1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4 2.4.21-47.0.1.EL (RHEL 3)
7 2.6.12.6 vanilla (kernel.org)
8 2.6.16.27-0.9 (SLES 10)
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 - 2.6.22 vanilla (kernel.org)
13 * Recommended e2fsprogs version: 1.39.cfs8
14 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Severity : enhancement
19 Description: add optional extra BUILD_VERSION info
20 Details : add a new environment variable (namely LUSTRE_VERS) which allows
21 to override the lustre version.
24 Frequency : 2.6.18 servers only
26 Description: ll_kern_mount() doesn't release the module reference
27 Details : The ldiskfs module reference count never drops down to 0
28 because ll_kern_mount() doesn't release the module reference.
33 Description: server LBUG when using old ost_num_threads parameter
34 Details : Accept the old ost_num_threads parameter but warn that it
35 is deprecated, and fix an off-by-one error that caused an LBUG.
40 Description: Transient SCSI error results in persistent IO issue
41 Details : iobuf->dr_error is not reinitialized to 0 between two
45 Frequency : sometimes when underlying device returns I/O errors
47 Description: OSTs not going read-only during write failures
48 Details : OSTs are not remounted read-only when the journal commit threads
49 get I/O errors because fsfilt_ext3 calls journal_start/stop()
50 instead of the ext3 wrappers.
54 Description: poor connect scaling with increasing client count
55 Details : Don't run filter_grant_sanity_check for more than 100 exports
56 to improve scaling for large numbers of clients.
59 Frequency : SLES10 only
61 Description: sanity-quota.sh quotacheck failed: rc = -22
62 Details : Quotas cannot be enabled on SLES10.
65 Frequency : liblustre clients only
67 Description: getdirentries does not give error when run on compute nodes
68 Details : getdirentries does not fail when the size specified as an argument
69 is too small to contain at least one entry
71 Severity : enhancement
73 Description: Add LNET router traceability for debug purposes
74 Details : If a checksum failure occurs with a router as part of the
75 IO path, the NID of the last router that forwarded the bulk data
76 is printed so it can be identified.
81 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
82 Details : Due to a race condition, liblustre clients were occasionally
85 Severity : enhancement
87 Description: lfs setstripe use optional parameters instead of postional
90 Severity : enhancement
92 Description: Nanosecond timestamp support for ldiskfs
93 Details : The on-disk ldiskfs filesystem has added support for nanosecond
94 resolution timestamps. There is not yet support for this at
95 the Lustre filesystem level.
98 Frequency : during server recovery
100 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
101 Details : request with rq_no_resend flag not awake l_wait_event if they get a
105 Frequency : nfs export on patchless client
107 Description: connectathon hang when test nfs export over patchless client
108 Details : Disconnected dentry cannot be found with lookup, so we do not need
109 to unhash it or make it invalid
112 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
113 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
114 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
119 Description: open req refcounting wrong on reconnect
120 Details : If reconnect happened between getting open reply from server and
121 call to mdc_set_replay_data in ll_file_open, we will schedule
122 replay for unreferenced request that we are about to free.
123 Subsequent close will crash in variety of ways.
124 Check that request is still eligible for replay in
125 mdc_set_replay_data().
130 Description: disable writes to filesystem when reading health_check file
131 Details : the default for reading the health_check proc file has changed
132 to NOT do a journal transaction and write to disk, because this
133 can cause reads of the /proc file to hang and block HA state
134 checking on a healthy but otherwise heavily loaded system. It
135 is possible to return to the previous behaviour during configure
136 with --enable-health-write.
138 Severity : enhancement
140 Description: 64-bit inode version
141 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
142 to the inode. This will be required for version-based recovery.
147 Description: MDS fails to start if a duplicate client export is detected
148 Details : in some rare cases it was possible for a client to connect to
149 an MDS multiple times. Upon recovery the MDS would detect this
150 and fail during startup. Handle this more gracefully.
152 Severity : enhancement
154 Description: Add -o localflock option to simulate old noflock
156 Details : This will achieve local-only flock/fcntl locks
162 Description: log_commit_thread vs filter_destroy race leads to crash
163 Details : Take import reference before releasing llog record semaphore
168 Description: Wrong request locking in request set processing
169 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
170 imp_delayed_list, in this place should be used imp_lock.
173 Frequency : when reconnection
175 Description: Grant Leak when osc reconnect to OST
176 Details : When osc reconnect ost, OST(filter) should check whether it
177 should grant more space to client by comparing fed_grant and
178 cl_avail_grant, and return the granted space to client instead
179 of "new granted" space, because client will call osc_init_grant
180 to update the client grant space info.
183 Frequency : when client reconnect to OST
185 Description: Grant Leak when osc do resend and replay bulk write
186 Details : When osc reconnect to OST, OST(filter)should clear grant info of
187 bulk write request, because the grant info while be sync between
188 OSC and OST when reconnect, and we should ignore the grant info
189 these of resend/replay write req.
194 Description: Grant space more than avaiable left space sometimes.
195 Details : When then OST is about to be full, if two bulk writing from
196 different clients came to OST. Accord the avaliable space of the
197 OST, the first req should be permitted, and the second one
198 should be denied by ENOSPC. But if the seconde arrived before
199 the first one is commited. The OST might wrongly permit second
200 writing, which will cause grant space > avaiable space.
203 Frequency : when client is evicted
205 Description: Grant might be wrongly erased when osc is evicted by OST
206 Details : when the import is evicted by server, it will fork another
207 thread ptlrpc_invalidate_import_thread to invalidate the
208 import, where the grant will be set to 0. While the original
209 thread will update the grant it got when connecting. So if
210 the former happened latter, the grant will be wrongly errased
211 because of this race.
216 Description: Checking Stale with correct fid
217 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
218 to check whether it is stale, because sometimes, we want the
219 enqueue happened anyway, and op_data.fid2 will not be initialized.
221 Severity : enhancement
223 Description: update patchless client
224 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
227 Frequency : only with 2.4 kernel
229 Description: random memory corruption
230 Details : size of struct ll_inode_info is to big for union inode.u and this
231 can be cause of random memory corruption.
236 Description: Memory leak in recovery
237 Details : Lov_mds_md was not free in an error handler in mds_create_object.
238 It should also check obd_fail before fsfilt_start, otherwise if
239 fsfilt_start return -EROFS,(failover mds during mds recovery).
240 then the req will return with repmsg->transno = 0 and rc = EROFS.
241 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
242 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
243 be freed no matter whether fsfilt_commit success or not.
246 Frequency : only with huge count clients
248 Description: Prevents from taking the superblock lock in llap_from_page for
250 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
251 for llap_from_page call in ll_removepage prevents from taking the
252 superblock lock for a soon died page.
257 Description: Not check open intent error before release open handle
258 Details : in some rare cases, the open intent error is not checked before
259 release open handle, which may cause
260 ASSERTION(open_req->rq_transno != 0), because it tries to release
261 the failed open handle.
266 Description: Set cat log bitmap only after create log success.
267 Details : in some rare cases, the cat log bitmap is set too early. and it
268 should be set only after create log success.
272 Description: Accessing a block bevice can re-enable I/O when Lustre is
273 tearing down a device.
274 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
279 Description: service threads may hog cpus when there are a lot of requests
280 Details : Insert cond_resched to give other threads a chance to use some CPU
285 Description: the cat log was not initialized in recovery
286 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
287 unlink log on mds will not be initialized until mds post
288 recovery. And also in mds post recovery, the unlink log will
289 initialization will be done asynchronausly, so there will be race
290 between add unlink log and unlink log initialization.
294 Description: brw_stats were being printed incorrectly
295 Details : brw_stats were being printed as log2 but all of them were not
296 recorded as log2. Also remove some code duplication arising from
297 filter_tally_{read,write}.
301 Frequency : rare, only in recovery.
302 Description: ASSERTION(req->rq_type != LI_POISON) failed
303 Details : imp_lock should be held while iterating over imp_sending_list for
304 prevent destroy request after get timeout in ptlrpc_queue_wait.
308 Description: replay-single.sh test 52 fails
309 Details : A lock's skiplist need to be cleanup when it being unlinked
310 from its resource list.
315 Description: Short directio read returns full requested size rather than
317 Details : Direct I/O operations should return actual amount of bytes
318 transferred rather than requested size.
320 Severity : enhancement
322 Description: metadata RPC reduction (e.g. for rm performance)
323 Details : decrease the amount of synchronous RPC between clients and servers
324 by canceling conflicing lock before the operation on the client
325 and packing thier handles into the main operation RPC to server.
327 Severity : enhancement
329 Description: add #ifdef HAVE_KERNEL_CONFIG_H
330 Details : kernels from 2.6.19 not need include linux/config.h, but add
331 include linux/autoconf.h in commpiler command line.
333 Severity : enhancement
335 Description: patchless client support for 2.6.22 kernel
336 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
337 removed. In this case we need drop using os depended interface to
338 kmem_cache and use cfs_mem_cache API.
343 Description: fix mal-formatted messages
344 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
348 Frequency : always in liblustre
349 Description: wrong IS_ERR implementation in liblustre.h
350 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
355 Description: Correct condition for output debug message.
356 Details : inode i_nlink equal zero is not enough for output message about
357 disk corruption, i_ctime and i_mode should be also checked.
361 Frequency : always in patchless client
362 Description: add configure check for truncate_complete_page
363 Details : improve checks for exported symbols. This allow run check without
364 sources, but with Module.symvers shipped with kernel distribution.
365 add check for truncate_complete_page used by patchless client.
370 Description: improve handling recoverable errors
371 Details : if request processig with error which can be recoverable on server
372 request should be resend, otherwise page released from cache and
377 Description: sanity.sh test_77h fails with "test_77h file compare failed"
378 Details : test_77h uses a file which was messed by other test case.
382 Description: Not Check whether lov_tgts is NULL in some lov functions
383 Details : Checking whether lov_tgts is NULL in some functions.
387 Description: replace obdo_alloc() with OBDO_ALLOC macro
388 Details : nothing special is done in obdo_alloc() function, and for
389 debugging purpose, it needs to be replaced with macros.
393 Description: bad return value and errno from fcntl call
394 Details : In liblustre API, errno should be a negative value if error
399 Description: ptlrpc_check_set() LBUG
400 Details : In case of positive reply from server and failed client bulk
401 callback after bulk transfer shouldn't LBUG, but process this
402 request as erroneous.
404 Severity : enhancement
406 Description: VFS operations stats tool.
407 Details : Tool which collects stats by tracking value written in pid,
408 ppid, gid and uses llstat to generate output to plot graph using
410 Updated lustre/utils/Makefile.am
411 Added lustre/utils/ltrack_stats.c
413 Severity : enhancement
415 Description: 2.6.18 server support (lustre 1.6.1)
416 Details : Support for 2.6.18 kernels on the server side.
421 Description: ASSERTION(imp->imp_conn_current) failed
422 Details : an assertion failure is hit if a client node boots and attempts to
423 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
426 Frequency : only for i686
428 Description: 1.4.11 RC1 build fails for RHEL 4, i686
429 Details : Fixed config variable for build.
434 Description: Updated patchess for new RHEL4 kernel
435 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
436 Updated patch jbd-stats-2.6.9.patch
437 Updated patch qsnet-rhel4-2.6.patch
438 Updated patch quota-deadlock-on-pagelock-core.patch
439 Updated patch vfs_intent-2.6-rhel4.patch
440 Updated patch vfs_races-2.6-rhel4.patch
441 Updated series file 2.6-rhel4-titech.series
442 Updated series file 2.6-rhel4.series
443 Updated kernel config files
448 Description: lquota slave complains LBUG when reconnecting with mds
450 Details : quota slave depends qctxt->lqc_import to send its quota request.
451 This pointer will be invalid if mds did failover or broke its
452 connect to osts, which leads to LBUG.
455 Frequency : when qunit size is too small(less than 20M)
457 Description: write is stopped by improper -EDQUOT
458 Details : If the master is busy and qunit size is small enough(let's say 1M),
459 the slave can not get quota from master on time, which will lead
460 slave to trigger a -EQUOTA to client.
465 Description: Deadlock during metadata tests
466 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
467 called with the per-dentry lock held.
470 Frequency : SLES9 only
472 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
473 Details : set CONFIG_NR_CPUS to 128 instead of 8.
475 Severity : enhancement
477 Description: remove fs_prep_san_write operation and related patches
478 Details : remove the ext3-san-jdike patches which are no longer useful.
483 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
484 Details : a disk corruption can cause the mballoc code to assert on a
485 double free or other extent corruptions. Handle these with
486 ext3_error() instead of assertions.
490 Description: ldiskfs_ext_search_right: bad header in inode #3981353: unexpected
492 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
493 can cause the filesystem to be remounted read-only.
495 --------------------------------------------------------------------------------
497 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
502 Frequency : on some architectures
504 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
505 Details : Uninitialized flags sometimes cause configuration commands to
509 Frequency : patchless clients only
511 Description: missing __iget() symbol export
512 Details : The __iget() symbol export is missing. To avoid the need for
513 this on patchless clients the deathrow inode reaper is turned
514 off, and we depend on the VM to clean up old inodes. This
515 dependency was during via the fix for bug 12181.
517 --------------------------------------------------------------------------------
519 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
521 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
522 INTEROPERATE with older versions automatically. In many cases a
523 special upgrade step is needed. Please read the
524 user documentation before upgrading any part of a 1.4.x system.
525 * WARNING: Lustre configuration and startup changes are required with
526 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
528 * Support for kernels:
529 2.4.21-47.0.1.EL (RHEL 3)
531 2.6.9-42.0.10.EL (RHEL 4)
532 2.6.12.6 vanilla (kernel.org)
533 2.6.16.27-0.9 (SLES10)
534 * Client support for unpatched kernels:
535 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
536 2.6.16 - 2.6.19 vanilla (kernel.org)
537 2.6.9-42.0.8EL (RHEL 4)
538 * Recommended e2fsprogs version: 1.39.cfs6
539 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
542 Severity : enhancement
544 Description: MountConf
545 Details : Lustre configuration is now managed via mkfs and mount
546 commands instead of lmc and lconf. New obd types (MGS, MGC)
547 are added for dynamic configuration management. See
548 https://mail.clusterfs.com/wikis/lustre/MountConf for
551 Severity : enhancement
553 Description: dynamic OST addition
554 Details : OSTs can now be added to a live filesystem
556 Severity : enhancement
558 Description: startup order invariance
559 Details : MDTs and OSTs can be started in any order. Clients only
560 require the MDT to complete startup.
562 Severity : enhancement
564 Description: parallel, asynchronous orphan cleanup
565 Details : orphan cleanup is now performed in separate threads for each
566 OST, allowing parallel non-blocking operation.
568 Severity : enhancement
570 Description: optimized stripe assignment
571 Details : stripe assignments are now made based on ost space available,
572 ost previous usage, and OSS previous usage, in order to try
573 to optimize storage space and networking resources.
575 Severity : enhancement
577 Description: Permanently set tunables
578 Details : All writable /proc/fs/lustre tunables can now be permanently
579 set on a per-server basis, at mkfs time or on a live
582 Severity : enhancement
584 Description: Lustre message v2
585 Details : Add lustre message format v2.
587 Severity : enhancement
589 Description: client OST exclusion list
590 Details : Clients can be started with a list of OSTs that should be
591 declared "inactive" for known non-responsive OSTs.
595 Description: ENOENT returned for valid filehandle during dbench.
596 Details : Check if a directory has children when invalidating dentries
597 associated with an inode during lock cancellation. This fixes
598 an incorrect ENOENT sometimes seen for valid filehandles during
602 Frequency : SFS test only (otherwise harmless)
604 Description: SPEC SFS validation failure on NFS v2 over lustre.
605 Details : Changes the blocksize for regular files to be 2x RPC size,
606 and not depend on stripe size.
608 Severity : enhancement
610 Description: fine-grained SMP locking inside DLM
611 Details : Improve DLM performance on SMP systems by removing the single
612 per-namespace lock and replace it with per-resource locks.
614 Severity : enhancement
616 Description: don't hold multiple extent locks at one time
617 Details : To avoid client eviction during large writes, locks are not
618 held on multiple stripes at one time or for very large writes.
619 Otherwise, clients can block waiting for a lock on a failed OST
620 while holding locks on other OSTs and be evicted.
622 Severity : enhancement
624 Description: Multiple MD RPCs in flight.
625 Details : Further unserialise some read-only MDT RPCs - learn about intents.
626 To avoid overly-overloading MDT, introduce a limit on number of
627 MDT RPCs in flight for a single client and add /proc controls
628 to adjust this limit.
630 Severity : enhancement
632 Description: client read/write statistics
633 Details : Add client read/write call usage stats for performance
634 analysis of user processes.
635 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
636 file access. extents_stats shows chunk size distribution.
637 extents_stats_per_process show chunk size distribution per
640 Severity : enhancement
642 Description: per-client statistics on server
643 Details : Add ldlm and operations statistics for each client in
644 /proc/fs/lustre/mds|obdfilter/*/exports/
646 Severity : enhancement
648 Description: improved MDT statistics
649 Details : Add detailed MDT operations statistics in
650 /proc/fs/lustre/mds/*/stats
652 Severity : enhancement
654 Description: VFS operations stats
655 Details : Add client VFS call stats, trackable by pid, ppid, or gid
656 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
661 Description: Fix client-side osc byte counters
662 Details : The osc read/write byte counters in
663 /proc/fs/lustre/osc/*/stats are now working
666 Frequency : always as root on SLES
668 Description: Failure of copying files with lustre special EAs.
669 Details : Client side always return success for setxattr call for lustre
670 special xattr (currently only "trusted.lov").
675 Description: Refcount LNET uuids
676 Details : The global LNET uuid list grew linearly with every startup;
677 refcount repeated list entries instead of always adding to
680 Severity : enhancement
682 Description: Dynamic service threads
683 Details : Within a small range, start extra service threads
684 automatically when the request queue builds up.
687 Frequency : mixed-endian client/server environments
689 Description: mixed-endian crashes
690 Details : The new msg_v2 system had some failures in mixed-endian
693 Severity : enhancement
695 Description: Easy OST removal
696 Details : OSTs can be permanently deactivated with e.g. 'lctl
697 conf_param lustre-OST0001.osc.active=0'
699 Severity : enhancement
701 Description: MGS proc entries
702 Details : Added basic proc entries for the MGS showing what filesystems
705 Severity : enhancement
707 Description: provide MGS failover
708 Details : Added config lock reacquisition after MGS server failover.
710 Severity : enhancement
712 Description: add Linux 2.4 support
713 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
717 Description: a large application tries to do I/O to the same resource and dies
719 Details : Check the req->rq_arrival time after the call to
720 ost_brw_lock_get(), but before we do anything about
721 processing it & sending the BULK transfer request. This
722 should help move old stale pending locks off the queue as
723 quickly as obd_timeout.
726 Frequency : when an incorrect nid is specified during startup
728 Description: ptlrpc connect to non-existant node causes kernel crash
729 Details : LNET can't be re-entered from an event callback, which
730 happened when we expire a message after the export has been
731 cleaned up. Instead, hand the zombie cleanup off to another
734 Severity : enhancement
736 Description: plain/inodebits lock performance improvement
737 Details : Grouping plain/inodebits in granted list by their request modes
738 and bits policy, thus improving the performance of search through
742 Frequency : only if OST filesystem is corrupted
744 Description: client incorrectly hits assertion in ptlrpc_replay_req()
745 Details : for a short time RPCs with bulk IO are in the replay list,
746 but replay of bulk IOs is unimplemented. If the OST filesystem
747 is corrupted due to disk cache incoherency and then replay is
748 started it is possible to trip an assertion. Avoid putting
749 committed RPCs into the replay list at all to avoid this issue.
752 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
754 Description: System hang on startup
755 Details : This bug allowed the liblustre (e.g. catamount) client to
756 return to the app before handling all startup RPCs. This
757 could leave the node unresponsive to lustre network traffic
758 and manifested as a server ptllnd timeout.
760 Severity : enhancement
762 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
763 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
764 boolean to control whether to print peer debug info when a
765 client's RPC times out.
768 Frequency : only for kernels with patches from Lustre below 1.4.3
770 Description: Remove old rdonly API
771 Details : Remove old rdonly API which unused from at least lustre 1.4.3
774 Frequency : only for devices with external journals
776 Description: Set external device read-only also
777 Details : During a commanded failover stop, we set the disk device
778 read-only while the server shuts down. We now also set any
779 external journal device read-only at the same time.
782 Frequency : when upgrading from 1.4 while trying to change parameters
784 Description: The wrong (new) MDC name was used when setting parameters for
785 upgraded MDT's. Also allows changing of OSC (and MDC)
786 parameters if --writeconf is specified at tunefs upgrade time.
789 Frequency : when setting specific ost indicies
791 Description: QOS code breaks on skipped indicies
792 Details : Add checks for missing OST indicies in the QOS code, so OSTs
793 created with --index need not be sequential.
795 Severity : enhancement
797 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
798 Details : The uninit_groups feature works in conjunction with the kernel
799 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
800 up the pass1 processing of e2fsck. This is a read-only feature
801 in ldiskfs2 only, so older kernels and current ldiskfs cannot
802 mount filesystems that have had this feature enabled.
804 Severity : enhancement
806 Description: Improve multi-block allocation algorithm to avoid fragmentation
807 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
808 allocation locality and avoid filesystem fragmentation.
810 ------------------------------------------------------------------------------
812 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
814 * Support for kernels:
815 2.4.21-47.0.1.EL (RHEL 3)
817 2.6.9-42.0.10.EL (RHEL 4)
818 2.6.12.6 vanilla (kernel.org)
819 2.6.16.27-0.9 (SLES 10)
820 * Recommended e2fsprogs version: 1.39.cfs5
822 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
826 Frequency : occasional, depends on client load and configuration
827 Bugzilla : 12181, 12203
828 Description: data loss for recently-modified files
830 Details : In some cases it is possible that recently written or created
831 files may not be written to disk in a timely manner (this should
832 normally be within 30s unless client IO load is very high).
833 The problem appears as zero-length files or files that are a
834 multiple of 1MB in size after a client crash or client eviction
835 that are missing data at the end of the file.
837 This problem is more likely to be hit on clients where files are
838 repeatedly created and unlinked in the same directory, clients
839 have a large amount of RAM, have many CPUs, the filesystem has
840 many OSTs, the clients are rebooted frequently, and/or the files
841 are not accessed by other nodes after being written.
843 The presence of the problem can be detected by looking at
844 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
845 smaller than the second (nr_unused) then dirty files will not
846 be flushed automatically to disk. "sync; sleep 10" should be
847 run several times on the node before unmounting it to update
848 Lustre (this is also safe to run on nodes without this problem).
850 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
851 that can cause this same problem, so customers using that kernel
852 also need to update the kernel in addition to Lustre. In order
853 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
855 It is normal that files written just before a client crash (less
856 than 30s) may not yet have been flushed to disk, even for local
860 Frequency : frequent on thin XT3 nodes
862 Description: UUID collision on thin XT3 Linux nodes
863 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
864 randomly, since we relied on an insufficiently-seeded PRNG.
869 Description: OSS hangs after "All ost request buffers busy"
870 Details : A deadlock between quota and journal operations caused OSS
871 hangs after printing "All ost request buffers busy."
874 Frequency : always on liblustre builds
876 Description: Cleanup compiler warnings on liblustre
879 Frequency : always on liblustre builds on XT3
881 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
885 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
886 > 32000 subdirectories
887 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
888 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
889 subdirectory count crosses 32000. This will aid e2fsck to
890 correctly handle more than 32000 subdirectories.
893 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
895 Description: System hang on startup
896 Details : This bug allowed the liblustre (e.g. catamount) client to
897 return to the app before handling all startup RPCs. This
898 could leave the node unresponsive to lustre network traffic
899 and manifested as a server ptllnd timeout.
901 Severity : enhancement
903 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
904 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
905 boolean to control whether to print peer debug info when a
906 client's RPC times out.
911 Description: make O_SYNC working on 2.6 kernels
912 Details : 2.6 kernels use different method for mark pages for write,
913 so need add a code to lustre for O_SYNC work.
918 Description: Failure to close file and release space on NFS
919 Details : Put inode details into lock acquired in ll_intent_file_open.
920 Use mdc_intent_lock in ll_intent_open to properly
921 detect all kind of errors unhandled by mdc_enqueue.
926 Description: proc file read during shutdown sometimes raced obd removal,
928 Details : Add lock to prevent obd access after proc file removal.
931 Frequency : Only for files larger than 4GB on 32-bit clients.
933 Description: improperly doing page alignment of locks
934 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
935 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
940 Description: RPCs being resent when they shouldn't be
941 Details : Some RPCs that should not be resent are being resent. This
942 can cause inconsistencies in the RPC state machine. Do not
943 resend such requests.
946 Frequency : rare, only with NFS export
948 Description: Crash on NFS re-export node
949 Details : under very unusual load conditions an assertion is hit in
950 ll_intent_file_open()
953 Frequency : only if OST filesystem is corrupted
955 Description: client incorrectly hits assertion in ptlrpc_replay_req()
956 Details : for a short time RPCs with bulk IO are in the replay list,
957 but replay of bulk IOs is unimplemented. If the OST filesystem
958 is corrupted due to disk cache incoherency and then replay is
959 started it is possible to trip an assertion. Avoid putting
960 committed RPCs into the replay list at all to avoid this issue.
965 Description: large O_DIRECT requests fail under memory pressure/fragmentation
966 Details : Large single O_DIRECT read and write calls can fail to allocate
967 a sufficiently large buffer to process the request. In case of
968 allocation failure the allocation is retried with a smaller
969 buffer and broken into smaller requests.
971 Severity : enhancement
973 Description: Add -o localflock option to simulate old noflock behaviour.
974 Details : This will achieve local-only flock/fcntl locks coherentness.
979 Description: versioning check is incomplete
980 Details : Checking the version difference of client vs. server, report
981 error if the gap is too big.
986 Description: add support PG_writeback bit
987 Details : add support for PG_writeback bit for Lustre, for more carefull
988 work with page cache in 2.6 kernel. This also fix some deadlocks
989 and remove hack for work O_SYNC with 2.6 kernel.
991 Severity : enhancement
993 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
994 Details : The uninit_groups feature works in conjunction with the kernel
995 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
996 up the pass1 processing of e2fsck. This is a read-only feature
997 in ldiskfs2 only, so older kernels and current ldiskfs cannot
998 mount filesystems that have had this feature enabled.
1000 Severity : enhancement
1002 Description: Improve multi-block allocation algorithm to avoid fragmentation
1003 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1004 allocation locality and avoid filesystem fragmentation.
1006 ------------------------------------------------------------------------------
1008 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1010 * Support for kernels:
1011 2.6.9-42.0.3.EL (RHEL 4)
1012 2.6.5-7.276 (SLES 9)
1013 2.4.21-47.0.1.EL (RHEL 3)
1014 2.6.12.6 vanilla (kernel.org)
1015 2.6.16.21-0.8 (SLES10)
1016 * Recommended e2fsprogs version: 1.39.cfs2-0
1018 * The backwards-compatible /proc/sys/portals symlink has been removed
1019 in this release. Before upgrading, please ensure that you change
1020 any configuration scripts or /etc/sysctl.conf files that access
1021 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1022 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1023 in advance of the upgrade on any system running Lustre 1.4.6 or
1024 newer, since /proc/sys/lnet was added in that version.
1025 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1029 Frequency : only when quota is used
1031 Description: avoid scanning export list for quota master
1032 Details : Change the algorithms to avoid scanning export list in order
1033 to improve the efficiency.
1036 Frequency : MDS failover only, very rarely
1038 Description: "went back in time" messages on mds failover
1039 Details : The greatest transno may be lost when the current operation
1040 finishes with an error (transno==0) and the client's last_rcvd
1041 record is over-written. Save the greatest transno in the
1042 mds_last_transno for this case.
1045 Frequency : always for specific kernels and striping counts
1047 Description: client may get "Matching packet too big" without ACL support
1048 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1049 when trying to access files in certain configurations. The
1050 clients should in fact be denied when mounting because they do
1051 not understand ACLs.
1054 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1056 Description: many clients connecting with IO in progress causes connect timeouts
1057 Details : Avoid synchronous journal commits to avoid delays caused by many
1058 clients connecting/disconnecting when bulk IO is in progress.
1059 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1060 OST_IO_PORTAL to avoid delays behind potentially many pending
1064 Frequency : occasionally with multiple writers to a single file
1066 Description: shared writes to file may result in wrong size reported by stat()
1067 Details : Allow growing of kms when extent lock is cancelled
1070 Frequency : always with random mmap IO to multi-striped file
1072 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1073 Details : Only if the hole is at the end of OST object so that kms is too
1074 small. Fix is to increase kms accordingly in ll_nopage.
1077 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1079 Description: writes to a missing object would leak memory on the OST
1080 Details : If there is an inconsistency between the MDS and OST filesystems,
1081 such that the MDS references an object that doesn't exist, writes
1082 to that object will leak memory due to incorrect cleanup in the
1083 error handling path, eventually running out of memory on the OST.
1088 Description: Creating too long symlink causes lustre errors
1089 Details : Check symlink and name lengths before sending requests to MDS.
1092 Frequency : only if flock is enabled (not on by default)
1094 Description: posix locks not released on fd closure on 2.6.9+
1095 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1096 kernels, this caused such locks not to be released on fd close and
1097 then assertions on fs unmount about still used locks.
1100 Frequency : MDS failover only, very rarely
1102 Description: clients may get ASSERTION(granted_lock != NULL)
1103 Details : When request was taking a long time, and a client was resending
1104 a getattr by name lock request. The were multiple lock requests
1105 with the same client lock handle and
1106 mds_getattr_name->fixup_handle_for_resent_request found one of the
1107 lock handles but later failed with ASSERTION(granted_lock != NULL).
1112 Description: handle->h_buffer_credits > 0, assertion failure
1113 Details : h_buffer_credits is zero after truncate, causing assertion
1114 failure. This patch extends the transaction or creates a new
1118 Frequency : NFS re-export or patchless client
1119 Bugzilla : 11179, 10796
1120 Description: Crash on NFS re-export node (__d_move)
1121 Details : We do not want to hash the dentry if we don't have a lock.
1122 But if this dentry is later used in d_move, we'd hit uninitialised
1123 list head d_hash, so we just do this to init d_hash field but
1124 leave dentry unhashed.
1127 Frequency : NFS re-export or patchless client
1129 Description: NFS exports has problem with symbolic link
1130 Details : lustre client didn't properly install dentry when re-exported
1131 to NFS or running patchless client.
1134 Frequency : NFS re-export or patchless client
1136 Description: Various nfs/patchless fixes.
1137 Details : fixes reuse disconected alias for lookup process - this fixes
1138 warning "find_exported_dentry: npd != pd",
1139 fix permission error with open files at nfs.
1140 fix apply umask when do revalidate.
1143 Frequency : occasional
1145 Description: Crash on NFS re-export node
1146 Details : calling clear_page() on the wrong pointer triggered oops in
1147 generic_mapping_read().
1150 Frequency : rarely, using O_DIRECT IO
1152 Description: unaligned directio crashes client with LASSERT
1153 Details : check for unaligned buffers before trying any requests.
1156 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1158 Description: stale data returned from RAID cache
1159 Details : If only a small amount of IO is done to the RAID device before
1160 reading it again it is possible to get stale data from the RAID
1161 cache instead of reading it from disk.
1164 Frequency : always for sles10 kernel
1166 Description: sles10 support
1167 Details : ll_follow_link: compile fixes and using of nd_set_link
1168 under newer kernels.
1171 Frequency : depends on arch, kernel and compiler version, always on sles10
1174 Description: recursive or deep enough symlinks cause stack overflow
1175 Details : getting rid of large stack-allocated variable in
1179 Frequency : depends on hardware
1181 Description: lustre write performance loss in the SLES10 kernel
1182 Details : the performance loss is caused by using of write barriers in the
1183 ext3 code. The SLES10 kernel turns barrier support on by
1184 default. The fix is to undo that change for ldiskfs.
1186 ------------------------------------------------------------------------------
1188 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1190 * Support for kernels:
1191 2.6.9-42.0.3EL (RHEL 4)
1192 2.6.5-7.276 (SLES 9)
1193 2.4.21-47.0.1.EL (RHEL 3)
1194 2.6.12.6 vanilla (kernel.org)
1198 Frequency : quota enabled and large files being deleted
1200 Description: releasing more than 4GB of quota at once hangs OST
1201 Details : If a user deletes more than 4GB of files on a single OST it
1202 will cause the OST to spin in an infinite loop. Release
1203 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1208 Description: statfs data retrieved from /proc may be stale or zero
1209 Details : When reading per-device statfs data from /proc, in the
1210 {kbytes,files}_{total,free,avail} files, it may appear
1211 as zero or be out of date.
1214 Frequency : systems with MD RAID1 external journal devices
1216 Description: lconf's call to blkid is confused by RAID1 journal devices
1217 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1218 of returning all block devices that match the journal UUID.
1221 Frequency : always, for aggregate stripe size over 4GB
1223 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1224 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1225 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1228 Frequency : always if "lfs find" used on a local file/directory
1230 Description: "lfs find" segfaults if used on a local file/directory
1231 Details : The case where a directory component was not specified wasn't
1232 handled correctly. Handle this properly.
1235 Frequency : always on ppc64
1237 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1238 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1239 perform endianness conversion.
1242 Frequency : rarely (truncate to non-zero file size after write under load)
1243 Bugzilla : 10730, 10687
1244 Description: Files padded with zeros to next 4K multiple
1245 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1246 it is possible that files that are truncated to a non-zero size
1247 immediately after being written are filled with zero bytes beyond
1248 the truncated size. No file data is lost.
1250 Severity : enhancement
1252 Description: Allow recovery/failover for liblustre clients.
1253 Details : liblustre clients were unaware of failover configurations until
1256 Severity : enhancement
1258 Description: user file locks should fail when not mounting with flock option
1259 Details : Set up an error-returning stub in ll_file_operations.lock field
1260 to prevent incorrect behaviour when client is mounted without
1261 flock option. Also, set up properly f_op->flock field for
1265 Frequency : always on ia64
1267 Description: "lfs df" loops on printing out MDS statfs information
1268 Details : The obd_ioctl_data was not initialized and in some systems
1269 this caused a failure during the ioctl that did not return
1270 an error. Initialize the struct and return an error on failure.
1273 Frequency : SLES 9 only
1275 Description: Error of copying files with lustre special EAs as root
1276 Details : Client side always return success for setxattr call for lustre
1277 special xattr (currently only "trusted.lov").
1280 Frequency : rarely on clusters with both ia64+i386 clients
1282 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1283 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1284 clients are doing concurrent writes to the same file it is
1285 possible that the ia64 clients may LASSERT because the OST
1286 extent locks are not PAGE_SIZE aligned. Ensure that grown
1287 locks are always aligned on the request boundary.
1290 Frequency : specific use, occasional
1292 Description: Overwriting in use executable truncates on-disk binary image
1293 Details : If one node attempts to overwrite an executable in use by
1294 another node, we now correctly return ETXTBSY instead of
1295 truncating the file.
1297 Severity : enhancement
1299 Description: Async OSC create to avoid the blocking unnecessarily.
1300 Details : If a OST has no remain object, system will block on the creating
1301 when need to create a new object on this OST. Now, ways use
1302 pre-created objects when available, instead of blocking on an
1303 empty osc while others are not empty. If we must block, we block
1304 for the shortest possible period of time.
1309 Description: chmod on Lustre root is propagated to other clients
1310 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1311 invalid by the follow_mount time.
1314 Frequency : liblustre clients only
1316 Description: Race in 'instant cancel' lock handling could lead to such locks
1317 never to be granted in case of SMP MDS
1318 Details : Do not destroy not yet granted but cbpending locks in
1322 Frequency : replay/resend of open
1324 Description: non null lock assetion failure in mds_intent_policy
1325 Details : Trying to replay/resend lockless open requests resulted in
1326 mds_open() returning 0 with no lock. Now it sets a flag if
1327 a lock is going to be returned.
1329 Severity : enhancement
1331 Description: Checksum enhancements
1332 Details : New checksum enhancements allow for resending RPCs that failed
1335 Severity : enhancement
1337 Description: Tunables on number of dirty pages in cacche
1338 Details : Allow to set limit on number of dirty pages cached.
1343 Description: client crash on unmount - lock still has references
1344 Details : In some error handling cases it was possible to leak a lock
1345 reference on a client while accessing a file. This was not
1346 harmful to the client during operation, but would cause the
1347 client to crash when the filesystem is unmounted.
1350 Frequency : specific case, rare
1352 Description: ETXTBSY on mds though file not in use by client
1353 Details : ETXTBSY is no longer incorrectly returned when attempting to
1354 chmod or chown a directory that the user previously tried to
1355 execute or a currently-executing binary.
1358 Frequency : extremely rare except on liblustre-based clients
1360 Description: Lustre space not freed when files are deleted
1361 Details : Clean up open-unlinked files after client eviction. Previously
1362 the unlink was skipped and the files remained as orphans.
1367 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1368 Details : In some cases it was possible to send a blocking callback to
1369 a client doing a glimpse, even though that client didn't get
1370 a lock granted. When the glimpse lock is cancelled on the OST
1371 the freed lock is left on the waiting list and corrupted the list.
1374 Frequency : all core dumps
1376 Description: Broke core dumps to lustre
1377 Details : Negative dentry may be unhashed if parent does not have UPDATE
1378 lock, but some callers, e.g. do_coredump, expect dentry to be
1379 hashed after successful create, hash it in ll_create_it.
1381 ------------------------------------------------------------------------------
1383 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1385 * Support for kernels:
1386 2.6.9-42.0.2.EL (RHEL 4)
1387 2.6.5-7.276 (SLES 9)
1388 2.4.21-40.EL (RHEL 3)
1389 2.6.12.6 vanilla (kernel.org)
1393 Frequency : always on RHEL 3
1395 Description: Number of open files grows over time
1396 Details : The number of open files grows over time, whether or not
1397 Lustre is started. This was due to a filp leak introduced
1398 by one of our kernel patches.
1400 ------------------------------------------------------------------------------
1402 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
1404 * Support for kernels:
1405 2.6.9-42.EL (RHEL 4)
1406 2.6.5-7.267 (SLES 9)
1407 2.4.21-40.EL (RHEL 3)
1408 2.6.12.6 vanilla (kernel.org)
1413 Bugzilla : 5719, 9635, 9792, 9684
1414 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1415 Details : If a server is under heavy load and cannot reply to new
1416 connection requests before the client resends the (re)connect,
1417 the connection handling code can behave badly if two service
1418 threads are concurrently handing separate (re)connections from
1419 the same client. Add better locking to the connection handling
1420 code, and ensure that only a single connection will be processed
1421 for a given client UUID, even if the lock is dropped.
1423 Severity : enhancement
1425 Description: add TCP zero-copy support to kernel
1426 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1427 sends if the hardware supports scatter-gather and checksumming.
1428 This allows socklnd to do client-write and server-read more
1429 efficiently and reduce CPU utilization from skbuf copying.
1432 Frequency : only if NFS exporting from client
1434 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1435 Details : If setting attributes on a file created under NFS that had
1436 never been opened it would be possible to oops the client
1437 if the file had no objects.
1441 Bugzilla : 9326, 10402, 10897
1442 Description: client crash in ptlrpcd_wake() thread when sending async RPC
1443 Details : It is possible that ptlrpcd_wake() dereferences a freed async
1444 RPC. In rare cases the ptlrpcd thread alread processed the RPC
1445 before ptlrpcd_wake() was called and the request was freed.
1448 Frequency : always for liblustre
1450 Description: liblustre client does MDS+OSTs setattr RPC for each write
1451 Details : When doing a write from a liblustre client, the client
1452 incorrectly issued an RPC to the MDS and each OST the file was
1453 striped over in order to update the timestamps. When writing
1454 with small chunks and many clients this could overwhelm the MDS
1455 with RPCs. In all cases it would slow down the write because
1456 these RPCs are unnecessary.
1458 Severity : enhancement
1460 Description: allow number of MDS service threads to be changed at module load
1461 Details : It is now possible to change the number of MDS service threads
1462 running. Adding "options mds mds_num_threads={N}" to the MDS's
1463 /etc/modprobe.conf will set the number of threads for the next
1464 time Lustre is restarted (assuming the "mds" module is also
1465 reloaded at that time). The default number of threads will
1466 stay the same, 32 for most systems.
1471 Description: OST crash if filesystem is unformatted or corrupt
1472 Details : If an OST is started on a device that has never been formatted
1473 or if the filesystem is corrupt and cannot even mount then the
1474 error handling cleanup routines would dereference a NULL pointer.
1479 Description: NULL pointer deref in llap_from_page.
1480 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1481 incorrect) mapping if the page was truncated/reclaimed while it was
1482 searched for. Check for this condition and skip such pages when
1483 doing readahead. Introduce extra check to llap_from_page() to
1484 verify page->mapping->host is non-NULL (so page is not anonymous).
1487 Frequency : Sometimes when using sys_sendfile
1489 Description: "page not covered by a lock" warnings from ll_readpage
1490 Details : sendfile called ll_readpage without right page locks present.
1491 Now we introduced ll_file_sendfile that does necessary locking
1492 around call to generic_file_sendfile() much like we do in
1496 Frequency : with certain MDS communication failures at client mount time
1498 Description: NULL pointer deref after failed client mount
1499 Details : a client connection request may delayed by the network layer
1500 and not be sent until after the PTLRPC layer has timed out the
1501 request. If the client fails the mount immediately it will try
1502 to clean up before the network times out the request. Add a
1503 reference from the request import to the obd device and delay
1504 the cleanup until the network drops the request.
1507 Frequency : occasionally during client (re)connect
1509 Description: assertion failure during client (re)connect
1510 Details : processing a client connection request may be delayed by the
1511 client or server longer than the client connect timeout. This
1512 causes the client to resend the connection request. If the
1513 original connection request is replied in this interval, the
1514 client may trip an assertion failure in ptlrpc_connect_interpret()
1515 which thought it would be the only running connect process.
1518 Frequency : only with obd_echo servers and clients that are rebooted
1520 Description: kernel BUG accessing uninitialized data structure
1521 Details : When running an obd_echo server it did not start the ping_evictor
1522 thread, and when a client was evicted an uninitialized data
1523 structure was accessed. Start the ping_evictor in the RPC
1524 service startup instead of the OBD startup.
1526 Severity : enhancement
1527 Bugzilla : 10193 (patchless)
1528 Description: Remove dependency on various unexported kernel interfaces.
1529 Details : No longer need reparent_to_init, exit_mm, exit_files,
1530 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1533 Frequency : rare (only users of deprecated and unsupported LDAP config)
1535 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1536 Details : LDAP apparently contains 'lustreName' attributes instead of
1537 'name'. A simple remapping of the name is sufficient.
1540 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1542 Description: waiting_locks_callback trips kernel BUG if client is evicted
1543 Details : Running with the dump_on_timeout debug flag turned on makes
1544 it possible that the waiting_locks_callback() can try to dump
1545 the Lustre kernel debug logs from an interrupt handler. Defer
1546 this log dumping to the expired_lock_main() thread.
1548 Severity : enhancement
1550 Description: Support NFS exporting on 2.6 kernels.
1551 Details : Implement non-rawops metadata methods for NFS server to use without
1552 changing NFS server code.
1555 Frequency : very rare (synthetic metadata workload only)
1557 Description: two racing renames might cause an MDS thread to deadlock
1558 Details : Running the "racer" program may cause one MDS thread to rename
1559 a file from being the source of a rename to being the target of
1560 a rename at exactly the same time that another thread is doing
1561 so, and the second thread has already enqueued these locks after
1562 doing a lookup of the target and is trying to relock them in
1563 order. Ensure that we don't try to re-lock the same resource.
1566 Frequency : only very large systems with liblustre clients
1568 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1569 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1570 all OSTs in parallel. This allows the eviction of stale liblustre
1571 clients to proceed much faster than if they were done in series,
1572 and also offers similar improvements for other set_info RPCs.
1577 Description: excessive CPU usage during initial read phase on client
1578 Details : During the initial read phase on a client, it would agressively
1579 retry readahead on the file, consuming too much CPU and impacting
1580 performance (since 1.4.5.8). Improve the readahead algorithm
1581 to avoid this, and also improve some other common cases (read
1582 of small files in particular, where "small" is files smaller than
1583 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1588 Description: MDS crash when receiving packet with unknown intent.
1589 Details : Do not LBUG in unknown intent case, just return -EFAULT
1591 Severity : enhancement
1592 Bugzilla : 9293, 9385
1593 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1594 Details : Do not serialize getattr (non-intent version) and statfs.
1597 Frequency : occasional, when OST network is overloaded/intermittent
1599 Description: client evicted by OST after bulk IO timeout
1600 Details : If a client sends a bulk IO request (read or write) the OST
1601 may evict the client if it is unresposive to its data GET/PUT
1602 request. This is incorrect if the network is overloaded (takes
1603 too long to transfer the RPC data) or dropped the OST GET/PUT
1604 request. There is no need to evict the client at all, since
1605 the pinger and/or lock callbacks will handle this, and the
1606 client can restart the bulk request.
1609 Frequency : Always when mmapping file with no objects
1611 Description: client crashes when mmapping file with no objects
1612 Details : Check that we actually have objects in a file before doing any
1613 operations on objects in ll_vm_open, ll_vm_close and
1619 Description: Request leak when working with deleted CWD
1620 Details : Introduce advanced request refcount tracking for requests
1621 referenced from lustre intent.
1623 Severity : Enhancement
1625 Description: Cache open file handles on client.
1626 Details : MDS now will return special lock along with openhandle, if
1627 requested and client is allowed to hold openhandle, even if unused,
1628 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1629 closing files for every read/write openration.
1631 Severity : Enhancement
1633 Description: Cache open negative dentries on client when possible.
1634 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1635 negative dentries on lock revocation.
1640 Description: Remounting a client read-only wasn't possible with a zconf mount
1641 Details : It wasn't possible to remount a client read-only with llmount.
1643 Severity : enhancement
1644 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1645 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1646 MPICH is included to add Lustre-specific ADIO interfaces.
1647 This is based closely on the UFS ADIO layer and only differs
1648 in file creation, in order to allow the OST striping to be set.
1649 This is user-contributed code and not supported by CFS.
1654 Description: extended inode attributes (immutable, append-only) work improperly
1655 when 2.4 and 2.6 kernels are used on client/server or vice versa
1656 Details : Introduce kernel-independent values for these flags.
1658 Severity : enhancement
1661 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1662 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1663 granularity. Now, Fractional MB granularity is be supported,
1664 this is very useful for low memory system.
1666 Severity : enhancement
1668 Description: Getattr by fid
1669 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1670 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1671 have locks and avoids getattr rpc requests for it on every stat.
1674 Frequency : Always, for filesystems larger than 2TB
1676 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1677 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1678 filesystems larger than 2TB to be mounted with mballoc enabled.
1681 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1683 Description: filesystem corruption for non-standard kernels and very large OSTs
1684 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1685 filesystems larger than 2TB is mounted then the kernel will
1686 silently corrupt the start of the filesystem. CONFIG_LBD is
1687 enabled for all CFS-supported kernels, but the possibility of
1688 this happening with a modified kernel config exists.
1690 Severity : enhancement
1692 Description: add client O_DIRECT support for 2.6 kernels
1693 Details : It is now possible to do O_DIRECT reads and writes to files
1694 in the Lustre client mountpoint on 2.6 kernel clients.
1696 Severity : enhancement
1698 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1699 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1700 parallel, not waiting for response from every OST before sending
1701 a rpc to the next OST.
1706 Description: setattr vs write race when updating file timestamps
1707 Details : Client processes that update a file timestamp into the past
1708 right after writing to the file (e.g. tar) it is possible that
1709 the updated file modification time can be reset to the current
1710 time due to a race between processing the setattr and write RPC.
1712 Severity : enhancement
1714 Description: Bring 'lfs find' closer in line with regular Linux find.
1715 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1716 -print0 options and obtains all the needed info through the lustre
1719 Severity : enhancement
1721 Description: support up to 1024 configured devices on one node
1722 Details : change obd_dev array from statically allocated to dynamically
1723 allocated structs as they are first used to reduce memory usage
1728 Description: Flush dirty partially truncated pages during truncate
1729 Details : Immediatelly flush partially truncated pages in filter_setattr,
1730 this way we completely avoid having any pages in page cache on OST
1731 and can retire ugly workarounds during writes to flush such pages.
1736 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1737 Details : protect inode from truncation within vfs_unlink() context
1738 just take a reference before calling vfs_unlink() and release it
1739 when parent's i_sem is free.
1742 Frequency : always, if extents are used on OSTs
1744 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1745 Details : OSTs using the extents format would not zero the high 16 bits of
1746 the index physical block number. This is not a problem for any
1747 OST filesystems smaller than 16TB, and no kernels support ext3
1748 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1749 new/modified files) and can be fixed for existing filesystems
1750 with e2fsprogs-1.39-cfs1.
1755 Description: import connection selection may be incorrect if timer wraps
1756 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1757 import connections to be ignored if the 32-bit jiffies counter
1758 wraps. Use a 64-bit jiffies counter.
1761 Frequency : during server recovery
1763 Description: crash after server is denying duplicate export
1764 Details : If clients are resending connect requests to the server, the
1765 server refuses to allow a client to connect multiple times.
1766 Fixed a bug in the handling of this case.
1769 Frequency : very large clusters immediately after boot
1771 Description: LNET request buffers exhausted under heavy short-term load
1772 Details : If a large number of client requests are generated on a service
1773 that has previously never seen so many requests it is possible
1774 that the request buffer growth cannot keep up with the spike in
1775 demand. Instead of dropping incoming requests, they are held in
1776 the LND until the RPC service can accept more requests.
1779 Frequency : Sometimes during replay
1781 Description: Assertion failure in ll_local_open after replay.
1782 Details : If replay happened on an open request reply before we were able
1783 to set replay handler, reply will become not swabbed tripping the
1784 assertion in ll_local_open. Now we set the handler right after
1785 recognising of open request
1788 Frequency : very rare
1790 Description: kernel reports "badness in vsnprintf"
1791 Details : Reading from the "recovery_status" /proc file in small chunks
1792 may cause a negative length in lprocfs_obd_rd_recovery_status()
1793 call to vsnprintf() (which is otherwise harmless). Exit early
1794 if there is no more space in the output buffer.
1796 Severity : enhancement
1798 Description: clear OBD RPC statistics by writing to them
1799 Details : It is now possible to clear the OBD RPC statistics by writing
1800 to the "stats" file.
1805 Description: Client mtime is not the same on different clients after utimes
1806 Details : In some cases, the client was using the utimes() syscall on
1807 a file cached on another node. The clients now validate the
1808 ctime from the MDS + OSTs to determine which one is right.
1813 Description: Inability to activate failout mode
1814 Details : lconf script incorrectly assumed that in python string's numeric
1815 value is used in comparisons.
1818 Frequency : always with multiple stripes per file
1820 Description: Inefficient object allocation for mutli-stripe files
1821 Details : When selecting which OSTs to stripe files over, for files with
1822 a stripe count that divides evenly into the number of OSTs,
1823 the MDS is always picking the same starting OST for each file.
1824 Return the OST selection heuristic to the original design.
1829 Description: mount failures may take full timeout to return an error
1830 Details : Under some heavy load conditions it is possible that a
1831 failed mount can wait for the full obd_timeout interval,
1832 possibly several minutes, before reporting an error.
1833 Instead return an error as soon as the status is known.
1835 ------------------------------------------------------------------------------
1837 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
1839 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1840 INTEROPERATE with older versions automatically. Please read the
1841 user documentation before upgrading any part of a live system.
1842 * WARNING: Lustre networking configuration changes are required with
1843 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1846 * Support for kernels:
1847 2.6.9-22.0.2.EL (RHEL 4)
1848 2.6.5-7.244 (SLES 9)
1849 2.6.12.6 vanilla (kernel.org)
1852 Severity : enhancement
1853 Bugzilla : 7981/8208
1854 Description: Introduced Lustre Networking (LNET)
1855 Details : LNET is new networking infrastructure for Lustre, it includes
1856 a reorganized network configuration mode (see the user
1857 documentation for full details) as well as support for routing
1858 between different network fabrics. Lustre Networking Devices
1859 (LNDs) for the supported network fabrics have also been
1860 created for this new infrastructure.
1862 Severity : enhancement
1863 Description: Introduced Access control lists
1864 Details : clients can set ACLs on files and directories in order to have
1865 more fine-grained permissions than the standard Unix UGO+RWX.
1866 The MDS must be started with the "-o acl" mount option.
1868 Severity : enhancement
1869 Description: Introduced filesystem quotas
1870 Details : Administrators may now establish per-user quotas on the
1873 Severity : enhancement
1875 Description: Configuration change for the XT3
1876 The PTLLND is now used to run Lustre over Portals on the XT3
1877 The configure option(s) --with-cray-portals are no longer used.
1878 Rather --with-portals=<path-to-portals-includes> is used to
1879 enable building on the XT3. In addition to enable XT3 specific
1880 features the option --enable-cray-xt3 must be used.
1885 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1886 Details : A race between allocating a new llcd and re-getting the llcd_lock
1887 allowed another thread to grab newly-allocated llcd.
1889 Severity : enhancement
1891 Description: 2.6 OST async journal commit and locking fix to improve performance
1892 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1893 2.6 kernels are now async as they already were in 2.4 kernels so
1894 that they can commit concurrently with the network bulk transfer.
1895 For block-allocated files the filter allocation semaphore is held
1896 to avoid filesystem fragmentation during allocation. BKL lock
1897 removed for 2.6 xattr operations where it is no longer needed.
1902 Description: lconf incorrectly determined whether two IP networks could talk
1903 Details : In some more complicated routing and multiple-network
1904 configurations, lconf will avoid trying to make a network
1905 connection to a disjoint part of the IP space. It was doing the
1906 math incorrectly for one set of cases.
1911 Description: Fix for potential infinite loop processing records in an llog.
1912 Details : If an llog record is corrupted/zeroed, it is possible to loop
1913 forever in llog_process(). Validate the llog record length
1914 and skip the remainder of the block on error.
1917 Frequency : occasional (liblustre only)
1919 Description: liblustre could not open files whose last component is a symlink
1920 Details : sysio_path_walk() would incorrectly pass the open intent to
1921 intermediate path components.
1924 Frequency : rare (liblustre only with non-standard tuning)
1925 Bugzilla : 7201 (7350)
1926 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1927 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1928 after it has been released, only do this for referenced locks
1930 Severity : enhancement
1932 Description: specifying an (invalid) directory default stripe_size of -1
1933 would reset the directory default striping
1934 Details : stripe_size -1 was used internally to signal directory stripe
1935 removal, now use "all default" to signal dir stripe removal
1936 as a directory striping of "all default" is not useful
1939 Frequency : common for large clusters running liblustre clients
1941 Description: doing an ls when liblustre clients are running is slow
1942 Details : sending a glimpse AST to a liblustre client waits for every AST
1943 to time out, as liblustre clients will not respond. Since they
1944 cannot cache data we refresh the OST lock LVB from disk instead.
1946 Severity : enhancement
1948 Description: doing an ls at the same time as file IO can be slow
1949 Details : enqueue and other "small" requests can be blocked behind many
1950 large IO requests. Create a new OST IO portal for non-IO
1951 requests so they can be processed faster.
1954 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1956 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1957 created file with mode 000
1958 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1959 allow an ftruncate() as a non-root user to a file with mode 000.
1960 Lustre now respects this flag to disable mode checks when
1961 truncating a file owned by the user
1964 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1967 Description: Revoking locks from clients that went dead or catatonic might take
1969 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1970 cancellation of such locks instant on servers without waiting for
1971 any reply from clients. Clients drops these locks when cancel
1972 notification from server is received without replying.
1975 Frequency : liblustre-only, when liblustre client dies or becomes busy
1977 Description: Doing ls on Linux clients can take a long time with active
1979 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1980 granting such locks to it in the first place if possible. Locks
1981 are taken by proxy on the OST during the read or write and
1982 dropped immediately afterward. Add connect flags handling, do
1983 not grant locks to liblustre clients for glimpse ASTs.
1985 Severity : enhancement
1987 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1988 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1989 sequential reads and behaves badly with applications doing random
1990 reads. Improve it by reading ahead at least read region, and
1991 avoiding excessive large RPC for small reads.
1993 Severity : enhancement
1995 Description: Creating more than 1000 files for a single job may cause a load
1996 imbalance on the OSTs if there are also a large number of OSTs.
1997 Details : qos_prep_create() uses an OST index reseed value that is an
1998 even multiple of the number of available OSTs so that if the
1999 reseed happens in the middle of the object allocation it will
2000 still utilize the OSTs as uniformly as possible.
2005 Description: OST or MDS may oops in ping_evictor_main()
2006 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2007 but doesn't restart at beginning of obd_exports_timed list
2010 Severity : enhancement
2012 Description: improve by-nid export eviction on the MDS and OST
2013 Details : allow multiple exports with the same NID to be evicted at one
2014 time without re-searching the exports list.
2017 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2019 Description: MDS may oops in groups_free()
2020 Details : in rare race conditions a newly allocated group_info struct is
2021 freed again, and this can be NULL. The 2.4 compatibility code
2022 for groups_free() checked for a NULL pointer, but 2.6 did not.
2025 Frequency : common for liblustre clients doing little filesystem IO
2026 Bugzilla : 9352, 7313
2027 Description: server may evict liblustre clients accessing contended locks
2028 Details : if a client is granted a lock or receives a completion AST
2029 with a blocking AST already set it would not reply to the AST
2030 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2033 Frequency : lfs setstripe, only systems with more than 160 OSTs
2035 Description: unable to set striping with a starting offset beyond OST 160
2036 Details : llapi_create_file() incorrectly limited the starting stripe
2037 index to the maximum single-file stripe count.
2040 Frequency : LDAP users only
2042 Description: lconf did not handle in-kernel recovery with LDAP properly
2043 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2045 Severity : enhancement
2047 Description: bind OST threads to NUMA nodes to improve performance
2048 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2049 node and do their allocations there to localize memory access
2051 Severity : enhancement
2053 Description: llmount can determine client NID directly from Myrinet (GM)
2054 Details : the client NID code from gmnalnid was moved directly into
2055 llmount, removing the need to use this or specifying the
2056 client NID explicitly when mounting GM clients with zeroconf
2059 Frequency : if client is started with down MDS
2061 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2062 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2063 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2064 ptlrpc_import_delayed_req() because we didn't check if the
2065 request was interrupted, and we also didn't break out of the
2066 event loop if there was no timeout
2071 Description: data loss during non-page-aligned writes to a single file from
2072 both multiple nodes and multiple threads on one node at same time
2073 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2074 inconsistency led to false short-reads, that were cached and later
2075 used by ->prepare_write() to fill in partially written page,
2076 leading to data loss.
2079 Frequency : always, if lconf --abort_recovery used
2081 Description: lconf --abort_recovery fails with 'Operation not supported'
2082 Details : lconf was attempting to abort recovery on the MDT device and not
2085 Severity : enhancement
2087 Description: remove cleanup logs
2088 Details : replace lconf-generated cleanup logs with lustre internal
2089 cleanup routines. Eliminates the need for client-cleanup and
2092 Severity : enhancement
2094 Description: add support for EAs (user and system) on lustre filesystems
2095 Details : it is now possible to store extended attributes in the Lustre
2096 client filesystem, and with the user_xattr mount option it
2097 is possible to allow users to store EAs on their files also
2099 Severity : enhancement
2101 Description: Add possibility (config option) to show minimal available OST free
2103 Details : When compiled with --enable-mindf configure option, statfs(2)
2104 (and so, df) will return least minimal free space available from
2105 all OSTs as amount of free space on FS, instead of summary of
2106 free spaces of all OSTs.
2108 Severity : enhancement
2110 Description: do not expand extent locks acquired on OST-side
2111 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2112 by server: they are not cached anyway.
2115 Frequency : when mmap is used/binaries executed from Lustre
2117 Description: Unmmap pages before throwing them away from read cache.
2118 Details : llap_shrink cache now attempts to unmap pages before discarding
2119 them (if unmapping failed - do not discard). SLES9 kernel has
2120 extra checks that trigger if this unmapping is not done first.
2125 Description: lconf didn't resolve symlinks before checking to see whether a
2126 given mountpoint was already in use
2129 Frequency : when migrating failover services
2130 Bugzilla : 6395, 9514
2131 Description: When migrating a subset of services from a node (e.g. failback
2132 from a failover service node) the remaining services would
2133 time out and evict clients.
2134 Details : lconf --force (implied by --failover) sets the global obd_timeout
2135 to 5 seconds in order to quickly disconnect, but this caused
2136 other RPCs to time out too quickly. Do not change the global
2137 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2139 Severity : enhancement
2140 Frequency : if MDS is started with down OST
2141 Bugzilla : 9439,5706
2142 Description: Allow startup/shutdown of an MDS without depending on the
2143 availability of the OSTs.
2144 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2145 Add appropriate locking and lov-osc refcounts for safe
2146 cleaning. Add osc abort_inflight calls in case the
2147 synchronize never started.
2150 Frequency : occasional (Cray XT3 only)
2152 Description: root not authorized to access files in CRAY_PORTALS environment
2153 Details : The client process capabilities were not honoured on the MDS in
2154 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2155 been accessed by an authorized user then root was able to access
2156 the file on the local client also. The root user capabilities
2157 are now allowed on the MDS, as this environment has secure UID.
2160 Frequency : occasional
2162 Description: ldiskfs "too long searching" message happens too often
2163 Details : A debugging message (otherwise harmless) prints too often on
2164 the OST console. This has been reduced to only happen when
2165 there are fragmentation problems on the filesystem.
2170 Description: Division by zero in statfs when all OSCs are inactive
2171 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2172 lov_statfs divides by value returned by lov_get_stripecnt().
2176 Bugzilla : 9489, 3273
2177 Description: First write from each client to each OST was only 4kB in size,
2178 to initialize client writeback cache, which caused sub-optimal
2179 RPCs and poor layout on disk for the first writen file.
2180 Details : Clients now request an initial cache grant at (re)connect time
2181 and so that they can start streaming writes to the cache right
2182 away and always do full-sized RPCs if there is enough data.
2183 If the OST is rebooted the client also re-establishes its grant
2184 so that client cached writes will be honoured under the grant.
2189 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2190 Details : Now I/O RPCs go to different portal number and (presumably) fast
2191 lock requests (and glimses) and other RPCs get their own service
2192 threads pool that should be able to service those RPCs
2195 Severity : enhancement
2197 Description: Ability to exchange lustre version between client and servers and
2198 issue warnings at client side if client is too old. Also for
2199 liblustre clients there is ability to refuse connection of too old
2201 Details : New 'version' field is added to connect data structure that is
2202 filled with version info. That info is later checked by server and
2206 Frequency : rare, liblustre only.
2207 Bugzilla : 9296, 9581
2208 Description: Two simultaneous writes from liblustre at offset within same page
2209 might proceed at the same time overwriting eachother with stale
2211 Details : I/O lock withing llu_file_prwv was released too early, before data
2212 actually was hitting the wire. Extended lock-holding time until
2213 server acknowledges receiving data.
2216 Frequency : extremely rare. Never observed in practice.
2218 Description: avoid generating lustre_handle cookie of 0.
2219 Details : class_handle_hash() generates handle cookies by incrementing
2220 global counter, and can hit 0 occasionaly (this is unlikely, but
2221 not impossible, because initial value of cookie counter is
2222 selected randonly). Value of 0 is used as a sentinel meaning
2223 "unassigned handle" --- avoid it. Also coalesce two critical
2224 sections in this function into one.
2226 Severity : enhancement
2228 Description: allow liblustre clients to delegate truncate locking to OST
2229 Details : To avoid overhead of locking, liblustre client instructs OST to
2230 take extent lock in ost_punch() on client's behalf. New connection
2231 flag is added to handle backward compatibility.
2233 Severity : enhancement
2234 Bugzilla : 4928, 7341, 9758
2235 Description: allow number of OST service threads to be specified
2236 Details : a module parameter allows the number of OST service threads
2237 to be specified via "options ost ost_num_threads={N}" in the
2238 OSS's /etc/modules.conf or /etc/modprobe.conf.
2242 Bugzilla : 6146, 9635, 9895
2243 Description: servers crash with bad pointer in target_handle_connect()
2244 Details : In rare cases when a client is reconnecting it was possible that
2245 the connection request was the last reference for that export.
2246 We would temporarily drop the export reference and get a new
2247 one, but this may have been the last reference and the export
2248 was just destroyed. Get new reference before dropping old one.
2250 Severity : enhancement
2251 Frequency : if client is started with failover MDS
2253 Description: Allow multiple MDS hostnames in the mount command
2254 Details : Try to read the configuration from all specified MDS
2255 hostnames during a client mount in case the "primary"
2258 Severity : enhancement
2260 Description: Stop sending data to evicted clients as soon as possible.
2261 Details : Check if the client we are about to send or are sending data to
2262 was evicted already. (Check is done every second of waiting,
2263 for which l_wait_event interface was extended to allow checking
2264 of exit condition at specified intervals).
2267 Frequency : rare, normally only when NFS exporting is done from client
2269 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2271 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2274 Frequency : always, if srand() is called before liblustre initialization
2276 Description: Liblustre uses system PRNG disturbing its usage by user application
2277 Details : Introduce internal to lustre fast and high-quality PRNG for
2278 lustre usage and make liblustre and some other places in generic
2279 lustre code to use it.
2281 Severity : enhancement
2282 Bugzilla : 9477, 9557, 9870
2283 Description: Verify that the MDS configuration logs are updated when xml is
2284 Details : Check if the .xml configuration logs are newer than the config
2285 logs stored on the MDS and report an error if this is the case.
2286 Request --write-conf, or allow starting with --old_conf.
2288 Severity : enhancement
2290 Description: Handle symlinks in the path when checking if Lustre is mounted.
2291 Details : Resolve intermediate symlinks when checking if a client has
2292 mounted a filesystem to avoid duplicate client mounts.
2297 Description: lconf can hit an error exception but still return success.
2298 Details : The lconf command catches the Command error exception at the top
2299 level script context and will exit with the associated exit
2300 status, but doesn't ensure that this exit status is non-zero.
2305 Description: failure of ptlrpc thread startup can cause oops
2306 Details : Starting a ptlrpc service thread can fail if there are a large
2307 number of threads or the server memory is very fragmented.
2308 Handle this without oopsing.
2311 Frequency : always, only if liblustre and non-default acceptor port was used
2313 Description: liblustre cannot connect to servers with non-default acceptor port
2314 Details : tcpnal_set_default_params() was not called and was therefore
2315 ignoring the environment varaible TCPNAL_PORT, as well as other
2316 TCPNAL_ environment variables
2321 Description: two objects could be created on the same OST for a single file
2322 Details : If an OST is down, in some cases it was possible to create two
2323 objects on a single OST for a single file. No problems other
2324 than potential performance impact and spurious error messages.
2328 Bugzilla : 5681, 9562
2329 Description: Client may oops in ll_unhash_aliases
2330 Details : Client dcache may become inconsistent in race condition.
2331 In some cases "getcwd" can fail if the current directory is
2337 Description: Inode refcounting problems in NFS export code
2338 Details : link_raw functions used to call d_instantiate without obtaining
2339 extra inode reference first.
2343 Bugzilla : 9942, 9903
2344 Description: Referencing freed requests leading to crash, memleaks with NFS.
2345 Details : We used to require that call to ll_revalidate_it was always
2346 followed by ll_lookup_it. Also with revalidate_special() it is
2347 possible to call ll_revalidate_it() twice for the same dentry
2348 even if first occurence returned success. This fix changes semantic
2349 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2350 reference on a request referred from the intent.
2351 ll_intent_release() then releases such a request.
2354 Frequency : rare, normally benchmark loads only
2356 Description: unlinked inodes were kept in memory on the client
2357 Details : If a client is repeatedly creating and unlinking files it
2358 can accumulate a lot of stale inodes in the inode slab cache.
2359 If there is no other client load running this can cause the
2360 client node to run out of memory. Instead flush old inodes
2361 from client cache that have the same inode number as a new inode.
2364 Frequency : SLES9 2.6.5 kernel and long filenames only
2365 Bugzilla : 9969, 10379
2366 Description: utime reports stale NFS file handle
2367 Details : SLES9 uses out-of-dentry names in some cases, which confused
2368 the lustre dentry revalidation. Change it to always use the
2372 Frequency : rare, unless heavy write-truncate concurrency is continuous
2373 Bugzilla : 4180, 6984, 7171, 9963, 9331
2374 Description: OST becomes very slow and/or deadlocked during object unlink
2375 Details : filter_destroy() was holding onto the parent directory lock
2376 while truncating+unlinking objects. For very large objects this
2377 may block other threads for a long time and slow overall OST
2378 responsiveness. It may also be possible to get a lock ordering
2379 deadlock in this case, or run out of journal credits because of
2380 the combined truncate+unlink. Solution is to do object truncate
2381 first in one transaction without parent lock, and then do the
2382 final unlink in a new transaction with the parent lock. This
2383 reduces the lock hold time dramatically.
2386 Frequency : rare, 2.4 kernels only
2388 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2389 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2390 before unmounting their filesystem in order to avoid deadlock.
2391 The kernel_locked() function in 2.4 kernels only checks whether
2392 the kernel lock is held, not whether it is this process that is
2393 holding it as 2.6 kernels do.
2398 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2399 Details : The client ptlrpc code may be trying to reconnect to a down
2400 server before a previous connection attempt has timed out.
2401 Increase the reconnect interval to be longer than the connection
2402 timeout interval to avoid sending duplicate connections to
2406 Frequency : echo_client brw_test command
2408 Description: fix echo_client to work with OST preallocated code
2409 Details : OST preallocation code (5137) didn't take echo_client IO path
2410 into account: echo_client calls filter methods outside of any
2411 OST thread and, hence, there is no per-thread preallocated
2412 pages and buffers to use. Solution: hijack pga pages for IO. As
2413 a byproduct, this avoids unnecessary data copying.
2417 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2418 Description: Client can oops in mdc_commit_close() after open replay
2419 Details : It was possible for the MDS to return an open request with no
2420 transaction number in mds_finish_transno() if the client was
2421 evicted, but without actually returning an error. Clients
2422 would later try to replay that open and may trip an assertion
2423 Simplify the client close codepath, and always return an error
2424 from the MDS in case the open is not successful.
2427 Frequency : rare, 2.6 OSTs only
2429 Description: OST may deadlock under high load on fragmented files
2430 Details : If there was a heavy load and highly-fragmented OST filesystems
2431 it was possible to have all the OST threads deadlock waiting on
2432 allocation of biovecs, because the biovecs were not released
2433 until the entire RPC IO was completed. Instead, release biovecs
2434 as soon as they are complete to ensure forward IO progress.
2436 Severity : enhancement
2438 Description: Support for specifying external journal device at mount
2439 Details : If an OST or MDS device is formatted with an external journal
2440 device, this device major/minor is stored in the ext3 superblock
2441 and may not be valid for failover. Allow detecting and
2442 specifying the external journal at mount time.
2447 Description: Mounting an MDS with pending unlinked files may cause oops
2448 Details : target_finish_recovery() calls mds_postrecov() which returned
2449 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2450 considers this an error and immediately begins cleaning up the
2451 lov, just after starting the mds_lov process
2453 Severity : enhancement
2455 Description: Implement 'lfs df' to report actual free space on per-OST basis
2456 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2457 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2458 usage of objects; '-h' to report in human readable format.
2460 ------------------------------------------------------------------------------
2462 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
2469 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2470 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2471 at startup during orphan unlinking. Instead, skip buddy update
2472 before setup, as it will be regenerated after recovery is complete.
2477 Description: If an OST is inactive, its locks might reference stale inodes.
2478 Details : lov_change_cbdata() must iterate over all namespaces, even if
2479 they are inactive to clear inode references from the lock.
2481 Severity : enhancement
2482 Frequency : occasional, if non-standard max_dirty_mb used
2484 Description: Client will block write RPCs if not enough grant
2485 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2486 then the client will block writes while waiting for another RPC
2487 to complete instead of consuming its dirty limit. With change
2488 we get improved performance when max_dirty_mb is small.
2490 Severity : enhancement
2491 Bugzilla : 3389, 6253
2492 Description: Add support for supplementary groups on the MDS.
2493 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2494 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2495 lookups for user supplementary groups into a cache.
2499 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2500 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2501 for liblustre O_EXCL is enforced by MDS.
2505 Description: readdir never returns NULL in liblustre.
2506 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2507 dentry in struct dirent.
2511 Frequency : liblustre only, depends on application IO pattern
2512 Description: liblustre clients evicted if not contacting servers
2513 Details : Don't put liblustre clients into the ping_evictor list, so
2514 they will not be evicted by the pinger ever.
2516 Severity : enhancement
2518 Description: Add ability to evict clients by NID from MDS.
2519 Details : By echoing "nid:$NID" string into
2520 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2521 $NID would be instantly evicted from this MDS and from all active
2522 OSTs connected to it.
2526 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2527 Details : lookup_it_finish() used to query file size from OSTs that was not
2532 Description: service threads change working directory to that of init
2533 Details : Starting lustre service threads may pin the working directory
2534 of the parent thread, making that filesystem busy. Threads
2535 now change to the working directory of init to avoid this.
2539 Frequency : during shutdown only
2540 Description: shutdown with a failed MDS or OST can cause unmount to hang
2541 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2546 Frequency : 2.6 only
2547 Description: chmod/chown may include an extra supplementary group
2548 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2549 supplementary group and if none is specified this is used.
2552 Bugzilla : 5479 (6816)
2554 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2555 Details : If lookup is in progress on a file that is unlinked we might try
2556 to revalidate the inode and fail in revalidate after lookup is
2557 complete and ll_file_open() enqueues the open again but
2558 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2561 Frequency : always, if lconf --abort_recovery used
2563 Description: lconf --abort_recovery fails with 'Operation not supported'
2564 Details : lconf was attempting to abort recovery on the MDT device and not
2567 ------------------------------------------------------------------------------
2569 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2574 Frequency : rare (only unsupported configurations with a node running as an
2576 Bugzilla : 6514, 5137
2577 Description: Mounting a Lustre file system on a node running as an OST could
2579 Details : OSTs now preallocates memory needed to write out data at
2580 startup, instead of when needed, to avoid having to
2581 allocate memory in possibly low memory situations.
2582 Specifically, if the file system is mounted on on OST,
2583 memory pressure could force it to try to write out data,
2584 which it needed to allocate memory to do. Due to the low
2585 memory, it would be unable to do so and the node would
2586 become unresponsive.
2588 Severity : enhancement
2590 Description: Addition of lconf --service command line option
2591 Details : lconf now accepts a '--service <arg>' option, which is
2592 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2594 Severity : enhancement
2596 Description: Failover mode is now the default for OSTs.
2597 Details : By default, OSTs will now run in failover mode. To return to
2598 the old behaviour, add '--failout' to the lmc line for OSTs.
2600 Severity : enhancement
2602 Description: Health checks are now provided for MDS and OSTs
2603 Details : Additional detailed health check information on MSD and OSTs
2604 is now provided through the procfs health_check value.
2607 Frequency : occasional, depends on IO load
2609 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2610 after numerous create/delete cycles
2611 Details : The ext3 inode allocation policy would not allocate new inodes
2612 very well on the OSTs because there are no new directories
2613 being created. Instead we look for groups with free space if
2614 the parent directories are nearly full.
2619 Description: Network or server problems during mount may cause partially
2620 mounted clients instead of returning an error.
2621 Details : The config llog parsing code may overwrite the error return
2622 code during mount error handling, returning success instead
2628 Description: MDS can fail to allocate large reply buffers
2629 Details : After long uptimes the MDS can fail to allocate large reply
2630 buffers (e.g. zconf client mount config records) due to memory
2631 fragmentation or consumption by the buffer cache. Preallocate
2632 some large reply buffers so that these replies can be sent even
2633 under memory pressure.
2637 Frequency : rare (liblustre)
2638 Description: fsx running with liblustre complained that using truncate() to
2639 extend the file doesn't work. This patch corrects that issue.
2640 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2641 ATTR_SIZE and lsm use in llu_setattr_raw.
2645 Frequency : rare, only 2.6 kernels
2646 Description: Unusual file access patterns on the MDS may result in inode
2647 data being lost in very rare circumstances.
2648 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2649 code in the 2.6 kernel caused the inode and/or EA data not to
2650 be read from disk, causing single-file corruption.
2654 Frequency : rare, only 2.6 filesystems using extents
2655 Description: Heavy concurrent write and delete load may cause data corruption.
2656 Details : It was possible under high-load situations to have an extent
2657 metadata block in the block device cache from a just-unlinked
2658 file overwrite a newly-allocated data block. We now unmap any
2659 metadata buffers that alias just-allocated data blocks.
2663 Frequency : filesystems with default stripe_count larger than 77
2664 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2665 Details : lconf specifies an inode size of 4096 bytes when the default
2666 stripe_count is larger than 77. This conflicts with the default
2667 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2668 this case to avoid pinning too much memory for large EAs.
2670 ------------------------------------------------------------------------------
2672 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2677 Frequency : rare (extremely heavy IO load with hundreds of clients)
2679 Description: Client is evicted, gets IO error writing to file
2680 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2681 caused clients to be evicted for AST timeouts. The fixes in
2682 bug 5192 mean we no longer need to have such short AST timeouts
2683 so ldlm_timeout has been increased.
2686 Frequency : occasional during --force or --failover shutdown under load
2687 Bugzilla : 5949, 4834
2688 Description: Server oops/LBUG if stopped with --force or --failover under load
2689 Details : a collection of import/export refcount and cleanup ordering
2690 issues fixed for safer force cleanup
2693 Frequency : only filesystems larger than 120 OSTs
2694 Bugzilla : 5990, 6223
2695 Description: lfs getstripe would oops on a very large filesystem
2696 Details : lov_getconfig used kfree on vmalloc'd memory
2699 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2700 Bugzilla : 6242, 6243
2701 Description: reading from files that had been truncated to a non-zero size
2702 but never opened returned no data
2703 Details : ll_file_read() reads zeros from no-object files to EOF
2708 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2709 Details : There's little harm in aborting MDS/OSS recovery and letting it
2710 try again, so I removed the LASSERT and return an error instead.
2712 Severity : enhancement
2714 Description: New debugging infrastructure for tracking down data corruption
2715 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2716 (b) cover more of the client-side code path, and (c) try to narrow
2717 down where problems occurred
2721 Bugzilla : 3819, 4364, 4397, 6313
2722 Description: Racing close and eviction MDS could cause assertion in mds_close
2723 Details : It was possible to get multiple mfd references during close and
2724 client eviction, leading to one thread referencing a freed mfd.
2726 Severity: : enhancement
2727 Bugzilla : 3262, 6359
2728 Description: Attempts to reconnect to servers are now more aggressive.
2729 Details : This builds on the enhanced upcall-less recovery that was added
2730 in 1.4.2. When trying to reconnect to servers, clients will
2731 now try each server in the failover group every 10 seconds. By
2732 default, clients would previously try one server every 25 seconds.
2737 Description: After recovery, certain operations trigger a failed
2738 assertion on a client.
2739 Details : Failing over an mds, using lconf -d --failover, while a
2740 client was doing a readdir() call would cause the client to
2741 LBUG after recovery completed and the readdir() was resent.
2743 Severity : enhancement
2745 Description: Default groups are now added by lconf
2746 Details : You can now run lconf --group <servicename> without having to
2747 manually add groups with lmc.
2750 Frequency : occasional
2752 Description: Nodes with an elan id of 0 trigger a failed assertion
2755 Frequency : always when accessing e.g. tty/console device nodes
2757 Description: tty and some other devices nodes cannot be used on lustre
2758 Details : file's private_data field is used by device data and lustre
2759 values in there got lost. New field was added to struct file to
2760 store fs-specific private data.
2763 Frequency : when exporting Lustre via NFS
2765 Description: NFSD failed occasionally when looking up a path component
2766 Details : NFSD is looking up ".." which was broken in ext3 directories
2767 that had grown large enough to become hashed.
2770 Frequency : Clusters with multiple interfaces not on the same subnet
2772 Description: Nodes will repeatedly try to reconnect to an interface which it
2773 cannot reach and report an error to the log.
2774 Details : Extra peer list entries will be created by lconf with some peers
2775 unreachable. lconf now validates the peer before adding it.
2778 Frequency : Only if a default stripe is set on the filesystem root.
2780 Description: Setting a default stripe on the filesystem root prevented the
2781 filesystem from being remounted.
2782 Details : The client was sending extra request flags in the root getattr
2783 request and did not allocate a reply buffer for the dir EA.
2786 Frequency : occasional, higher if lots of files are accessed by one client
2787 Bugzilla : 6159, 6097
2788 Description: Client trips assertion regarding lsm mismatch/magic
2789 Details : While revalidating inodes the VFS looks up inodes with ifind()
2790 and in rare cases can find an inode that is being freed.
2791 The ll_test_inode() code will free the lsm during ifind()
2792 when it finds an existing inode and then the VFS later attaches
2793 this free lsm to a new inode.
2797 Bugzilla : 6422, 7030
2798 Description: MDS deadlock between mkdir and client eviction
2799 Details : Creating a new file via mkdir or mknod (starting a transaction
2800 and getting the ns lock) can deadlock with client eviction
2801 (gets ns lock and trying to finish a synchronous transaction).
2804 Frequency : occasional
2805 Description: While starting a server, the fsfilt_ext3 module could not be
2807 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2808 kernels. Previously, lconf would still use the ext3 name
2809 when trying to load modules. Now, it will correctly use
2810 ext3 on 2.4 and ldiskfs on 2.6.
2812 Severity : enhancement
2813 Description: The default stripe count has been changed to 1
2814 Details : The interpretation of the default stripe count (0, to lfs
2815 or lmc) has been changed to mean striping across a single
2816 OST, rather than all available. For general usage we have
2817 found a stripe count of 1 or 2 works best.
2819 Severity : enhancement
2820 Description: Add support for compiling against Cray portals.
2821 Details : Conditional compiling for some areas that are different
2825 Frequency : occasional
2826 Bugzilla : 6409, 6834
2827 Description: Creating files with an explicit stripe count may lead to
2828 a failed assertion on the MDS
2829 Details : If some OSTs are full or unavailable, creating files may
2830 trigger a failed assertion on the MDS. Now, Lustre will
2831 try to use other servers or return an error to the
2835 Frequency : occasional
2837 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2838 caused lock timeout during readahead (since 1.4.2).
2839 Details : Processes doing readahead might match a lock that hasn't been
2840 granted yet if there are overlapping and conflicting lock
2841 requests. The readahead process waits on ungranted lock
2842 (original lock is CBPENDING), while OST waits for that process
2843 to cancel CBPENDING read lock and eventually evicts client.
2845 Severity : enhancement
2847 Description: Initial enabling of flock support for clients
2848 Details : Implements fcntl advisory locking and file status functions.
2849 This feature is provided as an optional mount flag (default
2850 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2851 locking are implemented yet, and those that are are not guaranteed
2852 to be completely correct in production environments.
2853 mount -t lustre -o [flock|noflock] ...
2856 Frequency : occasional
2858 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2859 assertion in the ext3 JBD (journaling) layer.
2860 Details : The b_committed_data struct is protected by the big kernel lock
2861 in 2.4 kernels, serializing journal_commit_transaction() and
2862 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2863 access to this struct. In 2.6 kernels there is finer grained
2864 locking to improve SMP performance of the JBD layer.
2868 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2870 -----------------------------------------------------------------------------
2872 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2874 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2875 versions of Lustre. Please update all servers and clients to
2876 version 1.4.2 or later at the same time. You must also run
2877 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2878 to update the configuration logs.
2880 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2881 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2882 - swab llog records of type '0' so we get proper header size/idx (5861)
2883 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2884 - fix rename of one directory over another leaking an inode (5953)
2885 - avoid SetPageDirty on 2.6 (5981)
2886 - don't re-add just-being-destroyed locks to the waiting list (5653)
2887 - when creating new directories, inherit the parent's custom
2888 striping settings if present parent (3048)
2889 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2890 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2891 - don't hold client locks on temporary worklist from l_lru (5666)
2892 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2893 - automatically evict dead clients (5921)
2894 - Update file size properly in create+truncate+fstat case (6196)
2895 - Do not unhash mountpoint dentries, do not allow removal of
2897 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2898 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2899 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2900 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2901 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2902 - Correctly handle reads of files with no objects (6243)
2903 - lctl recover will also mark a device active if deactivate used (5933)
2905 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2906 - allow --write-conf on an MDS with different nettype than client (5619)
2907 - don't write config llogs to MDS for mounts not from that MDS (5617)
2908 - lconf should create multiple TCP connections from a client (5201)
2909 - init scripts are now turned off by default; run chkconfig --on
2910 lustre and chkconfig --on lustrefs to use them
2911 - upcalls are no longer needed for clients to recover to failover
2913 - add --abort-recovery option to lconf to abort recovery on device
2915 - add support for an arbitrary number of OSTs (3026)
2916 - Quota support protocol changes.
2917 - forward compatibility changes to wire structs (6007)
2918 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2919 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2920 - improved "lustre status" script
2921 - initialize blocksize for non-regular files (6062)
2922 - added --disable-server and --disable-client configure options (5782)
2923 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2924 - Vanilla 2.4.29 support
2925 - increase maximum number of obd devices to 520 (6242)
2926 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2927 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2928 SLES 9 2.6.5 kernels
2929 - limit stripes per file to 160 (the maximum EA size) (6093)
2931 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2934 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2935 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2936 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2937 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2938 - small ext3 extents cleanups and fixes (5733)
2939 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2940 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2941 - avoid needless client->OST connect, fix handle mismatch (5317)
2942 - fix DLM error path that led to out-of-sync client, long delays (5779)
2943 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2944 - fix several locking issues related to i_size (5492,5624,5654,5672)
2945 - don't move pending lock onto export if it is already evicted (5683)
2946 - fix kernel oops when creating .foo in unlinked directory (5548)
2947 - fix deadlock in obdfilter statistics vs. object create (5811)
2948 - use time_{before,after} to avoid timer jiffies wrap (5882)
2949 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2950 - Do not leak request if server was not able to process it (5154)
2951 - If mds_open unable to find parent dir, make that negative lookup(5154)
2952 - don't create new directories with extent-mapping (5909, 5936)
2954 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2955 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2956 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2957 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2958 - support for RHEL4 kernel on i686 (5773)
2959 - provide error messages when incompatible logs are encountered (5898)
2961 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2962 * version 1.4.0.10 (1.4.1 release candidate 1)
2964 - don't keep a lock reference when lock is not granted (4238)
2965 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2966 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2967 - drop import inflight refcount on signal_completed_replay error (5255)
2968 - unlock page after async write error during send (3677)
2969 - handle missing objects in filter_preprw_read properly (5265)
2970 - no transno return for symlink open, don't save no-trasno open (3440)
2971 - don't try to complete elan receive that already failed (4012)
2972 - free RPC server reply state on error (5406)
2973 - clean up thread from ptlrpc_start_thread() on error (5160)
2974 - readahead could read extra page into cache that wasn't ejected (5388)
2975 - prevent races in class_attach/setup/cleanup/detach (5260)
2976 - don't dereference de->d_inode after l_dput of de (5458)
2977 - use "int" for stripe value returned from lock_to_stripe (5544)
2978 - mballoc allocation and error-checking fixes in 2.6 (5504)
2979 - block device patches to fix I/O request sizes in 2.6 (5482)
2980 - look up hostnames for IB nals (5602)
2981 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2982 - don't start multiple acceptors for the same port (5277)
2983 - fix incorrect LASSERT in mds_getattr_name (5635)
2984 - export a proc file for general "ping" checking (5628)
2985 - fix "lfs check" to not block when the MDS is down (5628)
2987 - service request history (4965)
2988 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2989 - create an "evict_client" /proc entry on OSTs, like the MDS has
2990 - fix mount usage message, return errors per mount(8) (5168)
2991 - change grep [] to grep "[]" in tests so they work in more UMLs
2992 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2993 - remove ancient LOV_MAGIC_V0 EA support (5047)
2994 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2995 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2996 - allow readahead allocations to fail when low on memory (5383)
2997 - mmap locking landed again, after considerable improvement (2828)
2998 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3000 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3003 - send OST transaction number in read/write reply to free req (4966)
3004 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3005 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3006 - the watchdog thread now runs as interruptible (5246)
3007 - flock/lockf fixes (but it's still disabled, pending 5135)
3008 - don't use EXT3 constants in llite code (5094)
3009 - memory shortage at startup could cause assertion (5176)
3011 - reorganization of lov code
3012 - single portals codebase
3014 - add extents/mballoc support (5025)
3015 - direct I/O reads in the obdfilter (4048)
3016 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3018 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3021 - send OST transaction number in read/write reply to free req (4966)
3022 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3023 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3024 - the watchdog thread now runs as interruptible (5246)
3025 - handle missing objects in filter_preprw_read properly (5265)
3026 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3027 - drop import inflight refcount on signal_completed_replay error (5255)
3028 - unlock page after async write error during send (3677)
3029 - return original error code on reconstructed replies (3761)
3030 - no transno return for symlink open, don't save no-trasno open (3440)
3032 - add pid to ldlm debugging output (4922)
3033 - bump the watchdog timeouts -- we can't handle 30sec yet
3034 - extra debugging for orphan dentry/inode bug (5259)
3036 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3039 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3040 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3041 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3042 - fix NULL dereference in /proc/sys/portals/routes (4827)
3043 - allow failed mdc_close() operations to be interrupted (4561)
3044 - stop precreate on OST before MDS would time out on it (4778)
3045 - don't send partial-page writes before EOF from client (4410)
3046 - discard client grant for sub-page writes on large-page clients (4520)
3047 - don't free dentries not owned by NFS code, check generation (4806)
3048 - fix lsm leak if mds_create_objects() fails (4801)
3049 - limit debug_daemon file size, always print CERROR messages (4789)
3050 - use transno after validating reply (3892)
3051 - process timed out requests if import state changes (3754)
3052 - update mtime on OST during writes, return in glimpse (4829)
3053 - add mkfsoptions to LDAP (4679)
3054 - use ->max_readahead method instead of zapping global ra (5039)
3055 - don't interrupt __l_wait_event() during strace
3057 - add software watchdogs to catch hung threads quickly (4941)
3058 - make lustrefs init script start after nfs is mounted
3059 - fix CWARN/ERROR duplication (4930)
3060 - return async write errors to application if possible (2248)
3061 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3062 - print NAL number in %x format (4645)
3063 - update barely-supported suse-2.4.21-171 series (4842)
3064 - support for sles 9 %post scripts
3065 - support for building 2.6 kernel-source packages
3066 - support for sles km_* packages
3068 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3071 - ignore -ENOENT errors in osc_destroy (3639)
3072 - notify osc create thread that OSC is being cleaned up (4600)
3073 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3074 - reconstruct ost_handle() like mds_handle() (4657)
3075 - create a new thread to do import eviction to avoid deadlock (3969)
3076 - let lconf resolve symlinked-to devices (4629)
3077 - don't unlink "objects" from directory with default EA (4554)
3078 - hold socknal file ref over connect in case target is down (4394)
3079 - allow more than 32000 subdirectories in a single directory (3244)
3080 - fix blocks count for O_DIRECT writes (3751)
3081 - OST returns ENOSPC from object create when no space left (4539)
3082 - don't send truncate RPC if file size isn't changing (4410)
3083 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3084 - bind to privileged port in socknal and tcpnal (3689)
3086 - rate limit CERROR/CWARN console message to avoid overload (4519)
3087 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3088 - basic mmap support (3918)
3089 - kernel patch series update from b1_4 (4711)
3091 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3094 - avoid crash during MDS cleanup with OST shut down (2775)
3095 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3096 - don't use bad inodes on the MDS (3744)
3097 - dynamic object preallocation to improve recovery speed (4236)
3098 - don't hold spinlock over lock dumping or change debug flags (4401)
3099 - don't zero obd_dev when it is force cleaned (3651)
3100 - print grants to console if they go negative (4431)
3101 - "lctl deactivate" will stop automatic recovery attempts (3406)
3102 - look for existing locks in ldlm_handle_enqueue() (3764)
3103 - don't resolve lock handle twice in recovery avoiding race (4401)
3104 - revalidate should check working dir is a directory (4134)
3106 - don't always mark "slow" obdfilter messages as errors (4418)
3108 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3111 - don't close LustreDB during write_conf until it is done (3860)
3112 - fix typo in lconf for_each_profile (3821)
3113 - allow dumping logs from multiple threads at one time (3820)
3114 - don't allow multiple threads in OSC recovery (3812)
3115 - fix debug_size parameters (3864)
3116 - fix mds_postrecov to initialize import for llog ctxt (3121)
3117 - replace config semaphore with spinlock (3306)
3118 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3119 - don't allow enqueue to complete on a destroyed export (3822)
3120 - down write_lock before checking llog header bitmap (3825)
3121 - recover from lock replay timeout (3764)
3122 - up llog sem before sending rpc (3652)
3123 - reduce ns lock hold times when setting kms (3267)
3124 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3125 - fix NULL deref and obd_dev leak on setup error (3312)
3126 - replace some LBUG about llog ops with error handling (3841)
3127 - don't match INVALID dentries from d_lookup and spin (3784)
3128 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3129 - fix invalid assertion in ptlrpc_set_wait (3880)
3131 - add libwrap support for the TCP acceptor (3996)
3132 - add /proc/sys/portals/routes for non-root route listing (3994)
3133 - allow setting MDS UUID in .xml (2580)
3134 - print the stack of a process that LBUGs (4228)
3136 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3139 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3140 - make sure to unset replay flag from failed open requests (3440)
3141 - if default stripe count is 0, use OST count for inode size (3636)
3142 - update parent mtime/ctime on client for create/unlink (2611)
3143 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3144 - free recovery state on server during a forced cleanup (3571)
3145 - unregister_reply for resent reqs (3063)
3146 - loop back devices mounting and status check on 2.6 (3563)
3147 - fix resource-creation race that can provoke i_size == 0 (3513)
3148 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3149 - more debugging for page-accounting assertion (3746)
3150 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3151 - avoid deadlock after precreation failure (3758)
3152 - fix race and lock order deadlock in orphan handling (3450, 3750)
3153 - add validity checks when grabbing inodes from l_ast_data (3599)
3155 - add /proc/.../recovery_status to obdfilter (3428)
3156 - lightweight CDEBUG infrastructure, debug daemon (3668)
3157 - change default OSC RPC parameters to be better on small clusters
3158 - turn off OST read cache for files smaller than 32MB
3159 - install man pages and include them in rpms (3100)
3160 - add new init script for (un)mounting lustre filesystems (2593)
3161 - run chkconfig in %post for init scripts (3701)
3162 - drop scimac NAL (unmaintained)
3164 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3167 - clean kiobufs before and after use (3485)
3168 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3169 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3170 - change init script to fail more clearly if not run as root (1528)
3171 - allow clients to reconnect during replay (1742)
3172 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3173 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3174 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3175 - fix cond_resched (3554)
3176 - only evict unfinished clients after recovery (3515)
3177 - allow bulk resend, prevent data loss (3570)
3178 - dynamic ptlrpc request buffer allocation (2102)
3179 - don't allow unlinking open directory if it isn't empty (2904)
3180 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3181 - remove extraneous obd dereference causing LASSERT failure (3334)
3182 - don't use get_cycles() when creating temp. files on the mds (3156)
3183 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3184 - handle EEXIST for set-stripe, set proper directory name (3336)
3186 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3187 - fix ksocknal_fmb_callback() error messages (2918)
3189 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3192 - don't copy lvb into (possibly NULL) reply on error (2983)
3193 - don't deref dentry after dput, don't free lvb on error (2922)
3194 - use the kms to determine writeback rpc length (2947)
3195 - increment oti_logcookies when osc is inactive (2948)
3196 - update client's i_blocks count via lvb messages (2543)
3197 - handle intent open/close of special files properly (1557)
3198 - mount MDS with errors=remount-ro, like obdfilter (2009)
3199 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3200 - don't use cancelling-locks' kms values (2947)
3201 - use highest lock extent for kms, not last one (2925)
3202 - don't dereference ERR_PTR() dentry in error handling path (3107)
3203 - fix thread race in portals_debug_dumplog() (3122)
3204 - create lprocfs device entries at setup instead of at attach (1519)
3205 - common AST error handler, don't evict client on completion race (3145)
3206 - zero nameidata in detach_mnt in 2.6 (3118)
3207 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3208 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3209 - initialize RPC timeout timer earlier for 2.6 (3219)
3210 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3211 - print nal/nid for unknown nid (3258)
3212 - additional checks for oscc recovery before doing precreate (3284)
3213 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3214 - don't crash in mdc_close for bad permissions on open (3285)
3215 - zero i_rdev for non-device files (3147)
3216 - clear page->private before handing to FS, better assertion (3119)
3217 - tune the read pipeline (3236)
3218 - fix incorrect decref of invalidated dentry (2350)
3219 - provide read-ahead stats and refine rpc in flight stats (3328)
3220 - don't hold journal transaction open across create RPC (3313)
3221 - update atime on MDS at close time (3265)
3222 - close LDAP connection when recovering to avoid server load (3315)
3223 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3224 - don't leak open file on MDS after open resend (3325)
3225 - serialize filter_precreate and filter_destroy_precreated (3329)
3226 - loop device shouldn't call sync_dev() for nul device (3092)
3227 - clear page cache after eviction (2766)
3228 - resynchronize MDS->OST in background (2824)
3229 - refuse to mount the same filesystem twice on same mountpoint (3394)
3230 - allow llmount to create routes for mounting behind routers (3320)
3231 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3232 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3233 - fix rare problem with rename on htree directories (3417)
3235 - allow default OST striping configuration per directory (1414)
3236 - fix compilation for qswnal for 2.6 kernels (3125)
3237 - increase maximum number of MDS request buffers for large systems
3238 - change liblustreapi to be useful for external progs like lfsck (3098)
3239 - increase local configuration timeout for slow disks (3353)
3240 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3242 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3245 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3246 - don't overwrite extent policy data in reply if lock was blocked (2901)
3247 - drop filter export grants atomically with removal from device (2663)
3248 - del obd_self_export from work_list in class_disconnect_exports (2908)
3249 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3250 - swab reply message in mdc_close, other PPC fixes (2464)
3251 - fix destroying of named logs (2325)
3252 - overwrite old logs when running lconf --write_conf (2264)
3253 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3254 - fix race in target_handle_connect (2898)
3255 - mds_reint_create() should take same inode create lock (2926)
3256 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3257 - don't close files for self_export to avoid uninitialized obd (2936)
3258 - allow MDS with the same name as client node (2939)
3259 - hold dentry reference for closed log files for unlink (2325)
3260 - reserve space for all logs during transactions (2059)
3261 - don't evict page beyond end of stripe extent (2925)
3262 - don't oops on a deleted current working directory (2399)
3263 - handle hard links to targets without a parent properly (2517)
3264 - don't dereference NULL lock when racing during eviction (2867)
3265 - don't grow lock extents when lots of conflicting locks (2919)
3267 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3270 - account for cache space usage on clients to avoid data loss (974)
3271 - lfsck support in lustre kernel code (2349)
3272 - reduce journal credits needed for BRW writes (2370)
3273 - orphan handling to avoid losing space on client/server crashes
3274 - ptlrpcd can be blocked, stopping ALL progress (2477)
3275 - use lock value blocks to assist in proper KMS, faster stat (1021)
3276 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3277 - recovery for initial connections (2355)
3278 - fixes for mds_cleanup_orphans (1934)
3279 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3280 - block all file creations until orphan recovery completes (1901)
3281 - client remove rq_connection from request struct (2423)
3282 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3283 - recovery timer race (2670)
3284 - mdc_close recovey bug (2532)
3285 - ptlrpc cleanup bug (2710)
3286 - mds timeout on local locks (2588)
3287 - namespace lock held during RPCs (2431)
3288 - handle interrupted sync write properly (2503)
3289 - don't try to handle a message that hasn't been replied to (2699)
3290 - client assert failure during cleanup after abort recovery (2701)
3291 - leak mdc device after failed mount (2712)
3292 - ptlrpc_check_set allows timedout requests to complete (2714)
3293 - wait for inflight reqs when ptlrpcd finishes (2710)
3294 - make sure unregistered services are removed from the srv_list
3295 - reset bulk XID's when resending them (caught by 1138 test)
3296 - unregister_bulk after timeout
3297 - fix lconf error (2694)
3298 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3299 - readahead locks pages, leaves pending causing memory pressure (2673)
3300 - increase OST request buffers to 4096 on large machines (2729)
3301 - fix up permission of existing directories in simple_mkdir (2661)
3302 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3303 - don't assign transno to errored transactions (2742)
3304 - don't delete objects on OST if given a bogus objid from MDS (2751)
3305 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3306 - if rq_no_resend, then timeout request after recovery (2432)
3307 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3308 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3309 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3310 - MDS recovery shouldn't delete objects that it has given out (2730)
3311 - if enqueue arrives after completion, don't clobber LVB (2819)
3312 - don't unlock pages twice when trigger_group_io returns error (2814)
3313 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3314 - don't write pages to disk if there was an error (1450)
3315 - don't ping imports that have recovery disabled (2676)
3316 - take buffered bytes into account when balancing socknal conn (2817)
3317 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3318 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3319 - remove little-endian swabbing of llog records (1987)
3320 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3321 - retry reposting request buffers if they fail (1191)
3322 - grow extent at grant time to avoid granting a revoked lock (2809)
3323 - lock revoke doesn't evict page if covered by a second lock (2765)
3324 - disable VM readahead to avoid reading outside lock extents (2805)
3326 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3327 - updated kernel patches for hp-2.4.20 kernel (2681)
3329 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3332 - fix truncated write corruption (2366)
3333 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3335 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3336 - ensure len(uuid) < 37 in lmc (1171)
3337 - fix ia64 OOPS in llog_test (2255)
3338 - zero end of page at obdfilter for partial page writes (2648)
3339 - don't leave stale dentries around after renames (bug 2428)
3340 - fix timeouts when evicting a client with a single lock held (2642)
3341 - set deadline for the initial HELLO message to drain (2634)
3342 - print out dotted-quad IP addresses in the socknal (2302)
3344 - additional debugging for MDS client eviction problem (2443)
3345 - fix mkfsoptions support for osts (2603, 2604)
3347 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3350 - add series for the vanilla 2.6.0 kernel
3351 - add series for the vanilla 2.4.24 kernel
3352 - add series for a cray x86/64 UL kernel drop
3353 - fix xattr patches for the vanilla 2.4.19 series
3355 - generate true UUIDs in lmc (1171)
3356 - have portals stack dumping break in UML (2466)
3357 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3358 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3359 - wake callback waiting threads on client eviction (2460)
3360 - Add --ptldebug and --subsystem to lmc (1719)
3361 - update assertion to allow safe interrupt allocation
3362 - set rq_no_resend for cancel requests (2432)
3363 - recalculate ptlrpcd timeout after resend (2494)
3364 - call vfs_rmdir when removing pending directories (2368)
3365 - fix renaming a file to itself (2429)
3366 - lmc creates a default one-stripe lov (2454)
3367 - expand procfs space to handle large clusters (2326)
3368 - increase UML stack to avoid overflow
3369 - update lconf's list of debug and subsystem masks
3370 - fix lfs find --obd (2510)
3371 - /proc tunable for disabling filter read caching (2591)
3372 - stop rpm packages from altering slapd.conf (2301)
3373 - disable nagle in the socknal under 0conf (2578)
3374 - choose mds inode size based on stripe count (2572)
3375 - fix kernel-source rpm problems (2516)
3377 - add --disable-doc to avoid pdf generation (2421)
3378 - update documentation, tests, type-os, comments
3379 - avoid format warnings on ia64
3380 - remove the TOE NAL
3381 - tiny code cleanups by removing unused fields
3383 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3386 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3387 - remove the most blatant lies from BUILDING file (2371)
3388 - change default debug level to reasonable production setting
3389 - reduce client side cache size to reduce cache flush time
3390 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3391 - make TCP zerocopy and pinger support enabled by default (2476)
3392 - sync writes completed after process exits caused crashes (2319)
3393 - maintain correct mount count on the MDS (2356)
3394 - backout 1557, because 2316 wasn't really fixed
3395 - better file I/O statistics gathering in /proc
3396 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3397 - another kernel patch to fix zero-copy TCP function export
3398 - don't take duplicate lock when processing re-sent getattr (2420)
3399 - lctl uses obd_self_export instead of creating new conn (2353)
3400 - MDS/OST recovery case which requires object creation asserted (2425)
3401 - move lfs from /usr/sbin to /usr/bin in packages
3402 - fix race between mds_client_add and mds_client_free (2417)
3403 - use kmalloc instead of slabs in portals (2430)
3404 - don't create duplicate records when a failover MDS is present (2442)
3405 - remove unnecessary mount age check (2332)
3406 - don't remove directory inodes from locks prematurely (2451)
3407 - don't break if MDS service name is the same as hostname (2103)
3408 - fix races in client write RPC generation when cache full (2482)
3410 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3413 - remove now-unused request->rq_obd (278)
3414 - if an allocation fails, print out how much memory we've used (1933)
3415 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3416 - add the "configurable stack size" patch to most series files (1256)
3417 - ability to write large log records, for 100+ OST configs (2306)
3418 - fix NULL deref when filter_prep fails (2314)
3419 - fix operator precedence error in filter_sync
3420 - dynamic allocation of socknal TX descriptors (2315)
3421 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3422 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3423 - ensure that CWARN messages in Portals always get to the syslog
3424 - __init/__exit are not for prototype decls (ldlm_init/exit)
3425 - x86-64 compile warning fixes
3426 - fix gateway LMC keyword conflict (2318)
3427 - fix MDS lock inversions in getattr/reint paths (1844)
3428 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3429 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3430 - DLM race condition prevented some lock evictions (2328)
3431 - ENOMEM detection and retry on socknal sends (2230)
3432 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3433 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3434 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3435 - better error messages when a client is rejected during recovery (1505)
3436 - avoid cancelling locks which were never granted, after failure (2330)
3437 - fix i_sem/journal inversion in mds_client_add (2333)
3438 - fix truncate/getattr lock cycle deadlock (2334)
3439 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3440 - fix two rare exit paths which could leak an l_lock() ref (2321)
3441 - fencepost error in MDS/OST orphan recovery (2226)
3442 - make log record alignment 8 bytes (1988)
3443 - lstripe now fails when requested offset > ost_count (2237)
3444 - ensure that all kernel series have a complete list.h (1607)
3445 - fix crashes in special-file operations (2316)
3446 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3448 - allow configurable automake binary, for testing new versions
3449 - small update to the lfs documentation
3451 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3453 * fix negative export reference count in fsfilt_sync (2312)
3455 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3456 * release candidate 0.9.1
3458 - orphans are moved into the PENDING directory for possible recovery
3459 - replayed opens now open by fid for orphan/rename safety (1042)
3460 - last close of an orphan inode generates a transno (683)
3461 - chdir() and mount() now pin the directory entry (1020)
3462 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3463 - discard very old requests without processing them (1502)
3464 - remove some common, well-understood CERRORs (1505)
3465 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3466 - clear "grant" flags in OST replies until OST grant code lands (1644)
3467 - fix read performance by not clobbering i_blksize on client (1598)
3468 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3469 - make LRU size a /proc tunable, clears locks when reduced (707)
3470 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3471 - server locks take references on exports now (1558)
3472 - build fixes for 2.4.20-rh trees (1663)
3473 - return an error from lov_create if all OSCs are inactive (1751)
3474 - fix import levels when a reconnect happens without a timeout (1597)
3475 - exit early from mds_open if we get a lookup error (1749)
3476 - partial page read at EOF wouldn't wait for disk before sending (1642)
3477 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3478 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3479 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3480 - write/truncate lock inversion (1639)
3481 - Don't auto-load obdclass, portals modules during cleanup (1495)
3482 - fix timestamps from jumping to "now" (1763)
3483 - extra journal assertions (1648)
3484 - add an extra multiunlink test (1771)
3485 - fix read_record/write_record API (1776)
3486 - fix leak of offset_extent, possible incorrect i_size later (1772)
3487 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3488 - Debugging for the kqswnal_get_idle_tx problems (1820)
3489 - Allow recovery to be attempted multiple times (1536)
3490 - Write out MDS last_rcvd file after it is first created (1600)
3491 - Fix tx_descriptor leak in failed transmit situations (1827)
3492 - ext3 journaling fixes for assertion failure after IO error (1871)
3493 - class_export_put() on freed export after completion AST error (1896)
3494 - Fix revalidate looping in VFS (1322)
3495 - Don't access a freed export during MDS_REINT timeout (1521)
3496 - Add open-unlink recovery support on the MDS (1673,1764)
3497 - Return an error if no MDS data was read from last_rcvd (1946)
3498 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3499 - Don't setup a disk device that doesn't match exported UUID (317)
3500 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3501 - avoid committing NULL handle in force close
3502 - local.sh is now a one-stripe LOV configuration
3503 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3504 - fix ext3 htree duplicate directory entry corruption (1516)
3505 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3506 - update to new LOV EA format (2097)
3507 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3509 2003-06-15 Phil Schwan <phil@clusterfs.com>
3512 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3513 - per-import/export recovery handling (958, 931, 959)
3514 - multiple last-rcvd slots, for serving multiple FSes (949)
3515 - connections are again shared between multiple imp/exports (963, 964)
3516 - "umount -f" would hang if any requests needed to be sent (393, 978)
3517 - avoid pinning large req buffer by copying for queued messages (989)
3518 - add "uuid" to "lctl device" command to help upcalls (991)
3519 - "open" RPCs with transnos would confuse recovery counters (1037)
3520 - do proper endian conversion of all wire messages (288, 340, 891)
3521 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3522 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3523 - fix ptlrpc_connection leak in target_handle_connect (1174)
3524 - fix import refcounting bug in OST and MDS cleanup (1134)
3525 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3526 - fix very unlikely obd_types race condition (501)
3527 - remove osc_open hack for echo_client (1187)
3528 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3529 - a failure in read_inode2 leads to deadlock (1139)
3530 - cancel ack-locks as soon as transaction is committed (1072)
3531 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3532 - make sure to commitrw after any preprw to avoid deadlock (1162)
3533 - failing to execute a file in a lustre FS would lock inode (1203)
3534 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3535 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3536 - obd_uuid2tgt would walk off the end of the list (1255)
3537 - on IA64 the osc would give portals incorrect bulk size (1258)
3538 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3539 - fix lock inversion caused by new llite matching code (1282)
3540 - limit the number of dirty pages on a client to 10MB (1286)
3541 - timed out locks were not being corrected cancelled (1289)
3542 - fix O_DIRECT above 4GB on IA-32 (1292)
3543 * major user-visible changes
3544 - fail out/fail over policy now controlled by the upcall (993)
3546 - add OBD_PING to check server availability and failure (954)
3547 - lustre messages are now sent in sending host order (288, 340, 891)
3548 - add eadatalen to MDS getattr reply (340)
3549 - OST read replies may contain second buffer, with per-page status (593)
3551 2003-03-11 Phil Schwan <phil@clusterfs.com>
3554 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3555 - failed open()s could cause deadlock; fixed (867, 869)
3556 - stop cancelling OST locks when files are closed (481)
3557 - overlapping XID spaces caused network corruption (851, 853)
3558 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3559 - setattr_raw added, to do single-RPC, server-side setattrs
3560 - lmc/lconf syntax change for OST UUIDs
3561 - fix crashy race condition between ptlrpc_free_req and osc_close
3562 - don't use request in mdc_enqueue if we hit a timeout (889)
3563 - don't set the inode i_size for regular files from the MDS (896)
3564 - handle out of order completion AST (842)
3565 - don't LBUG if a lock request times out after receiving AST (913)
3566 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3567 - if a bad lock AST arrives, send an error instead of dropping entirely
3568 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3569 - fix leak in bulk IO when only partially completed (899, 900, 926)
3570 - fix O_DIRECT for ia64 (55)
3571 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3572 - C-z after timeout could hang a process forever; fixed (977)
3574 - client-side I/O cache (678, 924, 929, 941, 970)
3576 - READPAGE and SETATTRs which don't take server-side locks get
3579 2003-02-11 Phil Schwan <phil@clusterfs.com>
3582 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3583 - Fix fsfilt_extN_readpage() to read a full page of directory
3584 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3585 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3586 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3587 - Most intent operations were removed, in favour of a new RPC mode
3588 that does a single RPC to the server and bypasses most of the VFS
3589 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3590 - Aggressively cancel local locks on DLM servers
3591 - mds_reint_unlink sends EA to the client if it's the last nlink.
3592 client uses that EA to unlink OST objects.
3593 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3594 - recursive symlinks were fixed (439)
3595 - fixed NULL deref in DEBUG_REQ
3596 - filter_update_lastobjid no longer calls sync, which annoyed extN
3597 - fixed multi-client small-writes to a single file problem (445)
3598 - fixed mtime updates during file writes (607)
3599 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3600 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3601 - fixed Linux OST size reporting problem (444, 656)
3602 - OST now updates object mtime with writes or setattr (607, 619)
3603 - client verifies file size before zeroing page past EOF (445)
3604 - OST now writes last allocated objid to disk with allocation (108)
3605 - LOV on echo now works (409)
3607 - mds_reint_unlink sends a new buffer, with the EA included. this
3608 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3609 set if a regular file was being unlinked, and it was the last link
3610 - use PtlGet from the target for bulk writes (315)
3611 - OST now updates object mtime with writes or setattr (607, 619)
3612 - LDLM now has a grant-time callback to revalidate locked items, if
3614 - Many MDS operations were reorganized to combat race conditions
3616 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3617 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3619 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3622 - Fully reactivate OST imports after reconnection (512, others)
3623 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3624 - More graceful error handling for truncating on dead OST (515)
3625 - Don't error out unless we're actually accessing dead stripes (474)
3626 - Fix garbage sizes when stripes are missing (410)
3627 - LRU counters were broken, causing constant lock purge (433, 432)
3628 - garbage on read from stripes with failed OSTs (441)
3629 - mark OSCs as active before reconnecting during recovery (438)
3630 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3631 - lfind did not preserve OST order in output (443)
3632 - symlinks cause hung clients, incorrect data (439)
3633 - stop dereferencing request after dropping refcount (457)
3634 - don't LASSERT(spin_is_locked) on non-SMP (455)
3635 - fixes for many rename() bugs
3636 - fstat didn't correctly synchronize attributes (399)
3637 - server must handle lock cancellation during blocking AST prep (487)
3638 - bulk descriptors were free()d too soon (511)
3639 - fix paths in lconf, which would load incorrect modules (451, 507)
3640 - fix confusing lconf 'host not found' error message (386)
3641 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3642 - fix race condition in mdc_blocking_ast() for inode access (526)
3643 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3644 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3645 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3646 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3647 - fix lstripe in several ways (kernel side) (527)
3648 - fix request leak in ldlm_cli_enqueue (262)
3649 - incorrect OSC was marked inactive after OST failure
3650 - call mds_fs_cleanup before unmounting filesystem (524)
3651 - fix races between taking ns_lock and ldlm_lock_change_resource
3652 - fix races updating LOV export open file list
3653 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3654 - fix recovery NULL deref in ldlm_cli_cancel_unused
3655 - fix some DLM races by using new hash table for lock handles (419)
3656 - permit the client to specify desired inodes, at replay
3657 - duplicate requests when we queue them for replay reintegration
3658 - fix last_rcvd offset calculation
3659 - sync after each recovered transaction, so we always make progress
3660 - never, not always, ERESTART requests without transnos
3661 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3663 - skip replay if the MDS says that the client is already connected
3664 - don't check for a recovery-enabled export to match lctl's UUID
3665 - don't INC_USE_COUNT for phantom exports
3666 - don't crash when cleaning up phantom exports (567)
3667 - don't double-finish or set replay data for errored mdc_open requests
3668 - abort requests when they time out, so we don't get old replies
3669 - send/receive replies for AST messages again
3670 - if the client says that it doesn't have the lock, cancel it on the
3672 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3673 mark it as destroyed, it will all work out when decref is called
3674 - fix module use counts (22, 581)
3676 - ASTs now expect a reply (server cancels lock on error reply)
3678 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3681 - fix many simultaneous client startup (392)
3682 - fix dentry->d_it clobbering
3683 - credentials weren't being shipped for readdir/getattr operations
3684 - remove invalid assertions triggered during some concurrent MD
3686 - proper Lustre versions added (336, 389)
3687 - fix memory leak for create error case (398)
3688 - fix LOV locking bug that would get cli/srv out of sync
3689 - fix echo client over LOV (409)
3690 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3691 - fix double-O_EXCL intent crash (424)
3692 - avoid sending multiple lock CANCELs (352)
3694 - MDS can do multi-client recovery (modulo bugs in new code)
3696 - many updates, edits, cleanups
3698 2002-11-18 Phil Schwan <phil@clusterfs.com>
3701 - fix null d_it dereference (346)
3702 - fix full OST/dbench hang (333)
3703 - fix permission problem with file removal (286)
3704 - fix removal of OSCs from LOV when they fail
3705 - fix NULL deref during bulk timeout (214)
3706 - fix problems related to multiple filesystems on one MDS (241)
3707 - fixed serious subtle metadata locking bugs
3708 - free locks on clients when inodes are removed due to memory
3710 - fix inode pointer in lock data (285)
3711 - partial support for multiple MDS on a single host (241)
3712 - data locks weren't cancelled at clear_inode time (290, 311)
3713 - intent locks could lead to unbounded lock growth (205)
3714 - added a maximum lock count, an LRU list, and a flusher
3715 - fix multiple rename (365)
3716 - properly abstracted the echo client
3717 - OSC locked 1 byte too many; fixed
3718 - rewrote brw callback code:
3719 - fixed recovery bugs related to LOVs (306)
3720 - fixed too-many-pages-in-one-write crash (191)
3721 - fixed (again) crash in sync_io_timeout (214)
3722 - probably fixed callback-related race (385)
3724 - Add capability to MDS protocol
3725 - LDLM cancellations and callbacks on different portals
3727 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3730 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3731 - defer open object destruction to close time (601981, 138)
3732 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3733 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3734 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3735 - don't call obd_setattr in ll_file_release for destroyed objects
3736 * protocol change to lustre_msg: move |version| and add |flags|
3737 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3738 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3739 * added replay of create, unlink, link and rename operations during
3740 MDS failover; recovery should be much more robust now
3741 * remove failed OSCs from LOVs (only lov_create uses this so far)
3742 * the lustre-HOWTO was brought (more) up to date (582544)
3744 2002-10-23 Phil Schwan <phil@clusterfs.com>
3747 - in-use dentries weren't being reused properly (617851)
3748 - prevent multiple LDLM setup (599178)
3749 - fix LOV size calculations for truncate (617853)
3750 - fix client handling of MDS intent errors (POSIX)
3751 - fix permission bug in lovstripe.c test (624321)
3752 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3753 - truncate past end of file could corrupt data
3754 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3755 - a race in recovery could return ETIMEDOUT to apps (623947)
3756 - building outside the source directory was fixed
3757 * the lustre-HOWTO was brought (more) up to date (582544)
3758 * major progress was made on recovery functionality
3760 2002-10-10 Phil Schwan <phil@clusterfs.com>
3763 - recovery deadlock fix
3764 - rm -rf causes LBUG fix (617817)
3765 - file open by multiple tasks fix (618962)
3766 - directory permissions bugs (602707 and 620007)
3767 - journal_stop fixed with locking (611313)
3768 - O_APPEND failures resolved (618273, perhaps 614459)
3769 - lconf PATH fix (619770)
3770 - IA64 build fix (621450)
3771 - RPC buffer sizes scale with amount of memory
3773 2002-10-01 Phil Schwan <phil@clusterfs.com>
3776 - locks would be cancelled without throwing away data pages,
3777 resulting in inconsistent data (605627)
3778 - inode attributes were not always being refreshed (605627, 612449)
3779 - lconf now continues to cleanup after lctl reports an error
3780 - MDS now enforces user permissions (602707)
3781 - lprocfs cleanup fixed, but not yet enabled (614157)
3782 - fixed infinite server hang, should a client not respond to an AST
3783 - avoid going into recovery if user calls readlink() with a buffer
3784 that's too small (613941)
3785 - AST RPCs no longer require replies (614867) -- this may be changed
3786 - don't crash server if client sends an IOV that's too big (611336)
3787 - fixed lock conversion deadlock (611892)
3788 - fixed the following of symlinks (614622)
3789 * recovery: the server can remove locks from a client that dies, other
3790 clients can make progress
3791 * more extN patch fixes
3792 * compile-time configurable ptlrpc buffer allocations
3794 - collaborative read cache document
3795 - Lustre Lite Performance CDR document-in-progress
3797 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3800 - fix typo in patch-2.4.18
3802 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3805 - clear ptlrpc request each time in handle_incoming_request()
3806 - unlink of files now destroys the object on the OST
3808 2002-09-19 Peter Braam <braam@clusterfs.com>
3810 * add hard link support
3811 * change obdfile creation method
3812 * kernel patch changed
3814 2002-09-19 Peter Braam <braam@clusterfs.com>
3817 - stack overflow bug in extN fixed
3819 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3821 * documentation updates
3822 - add man pages for config tools
3823 - update tests/README to describe testing with new config tools
3824 - finish metadata API descriptions
3825 * bug fixes and cleanups
3826 - statfs workaround for 16TB limit
3827 - LOV stripe allocation improved, can stripe on subset of OSTs
3828 - LOV file size/IO offset was wrong for files > 4GB in size
3829 - object EA data was being dropped, caused files to be unreadable
3830 - memory overflow with non-LOV OST caused memory corruption
3831 - fixed regression tests to work with new config tools, obdfilter
3832 - fixed bug when directory size became larger than 1 block
3833 - fixed bug (for single client case) when PWD was deleted
3834 - invalidate local directory pages when doing intent-based ops
3835 - avoid LDLM oops when lock callback contained bad data
3837 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3839 * documentation updates
3840 * bug fixes and cleanups
3841 - configuration tools
3844 - 64-bit compile warnings
3845 - 64-bit internal statfs data
3847 * test_brw on persistent OST devices
3849 * lprocfs (disabled)
3851 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3853 * documentation updates
3854 * bug fixes and cleanups
3855 * configuration tools
3857 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3860 * many small fixes to 0_5_4
3861 * io/network handling
3862 * thinkos in MDS operations
3864 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3867 * crucial basic fixes to 0.5.3
3868 * IOR, Iozone work over Elan
3871 2002-08-07 Phil Schwan <phil@clusterfs.com>
3872 * version 0_5_3, our first alpha
3873 * we use the new Portals iovs
3874 * documentation updates
3875 * bug fixes and cleanups
3876 * small changes in the DLM wire protocol
3878 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3879 * version 0_5_1 with some initial stability,
3880 * locking on MD and file I/O.
3881 * documentation updates
3882 * several bug fixes since 0.5.0
3883 * small changes in wire protocol
3885 2002-07-18 Phil Schwan <phil@clusterfs.com>
3887 * delivered as Lustre Light Alpha
3888 * fixed a crash after handling invalid MDS requests
3889 * fixed directory pages for architectures with non-4k pages sizes
3891 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3892 * release version v0_4_4
3893 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3895 * Fixes a number of interruption problems with OST operations.
3896 * Update documentation for portals header changes
3897 * Move all wire protocol structs/defines to lustre_idl.h
3898 * Fixes symlink length bug.
3899 * Add tcpdump to repository.
3901 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3902 * release version v0_4_3
3903 * Fixes statfs for inodes on extN.
3904 * Fixes bug in runtests which would delete /etc/hosts.
3905 * Use 64-bit object IDs wherever possible (not into VFS though)
3906 Remove ost_get_info, which is unused by lustre, and out of date.
3908 2002-07-03 Peter Braam <braam@clusterfs.com>
3909 * release version v0_4_2 Fixes a lookup error (type not passed)
3910 * move forward to head of Portals
3911 * move forward to latest Lustre kernel
3913 2002-06-25 Peter Braam <braam@clusterfs.com>
3914 * release version v0_4_1. Hopefully stable on single node use.