315)
3275 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3276 - don't leak open file on MDS after open resend (3325)
3277 - serialize filter_precreate and filter_destroy_precreated (3329)
3278 - loop device shouldn't call sync_dev() for nul device (3092)
3279 - clear page cache after eviction (2766)
3280 - resynchronize MDS->OST in background (2824)
3281 - refuse to mount the same filesystem twice on same mountpoint (3394)
3282 - allow llmount to create routes for mounting behind routers (3320)
3283 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3284 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3285 - fix rare problem with rename on htree directories (3417)
3287 - allow default OST striping configuration per directory (1414)
3288 - fix compilation for qswnal for 2.6 kernels (3125)
3289 - increase maximum number of MDS request buffers for large systems
3290 - change liblustreapi to be useful for external progs like lfsck (3098)
3291 - increase local configuration timeout for slow disks (3353)
3292 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3294 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3297 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3298 - don't overwrite extent policy data in reply if lock was blocked (2901)
3299 - drop filter export grants atomically with removal from device (2663)
3300 - del obd_self_export from work_list in class_disconnect_exports (2908)
3301 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3302 - swab reply message in mdc_close, other PPC fixes (2464)
3303 - fix destroying of named logs (2325)
3304 - overwrite old logs when running lconf --write_conf (2264)
3305 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3306 - fix race in target_handle_connect (2898)
3307 - mds_reint_create() should take same inode create lock (2926)
3308 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3309 - don't close files for self_export to avoid uninitialized obd (2936)
3310 - allow MDS with the same name as client node (2939)
3311 - hold dentry reference for closed log files for unlink (2325)
3312 - reserve space for all logs during transactions (2059)
3313 - don't evict page beyond end of stripe extent (2925)
3314 - don't oops on a deleted current working directory (2399)
3315 - handle hard links to targets without a parent properly (2517)
3316 - don't dereference NULL lock when racing during eviction (2867)
3317 - don't grow lock extents when lots of conflicting locks (2919)
3319 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3322 - account for cache space usage on clients to avoid data loss (974)
3323 - lfsck support in lustre kernel code (2349)
3324 - reduce journal credits needed for BRW writes (2370)
3325 - orphan handling to avoid losing space on client/server crashes
3326 - ptlrpcd can be blocked, stopping ALL progress (2477)
3327 - use lock value blocks to assist in proper KMS, faster stat (1021)
3328 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3329 - recovery for initial connections (2355)
3330 - fixes for mds_cleanup_orphans (1934)
3331 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3332 - block all file creations until orphan recovery completes (1901)
3333 - client remove rq_connection from request struct (2423)
3334 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3335 - recovery timer race (2670)
3336 - mdc_close recovey bug (2532)
3337 - ptlrpc cleanup bug (2710)
3338 - mds timeout on local locks (2588)
3339 - namespace lock held during RPCs (2431)
3340 - handle interrupted sync write properly (2503)
3341 - don't try to handle a message that hasn't been replied to (2699)
3342 - client assert failure during cleanup after abort recovery (2701)
3343 - leak mdc device after failed mount (2712)
3344 - ptlrpc_check_set allows timedout requests to complete (2714)
3345 - wait for inflight reqs when ptlrpcd finishes (2710)
3346 - make sure unregistered services are removed from the srv_list
3347 - reset bulk XID's when resending them (caught by 1138 test)
3348 - unregister_bulk after timeout
3349 - fix lconf error (2694)
3350 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3351 - readahead locks pages, leaves pending causing memory pressure (2673)
3352 - increase OST request buffers to 4096 on large machines (2729)
3353 - fix up permission of existing directories in simple_mkdir (2661)
3354 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3355 - don't assign transno to errored transactions (2742)
3356 - don't delete objects on OST if given a bogus objid from MDS (2751)
3357 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3358 - if rq_no_resend, then timeout request after recovery (2432)
3359 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3360 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3361 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3362 - MDS recovery shouldn't delete objects that it has given out (2730)
3363 - if enqueue arrives after completion, don't clobber LVB (2819)
3364 - don't unlock pages twice when trigger_group_io returns error (2814)
3365 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3366 - don't write pages to disk if there was an error (1450)
3367 - don't ping imports that have recovery disabled (2676)
3368 - take buffered bytes into account when balancing socknal conn (2817)
3369 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3370 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3371 - remove little-endian swabbing of llog records (1987)
3372 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3373 - retry reposting request buffers if they fail (1191)
3374 - grow extent at grant time to avoid granting a revoked lock (2809)
3375 - lock revoke doesn't evict page if covered by a second lock (2765)
3376 - disable VM readahead to avoid reading outside lock extents (2805)
3378 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3379 - updated kernel patches for hp-2.4.20 kernel (2681)
3381 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3384 - fix truncated write corruption (2366)
3385 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3387 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3388 - ensure len(uuid) < 37 in lmc (1171)
3389 - fix ia64 OOPS in llog_test (2255)
3390 - zero end of page at obdfilter for partial page writes (2648)
3391 - don't leave stale dentries around after renames (bug 2428)
3392 - fix timeouts when evicting a client with a single lock held (2642)
3393 - set deadline for the initial HELLO message to drain (2634)
3394 - print out dotted-quad IP addresses in the socknal (2302)
3396 - additional debugging for MDS client eviction problem (2443)
3397 - fix mkfsoptions support for osts (2603, 2604)
3399 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3402 - add series for the vanilla 2.6.0 kernel
3403 - add series for the vanilla 2.4.24 kernel
3404 - add series for a cray x86/64 UL kernel drop
3405 - fix xattr patches for the vanilla 2.4.19 series
3407 - generate true UUIDs in lmc (1171)
3408 - have portals stack dumping break in UML (2466)
3409 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3410 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3411 - wake callback waiting threads on client eviction (2460)
3412 - Add --ptldebug and --subsystem to lmc (1719)
3413 - update assertion to allow safe interrupt allocation
3414 - set rq_no_resend for cancel requests (2432)
3415 - recalculate ptlrpcd timeout after resend (2494)
3416 - call vfs_rmdir when removing pending directories (2368)
3417 - fix renaming a file to itself (2429)
3418 - lmc creates a default one-stripe lov (2454)
3419 - expand procfs space to handle large clusters (2326)
3420 - increase UML stack to avoid overflow
3421 - update lconf's list of debug and subsystem masks
3422 - fix lfs find --obd (2510)
3423 - /proc tunable for disabling filter read caching (2591)
3424 - stop rpm packages from altering slapd.conf (2301)
3425 - disable nagle in the socknal under 0conf (2578)
3426 - choose mds inode size based on stripe count (2572)
3427 - fix kernel-source rpm problems (2516)
3429 - add --disable-doc to avoid pdf generation (2421)
3430 - update documentation, tests, type-os, comments
3431 - avoid format warnings on ia64
3432 - remove the TOE NAL
3433 - tiny code cleanups by removing unused fields
3435 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3438 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3439 - remove the most blatant lies from BUILDING file (2371)
3440 - change default debug level to reasonable production setting
3441 - reduce client side cache size to reduce cache flush time
3442 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3443 - make TCP zerocopy and pinger support enabled by default (2476)
3444 - sync writes completed after process exits caused crashes (2319)
3445 - maintain correct mount count on the MDS (2356)
3446 - backout 1557, because 2316 wasn't really fixed
3447 - better file I/O statistics gathering in /proc
3448 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3449 - another kernel patch to fix zero-copy TCP function export
3450 - don't take duplicate lock when processing re-sent getattr (2420)
3451 - lctl uses obd_self_export instead of creating new conn (2353)
3452 - MDS/OST recovery case which requires object creation asserted (2425)
3453 - move lfs from /usr/sbin to /usr/bin in packages
3454 - fix race between mds_client_add and mds_client_free (2417)
3455 - use kmalloc instead of slabs in portals (2430)
3456 - don't create duplicate records when a failover MDS is present (2442)
3457 - remove unnecessary mount age check (2332)
3458 - don't remove directory inodes from locks prematurely (2451)
3459 - don't break if MDS service name is the same as hostname (2103)
3460 - fix races in client write RPC generation when cache full (2482)
3462 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3465 - remove now-unused request->rq_obd (278)
3466 - if an allocation fails, print out how much memory we've used (1933)
3467 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3468 - add the "configurable stack size" patch to most series files (1256)
3469 - ability to write large log records, for 100+ OST configs (2306)
3470 - fix NULL deref when filter_prep fails (2314)
3471 - fix operator precedence error in filter_sync
3472 - dynamic allocation of socknal TX descriptors (2315)
3473 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3474 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3475 - ensure that CWARN messages in Portals always get to the syslog
3476 - __init/__exit are not for prototype decls (ldlm_init/exit)
3477 - x86-64 compile warning fixes
3478 - fix gateway LMC keyword conflict (2318)
3479 - fix MDS lock inversions in getattr/reint paths (1844)
3480 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3481 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3482 - DLM race condition prevented some lock evictions (2328)
3483 - ENOMEM detection and retry on socknal sends (2230)
3484 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3485 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3486 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3487 - better error messages when a client is rejected during recovery (1505)
3488 - avoid cancelling locks which were never granted, after failure (2330)
3489 - fix i_sem/journal inversion in mds_client_add (2333)
3490 - fix truncate/getattr lock cycle deadlock (2334)
3491 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3492 - fix two rare exit paths which could leak an l_lock() ref (2321)
3493 - fencepost error in MDS/OST orphan recovery (2226)
3494 - make log record alignment 8 bytes (1988)
3495 - lstripe now fails when requested offset > ost_count (2237)
3496 - ensure that all kernel series have a complete list.h (1607)
3497 - fix crashes in special-file operations (2316)
3498 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3500 - allow configurable automake binary, for testing new versions
3501 - small update to the lfs documentation
3503 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3505 * fix negative export reference count in fsfilt_sync (2312)
3507 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3508 * release candidate 0.9.1
3510 - orphans are moved into the PENDING directory for possible recovery
3511 - replayed opens now open by fid for orphan/rename safety (1042)
3512 - last close of an orphan inode generates a transno (683)
3513 - chdir() and mount() now pin the directory entry (1020)
3514 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3515 - discard very old requests without processing them (1502)
3516 - remove some common, well-understood CERRORs (1505)
3517 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3518 - clear "grant" flags in OST replies until OST grant code lands (1644)
3519 - fix read performance by not clobbering i_blksize on client (1598)
3520 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3521 - make LRU size a /proc tunable, clears locks when reduced (707)
3522 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3523 - server locks take references on exports now (1558)
3524 - build fixes for 2.4.20-rh trees (1663)
3525 - return an error from lov_create if all OSCs are inactive (1751)
3526 - fix import levels when a reconnect happens without a timeout (1597)
3527 - exit early from mds_open if we get a lookup error (1749)
3528 - partial page read at EOF wouldn't wait for disk before sending (1642)
3529 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3530 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3531 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3532 - write/truncate lock inversion (1639)
3533 - Don't auto-load obdclass, portals modules during cleanup (1495)
3534 - fix timestamps from jumping to "now" (1763)
3535 - extra journal assertions (1648)
3536 - add an extra multiunlink test (1771)
3537 - fix read_record/write_record API (1776)
3538 - fix leak of offset_extent, possible incorrect i_size later (1772)
3539 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3540 - Debugging for the kqswnal_get_idle_tx problems (1820)
3541 - Allow recovery to be attempted multiple times (1536)
3542 - Write out MDS last_rcvd file after it is first created (1600)
3543 - Fix tx_descriptor leak in failed transmit situations (1827)
3544 - ext3 journaling fixes for assertion failure after IO error (1871)
3545 - class_export_put() on freed export after completion AST error (1896)
3546 - Fix revalidate looping in VFS (1322)
3547 - Don't access a freed export during MDS_REINT timeout (1521)
3548 - Add open-unlink recovery support on the MDS (1673,1764)
3549 - Return an error if no MDS data was read from last_rcvd (1946)
3550 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3551 - Don't setup a disk device that doesn't match exported UUID (317)
3552 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3553 - avoid committing NULL handle in force close
3554 - local.sh is now a one-stripe LOV configuration
3555 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3556 - fix ext3 htree duplicate directory entry corruption (1516)
3557 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3558 - update to new LOV EA format (2097)
3559 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3561 2003-06-15 Phil Schwan <phil@clusterfs.com>
3564 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3565 - per-import/export recovery handling (958, 931, 959)
3566 - multiple last-rcvd slots, for serving multiple FSes (949)
3567 - connections are again shared between multiple imp/exports (963, 964)
3568 - "umount -f" would hang if any requests needed to be sent (393, 978)
3569 - avoid pinning large req buffer by copying for queued messages (989)
3570 - add "uuid" to "lctl device" command to help upcalls (991)
3571 - "open" RPCs with transnos would confuse recovery counters (1037)
3572 - do proper endian conversion of all wire messages (288, 340, 891)
3573 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3574 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3575 - fix ptlrpc_connection leak in target_handle_connect (1174)
3576 - fix import refcounting bug in OST and MDS cleanup (1134)
3577 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3578 - fix very unlikely obd_types race condition (501)
3579 - remove osc_open hack for echo_client (1187)
3580 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3581 - a failure in read_inode2 leads to deadlock (1139)
3582 - cancel ack-locks as soon as transaction is committed (1072)
3583 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3584 - make sure to commitrw after any preprw to avoid deadlock (1162)
3585 - failing to execute a file in a lustre FS would lock inode (1203)
3586 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3587 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3588 - obd_uuid2tgt would walk off the end of the list (1255)
3589 - on IA64 the osc would give portals incorrect bulk size (1258)
3590 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3591 - fix lock inversion caused by new llite matching code (1282)
3592 - limit the number of dirty pages on a client to 10MB (1286)
3593 - timed out locks were not being corrected cancelled (1289)
3594 - fix O_DIRECT above 4GB on IA-32 (1292)
3595 * major user-visible changes
3596 - fail out/fail over policy now controlled by the upcall (993)
3598 - add OBD_PING to check server availability and failure (954)
3599 - lustre messages are now sent in sending host order (288, 340, 891)
3600 - add eadatalen to MDS getattr reply (340)
3601 - OST read replies may contain second buffer, with per-page status (593)
3603 2003-03-11 Phil Schwan <phil@clusterfs.com>
3606 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3607 - failed open()s could cause deadlock; fixed (867, 869)
3608 - stop cancelling OST locks when files are closed (481)
3609 - overlapping XID spaces caused network corruption (851, 853)
3610 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3611 - setattr_raw added, to do single-RPC, server-side setattrs
3612 - lmc/lconf syntax change for OST UUIDs
3613 - fix crashy race condition between ptlrpc_free_req and osc_close
3614 - don't use request in mdc_enqueue if we hit a timeout (889)
3615 - don't set the inode i_size for regular files from the MDS (896)
3616 - handle out of order completion AST (842)
3617 - don't LBUG if a lock request times out after receiving AST (913)
3618 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3619 - if a bad lock AST arrives, send an error instead of dropping entirely
3620 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3621 - fix leak in bulk IO when only partially completed (899, 900, 926)
3622 - fix O_DIRECT for ia64 (55)
3623 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3624 - C-z after timeout could hang a process forever; fixed (977)
3626 - client-side I/O cache (678, 924, 929, 941, 970)
3628 - READPAGE and SETATTRs which don't take server-side locks get
3631 2003-02-11 Phil Schwan <phil@clusterfs.com>
3634 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3635 - Fix fsfilt_extN_readpage() to read a full page of directory
3636 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3637 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3638 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3639 - Most intent operations were removed, in favour of a new RPC mode
3640 that does a single RPC to the server and bypasses most of the VFS
3641 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3642 - Aggressively cancel local locks on DLM servers
3643 - mds_reint_unlink sends EA to the client if it's the last nlink.
3644 client uses that EA to unlink OST objects.
3645 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3646 - recursive symlinks were fixed (439)
3647 - fixed NULL deref in DEBUG_REQ
3648 - filter_update_lastobjid no longer calls sync, which annoyed extN
3649 - fixed multi-client small-writes to a single file problem (445)
3650 - fixed mtime updates during file writes (607)
3651 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3652 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3653 - fixed Linux OST size reporting problem (444, 656)
3654 - OST now updates object mtime with writes or setattr (607, 619)
3655 - client verifies file size before zeroing page past EOF (445)
3656 - OST now writes last allocated objid to disk with allocation (108)
3657 - LOV on echo now works (409)
3659 - mds_reint_unlink sends a new buffer, with the EA included. this
3660 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3661 set if a regular file was being unlinked, and it was the last link
3662 - use PtlGet from the target for bulk writes (315)
3663 - OST now updates object mtime with writes or setattr (607, 619)
3664 - LDLM now has a grant-time callback to revalidate locked items, if
3666 - Many MDS operations were reorganized to combat race conditions
3668 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3669 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3671 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3674 - Fully reactivate OST imports after reconnection (512, others)
3675 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3676 - More graceful error handling for truncating on dead OST (515)
3677 - Don't error out unless we're actually accessing dead stripes (474)
3678 - Fix garbage sizes when stripes are missing (410)
3679 - LRU counters were broken, causing constant lock purge (433, 432)
3680 - garbage on read from stripes with failed OSTs (441)
3681 - mark OSCs as active before reconnecting during recovery (438)
3682 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3683 - lfind did not preserve OST order in output (443)
3684 - symlinks cause hung clients, incorrect data (439)
3685 - stop dereferencing request after dropping refcount (457)
3686 - don't LASSERT(spin_is_locked) on non-SMP (455)
3687 - fixes for many rename() bugs
3688 - fstat didn't correctly synchronize attributes (399)
3689 - server must handle lock cancellation during blocking AST prep (487)
3690 - bulk descriptors were free()d too soon (511)
3691 - fix paths in lconf, which would load incorrect modules (451, 507)
3692 - fix confusing lconf 'host not found' error message (386)
3693 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3694 - fix race condition in mdc_blocking_ast() for inode access (526)
3695 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3696 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3697 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3698 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3699 - fix lstripe in several ways (kernel side) (527)
3700 - fix request leak in ldlm_cli_enqueue (262)
3701 - incorrect OSC was marked inactive after OST failure
3702 - call mds_fs_cleanup before unmounting filesystem (524)
3703 - fix races between taking ns_lock and ldlm_lock_change_resource
3704 - fix races updating LOV export open file list
3705 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3706 - fix recovery NULL deref in ldlm_cli_cancel_unused
3707 - fix some DLM races by using new hash table for lock handles (419)
3708 - permit the client to specify desired inodes, at replay
3709 - duplicate requests when we queue them for replay reintegration
3710 - fix last_rcvd offset calculation
3711 - sync after each recovered transaction, so we always make progress
3712 - never, not always, ERESTART requests without transnos
3713 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3715 - skip replay if the MDS says that the client is already connected
3716 - don't check for a recovery-enabled export to match lctl's UUID
3717 - don't INC_USE_COUNT for phantom exports
3718 - don't crash when cleaning up phantom exports (567)
3719 - don't double-finish or set replay data for errored mdc_open requests
3720 - abort requests when they time out, so we don't get old replies
3721 - send/receive replies for AST messages again
3722 - if the client says that it doesn't have the lock, cancel it on the
3724 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3725 mark it as destroyed, it will all work out when decref is called
3726 - fix module use counts (22, 581)
3728 - ASTs now expect a reply (server cancels lock on error reply)
3730 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3733 - fix many simultaneous client startup (392)
3734 - fix dentry->d_it clobbering
3735 - credentials weren't being shipped for readdir/getattr operations
3736 - remove invalid assertions triggered during some concurrent MD
3738 - proper Lustre versions added (336, 389)
3739 - fix memory leak for create error case (398)
3740 - fix LOV locking bug that would get cli/srv out of sync
3741 - fix echo client over LOV (409)
3742 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3743 - fix double-O_EXCL intent crash (424)
3744 - avoid sending multiple lock CANCELs (352)
3746 - MDS can do multi-client recovery (modulo bugs in new code)
3748 - many updates, edits, cleanups
3750 2002-11-18 Phil Schwan <phil@clusterfs.com>
3753 - fix null d_it dereference (346)
3754 - fix full OST/dbench hang (333)
3755 - fix permission problem with file removal (286)
3756 - fix removal of OSCs from LOV when they fail
3757 - fix NULL deref during bulk timeout (214)
3758 - fix problems related to multiple filesystems on one MDS (241)
3759 - fixed serious subtle metadata locking bugs
3760 - free locks on clients when inodes are removed due to memory
3762 - fix inode pointer in lock data (285)
3763 - partial support for multiple MDS on a single host (241)
3764 - data locks weren't cancelled at clear_inode time (290, 311)
3765 - intent locks could lead to unbounded lock growth (205)
3766 - added a maximum lock count, an LRU list, and a flusher
3767 - fix multiple rename (365)
3768 - properly abstracted the echo client
3769 - OSC locked 1 byte too many; fixed
3770 - rewrote brw callback code:
3771 - fixed recovery bugs related to LOVs (306)
3772 - fixed too-many-pages-in-one-write crash (191)
3773 - fixed (again) crash in sync_io_timeout (214)
3774 - probably fixed callback-related race (385)
3776 - Add capability to MDS protocol
3777 - LDLM cancellations and callbacks on different portals
3779 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3782 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3783 - defer open object destruction to close time (601981, 138)
3784 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3785 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3786 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3787 - don't call obd_setattr in ll_file_release for destroyed objects
3788 * protocol change to lustre_msg: move |version| and add |flags|
3789 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3790 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3791 * added replay of create, unlink, link and rename operations during
3792 MDS failover; recovery should be much more robust now
3793 * remove failed OSCs from LOVs (only lov_create uses this so far)
3794 * the lustre-HOWTO was brought (more) up to date (582544)
3796 2002-10-23 Phil Schwan <phil@clusterfs.com>
3799 - in-use dentries weren't being reused properly (617851)
3800 - prevent multiple LDLM setup (599178)
3801 - fix LOV size calculations for truncate (617853)
3802 - fix client handling of MDS intent errors (POSIX)
3803 - fix permission bug in lovstripe.c test (624321)
3804 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3805 - truncate past end of file could corrupt data
3806 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3807 - a race in recovery could return ETIMEDOUT to apps (623947)
3808 - building outside the source directory was fixed
3809 * the lustre-HOWTO was brought (more) up to date (582544)
3810 * major progress was made on recovery functionality
3812 2002-10-10 Phil Schwan <phil@clusterfs.com>
3815 - recovery deadlock fix
3816 - rm -rf causes LBUG fix (617817)
3817 - file open by multiple tasks fix (618962)
3818 - directory permissions bugs (602707 and 620007)
3819 - journal_stop fixed with locking (611313)
3820 - O_APPEND failures resolved (618273, perhaps 614459)
3821 - lconf PATH fix (619770)
3822 - IA64 build fix (621450)
3823 - RPC buffer sizes scale with amount of memory
3825 2002-10-01 Phil Schwan <phil@clusterfs.com>
3828 - locks would be cancelled without throwing away data pages,
3829 resulting in inconsistent data (605627)
3830 - inode attributes were not always being refreshed (605627, 612449)
3831 - lconf now continues to cleanup after lctl reports an error
3832 - MDS now enforces user permissions (602707)
3833 - lprocfs cleanup fixed, but not yet enabled (614157)
3834 - fixed infinite server hang, should a client not respond to an AST
3835 - avoid going into recovery if user calls readlink() with a buffer
3836 that's too small (613941)
3837 - AST RPCs no longer require replies (614867) -- this may be changed
3838 - don't crash server if client sends an IOV that's too big (611336)
3839 - fixed lock conversion deadlock (611892)
3840 - fixed the following of symlinks (614622)
3841 * recovery: the server can remove locks from a client that dies, other
3842 clients can make progress
3843 * more extN patch fixes
3844 * compile-time configurable ptlrpc buffer allocations
3846 - collaborative read cache document
3847 - Lustre Lite Performance CDR document-in-progress
3849 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3852 - fix typo in patch-2.4.18
3854 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3857 - clear ptlrpc request each time in handle_incoming_request()
3858 - unlink of files now destroys the object on the OST
3860 2002-09-19 Peter Braam <braam@clusterfs.com>
3862 * add hard link support
3863 * change obdfile creation method
3864 * kernel patch changed
3866 2002-09-19 Peter Braam <braam@clusterfs.com>
3869 - stack overflow bug in extN fixed
3871 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3873 * documentation updates
3874 - add man pages for config tools
3875 - update tests/README to describe testing with new config tools
3876 - finish metadata API descriptions
3877 * bug fixes and cleanups
3878 - statfs workaround for 16TB limit
3879 - LOV stripe allocation improved, can stripe on subset of OSTs
3880 - LOV file size/IO offset was wrong for files > 4GB in size
3881 - object EA data was being dropped, caused files to be unreadable
3882 - memory overflow with non-LOV OST caused memory corruption
3883 - fixed regression tests to work with new config tools, obdfilter
3884 - fixed bug when directory size became larger than 1 block
3885 - fixed bug (for single client case) when PWD was deleted
3886 - invalidate local directory pages when doing intent-based ops
3887 - avoid LDLM oops when lock callback contained bad data
3889 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3891 * documentation updates
3892 * bug fixes and cleanups
3893 - configuration tools
3896 - 64-bit compile warnings
3897 - 64-bit internal statfs data
3899 * test_brw on persistent OST devices
3901 * lprocfs (disabled)
3903 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3905 * documentation updates
3906 * bug fixes and cleanups
3907 * configuration tools
3909 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3912 * many small fixes to 0_5_4
3913 * io/network handling
3914 * thinkos in MDS operations
3916 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3919 * crucial basic fixes to 0.5.3
3920 * IOR, Iozone work over Elan
3923 2002-08-07 Phil Schwan <phil@clusterfs.com>
3924 * version 0_5_3, our first alpha
3925 * we use the new Portals iovs
3926 * documentation updates
3927 * bug fixes and cleanups
3928 * small changes in the DLM wire protocol
3930 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3931 * version 0_5_1 with some initial stability,
3932 * locking on MD and file I/O.
3933 * documentation updates
3934 * several bug fixes since 0.5.0
3935 * small changes in wire protocol
3937 2002-07-18 Phil Schwan <phil@clusterfs.com>
3939 * delivered as Lustre Light Alpha
3940 * fixed a crash after handling invalid MDS requests
3941 * fixed directory pages for architectures with non-4k pages sizes
3943 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3944 * release version v0_4_4
3945 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3947 * Fixes a number of interruption problems with OST operations.
3948 * Update documentation for portals header changes
3949 * Move all wire protocol structs/defines to lustre_idl.h
3950 * Fixes symlink length bug.
3951 * Add tcpdump to repository.
3953 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3954 * release version v0_4_3
3955 * Fixes statfs for inodes on extN.
3956 * Fixes bug in runtests which would delete /etc/hosts.
3957 * Use 64-bit object IDs wherever possible (not into VFS though)
3958 Remove ost_get_info, which is unused by lustre, and out of date.
3960 2002-07-03 Peter Braam <braam@clusterfs.com>
3961 * release version v0_4_2 Fixes a lookup error (type not passed)
3962 * move forward to head of Portals
3963 * move forward to latest Lustre kernel
3965 2002-06-25 Peter Braam <braam@clusterfs.com>
3966 * release version v0_4_1. Hopefully stable on single node use.