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
220 Description: osts not allocated evenly to files
221 Details : change the condition to increase offset_idx
223 --------------------------------------------------------------------------------
225 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
227 * Support for kernels:
228 2.6.5-7.283 (SLES 9),
229 2.6.9-55.EL (RHEL 4),
230 2.6.16.46-0.14 (SLES 10),
231 2.6.18.8 vanilla (kernel.org)
232 * Client support for unpatched kernels:
233 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
234 2.6.16 - 2.6.22 vanilla (kernel.org)
235 * Due to recently discovered recovery problems, we do not recommend
236 using patchless RHEL 4 clients with this or any earlier release.
237 * Recommended e2fsprogs version: 1.39.cfs8
238 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
239 * Starting with this release, the ldiskfs backing filesystem required
240 by Lustre is now in its own package, lustre-ldiskfs. This package
241 should be installed. It is versioned separately from Lustre and
242 may be released separately in future.
246 Description: block reactivating mgc import until all deactivates complete
247 Details : Fix race when failing back MDT/MGS to itself (testing)
249 Severity : enhancement
251 Description: add optional extra BUILD_VERSION info
252 Details : add a new environment variable (namely LUSTRE_VERS) which allows
253 to override the lustre version.
256 Frequency : 2.6.18 servers only
258 Description: ll_kern_mount() doesn't release the module reference
259 Details : The ldiskfs module reference count never drops down to 0
260 because ll_kern_mount() doesn't release the module reference.
265 Description: server LBUG when using old ost_num_threads parameter
266 Details : Accept the old ost_num_threads parameter but warn that it
267 is deprecated, and fix an off-by-one error that caused an LBUG.
272 Description: Transient SCSI error results in persistent IO issue
273 Details : iobuf->dr_error is not reinitialized to 0 between two
277 Frequency : sometimes when underlying device returns I/O errors
279 Description: OSTs not going read-only during write failures
280 Details : OSTs are not remounted read-only when the journal commit threads
281 get I/O errors because fsfilt_ext3 calls journal_start/stop()
282 instead of the ext3 wrappers.
286 Description: poor connect scaling with increasing client count
287 Details : Don't run filter_grant_sanity_check for more than 100 exports
288 to improve scaling for large numbers of clients.
291 Frequency : SLES10 only
293 Description: sanity-quota.sh quotacheck failed: rc = -22
294 Details : Quotas cannot be enabled on SLES10.
297 Frequency : liblustre clients only
299 Description: getdirentries does not give error when run on compute nodes
300 Details : getdirentries does not fail when the size specified as an argument
301 is too small to contain at least one entry
303 Severity : enhancement
305 Description: Add LNET router traceability for debug purposes
306 Details : If a checksum failure occurs with a router as part of the
307 IO path, the NID of the last router that forwarded the bulk data
308 is printed so it can be identified.
313 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
314 Details : Due to a race condition, liblustre clients were occasionally
317 Severity : enhancement
319 Description: lfs setstripe use optional parameters instead of postional
322 Severity : enhancement
324 Description: Nanosecond timestamp support for ldiskfs
325 Details : The on-disk ldiskfs filesystem has added support for nanosecond
326 resolution timestamps. There is not yet support for this at
327 the Lustre filesystem level.
330 Frequency : during server recovery
332 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
333 Details : request with rq_no_resend flag not awake l_wait_event if they get a
337 Frequency : nfs export on patchless client
339 Description: connectathon hang when test nfs export over patchless client
340 Details : Disconnected dentry cannot be found with lookup, so we do not need
341 to unhash it or make it invalid
344 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
345 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
346 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
351 Description: open req refcounting wrong on reconnect
352 Details : If reconnect happened between getting open reply from server and
353 call to mdc_set_replay_data in ll_file_open, we will schedule
354 replay for unreferenced request that we are about to free.
355 Subsequent close will crash in variety of ways.
356 Check that request is still eligible for replay in
357 mdc_set_replay_data().
362 Description: disable writes to filesystem when reading health_check file
363 Details : the default for reading the health_check proc file has changed
364 to NOT do a journal transaction and write to disk, because this
365 can cause reads of the /proc file to hang and block HA state
366 checking on a healthy but otherwise heavily loaded system. It
367 is possible to return to the previous behaviour during configure
368 with --enable-health-write.
370 Severity : enhancement
372 Description: 64-bit inode version
373 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
374 to the inode. This will be required for version-based recovery.
379 Description: MDS fails to start if a duplicate client export is detected
380 Details : in some rare cases it was possible for a client to connect to
381 an MDS multiple times. Upon recovery the MDS would detect this
382 and fail during startup. Handle this more gracefully.
384 Severity : enhancement
386 Description: Add -o localflock option to simulate old noflock
388 Details : This will achieve local-only flock/fcntl locks
394 Description: log_commit_thread vs filter_destroy race leads to crash
395 Details : Take import reference before releasing llog record semaphore
400 Description: Wrong request locking in request set processing
401 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
402 imp_delayed_list, in this place should be used imp_lock.
405 Frequency : when reconnection
407 Description: Grant Leak when osc reconnect to OST
408 Details : When osc reconnect ost, OST(filter) should check whether it
409 should grant more space to client by comparing fed_grant and
410 cl_avail_grant, and return the granted space to client instead
411 of "new granted" space, because client will call osc_init_grant
412 to update the client grant space info.
415 Frequency : when client reconnect to OST
417 Description: Grant Leak when osc do resend and replay bulk write
418 Details : When osc reconnect to OST, OST(filter)should clear grant info of
419 bulk write request, because the grant info while be sync between
420 OSC and OST when reconnect, and we should ignore the grant info
421 these of resend/replay write req.
426 Description: Grant space more than avaiable left space sometimes.
427 Details : When then OST is about to be full, if two bulk writing from
428 different clients came to OST. Accord the avaliable space of the
429 OST, the first req should be permitted, and the second one
430 should be denied by ENOSPC. But if the seconde arrived before
431 the first one is commited. The OST might wrongly permit second
432 writing, which will cause grant space > avaiable space.
435 Frequency : when client is evicted
437 Description: Grant might be wrongly erased when osc is evicted by OST
438 Details : when the import is evicted by server, it will fork another
439 thread ptlrpc_invalidate_import_thread to invalidate the
440 import, where the grant will be set to 0. While the original
441 thread will update the grant it got when connecting. So if
442 the former happened latter, the grant will be wrongly errased
443 because of this race.
448 Description: Checking Stale with correct fid
449 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
450 to check whether it is stale, because sometimes, we want the
451 enqueue happened anyway, and op_data.fid2 will not be initialized.
453 Severity : enhancement
455 Description: update patchless client
456 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
459 Frequency : only with 2.4 kernel
461 Description: random memory corruption
462 Details : size of struct ll_inode_info is to big for union inode.u and this
463 can be cause of random memory corruption.
468 Description: Memory leak in recovery
469 Details : Lov_mds_md was not free in an error handler in mds_create_object.
470 It should also check obd_fail before fsfilt_start, otherwise if
471 fsfilt_start return -EROFS,(failover mds during mds recovery).
472 then the req will return with repmsg->transno = 0 and rc = EROFS.
473 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
474 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
475 be freed no matter whether fsfilt_commit success or not.
478 Frequency : only with huge count clients
480 Description: Prevents from taking the superblock lock in llap_from_page for
482 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
483 for llap_from_page call in ll_removepage prevents from taking the
484 superblock lock for a soon died page.
489 Description: Not check open intent error before release open handle
490 Details : in some rare cases, the open intent error is not checked before
491 release open handle, which may cause
492 ASSERTION(open_req->rq_transno != 0), because it tries to release
493 the failed open handle.
498 Description: Set cat log bitmap only after create log success.
499 Details : in some rare cases, the cat log bitmap is set too early. and it
500 should be set only after create log success.
504 Description: Accessing a block bevice can re-enable I/O when Lustre is
505 tearing down a device.
506 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
511 Description: service threads may hog cpus when there are a lot of requests
513 Details : Insert cond_resched to give other threads a chance to use some of
519 Description: the cat log was not initialized in recovery
520 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
521 unlink log on mds will not be initialized until mds post
522 recovery. And also in mds post recovery, the unlink log will
523 initialization will be done asynchronausly, so there will be race
524 between add unlink log and unlink log initialization.
528 Description: brw_stats were being printed incorrectly
529 Details : brw_stats were being printed as log2 but all of them were not
530 recorded as log2. Also remove some code duplication arising from
531 filter_tally_{read,write}.
535 Frequency : rare, only in recovery.
536 Description: ASSERTION(req->rq_type != LI_POISON) failed
537 Details : imp_lock should be held while iterating over imp_sending_list for
538 prevent destroy request after get timeout in ptlrpc_queue_wait.
542 Description: replay-single.sh test 52 fails
543 Details : A lock's skiplist need to be cleanup when it being unlinked
544 from its resource list.
548 Description: Short directio read returns full requested size rather than
550 Details : Direct I/O operations should return actual amount of bytes
551 transferred rather than requested size.
553 Severity : enhancement
555 Description: metadata RPC reduction (e.g. for rm performance)
556 Details : decrease the amount of synchronous RPC between clients and servers
557 by canceling conflicing lock before the operation on the client side
558 and packing thier handles into the main operation RPC to server.
560 Severity : enhancement
562 Description: Async OSC create to avoid the blocking unnecessarily.
563 Details : If a OST has no remain object, system will block on the creating
564 when need to create a new object on this OST. Now, ways use
565 pre-created objects when available, instead of blocking on an
566 empty osc while others are not empty. If we must block, we block
567 for the shortest possible period of time.
569 --------------------------------------------------------------------------------
571 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
576 Frequency : on some architectures
578 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
579 Details : Uninitialized flags sometimes cause configuration commands to
583 Frequency : patchless clients only
585 Description: missing __iget() symbol export
586 Details : The __iget() symbol export is missing. To avoid the need for
587 this on patchless clients the deathrow inode reaper is turned
588 off, and we depend on the VM to clean up old inodes. This
589 dependency was during via the fix for bug 12181.
591 --------------------------------------------------------------------------------
593 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
595 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
596 INTEROPERATE with older versions automatically. In many cases a
597 special upgrade step is needed. Please read the
598 user documentation before upgrading any part of a live system.
599 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
600 version will not interoperate with 1.6 betas before beta5 (1.5.95).
601 * WARNING: Lustre configuration and startup changes are required with
602 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
608 Severity : enhancement
610 Description: MountConf
611 Details : Lustre configuration is now managed via mkfs and mount
612 commands instead of lmc and lconf. New obd types (MGS, MGC)
613 are added for dynamic configuration management. See
614 https://mail.clusterfs.com/wikis/lustre/MountConf for
617 Severity : enhancement
619 Description: dynamic OST addition
620 Details : OSTs can now be added to a live filesystem
622 Severity : enhancement
624 Description: startup order invariance
625 Details : MDTs and OSTs can be started in any order. Clients only
626 require the MDT to complete startup.
628 Severity : enhancement
630 Description: parallel, asynchronous orphan cleanup
631 Details : orphan cleanup is now performed in separate threads for each
632 OST, allowing parallel non-blocking operation.
634 Severity : enhancement
636 Description: optimized stripe assignment
637 Details : stripe assignments are now made based on ost space available,
638 ost previous usage, and OSS previous usage, in order to try
639 to optimize storage space and networking resources.
641 Severity : enhancement
643 Description: Permanently set tunables
644 Details : All writable /proc/fs/lustre tunables can now be permanently
645 set on a per-server basis, at mkfs time or on a live
648 Severity : enhancement
650 Description: Lustre message v2
651 Details : Add lustre message format v2.
653 Severity : enhancement
655 Description: client OST exclusion list
656 Details : Clients can be started with a list of OSTs that should be
657 declared "inactive" for known non-responsive OSTs.
661 Description: SPEC SFS validation failure on NFS v2 over lustre.
662 Details : Changes the blocksize for regular files to be 2x RPC size,
663 and not depend on stripe size.
665 Severity : enhancement
667 Description: Multiple MD RPCs in flight.
668 Details : Further unserialise some read-only MDS RPCs - learn about intents.
669 To avoid overly-overloading MDS, introduce a limit on number of
670 MDS RPCs in flight for a single client and add /proc controls
671 to adjust this limit.
673 Severity : enhancement
675 Description: client read/write statistics
676 Details : Add client read/write call usage stats for performance
677 analysis of user processes.
678 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
679 file access. extents_stats shows chunk size distribution.
680 extents_stats_per_process show chunk size distribution per
683 Severity : enhancement
685 Description: mds statistics
686 Details : Add detailed mds operations statistics in
687 /proc/fs/lustre/mds/*/stats.
691 Description: Failure of copying files with lustre special EAs.
692 Details : Client side always return success for setxattr call for lustre
693 special xattr (currently only "trusted.lov").
698 Description: Refcount LNET uuids
699 Details : The global LNET uuid list grew linearly with every startup;
700 refcount repeated list entries instead of always adding to
703 Severity : enhancement
705 Description: Dynamic service threads
706 Details : Within a small range, start extra service threads
707 automatically when the request queue builds up.
710 Frequency : mixed-endian client/server environments
712 Description: mixed-endian crashes
713 Details : The new msg_v2 system had some failures in mixed-endian
716 Severity : enhancement
718 Description: Easy OST removal
719 Details : OSTs can be permanently deactivated with e.g. 'lctl
720 conf_param lustre-OST0001.osc.active=0'
722 Severity : enhancement
724 Description: MGS proc entries
725 Details : Added basic proc entries for the MGS showing what filesystems
728 Severity : enhancement
730 Description: provide MGS failover
731 Details : Added config lock reacquisition after MGS server failover.
733 Severity : enhancement
735 Description: add Linux 2.4 support
736 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
740 Description: a large application tries to do I/O to the same resource and dies
742 Details : Check the req->rq_arrival time after the call to
743 ost_brw_lock_get(), but before we do anything about
744 processing it & sending the BULK transfer request. This
745 should help move old stale pending locks off the queue as
746 quickly as obd_timeout.
749 Frequency : when an incorrect nid is specified during startup
751 Description: ptlrpc connect to non-existant node causes kernel crash
752 Details : LNET can't be re-entered from an event callback, which
753 happened when we expire a message after the export has been
754 cleaned up. Instead, hand the zombie cleanup off to another
757 Severity : enhancement
759 Description: plain/inodebits lock performance improvement
760 Details : Grouping plain/inodebits in granted list by their request modes
761 and bits policy, thus improving the performance of search through
765 Frequency : only if OST filesystem is corrupted
767 Description: client incorrectly hits assertion in ptlrpc_replay_req()
768 Details : for a short time RPCs with bulk IO are in the replay list,
769 but replay of bulk IOs is unimplemented. If the OST filesystem
770 is corrupted due to disk cache incoherency and then replay is
771 started it is possible to trip an assertion. Avoid putting
772 committed RPCs into the replay list at all to avoid this issue.
775 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
777 Description: System hang on startup
778 Details : This bug allowed the liblustre (e.g. catamount) client to
779 return to the app before handling all startup RPCs. This
780 could leave the node unresponsive to lustre network traffic
781 and manifested as a server ptllnd timeout.
783 Severity : enhancement
785 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
786 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
787 boolean to control whether to print peer debug info when a
788 client's RPC times out.
791 Frequency : only for kernels with patches from Lustre below 1.4.3
793 Description: Remove old rdonly API
794 Details : Remove old rdonly API which unused from at least lustre 1.4.3
797 Frequency : only for devices with external journals
799 Description: Set external device read-only also
800 Details : During a commanded failover stop, we set the disk device
801 read-only while the server shuts down. We now also set any
802 external journal device read-only at the same time.
805 Frequency : when upgrading from 1.4 while trying to change parameters
807 Description: The wrong (new) MDC name was used when setting parameters for
808 upgraded MDT's. Also allows changing of OSC (and MDC)
809 parameters if --writeconf is specified at tunefs upgrade time.
812 Frequency : when setting specific ost indicies
814 Description: QOS code breaks on skipped indicies
815 Details : Add checks for missing OST indicies in the QOS code, so OSTs
816 created with --index need not be sequential.
818 Severity : enhancement
820 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
821 Details : The uninit_groups feature works in conjunction with the kernel
822 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
823 up the pass1 processing of e2fsck. This is a read-only feature
824 in ldiskfs2 only, so older kernels and current ldiskfs cannot
825 mount filesystems that have had this feature enabled.
827 Severity : enhancement
829 Description: Improve multi-block allocation algorithm to avoid fragmentation
830 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
831 allocation locality and avoid filesystem fragmentation.
833 ------------------------------------------------------------------------------
835 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
837 * Support for kernels:
838 2.4.21-47.0.1.EL (RHEL 3)
840 2.6.9-42.0.10.EL (RHEL 4)
841 2.6.12.6 vanilla (kernel.org)
842 2.6.16.27-0.9 (SLES 10)
843 * Recommended e2fsprogs version: 1.39.cfs5
845 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
849 Frequency : occasional, depends on client load and configuration
850 Bugzilla : 12181, 12203
851 Description: data loss for recently-modified files
853 Details : In some cases it is possible that recently written or created
854 files may not be written to disk in a timely manner (this should
855 normally be within 30s unless client IO load is very high).
856 The problem appears as zero-length files or files that are a
857 multiple of 1MB in size after a client crash or client eviction
858 that are missing data at the end of the file.
860 This problem is more likely to be hit on clients where files are
861 repeatedly created and unlinked in the same directory, clients
862 have a large amount of RAM, have many CPUs, the filesystem has
863 many OSTs, the clients are rebooted frequently, and/or the files
864 are not accessed by other nodes after being written.
866 The presence of the problem can be detected by looking at
867 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
868 smaller than the second (nr_unused) then dirty files will not
869 be flushed automatically to disk. "sync; sleep 10" should be
870 run several times on the node before unmounting it to update
871 Lustre (this is also safe to run on nodes without this problem).
873 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
874 that can cause this same problem, so customers using that kernel
875 also need to update the kernel in addition to Lustre. In order
876 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
878 It is normal that files written just before a client crash (less
879 than 30s) may not yet have been flushed to disk, even for local
883 Frequency : frequent on thin XT3 nodes
885 Description: UUID collision on thin XT3 Linux nodes
886 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
887 randomly, since we relied on an insufficiently-seeded PRNG.
892 Description: OSS hangs after "All ost request buffers busy"
893 Details : A deadlock between quota and journal operations caused OSS
894 hangs after printing "All ost request buffers busy."
897 Frequency : always on liblustre builds
899 Description: Cleanup compiler warnings on liblustre
902 Frequency : always on liblustre builds on XT3
904 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
908 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
909 > 32000 subdirectories
910 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
911 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
912 subdirectory count crosses 32000. This will aid e2fsck to
913 correctly handle more than 32000 subdirectories.
916 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
918 Description: System hang on startup
919 Details : This bug allowed the liblustre (e.g. catamount) client to
920 return to the app before handling all startup RPCs. This
921 could leave the node unresponsive to lustre network traffic
922 and manifested as a server ptllnd timeout.
924 Severity : enhancement
926 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
927 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
928 boolean to control whether to print peer debug info when a
929 client's RPC times out.
934 Description: make O_SYNC working on 2.6 kernels
935 Details : 2.6 kernels use different method for mark pages for write,
936 so need add a code to lustre for O_SYNC work.
941 Description: Failure to close file and release space on NFS
942 Details : Put inode details into lock acquired in ll_intent_file_open.
943 Use mdc_intent_lock in ll_intent_open to properly
944 detect all kind of errors unhandled by mdc_enqueue.
949 Description: proc file read during shutdown sometimes raced obd removal,
951 Details : Add lock to prevent obd access after proc file removal.
954 Frequency : Only for files larger than 4GB on 32-bit clients.
956 Description: improperly doing page alignment of locks
957 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
958 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
963 Description: RPCs being resent when they shouldn't be
964 Details : Some RPCs that should not be resent are being resent. This
965 can cause inconsistencies in the RPC state machine. Do not
966 resend such requests.
969 Frequency : rare, only with NFS export
971 Description: Crash on NFS re-export node
972 Details : under very unusual load conditions an assertion is hit in
973 ll_intent_file_open()
976 Frequency : only if OST filesystem is corrupted
978 Description: client incorrectly hits assertion in ptlrpc_replay_req()
979 Details : for a short time RPCs with bulk IO are in the replay list,
980 but replay of bulk IOs is unimplemented. If the OST filesystem
981 is corrupted due to disk cache incoherency and then replay is
982 started it is possible to trip an assertion. Avoid putting
983 committed RPCs into the replay list at all to avoid this issue.
988 Description: large O_DIRECT requests fail under memory pressure/fragmentation
989 Details : Large single O_DIRECT read and write calls can fail to allocate
990 a sufficiently large buffer to process the request. In case of
991 allocation failure the allocation is retried with a smaller
992 buffer and broken into smaller requests.
994 Severity : enhancement
996 Description: Add -o localflock option to simulate old noflock behaviour.
997 Details : This will achieve local-only flock/fcntl locks coherentness.
1002 Description: versioning check is incomplete
1003 Details : Checking the version difference of client vs. server, report
1004 error if the gap is too big.
1009 Description: add support PG_writeback bit
1010 Details : add support for PG_writeback bit for Lustre, for more carefull
1011 work with page cache in 2.6 kernel. This also fix some deadlocks
1012 and remove hack for work O_SYNC with 2.6 kernel.
1014 Severity : enhancement
1016 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1017 Details : The uninit_groups feature works in conjunction with the kernel
1018 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1019 up the pass1 processing of e2fsck. This is a read-only feature
1020 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1021 mount filesystems that have had this feature enabled.
1023 Severity : enhancement
1025 Description: Improve multi-block allocation algorithm to avoid fragmentation
1026 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1027 allocation locality and avoid filesystem fragmentation.
1029 ------------------------------------------------------------------------------
1031 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1033 * Support for kernels:
1034 2.6.9-42.0.3.EL (RHEL 4)
1035 2.6.5-7.276 (SLES 9)
1036 2.4.21-47.0.1.EL (RHEL 3)
1037 2.6.12.6 vanilla (kernel.org)
1038 2.6.16.21-0.8 (SLES10)
1039 * Recommended e2fsprogs version: 1.39.cfs2-0
1041 * The backwards-compatible /proc/sys/portals symlink has been removed
1042 in this release. Before upgrading, please ensure that you change
1043 any configuration scripts or /etc/sysctl.conf files that access
1044 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1045 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1046 in advance of the upgrade on any system running Lustre 1.4.6 or
1047 newer, since /proc/sys/lnet was added in that version.
1048 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1052 Frequency : only when quota is used
1054 Description: avoid scanning export list for quota master
1055 Details : Change the algorithms to avoid scanning export list in order
1056 to improve the efficiency.
1059 Frequency : MDS failover only, very rarely
1061 Description: "went back in time" messages on mds failover
1062 Details : The greatest transno may be lost when the current operation
1063 finishes with an error (transno==0) and the client's last_rcvd
1064 record is over-written. Save the greatest transno in the
1065 mds_last_transno for this case.
1068 Frequency : always for specific kernels and striping counts
1070 Description: client may get "Matching packet too big" without ACL support
1071 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1072 when trying to access files in certain configurations. The
1073 clients should in fact be denied when mounting because they do
1074 not understand ACLs.
1077 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1079 Description: many clients connecting with IO in progress causes connect timeouts
1080 Details : Avoid synchronous journal commits to avoid delays caused by many
1081 clients connecting/disconnecting when bulk IO is in progress.
1082 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1083 OST_IO_PORTAL to avoid delays behind potentially many pending
1087 Frequency : occasionally with multiple writers to a single file
1089 Description: shared writes to file may result in wrong size reported by stat()
1090 Details : Allow growing of kms when extent lock is cancelled
1093 Frequency : always with random mmap IO to multi-striped file
1095 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1096 Details : Only if the hole is at the end of OST object so that kms is too
1097 small. Fix is to increase kms accordingly in ll_nopage.
1100 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1102 Description: writes to a missing object would leak memory on the OST
1103 Details : If there is an inconsistency between the MDS and OST filesystems,
1104 such that the MDS references an object that doesn't exist, writes
1105 to that object will leak memory due to incorrect cleanup in the
1106 error handling path, eventually running out of memory on the OST.
1111 Description: Creating too long symlink causes lustre errors
1112 Details : Check symlink and name lengths before sending requests to MDS.
1115 Frequency : only if flock is enabled (not on by default)
1117 Description: posix locks not released on fd closure on 2.6.9+
1118 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1119 kernels, this caused such locks not to be released on fd close and
1120 then assertions on fs unmount about still used locks.
1123 Frequency : MDS failover only, very rarely
1125 Description: clients may get ASSERTION(granted_lock != NULL)
1126 Details : When request was taking a long time, and a client was resending
1127 a getattr by name lock request. The were multiple lock requests
1128 with the same client lock handle and
1129 mds_getattr_name->fixup_handle_for_resent_request found one of the
1130 lock handles but later failed with ASSERTION(granted_lock != NULL).
1135 Description: handle->h_buffer_credits > 0, assertion failure
1136 Details : h_buffer_credits is zero after truncate, causing assertion
1137 failure. This patch extends the transaction or creates a new
1141 Frequency : NFS re-export or patchless client
1142 Bugzilla : 11179, 10796
1143 Description: Crash on NFS re-export node (__d_move)
1144 Details : We do not want to hash the dentry if we don't have a lock.
1145 But if this dentry is later used in d_move, we'd hit uninitialised
1146 list head d_hash, so we just do this to init d_hash field but
1147 leave dentry unhashed.
1150 Frequency : NFS re-export or patchless client
1152 Description: NFS exports has problem with symbolic link
1153 Details : lustre client didn't properly install dentry when re-exported
1154 to NFS or running patchless client.
1157 Frequency : NFS re-export or patchless client
1159 Description: Various nfs/patchless fixes.
1160 Details : fixes reuse disconected alias for lookup process - this fixes
1161 warning "find_exported_dentry: npd != pd",
1162 fix permission error with open files at nfs.
1163 fix apply umask when do revalidate.
1166 Frequency : occasional
1168 Description: Crash on NFS re-export node
1169 Details : calling clear_page() on the wrong pointer triggered oops in
1170 generic_mapping_read().
1173 Frequency : rarely, using O_DIRECT IO
1175 Description: unaligned directio crashes client with LASSERT
1176 Details : check for unaligned buffers before trying any requests.
1179 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1181 Description: stale data returned from RAID cache
1182 Details : If only a small amount of IO is done to the RAID device before
1183 reading it again it is possible to get stale data from the RAID
1184 cache instead of reading it from disk.
1187 Frequency : always for sles10 kernel
1189 Description: sles10 support
1190 Details : ll_follow_link: compile fixes and using of nd_set_link
1191 under newer kernels.
1194 Frequency : depends on arch, kernel and compiler version, always on sles10
1197 Description: recursive or deep enough symlinks cause stack overflow
1198 Details : getting rid of large stack-allocated variable in
1202 Frequency : depends on hardware
1204 Description: lustre write performance loss in the SLES10 kernel
1205 Details : the performance loss is caused by using of write barriers in the
1206 ext3 code. The SLES10 kernel turns barrier support on by
1207 default. The fix is to undo that change for ldiskfs.
1209 ------------------------------------------------------------------------------
1211 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1213 * Support for kernels:
1214 2.6.9-42.0.3EL (RHEL 4)
1215 2.6.5-7.276 (SLES 9)
1216 2.4.21-47.0.1.EL (RHEL 3)
1217 2.6.12.6 vanilla (kernel.org)
1221 Frequency : quota enabled and large files being deleted
1223 Description: releasing more than 4GB of quota at once hangs OST
1224 Details : If a user deletes more than 4GB of files on a single OST it
1225 will cause the OST to spin in an infinite loop. Release
1226 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1231 Description: statfs data retrieved from /proc may be stale or zero
1232 Details : When reading per-device statfs data from /proc, in the
1233 {kbytes,files}_{total,free,avail} files, it may appear
1234 as zero or be out of date.
1237 Frequency : systems with MD RAID1 external journal devices
1239 Description: lconf's call to blkid is confused by RAID1 journal devices
1240 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1241 of returning all block devices that match the journal UUID.
1244 Frequency : always, for aggregate stripe size over 4GB
1246 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1247 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1248 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1251 Frequency : always if "lfs find" used on a local file/directory
1253 Description: "lfs find" segfaults if used on a local file/directory
1254 Details : The case where a directory component was not specified wasn't
1255 handled correctly. Handle this properly.
1258 Frequency : always on ppc64
1260 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1261 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1262 perform endianness conversion.
1265 Frequency : rarely (truncate to non-zero file size after write under load)
1266 Bugzilla : 10730, 10687
1267 Description: Files padded with zeros to next 4K multiple
1268 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1269 it is possible that files that are truncated to a non-zero size
1270 immediately after being written are filled with zero bytes beyond
1271 the truncated size. No file data is lost.
1273 Severity : enhancement
1275 Description: Allow recovery/failover for liblustre clients.
1276 Details : liblustre clients were unaware of failover configurations until
1279 Severity : enhancement
1281 Description: user file locks should fail when not mounting with flock option
1282 Details : Set up an error-returning stub in ll_file_operations.lock field
1283 to prevent incorrect behaviour when client is mounted without
1284 flock option. Also, set up properly f_op->flock field for
1288 Frequency : always on ia64
1290 Description: "lfs df" loops on printing out MDS statfs information
1291 Details : The obd_ioctl_data was not initialized and in some systems
1292 this caused a failure during the ioctl that did not return
1293 an error. Initialize the struct and return an error on failure.
1296 Frequency : SLES 9 only
1298 Description: Error of copying files with lustre special EAs as root
1299 Details : Client side always return success for setxattr call for lustre
1300 special xattr (currently only "trusted.lov").
1303 Frequency : rarely on clusters with both ia64+i386 clients
1305 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1306 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1307 clients are doing concurrent writes to the same file it is
1308 possible that the ia64 clients may LASSERT because the OST
1309 extent locks are not PAGE_SIZE aligned. Ensure that grown
1310 locks are always aligned on the request boundary.
1313 Frequency : specific use, occasional
1315 Description: Overwriting in use executable truncates on-disk binary image
1316 Details : If one node attempts to overwrite an executable in use by
1317 another node, we now correctly return ETXTBSY instead of
1318 truncating the file.
1320 Severity : enhancement
1322 Description: Async OSC create to avoid the blocking unnecessarily.
1323 Details : If a OST has no remain object, system will block on the creating
1324 when need to create a new object on this OST. Now, ways use
1325 pre-created objects when available, instead of blocking on an
1326 empty osc while others are not empty. If we must block, we block
1327 for the shortest possible period of time.
1332 Description: chmod on Lustre root is propagated to other clients
1333 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1334 invalid by the follow_mount time.
1337 Frequency : liblustre clients only
1339 Description: Race in 'instant cancel' lock handling could lead to such locks
1340 never to be granted in case of SMP MDS
1341 Details : Do not destroy not yet granted but cbpending locks in
1345 Frequency : replay/resend of open
1347 Description: non null lock assetion failure in mds_intent_policy
1348 Details : Trying to replay/resend lockless open requests resulted in
1349 mds_open() returning 0 with no lock. Now it sets a flag if
1350 a lock is going to be returned.
1352 Severity : enhancement
1354 Description: Checksum enhancements
1355 Details : New checksum enhancements allow for resending RPCs that failed
1358 Severity : enhancement
1360 Description: Tunables on number of dirty pages in cacche
1361 Details : Allow to set limit on number of dirty pages cached.
1366 Description: client crash on unmount - lock still has references
1367 Details : In some error handling cases it was possible to leak a lock
1368 reference on a client while accessing a file. This was not
1369 harmful to the client during operation, but would cause the
1370 client to crash when the filesystem is unmounted.
1373 Frequency : specific case, rare
1375 Description: ETXTBSY on mds though file not in use by client
1376 Details : ETXTBSY is no longer incorrectly returned when attempting to
1377 chmod or chown a directory that the user previously tried to
1378 execute or a currently-executing binary.
1381 Frequency : extremely rare except on liblustre-based clients
1383 Description: Lustre space not freed when files are deleted
1384 Details : Clean up open-unlinked files after client eviction. Previously
1385 the unlink was skipped and the files remained as orphans.
1390 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1391 Details : In some cases it was possible to send a blocking callback to
1392 a client doing a glimpse, even though that client didn't get
1393 a lock granted. When the glimpse lock is cancelled on the OST
1394 the freed lock is left on the waiting list and corrupted the list.
1397 Frequency : all core dumps
1399 Description: Broke core dumps to lustre
1400 Details : Negative dentry may be unhashed if parent does not have UPDATE
1401 lock, but some callers, e.g. do_coredump, expect dentry to be
1402 hashed after successful create, hash it in ll_create_it.
1404 ------------------------------------------------------------------------------
1406 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1408 * Support for kernels:
1409 2.6.9-42.0.2.EL (RHEL 4)
1410 2.6.5-7.276 (SLES 9)
1411 2.4.21-40.EL (RHEL 3)
1412 2.6.12.6 vanilla (kernel.org)
1416 Frequency : always on RHEL 3
1418 Description: Number of open files grows over time
1419 Details : The number of open files grows over time, whether or not
1420 Lustre is started. This was due to a filp leak introduced
1421 by one of our kernel patches.
1423 ------------------------------------------------------------------------------
1425 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1427 * Support for kernels:
1428 2.6.9-42.EL (RHEL 4)
1429 2.6.5-7.276 (SLES 9)
1430 2.4.21-40.EL (RHEL 3)
1431 2.6.12.6 vanilla (kernel.org)
1436 Bugzilla : 5719, 9635, 9792, 9684
1437 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1438 Details : If a server is under heavy load and cannot reply to new
1439 connection requests before the client resends the (re)connect,
1440 the connection handling code can behave badly if two service
1441 threads are concurrently handing separate (re)connections from
1442 the same client. Add better locking to the connection handling
1443 code, and ensure that only a single connection will be processed
1444 for a given client UUID, even if the lock is dropped.
1446 Severity : enhancement
1448 Description: add TCP zero-copy support to kernel
1449 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1450 sends if the hardware supports scatter-gather and checksumming.
1451 This allows socklnd to do client-write and server-read more
1452 efficiently and reduce CPU utilization from skbuf copying.
1455 Frequency : only if NFS exporting from client
1457 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1458 Details : If setting attributes on a file created under NFS that had
1459 never been opened it would be possible to oops the client
1460 if the file had no objects.
1463 Frequency : always for liblustre
1465 Description: liblustre client does MDS+OSTs setattr RPC for each write
1466 Details : When doing a write from a liblustre client, the client
1467 incorrectly issued an RPC to the MDS and each OST the file was
1468 striped over in order to update the timestamps. When writing
1469 with small chunks and many clients this could overwhelm the MDS
1470 with RPCs. In all cases it would slow down the write because
1471 these RPCs are unnecessary.
1473 Severity : enhancement
1475 Description: allow number of MDS service threads to be changed at module load
1476 Details : It is now possible to change the number of MDS service threads
1477 running. Adding "options mds mds_num_threads={N}" to the MDS's
1478 /etc/modprobe.conf will set the number of threads for the next
1479 time Lustre is restarted (assuming the "mds" module is also
1480 reloaded at that time). The default number of threads will
1481 stay the same, 32 for most systems.
1486 Description: OST crash if filesystem is unformatted or corrupt
1487 Details : If an OST is started on a device that has never been formatted
1488 or if the filesystem is corrupt and cannot even mount then the
1489 error handling cleanup routines would dereference a NULL pointer.
1494 Description: NULL pointer deref in llap_from_page.
1495 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1496 incorrect) mapping if the page was truncated/reclaimed while it was
1497 searched for. Check for this condition and skip such pages when
1498 doing readahead. Introduce extra check to llap_from_page() to
1499 verify page->mapping->host is non-NULL (so page is not anonymous).
1502 Frequency : Sometimes when using sys_sendfile
1504 Description: "page not covered by a lock" warnings from ll_readpage
1505 Details : sendfile called ll_readpage without right page locks present.
1506 Now we introduced ll_file_sendfile that does necessary locking
1507 around call to generic_file_sendfile() much like we do in
1511 Frequency : with certain MDS communication failures at client mount time
1513 Description: NULL pointer deref after failed client mount
1514 Details : a client connection request may delayed by the network layer
1515 and not be sent until after the PTLRPC layer has timed out the
1516 request. If the client fails the mount immediately it will try
1517 to clean up before the network times out the request. Add a
1518 reference from the request import to the obd device and delay
1519 the cleanup until the network drops the request.
1522 Frequency : occasionally during client (re)connect
1524 Description: assertion failure during client (re)connect
1525 Details : processing a client connection request may be delayed by the
1526 client or server longer than the client connect timeout. This
1527 causes the client to resend the connection request. If the
1528 original connection request is replied in this interval, the
1529 client may trip an assertion failure in ptlrpc_connect_interpret()
1530 which thought it would be the only running connect process.
1533 Frequency : only with obd_echo servers and clients that are rebooted
1535 Description: kernel BUG accessing uninitialized data structure
1536 Details : When running an obd_echo server it did not start the ping_evictor
1537 thread, and when a client was evicted an uninitialized data
1538 structure was accessed. Start the ping_evictor in the RPC
1539 service startup instead of the OBD startup.
1541 Severity : enhancement
1542 Bugzilla : 10193 (patchless)
1543 Description: Remove dependency on various unexported kernel interfaces.
1544 Details : No longer need reparent_to_init, exit_mm, exit_files,
1545 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1548 Frequency : rare (only users of deprecated and unsupported LDAP config)
1550 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1551 Details : LDAP apparently contains 'lustreName' attributes instead of
1552 'name'. A simple remapping of the name is sufficient.
1555 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1557 Description: waiting_locks_callback trips kernel BUG if client is evicted
1558 Details : Running with the dump_on_timeout debug flag turned on makes
1559 it possible that the waiting_locks_callback() can try to dump
1560 the Lustre kernel debug logs from an interrupt handler. Defer
1561 this log dumping to the expired_lock_main() thread.
1563 Severity : enhancement
1565 Description: Support NFS exporting on 2.6 kernels.
1566 Details : Implement non-rawops metadata methods for NFS server to use without
1567 changing NFS server code.
1570 Frequency : very rare (synthetic metadata workload only)
1572 Description: two racing renames might cause an MDS thread to deadlock
1573 Details : Running the "racer" program may cause one MDS thread to rename
1574 a file from being the source of a rename to being the target of
1575 a rename at exactly the same time that another thread is doing
1576 so, and the second thread has already enqueued these locks after
1577 doing a lookup of the target and is trying to relock them in
1578 order. Ensure that we don't try to re-lock the same resource.
1581 Frequency : only very large systems with liblustre clients
1583 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1584 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1585 all OSTs in parallel. This allows the eviction of stale liblustre
1586 clients to proceed much faster than if they were done in series,
1587 and also offers similar improvements for other set_info RPCs.
1592 Description: excessive CPU usage during initial read phase on client
1593 Details : During the initial read phase on a client, it would agressively
1594 retry readahead on the file, consuming too much CPU and impacting
1595 performance (since 1.4.5.8). Improve the readahead algorithm
1596 to avoid this, and also improve some other common cases (read
1597 of small files in particular, where "small" is files smaller than
1598 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1603 Description: MDS crash when receiving packet with unknown intent.
1604 Details : Do not LBUG in unknown intent case, just return -EFAULT
1606 Severity : enhancement
1607 Bugzilla : 9293, 9385
1608 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1609 Details : Do not serialize getattr (non-intent version) and statfs.
1612 Frequency : occasional, when OST network is overloaded/intermittent
1614 Description: client evicted by OST after bulk IO timeout
1615 Details : If a client sends a bulk IO request (read or write) the OST
1616 may evict the client if it is unresposive to its data GET/PUT
1617 request. This is incorrect if the network is overloaded (takes
1618 too long to transfer the RPC data) or dropped the OST GET/PUT
1619 request. There is no need to evict the client at all, since
1620 the pinger and/or lock callbacks will handle this, and the
1621 client can restart the bulk request.
1624 Frequency : Always when mmapping file with no objects
1626 Description: client crashes when mmapping file with no objects
1627 Details : Check that we actually have objects in a file before doing any
1628 operations on objects in ll_vm_open, ll_vm_close and
1634 Description: Request leak when working with deleted CWD
1635 Details : Introduce advanced request refcount tracking for requests
1636 referenced from lustre intent.
1638 Severity : Enhancement
1640 Description: Cache open file handles on client.
1641 Details : MDS now will return special lock along with openhandle, if
1642 requested and client is allowed to hold openhandle, even if unused,
1643 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1644 closing files for every read/write openration.
1646 Severity : Enhancement
1648 Description: Cache open negative dentries on client when possible.
1649 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1650 negative dentries on lock revocation.
1655 Description: Remounting a client read-only wasn't possible with a zconf mount
1656 Details : It wasn't possible to remount a client read-only with llmount.
1658 Severity : enhancement
1659 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1660 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1661 MPICH is included to add Lustre-specific ADIO interfaces.
1662 This is based closely on the UFS ADIO layer and only differs
1663 in file creation, in order to allow the OST striping to be set.
1664 This is user-contributed code and not supported by CFS.
1669 Description: extended inode attributes (immutable, append-only) work improperly
1670 when 2.4 and 2.6 kernels are used on client/server or vice versa
1671 Details : Introduce kernel-independent values for these flags.
1673 Severity : enhancement
1676 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1677 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1678 granularity. Now, Fractional MB granularity is be supported,
1679 this is very useful for low memory system.
1681 Severity : enhancement
1683 Description: Getattr by fid
1684 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1685 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1686 have locks and avoids getattr rpc requests for it on every stat.
1689 Frequency : Always, for filesystems larger than 2TB
1691 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1692 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1693 filesystems larger than 2TB to be mounted with mballoc enabled.
1696 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1698 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1699 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1700 filesystems larger than 2TB is mounted then the kernel will
1701 silently corrupt the start of the filesystem. CONFIG_LBD is
1702 enabled for all CFS-supported kernels, but the possibility of
1703 this happening with a modified kernel config exists.
1705 Severity : enhancement
1707 Description: add client O_DIRECT support for 2.6 kernels
1708 Details : It is now possible to do O_DIRECT reads and writes to files
1709 in the Lustre client mountpoint on 2.6 kernel clients.
1711 Severity : enhancement
1713 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1714 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1715 parallel, not waiting for response from every OST before sending
1716 a rpc to the next OST.
1721 Description: setattr vs write race when updating file timestamps
1722 Details : Client processes that update a file timestamp into the past
1723 right after writing to the file (e.g. tar) it is possible that
1724 the updated file modification time can be reset to the current
1725 time due to a race between processing the setattr and write RPC.
1727 Severity : enhancement
1729 Description: Bring 'lfs find' closer in line with regular Linux find.
1730 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1731 -print0 options and obtains all the needed info through the lustre
1734 Severity : enhancement
1736 Description: support up to 1024 configured devices on one node
1737 Details : change obd_dev array from statically allocated to dynamically
1738 allocated structs as they are first used to reduce memory usage
1743 Description: Flush dirty partially truncated pages during truncate
1744 Details : Immediatelly flush partially truncated pages in filter_setattr,
1745 this way we completely avoid having any pages in page cache on OST
1746 and can retire ugly workarounds during writes to flush such pages.
1751 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1752 Details : protect inode from truncation within vfs_unlink() context
1753 just take a reference before calling vfs_unlink() and release it
1754 when parent's i_sem is free.
1759 Description: last_id value checked outside lock on OST caused LASSERT failure
1760 Details : If there were multiple MDS->OST object precreate requests in
1761 flight, it was possible that the OST's last object id was checked
1762 outside a lock and incorrectly tripped an assertion. Move checks
1763 inside locks, and discard old precreate requests.
1766 Frequency : always, if extents are used on OSTs
1768 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1769 Details : OSTs using the extents format would not zero the high 16 bits of
1770 the index physical block number. This is not a problem for any
1771 OST filesystems smaller than 16TB, and no kernels support ext3
1772 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1773 new/modified files) and can be fixed for existing filesystems
1774 with e2fsprogs-1.39-cfs1.
1779 Description: import connection selection may be incorrect if timer wraps
1780 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1781 import connections to be ignored if the 32-bit jiffies counter
1782 wraps. Use a 64-bit jiffies counter.
1785 Frequency : very large clusters immediately after boot
1787 Description: LNET request buffers exhausted under heavy short-term load
1788 Details : If a large number of client requests are generated on a service
1789 that has previously never seen so many requests it is possible
1790 that the request buffer growth cannot keep up with the spike in
1791 demand. Instead of dropping incoming requests, they are held in
1792 the LND until the RPC service can accept more requests.
1795 Frequency : Sometimes during replay
1797 Description: Assertion failure in ll_local_open after replay.
1798 Details : If replay happened on an open request reply before we were able
1799 to set replay handler, reply will become not swabbed tripping the
1800 assertion in ll_local_open. Now we set the handler right after
1801 recognising of open request
1804 Frequency : very rare
1806 Description: kernel reports "badness in vsnprintf"
1807 Details : Reading from the "recovery_status" /proc file in small chunks
1808 may cause a negative length in lprocfs_obd_rd_recovery_status()
1809 call to vsnprintf() (which is otherwise harmless). Exit early
1810 if there is no more space in the output buffer.
1812 Severity : enhancement
1814 Description: clear OBD RPC statistics by writing to them
1815 Details : It is now possible to clear the OBD RPC statistics by writing
1816 to the "stats" file.
1821 Description: Client mtime is not the same on different clients after utimes
1822 Details : In some cases, the client was using the utimes() syscall on
1823 a file cached on another node. The clients now validate the
1824 ctime from the MDS + OSTs to determine which one is right.
1829 Description: Inability to activate failout mode
1830 Details : lconf script incorrectly assumed that in pythong string's numeric
1831 value is used in comparisons.
1834 Frequency : always with multiple stripes per file
1836 Description: Inefficient object allocation for mutli-stripe files
1837 Details : When selecting which OSTs to stripe files over, for files with
1838 a stripe count that divides evenly into the number of OSTs,
1839 the MDS is always picking the same starting OST for each file.
1840 Return the OST selection heuristic to the original design.
1845 Description: mount failures may take full timeout to return an error
1846 Details : Under some heavy load conditions it is possible that a
1847 failed mount can wait for the full obd_timeout interval,
1848 possibly several minutes, before reporting an error.
1849 Instead return an error as soon as the status is known.
1851 Frequency : quota enabled and large files being deleted
1853 Description: releasing more than 4GB of quota at once hangs OST
1854 Details : If a user deletes more than 4GB of files on a single OST it
1855 will cause the OST to spin in an infinite loop. Release
1856 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1861 Description: statfs data retrieved from /proc may be stale or zero
1862 Details : When reading per-device statfs data from /proc, in the
1863 {kbytes,files}_{total,free,avail} files, it may appear
1864 as zero or be out of date.
1867 Frequency : systems with MD RAID1 external journal devices
1869 Description: lconf's call to blkid is confused by RAID1 journal devices
1870 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1871 of returning all block devices that match the journal UUID.
1874 Frequency : always, for aggregate stripe size over 4GB
1876 Description: assertion fails when trying to use 4GB stripe size
1877 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1878 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1881 Frequency : always on ppc64
1883 Description: the first write on an ext3 filesystem with mballoc got stuck
1884 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1885 perform endianness conversion.
1887 ------------------------------------------------------------------------------
1889 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1891 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1892 INTEROPERATE with older versions automatically. Please read the
1893 user documentation before upgrading any part of a live system.
1894 * WARNING: Lustre networking configuration changes are required with
1895 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1898 * Support for kernels:
1899 2.6.9-22.0.2.EL (RHEL 4)
1900 2.6.5-7.244 (SLES 9)
1901 2.6.12.6 vanilla (kernel.org)
1904 Severity : enhancement
1905 Bugzilla : 7981/8208
1906 Description: Introduced Lustre Networking (LNET)
1907 Details : LNET is new networking infrastructure for Lustre, it includes
1908 a reorganized network configuration mode (see the user
1909 documentation for full details) as well as support for routing
1910 between different network fabrics. Lustre Networking Devices
1911 (LNDs) for the supported network fabrics have also been
1912 created for this new infrastructure.
1914 Severity : enhancement
1915 Description: Introduced Access control lists
1916 Details : clients can set ACLs on files and directories in order to have
1917 more fine-grained permissions than the standard Unix UGO+RWX.
1918 The MDS must be started with the "-o acl" mount option.
1920 Severity : enhancement
1921 Description: Introduced filesystem quotas
1922 Details : Administrators may now establish per-user quotas on the
1925 Severity : enhancement
1927 Description: Configuration change for the XT3
1928 The PTLLND is now used to run Lustre over Portals on the XT3
1929 The configure option(s) --with-cray-portals are no longer used.
1930 Rather --with-portals=<path-to-portals-includes> is used to
1931 enable building on the XT3. In addition to enable XT3 specific
1932 features the option --enable-cray-xt3 must be used.
1937 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1938 Details : A race between allocating a new llcd and re-getting the llcd_lock
1939 allowed another thread to grab newly-allocated llcd.
1941 Severity : enhancement
1943 Description: 2.6 OST async journal commit and locking fix to improve performance
1944 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1945 2.6 kernels are now async as they already were in 2.4 kernels so
1946 that they can commit concurrently with the network bulk transfer.
1947 For block-allocated files the filter allocation semaphore is held
1948 to avoid filesystem fragmentation during allocation. BKL lock
1949 removed for 2.6 xattr operations where it is no longer needed.
1954 Description: lconf incorrectly determined whether two IP networks could talk
1955 Details : In some more complicated routing and multiple-network
1956 configurations, lconf will avoid trying to make a network
1957 connection to a disjoint part of the IP space. It was doing the
1958 math incorrectly for one set of cases.
1963 Description: Fix for potential infinite loop processing records in an llog.
1964 Details : If an llog record is corrupted/zeroed, it is possible to loop
1965 forever in llog_process(). Validate the llog record length
1966 and skip the remainder of the block on error.
1969 Frequency : occasional (liblustre only)
1971 Description: liblustre could not open files whose last component is a symlink
1972 Details : sysio_path_walk() would incorrectly pass the open intent to
1973 intermediate path components.
1976 Frequency : rare (liblustre only with non-standard tuning)
1977 Bugzilla : 7201 (7350)
1978 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1979 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1980 after it has been released, only do this for referenced locks
1982 Severity : enhancement
1984 Description: specifying an (invalid) directory default stripe_size of -1
1985 would reset the directory default striping
1986 Details : stripe_size -1 was used internally to signal directory stripe
1987 removal, now use "all default" to signal dir stripe removal
1988 as a directory striping of "all default" is not useful
1991 Frequency : common for large clusters running liblustre clients
1993 Description: doing an ls when liblustre clients are running is slow
1994 Details : sending a glimpse AST to a liblustre client waits for every AST
1995 to time out, as liblustre clients will not respond. Since they
1996 cannot cache data we refresh the OST lock LVB from disk instead.
1998 Severity : enhancement
2000 Description: doing an ls at the same time as file IO can be slow
2001 Details : enqueue and other "small" requests can be blocked behind many
2002 large IO requests. Create a new OST IO portal for non-IO
2003 requests so they can be processed faster.
2006 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2008 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2009 created file with mode 000
2010 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2011 allow an ftruncate() as a non-root user to a file with mode 000.
2012 Lustre now respects this flag to disable mode checks when
2013 truncating a file owned by the user
2016 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2019 Description: Revoking locks from clients that went dead or catatonic might take
2021 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2022 cancellation of such locks instant on servers without waiting for
2023 any reply from clients. Clients drops these locks when cancel
2024 notification from server is received without replying.
2027 Frequency : liblustre-only, when liblustre client dies or becomes busy
2029 Description: Doing ls on Linux clients can take a long time with active
2031 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2032 granting such locks to it in the first place if possible. Locks
2033 are taken by proxy on the OST during the read or write and
2034 dropped immediately afterward. Add connect flags handling, do
2035 not grant locks to liblustre clients for glimpse ASTs.
2037 Severity : enhancement
2039 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2040 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2041 sequential reads and behaves badly with applications doing random
2042 reads. Improve it by reading ahead at least read region, and
2043 avoiding excessive large RPC for small reads.
2045 Severity : enhancement
2047 Description: Creating more than 1000 files for a single job may cause a load
2048 imbalance on the OSTs if there are also a large number of OSTs.
2049 Details : qos_prep_create() uses an OST index reseed value that is an
2050 even multiple of the number of available OSTs so that if the
2051 reseed happens in the middle of the object allocation it will
2052 still utilize the OSTs as uniformly as possible.
2057 Description: OST or MDS may oops in ping_evictor_main()
2058 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2059 but doesn't restart at beginning of obd_exports_timed list
2062 Severity : enhancement
2064 Description: improve by-nid export eviction on the MDS and OST
2065 Details : allow multiple exports with the same NID to be evicted at one
2066 time without re-searching the exports list.
2069 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2071 Description: MDS may oops in groups_free()
2072 Details : in rare race conditions a newly allocated group_info struct is
2073 freed again, and this can be NULL. The 2.4 compatibility code
2074 for groups_free() checked for a NULL pointer, but 2.6 did not.
2077 Frequency : common for liblustre clients doing little filesystem IO
2078 Bugzilla : 9352, 7313
2079 Description: server may evict liblustre clients accessing contended locks
2080 Details : if a client is granted a lock or receives a completion AST
2081 with a blocking AST already set it would not reply to the AST
2082 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2085 Frequency : lfs setstripe, only systems with more than 160 OSTs
2087 Description: unable to set striping with a starting offset beyond OST 160
2088 Details : llapi_create_file() incorrectly limited the starting stripe
2089 index to the maximum single-file stripe count.
2092 Frequency : LDAP users only
2094 Description: lconf did not handle in-kernel recovery with LDAP properly
2095 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2097 Severity : enhancement
2099 Description: bind OST threads to NUMA nodes to improve performance
2100 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2101 node and do their allocations there to localize memory access
2103 Severity : enhancement
2105 Description: llmount can determine client NID directly from Myrinet (GM)
2106 Details : the client NID code from gmnalnid was moved directly into
2107 llmount, removing the need to use this or specifying the
2108 client NID explicitly when mounting GM clients with zeroconf
2111 Frequency : if client is started with down MDS
2113 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2114 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2115 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2116 ptlrpc_import_delayed_req() because we didn't check if the
2117 request was interrupted, and we also didn't break out of the
2118 event loop if there was no timeout
2123 Description: data loss during non-page-aligned writes to a single file from
2124 both multiple nodes and multiple threads on one node at same time
2125 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2126 inconsistency led to false short-reads, that were cached and later
2127 used by ->prepare_write() to fill in partially written page,
2128 leading to data loss.
2131 Frequency : always, if lconf --abort_recovery used
2133 Description: lconf --abort_recovery fails with 'Operation not supported'
2134 Details : lconf was attempting to abort recovery on the MDT device and not
2137 Severity : enhancement
2139 Description: remove cleanup logs
2140 Details : replace lconf-generated cleanup logs with lustre internal
2141 cleanup routines. Eliminates the need for client-cleanup and
2144 Severity : enhancement
2146 Description: add support for EAs (user and system) on lustre filesystems
2147 Details : it is now possible to store extended attributes in the Lustre
2148 client filesystem, and with the user_xattr mount option it
2149 is possible to allow users to store EAs on their files also
2151 Severity : enhancement
2153 Description: Add possibility (config option) to show minimal available OST free
2155 Details : When compiled with --enable-mindf configure option, statfs(2)
2156 (and so, df) will return least minimal free space available from
2157 all OSTs as amount of free space on FS, instead of summary of
2158 free spaces of all OSTs.
2160 Severity : enhancement
2162 Description: do not expand extent locks acquired on OST-side
2163 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2164 by server: they are not cached anyway.
2167 Frequency : when mmap is used/binaries executed from Lustre
2169 Description: Unmmap pages before throwing them away from read cache.
2170 Details : llap_shrink cache now attempts to unmap pages before discarding
2171 them (if unmapping failed - do not discard). SLES9 kernel has
2172 extra checks that trigger if this unmapping is not done first.
2177 Description: lconf didn't resolve symlinks before checking to see whether a
2178 given mountpoint was already in use
2181 Frequency : when migrating failover services
2182 Bugzilla : 6395, 9514
2183 Description: When migrating a subset of services from a node (e.g. failback
2184 from a failover service node) the remaining services would
2185 time out and evict clients.
2186 Details : lconf --force (implied by --failover) sets the global obd_timeout
2187 to 5 seconds in order to quickly disconnect, but this caused
2188 other RPCs to time out too quickly. Do not change the global
2189 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2191 Severity : enhancement
2192 Frequency : if MDS is started with down OST
2193 Bugzilla : 9439,5706
2194 Description: Allow startup/shutdown of an MDS without depending on the
2195 availability of the OSTs.
2196 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2197 Add appropriate locking and lov-osc refcounts for safe
2198 cleaning. Add osc abort_inflight calls in case the
2199 synchronize never started.
2202 Frequency : occasional (Cray XT3 only)
2204 Description: root not authorized to access files in CRAY_PORTALS environment
2205 Details : The client process capabilities were not honoured on the MDS in
2206 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2207 been accessed by an authorized user then root was able to access
2208 the file on the local client also. The root user capabilities
2209 are now allowed on the MDS, as this environment has secure UID.
2212 Frequency : occasional
2214 Description: ldiskfs "too long searching" message happens too often
2215 Details : A debugging message (otherwise harmless) prints too often on
2216 the OST console. This has been reduced to only happen when
2217 there are fragmentation problems on the filesystem.
2222 Description: Division by zero in statfs when all OSCs are inactive
2223 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2224 lov_statfs divides by value returned by lov_get_stripecnt().
2228 Bugzilla : 9489, 3273
2229 Description: First write from each client to each OST was only 4kB in size,
2230 to initialize client writeback cache, which caused sub-optimal
2231 RPCs and poor layout on disk for the first writen file.
2232 Details : Clients now request an initial cache grant at (re)connect time
2233 and so that they can start streaming writes to the cache right
2234 away and always do full-sized RPCs if there is enough data.
2235 If the OST is rebooted the client also re-establishes its grant
2236 so that client cached writes will be honoured under the grant.
2241 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2242 Details : Now I/O RPCs go to different portal number and (presumably) fast
2243 lock requests (and glimses) and other RPCs get their own service
2244 threads pool that should be able to service those RPCs
2247 Severity : enhancement
2249 Description: Ability to exchange lustre version between client and servers and
2250 issue warnings at client side if client is too old. Also for
2251 liblustre clients there is ability to refuse connection of too old
2253 Details : New 'version' field is added to connect data structure that is
2254 filled with version info. That info is later checked by server and
2258 Frequency : rare, liblustre only.
2259 Bugzilla : 9296, 9581
2260 Description: Two simultaneous writes from liblustre at offset within same page
2261 might proceed at the same time overwriting eachother with stale
2263 Details : I/O lock withing llu_file_prwv was released too early, before data
2264 actually was hitting the wire. Extended lock-holding time until
2265 server acknowledges receiving data.
2268 Frequency : extremely rare. Never observed in practice.
2270 Description: avoid generating lustre_handle cookie of 0.
2271 Details : class_handle_hash() generates handle cookies by incrementing
2272 global counter, and can hit 0 occasionaly (this is unlikely, but
2273 not impossible, because initial value of cookie counter is
2274 selected randonly). Value of 0 is used as a sentinel meaning
2275 "unassigned handle" --- avoid it. Also coalesce two critical
2276 sections in this function into one.
2278 Severity : enhancement
2280 Description: allow liblustre clients to delegate truncate locking to OST
2281 Details : To avoid overhead of locking, liblustre client instructs OST to
2282 take extent lock in ost_punch() on client's behalf. New connection
2283 flag is added to handle backward compatibility.
2285 Severity : enhancement
2286 Bugzilla : 4928, 7341, 9758
2287 Description: allow number of OST service threads to be specified
2288 Details : a module parameter allows the number of OST service threads
2289 to be specified via "options ost ost_num_threads={N}" in the
2290 OSS's /etc/modules.conf or /etc/modprobe.conf.
2294 Bugzilla : 6146, 9635, 9895
2295 Description: servers crash with bad pointer in target_handle_connect()
2296 Details : In rare cases when a client is reconnecting it was possible that
2297 the connection request was the last reference for that export.
2298 We would temporarily drop the export reference and get a new
2299 one, but this may have been the last reference and the export
2300 was just destroyed. Get new reference before dropping old one.
2302 Severity : enhancement
2303 Frequency : if client is started with failover MDS
2305 Description: Allow multiple MDS hostnames in the mount command
2306 Details : Try to read the configuration from all specified MDS
2307 hostnames during a client mount in case the "primary"
2310 Severity : enhancement
2312 Description: Stop sending data to evicted clients as soon as possible.
2313 Details : Check if the client we are about to send or are sending data to
2314 was evicted already. (Check is done every second of waiting,
2315 for which l_wait_event interface was extended to allow checking
2316 of exit condition at specified intervals).
2319 Frequency : rare, normally only when NFS exporting is done from client
2321 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2323 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2326 Frequency : always, if srand() is called before liblustre initialization
2328 Description: Liblustre uses system PRNG disturbing its usage by user application
2329 Details : Introduce internal to lustre fast and high-quality PRNG for
2330 lustre usage and make liblustre and some other places in generic
2331 lustre code to use it.
2333 Severity : enhancement
2334 Bugzilla : 9477, 9557, 9870
2335 Description: Verify that the MDS configuration logs are updated when xml is
2336 Details : Check if the .xml configuration logs are newer than the config
2337 logs stored on the MDS and report an error if this is the case.
2338 Request --write-conf, or allow starting with --old_conf.
2340 Severity : enhancement
2342 Description: Handle symlinks in the path when checking if Lustre is mounted.
2343 Details : Resolve intermediate symlinks when checking if a client has
2344 mounted a filesystem to avoid duplicate client mounts.
2349 Description: lconf can hit an error exception but still return success.
2350 Details : The lconf command catches the Command error exception at the top
2351 level script context and will exit with the associated exit
2352 status, but doesn't ensure that this exit status is non-zero.
2357 Description: failure of ptlrpc thread startup can cause oops
2358 Details : Starting a ptlrpc service thread can fail if there are a large
2359 number of threads or the server memory is very fragmented.
2360 Handle this without oopsing.
2363 Frequency : always, only if liblustre and non-default acceptor port was used
2365 Description: liblustre cannot connect to servers with non-default acceptor port
2366 Details : tcpnal_set_default_params() was not called and was therefore
2367 ignoring the environment varaible TCPNAL_PORT, as well as other
2368 TCPNAL_ environment variables
2373 Description: two objects could be created on the same OST for a single file
2374 Details : If an OST is down, in some cases it was possible to create two
2375 objects on a single OST for a single file. No problems other
2376 than potential performance impact and spurious error messages.
2380 Bugzilla : 5681, 9562
2381 Description: Client may oops in ll_unhash_aliases
2382 Details : Client dcache may become inconsistent in race condition.
2383 In some cases "getcwd" can fail if the current directory is
2389 Description: Inode refcounting problems in NFS export code
2390 Details : link_raw functions used to call d_instantiate without obtaining
2391 extra inode reference first.
2395 Bugzilla : 9942, 9903
2396 Description: Referencing freed requests leading to crash, memleaks with NFS.
2397 Details : We used to require that call to ll_revalidate_it was always
2398 followed by ll_lookup_it. Also with revalidate_special() it is
2399 possible to call ll_revalidate_it() twice for the same dentry
2400 even if first occurence returned success. This fix changes semantic
2401 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2402 reference on a request referred from the intent.
2403 ll_intent_release() then releases such a request.
2406 Frequency : rare, normally benchmark loads only
2408 Description: unlinked inodes were kept in memory on the client
2409 Details : If a client is repeatedly creating and unlinking files it
2410 can accumulate a lot of stale inodes in the inode slab cache.
2411 If there is no other client load running this can cause the
2412 client node to run out of memory. Instead flush old inodes
2413 from client cache that have the same inode number as a new inode.
2416 Frequency : SLES9 2.6.5 kernel and long filenames only
2417 Bugzilla : 9969, 10379
2418 Description: utime reports stale NFS file handle
2419 Details : SLES9 uses out-of-dentry names in some cases, which confused
2420 the lustre dentry revalidation. Change it to always use the
2424 Frequency : rare, unless heavy write-truncate concurrency is continuous
2425 Bugzilla : 4180, 6984, 7171, 9963, 9331
2426 Description: OST becomes very slow and/or deadlocked during object unlink
2427 Details : filter_destroy() was holding onto the parent directory lock
2428 while truncating+unlinking objects. For very large objects this
2429 may block other threads for a long time and slow overall OST
2430 responsiveness. It may also be possible to get a lock ordering
2431 deadlock in this case, or run out of journal credits because of
2432 the combined truncate+unlink. Solution is to do object truncate
2433 first in one transaction without parent lock, and then do the
2434 final unlink in a new transaction with the parent lock. This
2435 reduces the lock hold time dramatically.
2438 Frequency : rare, 2.4 kernels only
2440 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2441 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2442 before unmounting their filesystem in order to avoid deadlock.
2443 The kernel_locked() function in 2.4 kernels only checks whether
2444 the kernel lock is held, not whether it is this process that is
2445 holding it as 2.6 kernels do.
2450 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2451 Details : The client ptlrpc code may be trying to reconnect to a down
2452 server before a previous connection attempt has timed out.
2453 Increase the reconnect interval to be longer than the connection
2454 timeout interval to avoid sending duplicate connections to
2458 Frequency : echo_client brw_test command
2460 Description: fix echo_client to work with OST preallocated code
2461 Details : OST preallocation code (5137) didn't take echo_client IO path
2462 into account: echo_client calls filter methods outside of any
2463 OST thread and, hence, there is no per-thread preallocated
2464 pages and buffers to use. Solution: hijack pga pages for IO. As
2465 a byproduct, this avoids unnecessary data copying.
2469 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2470 Description: Client can oops in mdc_commit_close() after open replay
2471 Details : It was possible for the MDS to return an open request with no
2472 transaction number in mds_finish_transno() if the client was
2473 evicted, but without actually returning an error. Clients
2474 would later try to replay that open and may trip an assertion
2475 Simplify the client close codepath, and always return an error
2476 from the MDS in case the open is not successful.
2479 Frequency : rare, 2.6 OSTs only
2481 Description: OST may deadlock under high load on fragmented files
2482 Details : If there was a heavy load and highly-fragmented OST filesystems
2483 it was possible to have all the OST threads deadlock waiting on
2484 allocation of biovecs, because the biovecs were not released
2485 until the entire RPC IO was completed. Instead, release biovecs
2486 as soon as they are complete to ensure forward IO progress.
2488 Severity : enhancement
2490 Description: Support for specifying external journal device at mount
2491 Details : If an OST or MDS device is formatted with an external journal
2492 device, this device major/minor is stored in the ext3 superblock
2493 and may not be valid for failover. Allow detecting and
2494 specifying the external journal at mount time.
2499 Description: Mounting an MDS with pending unlinked files may cause oops
2500 Details : target_finish_recovery() calls mds_postrecov() which returned
2501 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2502 considers this an error and immediately begins cleaning up the
2503 lov, just after starting the mds_lov process
2505 Severity : enhancement
2507 Description: Implement 'lfs df' to report actual free space on per-OST basis
2508 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2509 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2510 usage of objects; '-h' to report in human readable format.
2512 ------------------------------------------------------------------------------
2514 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2521 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2522 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2523 at startup during orphan unlinking. Instead, skip buddy update
2524 before setup, as it will be regenerated after recovery is complete.
2529 Description: If an OST is inactive, its locks might reference stale inodes.
2530 Details : lov_change_cbdata() must iterate over all namespaces, even if
2531 they are inactive to clear inode references from the lock.
2533 Severity : enhancement
2534 Frequency : occasional, if non-standard max_dirty_mb used
2536 Description: Client will block write RPCs if not enough grant
2537 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2538 then the client will block writes while waiting for another RPC
2539 to complete instead of consuming its dirty limit. With change
2540 we get improved performance when max_dirty_mb is small.
2542 Severity : enhancement
2543 Bugzilla : 3389, 6253
2544 Description: Add support for supplementary groups on the MDS.
2545 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2546 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2547 lookups for user supplementary groups into a cache.
2551 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2552 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2553 for liblustre O_EXCL is enforced by MDS.
2557 Description: readdir never returns NULL in liblustre.
2558 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2559 dentry in struct dirent.
2563 Frequency : liblustre only, depends on application IO pattern
2564 Description: liblustre clients evicted if not contacting servers
2565 Details : Don't put liblustre clients into the ping_evictor list, so
2566 they will not be evicted by the pinger ever.
2568 Severity : enhancement
2570 Description: Add ability to evict clients by NID from MDS.
2571 Details : By echoing "nid:$NID" string into
2572 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2573 $NID would be instantly evicted from this MDS and from all active
2574 OSTs connected to it.
2578 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2579 Details : lookup_it_finish() used to query file size from OSTs that was not
2584 Description: service threads change working directory to that of init
2585 Details : Starting lustre service threads may pin the working directory
2586 of the parent thread, making that filesystem busy. Threads
2587 now change to the working directory of init to avoid this.
2591 Frequency : during shutdown only
2592 Description: shutdown with a failed MDS or OST can cause unmount to hang
2593 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2598 Frequency : 2.6 only
2599 Description: chmod/chown may include an extra supplementary group
2600 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2601 supplementary group and if none is specified this is used.
2604 Bugzilla : 5479 (6816)
2606 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2607 Details : If lookup is in progress on a file that is unlinked we might try
2608 to revalidate the inode and fail in revalidate after lookup is
2609 complete and ll_file_open() enqueues the open again but
2610 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2613 Frequency : always, if lconf --abort_recovery used
2615 Description: lconf --abort_recovery fails with 'Operation not supported'
2616 Details : lconf was attempting to abort recovery on the MDT device and not
2619 ------------------------------------------------------------------------------
2621 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2626 Frequency : rare (only unsupported configurations with a node running as an
2628 Bugzilla : 6514, 5137
2629 Description: Mounting a Lustre file system on a node running as an OST could
2631 Details : OSTs now preallocates memory needed to write out data at
2632 startup, instead of when needed, to avoid having to
2633 allocate memory in possibly low memory situations.
2634 Specifically, if the file system is mounted on on OST,
2635 memory pressure could force it to try to write out data,
2636 which it needed to allocate memory to do. Due to the low
2637 memory, it would be unable to do so and the node would
2638 become unresponsive.
2640 Severity : enhancement
2642 Description: Addition of lconf --service command line option
2643 Details : lconf now accepts a '--service <arg>' option, which is
2644 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2646 Severity : enhancement
2648 Description: Failover mode is now the default for OSTs.
2649 Details : By default, OSTs will now run in failover mode. To return to
2650 the old behaviour, add '--failout' to the lmc line for OSTs.
2652 Severity : enhancement
2654 Description: Health checks are now provided for MDS and OSTs
2655 Details : Additional detailed health check information on MSD and OSTs
2656 is now provided through the procfs health_check value.
2659 Frequency : occasional, depends on IO load
2661 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2662 after numerous create/delete cycles
2663 Details : The ext3 inode allocation policy would not allocate new inodes
2664 very well on the OSTs because there are no new directories
2665 being created. Instead we look for groups with free space if
2666 the parent directories are nearly full.
2671 Description: Network or server problems during mount may cause partially
2672 mounted clients instead of returning an error.
2673 Details : The config llog parsing code may overwrite the error return
2674 code during mount error handling, returning success instead
2680 Description: MDS can fail to allocate large reply buffers
2681 Details : After long uptimes the MDS can fail to allocate large reply
2682 buffers (e.g. zconf client mount config records) due to memory
2683 fragmentation or consumption by the buffer cache. Preallocate
2684 some large reply buffers so that these replies can be sent even
2685 under memory pressure.
2689 Frequency : rare (liblustre)
2690 Description: fsx running with liblustre complained that using truncate() to
2691 extend the file doesn't work. This patch corrects that issue.
2692 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2693 ATTR_SIZE and lsm use in llu_setattr_raw.
2697 Frequency : rare, only 2.6 kernels
2698 Description: Unusual file access patterns on the MDS may result in inode
2699 data being lost in very rare circumstances.
2700 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2701 code in the 2.6 kernel caused the inode and/or EA data not to
2702 be read from disk, causing single-file corruption.
2706 Frequency : rare, only 2.6 filesystems using extents
2707 Description: Heavy concurrent write and delete load may cause data corruption.
2708 Details : It was possible under high-load situations to have an extent
2709 metadata block in the block device cache from a just-unlinked
2710 file overwrite a newly-allocated data block. We now unmap any
2711 metadata buffers that alias just-allocated data blocks.
2715 Frequency : filesystems with default stripe_count larger than 77
2716 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2717 Details : lconf specifies an inode size of 4096 bytes when the default
2718 stripe_count is larger than 77. This conflicts with the default
2719 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2720 this case to avoid pinning too much memory for large EAs.
2722 ------------------------------------------------------------------------------
2724 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2729 Frequency : rare (extremely heavy IO load with hundreds of clients)
2731 Description: Client is evicted, gets IO error writing to file
2732 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2733 caused clients to be evicted for AST timeouts. The fixes in
2734 bug 5192 mean we no longer need to have such short AST timeouts
2735 so ldlm_timeout has been increased.
2738 Frequency : occasional during --force or --failover shutdown under load
2739 Bugzilla : 5949, 4834
2740 Description: Server oops/LBUG if stopped with --force or --failover under load
2741 Details : a collection of import/export refcount and cleanup ordering
2742 issues fixed for safer force cleanup
2745 Frequency : only filesystems larger than 120 OSTs
2746 Bugzilla : 5990, 6223
2747 Description: lfs getstripe would oops on a very large filesystem
2748 Details : lov_getconfig used kfree on vmalloc'd memory
2751 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2752 Bugzilla : 6242, 6243
2753 Description: reading from files that had been truncated to a non-zero size
2754 but never opened returned no data
2755 Details : ll_file_read() reads zeros from no-object files to EOF
2760 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2761 Details : There's little harm in aborting MDS/OSS recovery and letting it
2762 try again, so I removed the LASSERT and return an error instead.
2764 Severity : enhancement
2766 Description: New debugging infrastructure for tracking down data corruption
2767 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2768 (b) cover more of the client-side code path, and (c) try to narrow
2769 down where problems occurred
2773 Bugzilla : 3819, 4364, 4397, 6313
2774 Description: Racing close and eviction MDS could cause assertion in mds_close
2775 Details : It was possible to get multiple mfd references during close and
2776 client eviction, leading to one thread referencing a freed mfd.
2778 Severity: : enhancement
2779 Bugzilla : 3262, 6359
2780 Description: Attempts to reconnect to servers are now more aggressive.
2781 Details : This builds on the enhanced upcall-less recovery that was added
2782 in 1.4.2. When trying to reconnect to servers, clients will
2783 now try each server in the failover group every 10 seconds. By
2784 default, clients would previously try one server every 25 seconds.
2789 Description: After recovery, certain operations trigger a failed
2790 assertion on a client.
2791 Details : Failing over an mds, using lconf -d --failover, while a
2792 client was doing a readdir() call would cause the client to
2793 LBUG after recovery completed and the readdir() was resent.
2795 Severity : enhancement
2797 Description: Default groups are now added by lconf
2798 Details : You can now run lconf --group <servicename> without having to
2799 manually add groups with lmc.
2802 Frequency : occasional
2804 Description: Nodes with an elan id of 0 trigger a failed assertion
2807 Frequency : always when accessing e.g. tty/console device nodes
2809 Description: tty and some other devices nodes cannot be used on lustre
2810 Details : file's private_data field is used by device data and lustre
2811 values in there got lost. New field was added to struct file to
2812 store fs-specific private data.
2815 Frequency : when exporting Lustre via NFS
2817 Description: NFSD failed occasionally when looking up a path component
2818 Details : NFSD is looking up ".." which was broken in ext3 directories
2819 that had grown large enough to become hashed.
2822 Frequency : Clusters with multiple interfaces not on the same subnet
2824 Description: Nodes will repeatedly try to reconnect to an interface which it
2825 cannot reach and report an error to the log.
2826 Details : Extra peer list entries will be created by lconf with some peers
2827 unreachable. lconf now validates the peer before adding it.
2830 Frequency : Only if a default stripe is set on the filesystem root.
2832 Description: Setting a default stripe on the filesystem root prevented the
2833 filesystem from being remounted.
2834 Details : The client was sending extra request flags in the root getattr
2835 request and did not allocate a reply buffer for the dir EA.
2838 Frequency : occasional, higher if lots of files are accessed by one client
2839 Bugzilla : 6159, 6097
2840 Description: Client trips assertion regarding lsm mismatch/magic
2841 Details : While revalidating inodes the VFS looks up inodes with ifind()
2842 and in rare cases can find an inode that is being freed.
2843 The ll_test_inode() code will free the lsm during ifind()
2844 when it finds an existing inode and then the VFS later attaches
2845 this free lsm to a new inode.
2849 Bugzilla : 6422, 7030
2850 Description: MDS deadlock between mkdir and client eviction
2851 Details : Creating a new file via mkdir or mknod (starting a transaction
2852 and getting the ns lock) can deadlock with client eviction
2853 (gets ns lock and trying to finish a synchronous transaction).
2856 Frequency : occasional
2857 Description: While starting a server, the fsfilt_ext3 module could not be
2859 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2860 kernels. Previously, lconf would still use the ext3 name
2861 when trying to load modules. Now, it will correctly use
2862 ext3 on 2.4 and ldiskfs on 2.6.
2864 Severity : enhancement
2865 Description: The default stripe count has been changed to 1
2866 Details : The interpretation of the default stripe count (0, to lfs
2867 or lmc) has been changed to mean striping across a single
2868 OST, rather than all available. For general usage we have
2869 found a stripe count of 1 or 2 works best.
2871 Severity : enhancement
2872 Description: Add support for compiling against Cray portals.
2873 Details : Conditional compiling for some areas that are different
2877 Frequency : occasional
2878 Bugzilla : 6409, 6834
2879 Description: Creating files with an explicit stripe count may lead to
2880 a failed assertion on the MDS
2881 Details : If some OSTs are full or unavailable, creating files may
2882 trigger a failed assertion on the MDS. Now, Lustre will
2883 try to use other servers or return an error to the
2887 Frequency : occasional
2889 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2890 caused lock timeout during readahead (since 1.4.2).
2891 Details : Processes doing readahead might match a lock that hasn't been
2892 granted yet if there are overlapping and conflicting lock
2893 requests. The readahead process waits on ungranted lock
2894 (original lock is CBPENDING), while OST waits for that process
2895 to cancel CBPENDING read lock and eventually evicts client.
2897 Severity : enhancement
2899 Description: Initial enabling of flock support for clients
2900 Details : Implements fcntl advisory locking and file status functions.
2901 This feature is provided as an optional mount flag (default
2902 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2903 locking are implemented yet, and those that are are not guaranteed
2904 to be completely correct in production environments.
2905 mount -t lustre -o [flock|noflock] ...
2908 Frequency : occasional
2910 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2911 assertion in the ext3 JBD (journaling) layer.
2912 Details : The b_committed_data struct is protected by the big kernel lock
2913 in 2.4 kernels, serializing journal_commit_transaction() and
2914 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2915 access to this struct. In 2.6 kernels there is finer grained
2916 locking to improve SMP performance of the JBD layer.
2920 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2925 Description: merge and cleanup kernel patches.
2928 -----------------------------------------------------------------------------
2930 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2932 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2933 versions of Lustre. Please update all servers and clients to
2934 version 1.4.2 or later at the same time. You must also run
2935 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2936 to update the configuration logs.
2938 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2939 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2940 - swab llog records of type '0' so we get proper header size/idx (5861)
2941 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2942 - fix rename of one directory over another leaking an inode (5953)
2943 - avoid SetPageDirty on 2.6 (5981)
2944 - don't re-add just-being-destroyed locks to the waiting list (5653)
2945 - when creating new directories, inherit the parent's custom
2946 striping settings if present parent (3048)
2947 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2948 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2949 - don't hold client locks on temporary worklist from l_lru (5666)
2950 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2951 - automatically evict dead clients (5921)
2952 - Update file size properly in create+truncate+fstat case (6196)
2953 - Do not unhash mountpoint dentries, do not allow removal of
2955 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2956 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2957 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2958 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2959 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2960 - Correctly handle reads of files with no objects (6243)
2961 - lctl recover will also mark a device active if deactivate used (5933)
2963 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2964 - allow --write-conf on an MDS with different nettype than client (5619)
2965 - don't write config llogs to MDS for mounts not from that MDS (5617)
2966 - lconf should create multiple TCP connections from a client (5201)
2967 - init scripts are now turned off by default; run chkconfig --on
2968 lustre and chkconfig --on lustrefs to use them
2969 - upcalls are no longer needed for clients to recover to failover
2971 - add --abort-recovery option to lconf to abort recovery on device
2973 - add support for an arbitrary number of OSTs (3026)
2974 - Quota support protocol changes.
2975 - forward compatibility changes to wire structs (6007)
2976 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2977 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2978 - improved "lustre status" script
2979 - initialize blocksize for non-regular files (6062)
2980 - added --disable-server and --disable-client configure options (5782)
2981 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2982 - Vanilla 2.4.29 support
2983 - increase maximum number of obd devices to 520 (6242)
2984 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2985 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2986 SLES 9 2.6.5 kernels
2987 - limit stripes per file to 160 (the maximum EA size) (6093)
2989 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2992 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2993 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2994 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2995 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2996 - small ext3 extents cleanups and fixes (5733)
2997 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2998 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2999 - avoid needless client->OST connect, fix handle mismatch (5317)
3000 - fix DLM error path that led to out-of-sync client, long delays (5779)
3001 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3002 - fix several locking issues related to i_size (5492,5624,5654,5672)
3003 - don't move pending lock onto export if it is already evicted (5683)
3004 - fix kernel oops when creating .foo in unlinked directory (5548)
3005 - fix deadlock in obdfilter statistics vs. object create (5811)
3006 - use time_{before,after} to avoid timer jiffies wrap (5882)
3007 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3008 - Do not leak request if server was not able to process it (5154)
3009 - If mds_open unable to find parent dir, make that negative lookup(5154)
3010 - don't create new directories with extent-mapping (5909, 5936)
3012 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3013 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3014 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3015 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3016 - support for RHEL4 kernel on i686 (5773)
3017 - provide error messages when incompatible logs are encountered (5898)
3019 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3020 * version 1.4.0.10 (1.4.1 release candidate 1)
3022 - don't keep a lock reference when lock is not granted (4238)
3023 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3024 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3025 - drop import inflight refcount on signal_completed_replay error (5255)
3026 - unlock page after async write error during send (3677)
3027 - handle missing objects in filter_preprw_read properly (5265)
3028 - no transno return for symlink open, don't save no-trasno open (3440)
3029 - don't try to complete elan receive that already failed (4012)
3030 - free RPC server reply state on error (5406)
3031 - clean up thread from ptlrpc_start_thread() on error (5160)
3032 - readahead could read extra page into cache that wasn't ejected (5388)
3033 - prevent races in class_attach/setup/cleanup/detach (5260)
3034 - don't dereference de->d_inode after l_dput of de (5458)
3035 - use "int" for stripe value returned from lock_to_stripe (5544)
3036 - mballoc allocation and error-checking fixes in 2.6 (5504)
3037 - block device patches to fix I/O request sizes in 2.6 (5482)
3038 - look up hostnames for IB nals (5602)
3039 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3040 - don't start multiple acceptors for the same port (5277)
3041 - fix incorrect LASSERT in mds_getattr_name (5635)
3042 - export a proc file for general "ping" checking (5628)
3043 - fix "lfs check" to not block when the MDS is down (5628)
3045 - service request history (4965)
3046 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3047 - create an "evict_client" /proc entry on OSTs, like the MDS has
3048 - fix mount usage message, return errors per mount(8) (5168)
3049 - change grep [] to grep "[]" in tests so they work in more UMLs
3050 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3051 - remove ancient LOV_MAGIC_V0 EA support (5047)
3052 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3053 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3054 - allow readahead allocations to fail when low on memory (5383)
3055 - mmap locking landed again, after considerable improvement (2828)
3056 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3058 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3061 - send OST transaction number in read/write reply to free req (4966)
3062 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3063 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3064 - the watchdog thread now runs as interruptible (5246)
3065 - flock/lockf fixes (but it's still disabled, pending 5135)
3066 - don't use EXT3 constants in llite code (5094)
3067 - memory shortage at startup could cause assertion (5176)
3069 - reorganization of lov code
3070 - single portals codebase
3072 - add extents/mballoc support (5025)
3073 - direct I/O reads in the obdfilter (4048)
3074 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3076 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3079 - send OST transaction number in read/write reply to free req (4966)
3080 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3081 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3082 - the watchdog thread now runs as interruptible (5246)
3083 - handle missing objects in filter_preprw_read properly (5265)
3084 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3085 - drop import inflight refcount on signal_completed_replay error (5255)
3086 - unlock page after async write error during send (3677)
3087 - return original error code on reconstructed replies (3761)
3088 - no transno return for symlink open, don't save no-trasno open (3440)
3090 - add pid to ldlm debugging output (4922)
3091 - bump the watchdog timeouts -- we can't handle 30sec yet
3092 - extra debugging for orphan dentry/inode bug (5259)
3094 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3097 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3098 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3099 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3100 - fix NULL dereference in /proc/sys/portals/routes (4827)
3101 - allow failed mdc_close() operations to be interrupted (4561)
3102 - stop precreate on OST before MDS would time out on it (4778)
3103 - don't send partial-page writes before EOF from client (4410)
3104 - discard client grant for sub-page writes on large-page clients (4520)
3105 - don't free dentries not owned by NFS code, check generation (4806)
3106 - fix lsm leak if mds_create_objects() fails (4801)
3107 - limit debug_daemon file size, always print CERROR messages (4789)
3108 - use transno after validating reply (3892)
3109 - process timed out requests if import state changes (3754)
3110 - update mtime on OST during writes, return in glimpse (4829)
3111 - add mkfsoptions to LDAP (4679)
3112 - use ->max_readahead method instead of zapping global ra (5039)
3113 - don't interrupt __l_wait_event() during strace
3115 - add software watchdogs to catch hung threads quickly (4941)
3116 - make lustrefs init script start after nfs is mounted
3117 - fix CWARN/ERROR duplication (4930)
3118 - return async write errors to application if possible (2248)
3119 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3120 - print NAL number in %x format (4645)
3121 - update barely-supported suse-2.4.21-171 series (4842)
3122 - support for sles 9 %post scripts
3123 - support for building 2.6 kernel-source packages
3124 - support for sles km_* packages
3126 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3129 - ignore -ENOENT errors in osc_destroy (3639)
3130 - notify osc create thread that OSC is being cleaned up (4600)
3131 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3132 - reconstruct ost_handle() like mds_handle() (4657)
3133 - create a new thread to do import eviction to avoid deadlock (3969)
3134 - let lconf resolve symlinked-to devices (4629)
3135 - don't unlink "objects" from directory with default EA (4554)
3136 - hold socknal file ref over connect in case target is down (4394)
3137 - allow more than 32000 subdirectories in a single directory (3244)
3138 - fix blocks count for O_DIRECT writes (3751)
3139 - OST returns ENOSPC from object create when no space left (4539)
3140 - don't send truncate RPC if file size isn't changing (4410)
3141 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3142 - bind to privileged port in socknal and tcpnal (3689)
3144 - rate limit CERROR/CWARN console message to avoid overload (4519)
3145 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3146 - basic mmap support (3918)
3147 - kernel patch series update from b1_4 (4711)
3149 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3152 - avoid crash during MDS cleanup with OST shut down (2775)
3153 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3154 - don't use bad inodes on the MDS (3744)
3155 - dynamic object preallocation to improve recovery speed (4236)
3156 - don't hold spinlock over lock dumping or change debug flags (4401)
3157 - don't zero obd_dev when it is force cleaned (3651)
3158 - print grants to console if they go negative (4431)
3159 - "lctl deactivate" will stop automatic recovery attempts (3406)
3160 - look for existing locks in ldlm_handle_enqueue() (3764)
3161 - don't resolve lock handle twice in recovery avoiding race (4401)
3162 - revalidate should check working dir is a directory (4134)
3164 - don't always mark "slow" obdfilter messages as errors (4418)
3166 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3169 - don't close LustreDB during write_conf until it is done (3860)
3170 - fix typo in lconf for_each_profile (3821)
3171 - allow dumping logs from multiple threads at one time (3820)
3172 - don't allow multiple threads in OSC recovery (3812)
3173 - fix debug_size parameters (3864)
3174 - fix mds_postrecov to initialize import for llog ctxt (3121)
3175 - replace config semaphore with spinlock (3306)
3176 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3177 - don't allow enqueue to complete on a destroyed export (3822)
3178 - down write_lock before checking llog header bitmap (3825)
3179 - recover from lock replay timeout (3764)
3180 - up llog sem before sending rpc (3652)
3181 - reduce ns lock hold times when setting kms (3267)
3182 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3183 - fix NULL deref and obd_dev leak on setup error (3312)
3184 - replace some LBUG about llog ops with error handling (3841)
3185 - don't match INVALID dentries from d_lookup and spin (3784)
3186 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3187 - fix invalid assertion in ptlrpc_set_wait (3880)
3189 - add libwrap support for the TCP acceptor (3996)
3190 - add /proc/sys/portals/routes for non-root route listing (3994)
3191 - allow setting MDS UUID in .xml (2580)
3192 - print the stack of a process that LBUGs (4228)
3194 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3197 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3198 - make sure to unset replay flag from failed open requests (3440)
3199 - if default stripe count is 0, use OST count for inode size (3636)
3200 - update parent mtime/ctime on client for create/unlink (2611)
3201 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3202 - free recovery state on server during a forced cleanup (3571)
3203 - unregister_reply for resent reqs (3063)
3204 - loop back devices mounting and status check on 2.6 (3563)
3205 - fix resource-creation race that can provoke i_size == 0 (3513)
3206 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3207 - more debugging for page-accounting assertion (3746)
3208 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3209 - avoid deadlock after precreation failure (3758)
3210 - fix race and lock order deadlock in orphan handling (3450, 3750)
3211 - add validity checks when grabbing inodes from l_ast_data (3599)
3213 - add /proc/.../recovery_status to obdfilter (3428)
3214 - lightweight CDEBUG infrastructure, debug daemon (3668)
3215 - change default OSC RPC parameters to be better on small clusters
3216 - turn off OST read cache for files smaller than 32MB
3217 - install man pages and include them in rpms (3100)
3218 - add new init script for (un)mounting lustre filesystems (2593)
3219 - run chkconfig in %post for init scripts (3701)
3220 - drop scimac NAL (unmaintained)
3222 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3225 - clean kiobufs before and after use (3485)
3226 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3227 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3228 - change init script to fail more clearly if not run as root (1528)
3229 - allow clients to reconnect during replay (1742)
3230 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3231 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3232 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3233 - fix cond_resched (3554)
3234 - only evict unfinished clients after recovery (3515)
3235 - allow bulk resend, prevent data loss (3570)
3236 - dynamic ptlrpc request buffer allocation (2102)
3237 - don't allow unlinking open directory if it isn't empty (2904)
3238 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3239 - remove extraneous obd dereference causing LASSERT failure (3334)
3240 - don't use get_cycles() when creating temp. files on the mds (3156)
3241 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3242 - handle EEXIST for set-stripe, set proper directory name (3336)
3244 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3245 - fix ksocknal_fmb_callback() error messages (2918)
3247 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3250 - don't copy lvb into (possibly NULL) reply on error (2983)
3251 - don't deref dentry after dput, don't free lvb on error (2922)
3252 - use the kms to determine writeback rpc length (2947)
3253 - increment oti_logcookies when osc is inactive (2948)
3254 - update client's i_blocks count via lvb messages (2543)
3255 - handle intent open/close of special files properly (1557)
3256 - mount MDS with errors=remount-ro, like obdfilter (2009)
3257 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3258 - don't use cancelling-locks' kms values (2947)
3259 - use highest lock extent for kms, not last one (2925)
3260 - don't dereference ERR_PTR() dentry in error handling path (3107)
3261 - fix thread race in portals_debug_dumplog() (3122)
3262 - create lprocfs device entries at setup instead of at attach (1519)
3263 - common AST error handler, don't evict client on completion race (3145)
3264 - zero nameidata in detach_mnt in 2.6 (3118)
3265 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3266 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3267 - initialize RPC timeout timer earlier for 2.6 (3219)
3268 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3269 - print nal/nid for unknown nid (3258)
3270 - additional checks for oscc recovery before doing precreate (3284)
3271 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3272 - don't crash in mdc_close for bad permissions on open (3285)
3273 - zero i_rdev for non-device files (3147)
3274 - clear page->private before handing to FS, better assertion (3119)
3275 - tune the read pipeline (3236)
3276 - fix incorrect decref of invalidated dentry (2350)
3277 - provide read-ahead stats and refine rpc in flight stats (3328)
3278 - don't hold journal transaction open across create RPC (3313)
3279 - update atime on MDS at close time (3265)
3280 - close LDAP connection when recovering to avoid server load (3315)
3281 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3282 - don't leak open file on MDS after open resend (3325)
3283 - serialize filter_precreate and filter_destroy_precreated (3329)
3284 - loop device shouldn't call sync_dev() for nul device (3092)
3285 - clear page cache after eviction (2766)
3286 - resynchronize MDS->OST in background (2824)
3287 - refuse to mount the same filesystem twice on same mountpoint (3394)
3288 - allow llmount to create routes for mounting behind routers (3320)
3289 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3290 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3291 - fix rare problem with rename on htree directories (3417)
3293 - allow default OST striping configuration per directory (1414)
3294 - fix compilation for qswnal for 2.6 kernels (3125)
3295 - increase maximum number of MDS request buffers for large systems
3296 - change liblustreapi to be useful for external progs like lfsck (3098)
3297 - increase local configuration timeout for slow disks (3353)
3298 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3300 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3303 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3304 - don't overwrite extent policy data in reply if lock was blocked (2901)
3305 - drop filter export grants atomically with removal from device (2663)
3306 - del obd_self_export from work_list in class_disconnect_exports (2908)
3307 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3308 - swab reply message in mdc_close, other PPC fixes (2464)
3309 - fix destroying of named logs (2325)
3310 - overwrite old logs when running lconf --write_conf (2264)
3311 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3312 - fix race in target_handle_connect (2898)
3313 - mds_reint_create() should take same inode create lock (2926)
3314 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3315 - don't close files for self_export to avoid uninitialized obd (2936)
3316 - allow MDS with the same name as client node (2939)
3317 - hold dentry reference for closed log files for unlink (2325)
3318 - reserve space for all logs during transactions (2059)
3319 - don't evict page beyond end of stripe extent (2925)
3320 - don't oops on a deleted current working directory (2399)
3321 - handle hard links to targets without a parent properly (2517)
3322 - don't dereference NULL lock when racing during eviction (2867)
3323 - don't grow lock extents when lots of conflicting locks (2919)
3325 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3328 - account for cache space usage on clients to avoid data loss (974)
3329 - lfsck support in lustre kernel code (2349)
3330 - reduce journal credits needed for BRW writes (2370)
3331 - orphan handling to avoid losing space on client/server crashes
3332 - ptlrpcd can be blocked, stopping ALL progress (2477)
3333 - use lock value blocks to assist in proper KMS, faster stat (1021)
3334 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3335 - recovery for initial connections (2355)
3336 - fixes for mds_cleanup_orphans (1934)
3337 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3338 - block all file creations until orphan recovery completes (1901)
3339 - client remove rq_connection from request struct (2423)
3340 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3341 - recovery timer race (2670)
3342 - mdc_close recovey bug (2532)
3343 - ptlrpc cleanup bug (2710)
3344 - mds timeout on local locks (2588)
3345 - namespace lock held during RPCs (2431)
3346 - handle interrupted sync write properly (2503)
3347 - don't try to handle a message that hasn't been replied to (2699)
3348 - client assert failure during cleanup after abort recovery (2701)
3349 - leak mdc device after failed mount (2712)
3350 - ptlrpc_check_set allows timedout requests to complete (2714)
3351 - wait for inflight reqs when ptlrpcd finishes (2710)
3352 - make sure unregistered services are removed from the srv_list
3353 - reset bulk XID's when resending them (caught by 1138 test)
3354 - unregister_bulk after timeout
3355 - fix lconf error (2694)
3356 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3357 - readahead locks pages, leaves pending causing memory pressure (2673)
3358 - increase OST request buffers to 4096 on large machines (2729)
3359 - fix up permission of existing directories in simple_mkdir (2661)
3360 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3361 - don't assign transno to errored transactions (2742)
3362 - don't delete objects on OST if given a bogus objid from MDS (2751)
3363 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3364 - if rq_no_resend, then timeout request after recovery (2432)
3365 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3366 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3367 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3368 - MDS recovery shouldn't delete objects that it has given out (2730)
3369 - if enqueue arrives after completion, don't clobber LVB (2819)
3370 - don't unlock pages twice when trigger_group_io returns error (2814)
3371 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3372 - don't write pages to disk if there was an error (1450)
3373 - don't ping imports that have recovery disabled (2676)
3374 - take buffered bytes into account when balancing socknal conn (2817)
3375 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3376 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3377 - remove little-endian swabbing of llog records (1987)
3378 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3379 - retry reposting request buffers if they fail (1191)
3380 - grow extent at grant time to avoid granting a revoked lock (2809)
3381 - lock revoke doesn't evict page if covered by a second lock (2765)
3382 - disable VM readahead to avoid reading outside lock extents (2805)
3384 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3385 - updated kernel patches for hp-2.4.20 kernel (2681)
3387 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3390 - fix truncated write corruption (2366)
3391 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3393 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3394 - ensure len(uuid) < 37 in lmc (1171)
3395 - fix ia64 OOPS in llog_test (2255)
3396 - zero end of page at obdfilter for partial page writes (2648)
3397 - don't leave stale dentries around after renames (bug 2428)
3398 - fix timeouts when evicting a client with a single lock held (2642)
3399 - set deadline for the initial HELLO message to drain (2634)
3400 - print out dotted-quad IP addresses in the socknal (2302)
3402 - additional debugging for MDS client eviction problem (2443)
3403 - fix mkfsoptions support for osts (2603, 2604)
3405 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3408 - add series for the vanilla 2.6.0 kernel
3409 - add series for the vanilla 2.4.24 kernel
3410 - add series for a cray x86/64 UL kernel drop
3411 - fix xattr patches for the vanilla 2.4.19 series
3413 - generate true UUIDs in lmc (1171)
3414 - have portals stack dumping break in UML (2466)
3415 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3416 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3417 - wake callback waiting threads on client eviction (2460)
3418 - Add --ptldebug and --subsystem to lmc (1719)
3419 - update assertion to allow safe interrupt allocation
3420 - set rq_no_resend for cancel requests (2432)
3421 - recalculate ptlrpcd timeout after resend (2494)
3422 - call vfs_rmdir when removing pending directories (2368)
3423 - fix renaming a file to itself (2429)
3424 - lmc creates a default one-stripe lov (2454)
3425 - expand procfs space to handle large clusters (2326)
3426 - increase UML stack to avoid overflow
3427 - update lconf's list of debug and subsystem masks
3428 - fix lfs find --obd (2510)
3429 - /proc tunable for disabling filter read caching (2591)
3430 - stop rpm packages from altering slapd.conf (2301)
3431 - disable nagle in the socknal under 0conf (2578)
3432 - choose mds inode size based on stripe count (2572)
3433 - fix kernel-source rpm problems (2516)
3435 - add --disable-doc to avoid pdf generation (2421)
3436 - update documentation, tests, type-os, comments
3437 - avoid format warnings on ia64
3438 - remove the TOE NAL
3439 - tiny code cleanups by removing unused fields
3441 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3444 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3445 - remove the most blatant lies from BUILDING file (2371)
3446 - change default debug level to reasonable production setting
3447 - reduce client side cache size to reduce cache flush time
3448 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3449 - make TCP zerocopy and pinger support enabled by default (2476)
3450 - sync writes completed after process exits caused crashes (2319)
3451 - maintain correct mount count on the MDS (2356)
3452 - backout 1557, because 2316 wasn't really fixed
3453 - better file I/O statistics gathering in /proc
3454 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3455 - another kernel patch to fix zero-copy TCP function export
3456 - don't take duplicate lock when processing re-sent getattr (2420)
3457 - lctl uses obd_self_export instead of creating new conn (2353)
3458 - MDS/OST recovery case which requires object creation asserted (2425)
3459 - move lfs from /usr/sbin to /usr/bin in packages
3460 - fix race between mds_client_add and mds_client_free (2417)
3461 - use kmalloc instead of slabs in portals (2430)
3462 - don't create duplicate records when a failover MDS is present (2442)
3463 - remove unnecessary mount age check (2332)
3464 - don't remove directory inodes from locks prematurely (2451)
3465 - don't break if MDS service name is the same as hostname (2103)
3466 - fix races in client write RPC generation when cache full (2482)
3468 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3471 - remove now-unused request->rq_obd (278)
3472 - if an allocation fails, print out how much memory we've used (1933)
3473 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3474 - add the "configurable stack size" patch to most series files (1256)
3475 - ability to write large log records, for 100+ OST configs (2306)
3476 - fix NULL deref when filter_prep fails (2314)
3477 - fix operator precedence error in filter_sync
3478 - dynamic allocation of socknal TX descriptors (2315)
3479 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3480 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3481 - ensure that CWARN messages in Portals always get to the syslog
3482 - __init/__exit are not for prototype decls (ldlm_init/exit)
3483 - x86-64 compile warning fixes
3484 - fix gateway LMC keyword conflict (2318)
3485 - fix MDS lock inversions in getattr/reint paths (1844)
3486 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3487 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3488 - DLM race condition prevented some lock evictions (2328)
3489 - ENOMEM detection and retry on socknal sends (2230)
3490 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3491 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3492 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3493 - better error messages when a client is rejected during recovery (1505)
3494 - avoid cancelling locks which were never granted, after failure (2330)
3495 - fix i_sem/journal inversion in mds_client_add (2333)
3496 - fix truncate/getattr lock cycle deadlock (2334)
3497 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3498 - fix two rare exit paths which could leak an l_lock() ref (2321)
3499 - fencepost error in MDS/OST orphan recovery (2226)
3500 - make log record alignment 8 bytes (1988)
3501 - lstripe now fails when requested offset > ost_count (2237)
3502 - ensure that all kernel series have a complete list.h (1607)
3503 - fix crashes in special-file operations (2316)
3504 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3506 - allow configurable automake binary, for testing new versions
3507 - small update to the lfs documentation
3509 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3511 * fix negative export reference count in fsfilt_sync (2312)
3513 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3514 * release candidate 0.9.1
3516 - orphans are moved into the PENDING directory for possible recovery
3517 - replayed opens now open by fid for orphan/rename safety (1042)
3518 - last close of an orphan inode generates a transno (683)
3519 - chdir() and mount() now pin the directory entry (1020)
3520 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3521 - discard very old requests without processing them (1502)
3522 - remove some common, well-understood CERRORs (1505)
3523 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3524 - clear "grant" flags in OST replies until OST grant code lands (1644)
3525 - fix read performance by not clobbering i_blksize on client (1598)
3526 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3527 - make LRU size a /proc tunable, clears locks when reduced (707)
3528 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3529 - server locks take references on exports now (1558)
3530 - build fixes for 2.4.20-rh trees (1663)
3531 - return an error from lov_create if all OSCs are inactive (1751)
3532 - fix import levels when a reconnect happens without a timeout (1597)
3533 - exit early from mds_open if we get a lookup error (1749)
3534 - partial page read at EOF wouldn't wait for disk before sending (1642)
3535 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3536 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3537 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3538 - write/truncate lock inversion (1639)
3539 - Don't auto-load obdclass, portals modules during cleanup (1495)
3540 - fix timestamps from jumping to "now" (1763)
3541 - extra journal assertions (1648)
3542 - add an extra multiunlink test (1771)
3543 - fix read_record/write_record API (1776)
3544 - fix leak of offset_extent, possible incorrect i_size later (1772)
3545 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3546 - Debugging for the kqswnal_get_idle_tx problems (1820)
3547 - Allow recovery to be attempted multiple times (1536)
3548 - Write out MDS last_rcvd file after it is first created (1600)
3549 - Fix tx_descriptor leak in failed transmit situations (1827)
3550 - ext3 journaling fixes for assertion failure after IO error (1871)
3551 - class_export_put() on freed export after completion AST error (1896)
3552 - Fix revalidate looping in VFS (1322)
3553 - Don't access a freed export during MDS_REINT timeout (1521)
3554 - Add open-unlink recovery support on the MDS (1673,1764)
3555 - Return an error if no MDS data was read from last_rcvd (1946)
3556 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3557 - Don't setup a disk device that doesn't match exported UUID (317)
3558 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3559 - avoid committing NULL handle in force close
3560 - local.sh is now a one-stripe LOV configuration
3561 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3562 - fix ext3 htree duplicate directory entry corruption (1516)
3563 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3564 - update to new LOV EA format (2097)
3565 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3567 2003-06-15 Phil Schwan <phil@clusterfs.com>
3570 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3571 - per-import/export recovery handling (958, 931, 959)
3572 - multiple last-rcvd slots, for serving multiple FSes (949)
3573 - connections are again shared between multiple imp/exports (963, 964)
3574 - "umount -f" would hang if any requests needed to be sent (393, 978)
3575 - avoid pinning large req buffer by copying for queued messages (989)
3576 - add "uuid" to "lctl device" command to help upcalls (991)
3577 - "open" RPCs with transnos would confuse recovery counters (1037)
3578 - do proper endian conversion of all wire messages (288, 340, 891)
3579 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3580 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3581 - fix ptlrpc_connection leak in target_handle_connect (1174)
3582 - fix import refcounting bug in OST and MDS cleanup (1134)
3583 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3584 - fix very unlikely obd_types race condition (501)
3585 - remove osc_open hack for echo_client (1187)
3586 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3587 - a failure in read_inode2 leads to deadlock (1139)
3588 - cancel ack-locks as soon as transaction is committed (1072)
3589 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3590 - make sure to commitrw after any preprw to avoid deadlock (1162)
3591 - failing to execute a file in a lustre FS would lock inode (1203)
3592 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3593 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3594 - obd_uuid2tgt would walk off the end of the list (1255)
3595 - on IA64 the osc would give portals incorrect bulk size (1258)
3596 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3597 - fix lock inversion caused by new llite matching code (1282)
3598 - limit the number of dirty pages on a client to 10MB (1286)
3599 - timed out locks were not being corrected cancelled (1289)
3600 - fix O_DIRECT above 4GB on IA-32 (1292)
3601 * major user-visible changes
3602 - fail out/fail over policy now controlled by the upcall (993)
3604 - add OBD_PING to check server availability and failure (954)
3605 - lustre messages are now sent in sending host order (288, 340, 891)
3606 - add eadatalen to MDS getattr reply (340)
3607 - OST read replies may contain second buffer, with per-page status (593)
3609 2003-03-11 Phil Schwan <phil@clusterfs.com>
3612 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3613 - failed open()s could cause deadlock; fixed (867, 869)
3614 - stop cancelling OST locks when files are closed (481)
3615 - overlapping XID spaces caused network corruption (851, 853)
3616 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3617 - setattr_raw added, to do single-RPC, server-side setattrs
3618 - lmc/lconf syntax change for OST UUIDs
3619 - fix crashy race condition between ptlrpc_free_req and osc_close
3620 - don't use request in mdc_enqueue if we hit a timeout (889)
3621 - don't set the inode i_size for regular files from the MDS (896)
3622 - handle out of order completion AST (842)
3623 - don't LBUG if a lock request times out after receiving AST (913)
3624 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3625 - if a bad lock AST arrives, send an error instead of dropping entirely
3626 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3627 - fix leak in bulk IO when only partially completed (899, 900, 926)
3628 - fix O_DIRECT for ia64 (55)
3629 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3630 - C-z after timeout could hang a process forever; fixed (977)
3632 - client-side I/O cache (678, 924, 929, 941, 970)
3634 - READPAGE and SETATTRs which don't take server-side locks get
3637 2003-02-11 Phil Schwan <phil@clusterfs.com>
3640 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3641 - Fix fsfilt_extN_readpage() to read a full page of directory
3642 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3643 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3644 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3645 - Most intent operations were removed, in favour of a new RPC mode
3646 that does a single RPC to the server and bypasses most of the VFS
3647 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3648 - Aggressively cancel local locks on DLM servers
3649 - mds_reint_unlink sends EA to the client if it's the last nlink.
3650 client uses that EA to unlink OST objects.
3651 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3652 - recursive symlinks were fixed (439)
3653 - fixed NULL deref in DEBUG_REQ
3654 - filter_update_lastobjid no longer calls sync, which annoyed extN
3655 - fixed multi-client small-writes to a single file problem (445)
3656 - fixed mtime updates during file writes (607)
3657 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3658 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3659 - fixed Linux OST size reporting problem (444, 656)
3660 - OST now updates object mtime with writes or setattr (607, 619)
3661 - client verifies file size before zeroing page past EOF (445)
3662 - OST now writes last allocated objid to disk with allocation (108)
3663 - LOV on echo now works (409)
3665 - mds_reint_unlink sends a new buffer, with the EA included. this
3666 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3667 set if a regular file was being unlinked, and it was the last link
3668 - use PtlGet from the target for bulk writes (315)
3669 - OST now updates object mtime with writes or setattr (607, 619)
3670 - LDLM now has a grant-time callback to revalidate locked items, if
3672 - Many MDS operations were reorganized to combat race conditions
3674 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3675 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3677 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3680 - Fully reactivate OST imports after reconnection (512, others)
3681 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3682 - More graceful error handling for truncating on dead OST (515)
3683 - Don't error out unless we're actually accessing dead stripes (474)
3684 - Fix garbage sizes when stripes are missing (410)
3685 - LRU counters were broken, causing constant lock purge (433, 432)
3686 - garbage on read from stripes with failed OSTs (441)
3687 - mark OSCs as active before reconnecting during recovery (438)
3688 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3689 - lfind did not preserve OST order in output (443)
3690 - symlinks cause hung clients, incorrect data (439)
3691 - stop dereferencing request after dropping refcount (457)
3692 - don't LASSERT(spin_is_locked) on non-SMP (455)
3693 - fixes for many rename() bugs
3694 - fstat didn't correctly synchronize attributes (399)
3695 - server must handle lock cancellation during blocking AST prep (487)
3696 - bulk descriptors were free()d too soon (511)
3697 - fix paths in lconf, which would load incorrect modules (451, 507)
3698 - fix confusing lconf 'host not found' error message (386)
3699 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3700 - fix race condition in mdc_blocking_ast() for inode access (526)
3701 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3702 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3703 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3704 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3705 - fix lstripe in several ways (kernel side) (527)
3706 - fix request leak in ldlm_cli_enqueue (262)
3707 - incorrect OSC was marked inactive after OST failure
3708 - call mds_fs_cleanup before unmounting filesystem (524)
3709 - fix races between taking ns_lock and ldlm_lock_change_resource
3710 - fix races updating LOV export open file list
3711 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3712 - fix recovery NULL deref in ldlm_cli_cancel_unused
3713 - fix some DLM races by using new hash table for lock handles (419)
3714 - permit the client to specify desired inodes, at replay
3715 - duplicate requests when we queue them for replay reintegration
3716 - fix last_rcvd offset calculation
3717 - sync after each recovered transaction, so we always make progress
3718 - never, not always, ERESTART requests without transnos
3719 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3721 - skip replay if the MDS says that the client is already connected
3722 - don't check for a recovery-enabled export to match lctl's UUID
3723 - don't INC_USE_COUNT for phantom exports
3724 - don't crash when cleaning up phantom exports (567)
3725 - don't double-finish or set replay data for errored mdc_open requests
3726 - abort requests when they time out, so we don't get old replies
3727 - send/receive replies for AST messages again
3728 - if the client says that it doesn't have the lock, cancel it on the
3730 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3731 mark it as destroyed, it will all work out when decref is called
3732 - fix module use counts (22, 581)
3734 - ASTs now expect a reply (server cancels lock on error reply)
3736 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3739 - fix many simultaneous client startup (392)
3740 - fix dentry->d_it clobbering
3741 - credentials weren't being shipped for readdir/getattr operations
3742 - remove invalid assertions triggered during some concurrent MD
3744 - proper Lustre versions added (336, 389)
3745 - fix memory leak for create error case (398)
3746 - fix LOV locking bug that would get cli/srv out of sync
3747 - fix echo client over LOV (409)
3748 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3749 - fix double-O_EXCL intent crash (424)
3750 - avoid sending multiple lock CANCELs (352)
3752 - MDS can do multi-client recovery (modulo bugs in new code)
3754 - many updates, edits, cleanups
3756 2002-11-18 Phil Schwan <phil@clusterfs.com>
3759 - fix null d_it dereference (346)
3760 - fix full OST/dbench hang (333)
3761 - fix permission problem with file removal (286)
3762 - fix removal of OSCs from LOV when they fail
3763 - fix NULL deref during bulk timeout (214)
3764 - fix problems related to multiple filesystems on one MDS (241)
3765 - fixed serious subtle metadata locking bugs
3766 - free locks on clients when inodes are removed due to memory
3768 - fix inode pointer in lock data (285)
3769 - partial support for multiple MDS on a single host (241)
3770 - data locks weren't cancelled at clear_inode time (290, 311)
3771 - intent locks could lead to unbounded lock growth (205)
3772 - added a maximum lock count, an LRU list, and a flusher
3773 - fix multiple rename (365)
3774 - properly abstracted the echo client
3775 - OSC locked 1 byte too many; fixed
3776 - rewrote brw callback code:
3777 - fixed recovery bugs related to LOVs (306)
3778 - fixed too-many-pages-in-one-write crash (191)
3779 - fixed (again) crash in sync_io_timeout (214)
3780 - probably fixed callback-related race (385)
3782 - Add capability to MDS protocol
3783 - LDLM cancellations and callbacks on different portals
3785 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3788 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3789 - defer open object destruction to close time (601981, 138)
3790 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3791 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3792 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3793 - don't call obd_setattr in ll_file_release for destroyed objects
3794 * protocol change to lustre_msg: move |version| and add |flags|
3795 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3796 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3797 * added replay of create, unlink, link and rename operations during
3798 MDS failover; recovery should be much more robust now
3799 * remove failed OSCs from LOVs (only lov_create uses this so far)
3800 * the lustre-HOWTO was brought (more) up to date (582544)
3802 2002-10-23 Phil Schwan <phil@clusterfs.com>
3805 - in-use dentries weren't being reused properly (617851)
3806 - prevent multiple LDLM setup (599178)
3807 - fix LOV size calculations for truncate (617853)
3808 - fix client handling of MDS intent errors (POSIX)
3809 - fix permission bug in lovstripe.c test (624321)
3810 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3811 - truncate past end of file could corrupt data
3812 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3813 - a race in recovery could return ETIMEDOUT to apps (623947)
3814 - building outside the source directory was fixed
3815 * the lustre-HOWTO was brought (more) up to date (582544)
3816 * major progress was made on recovery functionality
3818 2002-10-10 Phil Schwan <phil@clusterfs.com>
3821 - recovery deadlock fix
3822 - rm -rf causes LBUG fix (617817)
3823 - file open by multiple tasks fix (618962)
3824 - directory permissions bugs (602707 and 620007)
3825 - journal_stop fixed with locking (611313)
3826 - O_APPEND failures resolved (618273, perhaps 614459)
3827 - lconf PATH fix (619770)
3828 - IA64 build fix (621450)
3829 - RPC buffer sizes scale with amount of memory
3831 2002-10-01 Phil Schwan <phil@clusterfs.com>
3834 - locks would be cancelled without throwing away data pages,
3835 resulting in inconsistent data (605627)
3836 - inode attributes were not always being refreshed (605627, 612449)
3837 - lconf now continues to cleanup after lctl reports an error
3838 - MDS now enforces user permissions (602707)
3839 - lprocfs cleanup fixed, but not yet enabled (614157)
3840 - fixed infinite server hang, should a client not respond to an AST
3841 - avoid going into recovery if user calls readlink() with a buffer
3842 that's too small (613941)
3843 - AST RPCs no longer require replies (614867) -- this may be changed
3844 - don't crash server if client sends an IOV that's too big (611336)
3845 - fixed lock conversion deadlock (611892)
3846 - fixed the following of symlinks (614622)
3847 * recovery: the server can remove locks from a client that dies, other
3848 clients can make progress
3849 * more extN patch fixes
3850 * compile-time configurable ptlrpc buffer allocations
3852 - collaborative read cache document
3853 - Lustre Lite Performance CDR document-in-progress
3855 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3858 - fix typo in patch-2.4.18
3860 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3863 - clear ptlrpc request each time in handle_incoming_request()
3864 - unlink of files now destroys the object on the OST
3866 2002-09-19 Peter Braam <braam@clusterfs.com>
3868 * add hard link support
3869 * change obdfile creation method
3870 * kernel patch changed
3872 2002-09-19 Peter Braam <braam@clusterfs.com>
3875 - stack overflow bug in extN fixed
3877 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3879 * documentation updates
3880 - add man pages for config tools
3881 - update tests/README to describe testing with new config tools
3882 - finish metadata API descriptions
3883 * bug fixes and cleanups
3884 - statfs workaround for 16TB limit
3885 - LOV stripe allocation improved, can stripe on subset of OSTs
3886 - LOV file size/IO offset was wrong for files > 4GB in size
3887 - object EA data was being dropped, caused files to be unreadable
3888 - memory overflow with non-LOV OST caused memory corruption
3889 - fixed regression tests to work with new config tools, obdfilter
3890 - fixed bug when directory size became larger than 1 block
3891 - fixed bug (for single client case) when PWD was deleted
3892 - invalidate local directory pages when doing intent-based ops
3893 - avoid LDLM oops when lock callback contained bad data
3895 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3897 * documentation updates
3898 * bug fixes and cleanups
3899 - configuration tools
3902 - 64-bit compile warnings
3903 - 64-bit internal statfs data
3905 * test_brw on persistent OST devices
3907 * lprocfs (disabled)
3909 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3911 * documentation updates
3912 * bug fixes and cleanups
3913 * configuration tools
3915 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3918 * many small fixes to 0_5_4
3919 * io/network handling
3920 * thinkos in MDS operations
3922 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3925 * crucial basic fixes to 0.5.3
3926 * IOR, Iozone work over Elan
3929 2002-08-07 Phil Schwan <phil@clusterfs.com>
3930 * version 0_5_3, our first alpha
3931 * we use the new Portals iovs
3932 * documentation updates
3933 * bug fixes and cleanups
3934 * small changes in the DLM wire protocol
3936 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3937 * version 0_5_1 with some initial stability,
3938 * locking on MD and file I/O.
3939 * documentation updates
3940 * several bug fixes since 0.5.0
3941 * small changes in wire protocol
3943 2002-07-18 Phil Schwan <phil@clusterfs.com>
3945 * delivered as Lustre Light Alpha
3946 * fixed a crash after handling invalid MDS requests
3947 * fixed directory pages for architectures with non-4k pages sizes
3949 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3950 * release version v0_4_4
3951 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3953 * Fixes a number of interruption problems with OST operations.
3954 * Update documentation for portals header changes
3955 * Move all wire protocol structs/defines to lustre_idl.h
3956 * Fixes symlink length bug.
3957 * Add tcpdump to repository.
3959 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3960 * release version v0_4_3
3961 * Fixes statfs for inodes on extN.
3962 * Fixes bug in runtests which would delete /etc/hosts.
3963 * Use 64-bit object IDs wherever possible (not into VFS though)
3964 Remove ost_get_info, which is unused by lustre, and out of date.
3966 2002-07-03 Peter Braam <braam@clusterfs.com>
3967 * release version v0_4_2 Fixes a lookup error (type not passed)
3968 * move forward to head of Portals
3969 * move forward to latest Lustre kernel
3971 2002-06-25 Peter Braam <braam@clusterfs.com>
3972 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob