1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 2.6.16.46-0.14 (SLES 10),
7 2.6.18-8.1.8.EL5 (RHEL 5).
8 2.6.18.8 vanilla (kernel.org)
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.9-42.0.10.EL (RHEL 4),
12 2.6.16 - 2.6.21 vanilla (kernel.org)
13 * Recommended e2fsprogs version: 1.40.2-cfs1
14 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
18 Description: don't use GFP_* in generic Lustre code.
19 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
23 Description: obdclass is limited by single OBD_ALLOC(idarray)
24 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
28 Description: warnings with build patchless client with vanila 2.6.19 and up
29 Details : change old ctl_table style and replace ctl_table/ctl_table_header
30 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
34 Frequency : Only for SLES
35 Description: sanity_quota fail test_1
36 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
37 file, which makes TSTID[2] inunique.
41 Frequency : Only for SLES9
42 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
43 Details : Update target/ChangeLog/which_patch .
47 Frequency : Only for test_10 in sanity-quota.sh
48 Description: a bug in quota test code
49 Details : checking a test flag in a wrong place
53 Frequency : Only for test_13 in sanity-quota.sh
54 Description: a bug in quota test script
55 Details : shouldn't check overfull allocated quota
59 Description: jbd statistics
60 Details : Port older jbd statistics patch for sles10
62 Severity : enhancement
64 Description: Add printing inode info into message about error in writepage.
69 Description: reply_lock_interpret crash due to race with it and lock cancel.
70 Details : Do not replay locks that are being cancelled. Do not reference
71 locks by their address during replay, just by their handle.
75 Frequency : When flocks are used.
76 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
77 Details : Flock locks might destroy just granted lock if it could be merged
78 with another existing flock, this is done in completion handler,
79 so teach ldlm_cli_enquque_fini that this is a valid case for
85 Description: Oops in read and write path when failing to allocate lock.
86 Details : Check if lock allocation failed and return error back.
90 Description: lstripe command fails for valid OST index
91 Details : The stripe offset is compared to lov->desc.ld_tgt_count
92 instead of lov->desc.ld_active_tgt_count.
96 Description: sanity.sh failed test 103
97 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
102 Description: df doesn't work properly if diskfs blocksize != 4K
103 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
107 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
108 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
112 Description: obd_health_check_timeout too short
113 Details : set obd_health_check_timeout as 1.5x of obd_timeout
117 Description: llapi_file_create() does not allow some changes
118 Details : add llapi_file_open() that allows specifying the mode and
119 open flags, and also returns an open file handle.
123 Description: merge and cleanup kernel patches.
124 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
128 Description: Add journal checksum support.(Kernel part)
129 Details : The journal checksum feature adds two new flags i.e
130 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
131 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
132 indicates that the commit block contains the checksum for
133 the blocks described by the descriptor blocks. Now commit
134 record can be sent to disk without waiting for descriptor
135 blocks to be written to disk. This behavior is controlled
136 using JBD2_FEATURE_ASYNC_COMMIT flag.
140 Description: OSS needs mutliple precreate threads
141 Details : Add ability to start more than one create thread per OSS.
145 Description: Sanity.sh test_65k failed: setstripe should have succeeded
146 Details : execute lfs setstripe on client
150 Description: mds_obd_create error creating tmp object
151 Details : When the user sets quota on root, llog will be affected and can't
152 create files and write files.
155 Frequency : Always on ia64 patchless client, and possibly others.
157 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
158 Details : This allows the patchless client to be loaded on architectures
163 Description: RedHat Update kernel for RHEL5
164 Details : Add the kernel config file for RHEL5.
168 Description: RedHat Update kernel for RHEL5
169 Details : Modify the kernel config file more closer RHEL5.
173 Description: Build failure against Centos5 (RHEL5)
174 Details : Define PAGE_SIZE when it isn't present.
178 Description: test_7 on sanity-quota.sh can't be used on separate machines
179 Details : add facet to handle it
183 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
184 Details : with NFS, the anon dentry's parent was set to itself in
185 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
186 corresponding dentry other than use rec->ur_name.
188 Severity : enhancement
190 Description: lfs setstripe enhancement
191 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
195 Description: testing performance impact of enabling checksumming
196 Details : enable checksum by default, allow --disable-checksum
197 configure option and "-o nochecksum" mount option
201 Description: lustre support for RHEL5
202 Details : Add support for RHEL5.
206 Description: Client eviction due to failover config
207 Details : after a connection loss, the lustre client should attempt to
208 reconnect to the last active server first before trying the
209 other potential connections.
211 --------------------------------------------------------------------------------
213 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
215 * Support for kernels:
216 2.6.5-7.283 (SLES 9),
217 2.6.9-55.EL (RHEL 4),
218 2.6.16.46-0.14 (SLES 10),
219 2.6.18.8 vanilla (kernel.org)
220 * Client support for unpatched kernels:
221 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
222 2.6.16 - 2.6.22 vanilla (kernel.org)
223 * Due to recently discovered recovery problems, we do not recommend
224 using patchless RHEL 4 clients with this or any earlier release.
225 * Recommended e2fsprogs version: 1.39.cfs8
226 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
227 * Starting with this release, the ldiskfs backing filesystem required
228 by Lustre is now in its own package, lustre-ldiskfs. This package
229 should be installed. It is versioned separately from Lustre and
230 may be released separately in future.
234 Description: block reactivating mgc import until all deactivates complete
235 Details : Fix race when failing back MDT/MGS to itself (testing)
237 Severity : enhancement
239 Description: add optional extra BUILD_VERSION info
240 Details : add a new environment variable (namely LUSTRE_VERS) which allows
241 to override the lustre version.
244 Frequency : 2.6.18 servers only
246 Description: ll_kern_mount() doesn't release the module reference
247 Details : The ldiskfs module reference count never drops down to 0
248 because ll_kern_mount() doesn't release the module reference.
253 Description: server LBUG when using old ost_num_threads parameter
254 Details : Accept the old ost_num_threads parameter but warn that it
255 is deprecated, and fix an off-by-one error that caused an LBUG.
260 Description: Transient SCSI error results in persistent IO issue
261 Details : iobuf->dr_error is not reinitialized to 0 between two
265 Frequency : sometimes when underlying device returns I/O errors
267 Description: OSTs not going read-only during write failures
268 Details : OSTs are not remounted read-only when the journal commit threads
269 get I/O errors because fsfilt_ext3 calls journal_start/stop()
270 instead of the ext3 wrappers.
274 Description: poor connect scaling with increasing client count
275 Details : Don't run filter_grant_sanity_check for more than 100 exports
276 to improve scaling for large numbers of clients.
279 Frequency : SLES10 only
281 Description: sanity-quota.sh quotacheck failed: rc = -22
282 Details : Quotas cannot be enabled on SLES10.
285 Frequency : liblustre clients only
287 Description: getdirentries does not give error when run on compute nodes
288 Details : getdirentries does not fail when the size specified as an argument
289 is too small to contain at least one entry
291 Severity : enhancement
293 Description: Add LNET router traceability for debug purposes
294 Details : If a checksum failure occurs with a router as part of the
295 IO path, the NID of the last router that forwarded the bulk data
296 is printed so it can be identified.
301 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
302 Details : Due to a race condition, liblustre clients were occasionally
305 Severity : enhancement
307 Description: lfs setstripe use optional parameters instead of postional
310 Severity : enhancement
312 Description: Nanosecond timestamp support for ldiskfs
313 Details : The on-disk ldiskfs filesystem has added support for nanosecond
314 resolution timestamps. There is not yet support for this at
315 the Lustre filesystem level.
318 Frequency : during server recovery
320 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
321 Details : request with rq_no_resend flag not awake l_wait_event if they get a
325 Frequency : nfs export on patchless client
327 Description: connectathon hang when test nfs export over patchless client
328 Details : Disconnected dentry cannot be found with lookup, so we do not need
329 to unhash it or make it invalid
332 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
333 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
334 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
339 Description: open req refcounting wrong on reconnect
340 Details : If reconnect happened between getting open reply from server and
341 call to mdc_set_replay_data in ll_file_open, we will schedule
342 replay for unreferenced request that we are about to free.
343 Subsequent close will crash in variety of ways.
344 Check that request is still eligible for replay in
345 mdc_set_replay_data().
350 Description: disable writes to filesystem when reading health_check file
351 Details : the default for reading the health_check proc file has changed
352 to NOT do a journal transaction and write to disk, because this
353 can cause reads of the /proc file to hang and block HA state
354 checking on a healthy but otherwise heavily loaded system. It
355 is possible to return to the previous behaviour during configure
356 with --enable-health-write.
358 Severity : enhancement
360 Description: 64-bit inode version
361 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
362 to the inode. This will be required for version-based recovery.
367 Description: MDS fails to start if a duplicate client export is detected
368 Details : in some rare cases it was possible for a client to connect to
369 an MDS multiple times. Upon recovery the MDS would detect this
370 and fail during startup. Handle this more gracefully.
372 Severity : enhancement
374 Description: Add -o localflock option to simulate old noflock
376 Details : This will achieve local-only flock/fcntl locks
382 Description: log_commit_thread vs filter_destroy race leads to crash
383 Details : Take import reference before releasing llog record semaphore
388 Description: Wrong request locking in request set processing
389 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
390 imp_delayed_list, in this place should be used imp_lock.
393 Frequency : when reconnection
395 Description: Grant Leak when osc reconnect to OST
396 Details : When osc reconnect ost, OST(filter) should check whether it
397 should grant more space to client by comparing fed_grant and
398 cl_avail_grant, and return the granted space to client instead
399 of "new granted" space, because client will call osc_init_grant
400 to update the client grant space info.
403 Frequency : when client reconnect to OST
405 Description: Grant Leak when osc do resend and replay bulk write
406 Details : When osc reconnect to OST, OST(filter)should clear grant info of
407 bulk write request, because the grant info while be sync between
408 OSC and OST when reconnect, and we should ignore the grant info
409 these of resend/replay write req.
414 Description: Grant space more than avaiable left space sometimes.
415 Details : When then OST is about to be full, if two bulk writing from
416 different clients came to OST. Accord the avaliable space of the
417 OST, the first req should be permitted, and the second one
418 should be denied by ENOSPC. But if the seconde arrived before
419 the first one is commited. The OST might wrongly permit second
420 writing, which will cause grant space > avaiable space.
423 Frequency : when client is evicted
425 Description: Grant might be wrongly erased when osc is evicted by OST
426 Details : when the import is evicted by server, it will fork another
427 thread ptlrpc_invalidate_import_thread to invalidate the
428 import, where the grant will be set to 0. While the original
429 thread will update the grant it got when connecting. So if
430 the former happened latter, the grant will be wrongly errased
431 because of this race.
436 Description: Checking Stale with correct fid
437 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
438 to check whether it is stale, because sometimes, we want the
439 enqueue happened anyway, and op_data.fid2 will not be initialized.
441 Severity : enhancement
443 Description: update patchless client
444 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
447 Frequency : only with 2.4 kernel
449 Description: random memory corruption
450 Details : size of struct ll_inode_info is to big for union inode.u and this
451 can be cause of random memory corruption.
456 Description: Memory leak in recovery
457 Details : Lov_mds_md was not free in an error handler in mds_create_object.
458 It should also check obd_fail before fsfilt_start, otherwise if
459 fsfilt_start return -EROFS,(failover mds during mds recovery).
460 then the req will return with repmsg->transno = 0 and rc = EROFS.
461 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
462 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
463 be freed no matter whether fsfilt_commit success or not.
466 Frequency : only with huge count clients
468 Description: Prevents from taking the superblock lock in llap_from_page for
470 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
471 for llap_from_page call in ll_removepage prevents from taking the
472 superblock lock for a soon died page.
477 Description: Not check open intent error before release open handle
478 Details : in some rare cases, the open intent error is not checked before
479 release open handle, which may cause
480 ASSERTION(open_req->rq_transno != 0), because it tries to release
481 the failed open handle.
486 Description: Set cat log bitmap only after create log success.
487 Details : in some rare cases, the cat log bitmap is set too early. and it
488 should be set only after create log success.
492 Description: Accessing a block bevice can re-enable I/O when Lustre is
493 tearing down a device.
494 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
499 Description: service threads may hog cpus when there are a lot of requests
501 Details : Insert cond_resched to give other threads a chance to use some of
507 Description: the cat log was not initialized in recovery
508 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
509 unlink log on mds will not be initialized until mds post
510 recovery. And also in mds post recovery, the unlink log will
511 initialization will be done asynchronausly, so there will be race
512 between add unlink log and unlink log initialization.
516 Description: brw_stats were being printed incorrectly
517 Details : brw_stats were being printed as log2 but all of them were not
518 recorded as log2. Also remove some code duplication arising from
519 filter_tally_{read,write}.
523 Frequency : rare, only in recovery.
524 Description: ASSERTION(req->rq_type != LI_POISON) failed
525 Details : imp_lock should be held while iterating over imp_sending_list for
526 prevent destroy request after get timeout in ptlrpc_queue_wait.
530 Description: replay-single.sh test 52 fails
531 Details : A lock's skiplist need to be cleanup when it being unlinked
532 from its resource list.
536 Description: Short directio read returns full requested size rather than
538 Details : Direct I/O operations should return actual amount of bytes
539 transferred rather than requested size.
541 Severity : enhancement
543 Description: metadata RPC reduction (e.g. for rm performance)
544 Details : decrease the amount of synchronous RPC between clients and servers
545 by canceling conflicing lock before the operation on the client side
546 and packing thier handles into the main operation RPC to server.
548 Severity : enhancement
550 Description: Async OSC create to avoid the blocking unnecessarily.
551 Details : If a OST has no remain object, system will block on the creating
552 when need to create a new object on this OST. Now, ways use
553 pre-created objects when available, instead of blocking on an
554 empty osc while others are not empty. If we must block, we block
555 for the shortest possible period of time.
557 --------------------------------------------------------------------------------
559 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
564 Frequency : on some architectures
566 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
567 Details : Uninitialized flags sometimes cause configuration commands to
571 Frequency : patchless clients only
573 Description: missing __iget() symbol export
574 Details : The __iget() symbol export is missing. To avoid the need for
575 this on patchless clients the deathrow inode reaper is turned
576 off, and we depend on the VM to clean up old inodes. This
577 dependency was during via the fix for bug 12181.
579 --------------------------------------------------------------------------------
581 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
583 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
584 INTEROPERATE with older versions automatically. In many cases a
585 special upgrade step is needed. Please read the
586 user documentation before upgrading any part of a live system.
587 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
588 version will not interoperate with 1.6 betas before beta5 (1.5.95).
589 * WARNING: Lustre configuration and startup changes are required with
590 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
596 Severity : enhancement
598 Description: MountConf
599 Details : Lustre configuration is now managed via mkfs and mount
600 commands instead of lmc and lconf. New obd types (MGS, MGC)
601 are added for dynamic configuration management. See
602 https://mail.clusterfs.com/wikis/lustre/MountConf for
605 Severity : enhancement
607 Description: dynamic OST addition
608 Details : OSTs can now be added to a live filesystem
610 Severity : enhancement
612 Description: startup order invariance
613 Details : MDTs and OSTs can be started in any order. Clients only
614 require the MDT to complete startup.
616 Severity : enhancement
618 Description: parallel, asynchronous orphan cleanup
619 Details : orphan cleanup is now performed in separate threads for each
620 OST, allowing parallel non-blocking operation.
622 Severity : enhancement
624 Description: optimized stripe assignment
625 Details : stripe assignments are now made based on ost space available,
626 ost previous usage, and OSS previous usage, in order to try
627 to optimize storage space and networking resources.
629 Severity : enhancement
631 Description: Permanently set tunables
632 Details : All writable /proc/fs/lustre tunables can now be permanently
633 set on a per-server basis, at mkfs time or on a live
636 Severity : enhancement
638 Description: Lustre message v2
639 Details : Add lustre message format v2.
641 Severity : enhancement
643 Description: client OST exclusion list
644 Details : Clients can be started with a list of OSTs that should be
645 declared "inactive" for known non-responsive OSTs.
649 Description: SPEC SFS validation failure on NFS v2 over lustre.
650 Details : Changes the blocksize for regular files to be 2x RPC size,
651 and not depend on stripe size.
653 Severity : enhancement
655 Description: Multiple MD RPCs in flight.
656 Details : Further unserialise some read-only MDS RPCs - learn about intents.
657 To avoid overly-overloading MDS, introduce a limit on number of
658 MDS RPCs in flight for a single client and add /proc controls
659 to adjust this limit.
661 Severity : enhancement
663 Description: client read/write statistics
664 Details : Add client read/write call usage stats for performance
665 analysis of user processes.
666 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
667 file access. extents_stats shows chunk size distribution.
668 extents_stats_per_process show chunk size distribution per
671 Severity : enhancement
673 Description: mds statistics
674 Details : Add detailed mds operations statistics in
675 /proc/fs/lustre/mds/*/stats.
679 Description: Failure of copying files with lustre special EAs.
680 Details : Client side always return success for setxattr call for lustre
681 special xattr (currently only "trusted.lov").
686 Description: Refcount LNET uuids
687 Details : The global LNET uuid list grew linearly with every startup;
688 refcount repeated list entries instead of always adding to
691 Severity : enhancement
693 Description: Dynamic service threads
694 Details : Within a small range, start extra service threads
695 automatically when the request queue builds up.
698 Frequency : mixed-endian client/server environments
700 Description: mixed-endian crashes
701 Details : The new msg_v2 system had some failures in mixed-endian
704 Severity : enhancement
706 Description: Easy OST removal
707 Details : OSTs can be permanently deactivated with e.g. 'lctl
708 conf_param lustre-OST0001.osc.active=0'
710 Severity : enhancement
712 Description: MGS proc entries
713 Details : Added basic proc entries for the MGS showing what filesystems
716 Severity : enhancement
718 Description: provide MGS failover
719 Details : Added config lock reacquisition after MGS server failover.
721 Severity : enhancement
723 Description: add Linux 2.4 support
724 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
728 Description: a large application tries to do I/O to the same resource and dies
730 Details : Check the req->rq_arrival time after the call to
731 ost_brw_lock_get(), but before we do anything about
732 processing it & sending the BULK transfer request. This
733 should help move old stale pending locks off the queue as
734 quickly as obd_timeout.
737 Frequency : when an incorrect nid is specified during startup
739 Description: ptlrpc connect to non-existant node causes kernel crash
740 Details : LNET can't be re-entered from an event callback, which
741 happened when we expire a message after the export has been
742 cleaned up. Instead, hand the zombie cleanup off to another
745 Severity : enhancement
747 Description: plain/inodebits lock performance improvement
748 Details : Grouping plain/inodebits in granted list by their request modes
749 and bits policy, thus improving the performance of search through
753 Frequency : only if OST filesystem is corrupted
755 Description: client incorrectly hits assertion in ptlrpc_replay_req()
756 Details : for a short time RPCs with bulk IO are in the replay list,
757 but replay of bulk IOs is unimplemented. If the OST filesystem
758 is corrupted due to disk cache incoherency and then replay is
759 started it is possible to trip an assertion. Avoid putting
760 committed RPCs into the replay list at all to avoid this issue.
763 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
765 Description: System hang on startup
766 Details : This bug allowed the liblustre (e.g. catamount) client to
767 return to the app before handling all startup RPCs. This
768 could leave the node unresponsive to lustre network traffic
769 and manifested as a server ptllnd timeout.
771 Severity : enhancement
773 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
774 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
775 boolean to control whether to print peer debug info when a
776 client's RPC times out.
779 Frequency : only for kernels with patches from Lustre below 1.4.3
781 Description: Remove old rdonly API
782 Details : Remove old rdonly API which unused from at least lustre 1.4.3
785 Frequency : only for devices with external journals
787 Description: Set external device read-only also
788 Details : During a commanded failover stop, we set the disk device
789 read-only while the server shuts down. We now also set any
790 external journal device read-only at the same time.
793 Frequency : when upgrading from 1.4 while trying to change parameters
795 Description: The wrong (new) MDC name was used when setting parameters for
796 upgraded MDT's. Also allows changing of OSC (and MDC)
797 parameters if --writeconf is specified at tunefs upgrade time.
800 Frequency : when setting specific ost indicies
802 Description: QOS code breaks on skipped indicies
803 Details : Add checks for missing OST indicies in the QOS code, so OSTs
804 created with --index need not be sequential.
806 Severity : enhancement
808 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
809 Details : The uninit_groups feature works in conjunction with the kernel
810 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
811 up the pass1 processing of e2fsck. This is a read-only feature
812 in ldiskfs2 only, so older kernels and current ldiskfs cannot
813 mount filesystems that have had this feature enabled.
815 Severity : enhancement
817 Description: Improve multi-block allocation algorithm to avoid fragmentation
818 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
819 allocation locality and avoid filesystem fragmentation.
821 ------------------------------------------------------------------------------
823 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
825 * Support for kernels:
826 2.4.21-47.0.1.EL (RHEL 3)
828 2.6.9-42.0.10.EL (RHEL 4)
829 2.6.12.6 vanilla (kernel.org)
830 2.6.16.27-0.9 (SLES 10)
831 * Recommended e2fsprogs version: 1.39.cfs5
833 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
837 Frequency : occasional, depends on client load and configuration
838 Bugzilla : 12181, 12203
839 Description: data loss for recently-modified files
841 Details : In some cases it is possible that recently written or created
842 files may not be written to disk in a timely manner (this should
843 normally be within 30s unless client IO load is very high).
844 The problem appears as zero-length files or files that are a
845 multiple of 1MB in size after a client crash or client eviction
846 that are missing data at the end of the file.
848 This problem is more likely to be hit on clients where files are
849 repeatedly created and unlinked in the same directory, clients
850 have a large amount of RAM, have many CPUs, the filesystem has
851 many OSTs, the clients are rebooted frequently, and/or the files
852 are not accessed by other nodes after being written.
854 The presence of the problem can be detected by looking at
855 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
856 smaller than the second (nr_unused) then dirty files will not
857 be flushed automatically to disk. "sync; sleep 10" should be
858 run several times on the node before unmounting it to update
859 Lustre (this is also safe to run on nodes without this problem).
861 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
862 that can cause this same problem, so customers using that kernel
863 also need to update the kernel in addition to Lustre. In order
864 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
866 It is normal that files written just before a client crash (less
867 than 30s) may not yet have been flushed to disk, even for local
871 Frequency : frequent on thin XT3 nodes
873 Description: UUID collision on thin XT3 Linux nodes
874 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
875 randomly, since we relied on an insufficiently-seeded PRNG.
880 Description: OSS hangs after "All ost request buffers busy"
881 Details : A deadlock between quota and journal operations caused OSS
882 hangs after printing "All ost request buffers busy."
885 Frequency : always on liblustre builds
887 Description: Cleanup compiler warnings on liblustre
890 Frequency : always on liblustre builds on XT3
892 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
896 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
897 > 32000 subdirectories
898 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
899 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
900 subdirectory count crosses 32000. This will aid e2fsck to
901 correctly handle more than 32000 subdirectories.
904 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
906 Description: System hang on startup
907 Details : This bug allowed the liblustre (e.g. catamount) client to
908 return to the app before handling all startup RPCs. This
909 could leave the node unresponsive to lustre network traffic
910 and manifested as a server ptllnd timeout.
912 Severity : enhancement
914 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
915 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
916 boolean to control whether to print peer debug info when a
917 client's RPC times out.
922 Description: make O_SYNC working on 2.6 kernels
923 Details : 2.6 kernels use different method for mark pages for write,
924 so need add a code to lustre for O_SYNC work.
929 Description: Failure to close file and release space on NFS
930 Details : Put inode details into lock acquired in ll_intent_file_open.
931 Use mdc_intent_lock in ll_intent_open to properly
932 detect all kind of errors unhandled by mdc_enqueue.
937 Description: proc file read during shutdown sometimes raced obd removal,
939 Details : Add lock to prevent obd access after proc file removal.
942 Frequency : Only for files larger than 4GB on 32-bit clients.
944 Description: improperly doing page alignment of locks
945 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
946 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
951 Description: RPCs being resent when they shouldn't be
952 Details : Some RPCs that should not be resent are being resent. This
953 can cause inconsistencies in the RPC state machine. Do not
954 resend such requests.
957 Frequency : rare, only with NFS export
959 Description: Crash on NFS re-export node
960 Details : under very unusual load conditions an assertion is hit in
961 ll_intent_file_open()
964 Frequency : only if OST filesystem is corrupted
966 Description: client incorrectly hits assertion in ptlrpc_replay_req()
967 Details : for a short time RPCs with bulk IO are in the replay list,
968 but replay of bulk IOs is unimplemented. If the OST filesystem
969 is corrupted due to disk cache incoherency and then replay is
970 started it is possible to trip an assertion. Avoid putting
971 committed RPCs into the replay list at all to avoid this issue.
976 Description: large O_DIRECT requests fail under memory pressure/fragmentation
977 Details : Large single O_DIRECT read and write calls can fail to allocate
978 a sufficiently large buffer to process the request. In case of
979 allocation failure the allocation is retried with a smaller
980 buffer and broken into smaller requests.
982 Severity : enhancement
984 Description: Add -o localflock option to simulate old noflock behaviour.
985 Details : This will achieve local-only flock/fcntl locks coherentness.
990 Description: versioning check is incomplete
991 Details : Checking the version difference of client vs. server, report
992 error if the gap is too big.
997 Description: add support PG_writeback bit
998 Details : add support for PG_writeback bit for Lustre, for more carefull
999 work with page cache in 2.6 kernel. This also fix some deadlocks
1000 and remove hack for work O_SYNC with 2.6 kernel.
1002 Severity : enhancement
1004 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1005 Details : The uninit_groups feature works in conjunction with the kernel
1006 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1007 up the pass1 processing of e2fsck. This is a read-only feature
1008 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1009 mount filesystems that have had this feature enabled.
1011 Severity : enhancement
1013 Description: Improve multi-block allocation algorithm to avoid fragmentation
1014 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1015 allocation locality and avoid filesystem fragmentation.
1017 ------------------------------------------------------------------------------
1019 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1021 * Support for kernels:
1022 2.6.9-42.0.3.EL (RHEL 4)
1023 2.6.5-7.276 (SLES 9)
1024 2.4.21-47.0.1.EL (RHEL 3)
1025 2.6.12.6 vanilla (kernel.org)
1026 2.6.16.21-0.8 (SLES10)
1027 * Recommended e2fsprogs version: 1.39.cfs2-0
1029 * The backwards-compatible /proc/sys/portals symlink has been removed
1030 in this release. Before upgrading, please ensure that you change
1031 any configuration scripts or /etc/sysctl.conf files that access
1032 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1033 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1034 in advance of the upgrade on any system running Lustre 1.4.6 or
1035 newer, since /proc/sys/lnet was added in that version.
1036 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1040 Frequency : only when quota is used
1042 Description: avoid scanning export list for quota master
1043 Details : Change the algorithms to avoid scanning export list in order
1044 to improve the efficiency.
1047 Frequency : MDS failover only, very rarely
1049 Description: "went back in time" messages on mds failover
1050 Details : The greatest transno may be lost when the current operation
1051 finishes with an error (transno==0) and the client's last_rcvd
1052 record is over-written. Save the greatest transno in the
1053 mds_last_transno for this case.
1056 Frequency : always for specific kernels and striping counts
1058 Description: client may get "Matching packet too big" without ACL support
1059 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1060 when trying to access files in certain configurations. The
1061 clients should in fact be denied when mounting because they do
1062 not understand ACLs.
1065 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1067 Description: many clients connecting with IO in progress causes connect timeouts
1068 Details : Avoid synchronous journal commits to avoid delays caused by many
1069 clients connecting/disconnecting when bulk IO is in progress.
1070 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1071 OST_IO_PORTAL to avoid delays behind potentially many pending
1075 Frequency : occasionally with multiple writers to a single file
1077 Description: shared writes to file may result in wrong size reported by stat()
1078 Details : Allow growing of kms when extent lock is cancelled
1081 Frequency : always with random mmap IO to multi-striped file
1083 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1084 Details : Only if the hole is at the end of OST object so that kms is too
1085 small. Fix is to increase kms accordingly in ll_nopage.
1088 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1090 Description: writes to a missing object would leak memory on the OST
1091 Details : If there is an inconsistency between the MDS and OST filesystems,
1092 such that the MDS references an object that doesn't exist, writes
1093 to that object will leak memory due to incorrect cleanup in the
1094 error handling path, eventually running out of memory on the OST.
1099 Description: Creating too long symlink causes lustre errors
1100 Details : Check symlink and name lengths before sending requests to MDS.
1103 Frequency : only if flock is enabled (not on by default)
1105 Description: posix locks not released on fd closure on 2.6.9+
1106 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1107 kernels, this caused such locks not to be released on fd close and
1108 then assertions on fs unmount about still used locks.
1111 Frequency : MDS failover only, very rarely
1113 Description: clients may get ASSERTION(granted_lock != NULL)
1114 Details : When request was taking a long time, and a client was resending
1115 a getattr by name lock request. The were multiple lock requests
1116 with the same client lock handle and
1117 mds_getattr_name->fixup_handle_for_resent_request found one of the
1118 lock handles but later failed with ASSERTION(granted_lock != NULL).
1123 Description: handle->h_buffer_credits > 0, assertion failure
1124 Details : h_buffer_credits is zero after truncate, causing assertion
1125 failure. This patch extends the transaction or creates a new
1129 Frequency : NFS re-export or patchless client
1130 Bugzilla : 11179, 10796
1131 Description: Crash on NFS re-export node (__d_move)
1132 Details : We do not want to hash the dentry if we don't have a lock.
1133 But if this dentry is later used in d_move, we'd hit uninitialised
1134 list head d_hash, so we just do this to init d_hash field but
1135 leave dentry unhashed.
1138 Frequency : NFS re-export or patchless client
1140 Description: NFS exports has problem with symbolic link
1141 Details : lustre client didn't properly install dentry when re-exported
1142 to NFS or running patchless client.
1145 Frequency : NFS re-export or patchless client
1147 Description: Various nfs/patchless fixes.
1148 Details : fixes reuse disconected alias for lookup process - this fixes
1149 warning "find_exported_dentry: npd != pd",
1150 fix permission error with open files at nfs.
1151 fix apply umask when do revalidate.
1154 Frequency : occasional
1156 Description: Crash on NFS re-export node
1157 Details : calling clear_page() on the wrong pointer triggered oops in
1158 generic_mapping_read().
1161 Frequency : rarely, using O_DIRECT IO
1163 Description: unaligned directio crashes client with LASSERT
1164 Details : check for unaligned buffers before trying any requests.
1167 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1169 Description: stale data returned from RAID cache
1170 Details : If only a small amount of IO is done to the RAID device before
1171 reading it again it is possible to get stale data from the RAID
1172 cache instead of reading it from disk.
1175 Frequency : always for sles10 kernel
1177 Description: sles10 support
1178 Details : ll_follow_link: compile fixes and using of nd_set_link
1179 under newer kernels.
1182 Frequency : depends on arch, kernel and compiler version, always on sles10
1185 Description: recursive or deep enough symlinks cause stack overflow
1186 Details : getting rid of large stack-allocated variable in
1190 Frequency : depends on hardware
1192 Description: lustre write performance loss in the SLES10 kernel
1193 Details : the performance loss is caused by using of write barriers in the
1194 ext3 code. The SLES10 kernel turns barrier support on by
1195 default. The fix is to undo that change for ldiskfs.
1197 ------------------------------------------------------------------------------
1199 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1201 * Support for kernels:
1202 2.6.9-42.0.3EL (RHEL 4)
1203 2.6.5-7.276 (SLES 9)
1204 2.4.21-47.0.1.EL (RHEL 3)
1205 2.6.12.6 vanilla (kernel.org)
1209 Frequency : quota enabled and large files being deleted
1211 Description: releasing more than 4GB of quota at once hangs OST
1212 Details : If a user deletes more than 4GB of files on a single OST it
1213 will cause the OST to spin in an infinite loop. Release
1214 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1219 Description: statfs data retrieved from /proc may be stale or zero
1220 Details : When reading per-device statfs data from /proc, in the
1221 {kbytes,files}_{total,free,avail} files, it may appear
1222 as zero or be out of date.
1225 Frequency : systems with MD RAID1 external journal devices
1227 Description: lconf's call to blkid is confused by RAID1 journal devices
1228 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1229 of returning all block devices that match the journal UUID.
1232 Frequency : always, for aggregate stripe size over 4GB
1234 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1235 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1236 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1239 Frequency : always if "lfs find" used on a local file/directory
1241 Description: "lfs find" segfaults if used on a local file/directory
1242 Details : The case where a directory component was not specified wasn't
1243 handled correctly. Handle this properly.
1246 Frequency : always on ppc64
1248 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1249 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1250 perform endianness conversion.
1253 Frequency : rarely (truncate to non-zero file size after write under load)
1254 Bugzilla : 10730, 10687
1255 Description: Files padded with zeros to next 4K multiple
1256 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1257 it is possible that files that are truncated to a non-zero size
1258 immediately after being written are filled with zero bytes beyond
1259 the truncated size. No file data is lost.
1261 Severity : enhancement
1263 Description: Allow recovery/failover for liblustre clients.
1264 Details : liblustre clients were unaware of failover configurations until
1267 Severity : enhancement
1269 Description: user file locks should fail when not mounting with flock option
1270 Details : Set up an error-returning stub in ll_file_operations.lock field
1271 to prevent incorrect behaviour when client is mounted without
1272 flock option. Also, set up properly f_op->flock field for
1276 Frequency : always on ia64
1278 Description: "lfs df" loops on printing out MDS statfs information
1279 Details : The obd_ioctl_data was not initialized and in some systems
1280 this caused a failure during the ioctl that did not return
1281 an error. Initialize the struct and return an error on failure.
1284 Frequency : SLES 9 only
1286 Description: Error of copying files with lustre special EAs as root
1287 Details : Client side always return success for setxattr call for lustre
1288 special xattr (currently only "trusted.lov").
1291 Frequency : rarely on clusters with both ia64+i386 clients
1293 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1294 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1295 clients are doing concurrent writes to the same file it is
1296 possible that the ia64 clients may LASSERT because the OST
1297 extent locks are not PAGE_SIZE aligned. Ensure that grown
1298 locks are always aligned on the request boundary.
1301 Frequency : specific use, occasional
1303 Description: Overwriting in use executable truncates on-disk binary image
1304 Details : If one node attempts to overwrite an executable in use by
1305 another node, we now correctly return ETXTBSY instead of
1306 truncating the file.
1308 Severity : enhancement
1310 Description: Async OSC create to avoid the blocking unnecessarily.
1311 Details : If a OST has no remain object, system will block on the creating
1312 when need to create a new object on this OST. Now, ways use
1313 pre-created objects when available, instead of blocking on an
1314 empty osc while others are not empty. If we must block, we block
1315 for the shortest possible period of time.
1320 Description: chmod on Lustre root is propagated to other clients
1321 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1322 invalid by the follow_mount time.
1325 Frequency : liblustre clients only
1327 Description: Race in 'instant cancel' lock handling could lead to such locks
1328 never to be granted in case of SMP MDS
1329 Details : Do not destroy not yet granted but cbpending locks in
1333 Frequency : replay/resend of open
1335 Description: non null lock assetion failure in mds_intent_policy
1336 Details : Trying to replay/resend lockless open requests resulted in
1337 mds_open() returning 0 with no lock. Now it sets a flag if
1338 a lock is going to be returned.
1340 Severity : enhancement
1342 Description: Checksum enhancements
1343 Details : New checksum enhancements allow for resending RPCs that failed
1346 Severity : enhancement
1348 Description: Tunables on number of dirty pages in cacche
1349 Details : Allow to set limit on number of dirty pages cached.
1354 Description: client crash on unmount - lock still has references
1355 Details : In some error handling cases it was possible to leak a lock
1356 reference on a client while accessing a file. This was not
1357 harmful to the client during operation, but would cause the
1358 client to crash when the filesystem is unmounted.
1361 Frequency : specific case, rare
1363 Description: ETXTBSY on mds though file not in use by client
1364 Details : ETXTBSY is no longer incorrectly returned when attempting to
1365 chmod or chown a directory that the user previously tried to
1366 execute or a currently-executing binary.
1369 Frequency : extremely rare except on liblustre-based clients
1371 Description: Lustre space not freed when files are deleted
1372 Details : Clean up open-unlinked files after client eviction. Previously
1373 the unlink was skipped and the files remained as orphans.
1378 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1379 Details : In some cases it was possible to send a blocking callback to
1380 a client doing a glimpse, even though that client didn't get
1381 a lock granted. When the glimpse lock is cancelled on the OST
1382 the freed lock is left on the waiting list and corrupted the list.
1385 Frequency : all core dumps
1387 Description: Broke core dumps to lustre
1388 Details : Negative dentry may be unhashed if parent does not have UPDATE
1389 lock, but some callers, e.g. do_coredump, expect dentry to be
1390 hashed after successful create, hash it in ll_create_it.
1392 ------------------------------------------------------------------------------
1394 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1396 * Support for kernels:
1397 2.6.9-42.0.2.EL (RHEL 4)
1398 2.6.5-7.276 (SLES 9)
1399 2.4.21-40.EL (RHEL 3)
1400 2.6.12.6 vanilla (kernel.org)
1404 Frequency : always on RHEL 3
1406 Description: Number of open files grows over time
1407 Details : The number of open files grows over time, whether or not
1408 Lustre is started. This was due to a filp leak introduced
1409 by one of our kernel patches.
1411 ------------------------------------------------------------------------------
1413 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1415 * Support for kernels:
1416 2.6.9-42.EL (RHEL 4)
1417 2.6.5-7.276 (SLES 9)
1418 2.4.21-40.EL (RHEL 3)
1419 2.6.12.6 vanilla (kernel.org)
1424 Bugzilla : 5719, 9635, 9792, 9684
1425 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1426 Details : If a server is under heavy load and cannot reply to new
1427 connection requests before the client resends the (re)connect,
1428 the connection handling code can behave badly if two service
1429 threads are concurrently handing separate (re)connections from
1430 the same client. Add better locking to the connection handling
1431 code, and ensure that only a single connection will be processed
1432 for a given client UUID, even if the lock is dropped.
1434 Severity : enhancement
1436 Description: add TCP zero-copy support to kernel
1437 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1438 sends if the hardware supports scatter-gather and checksumming.
1439 This allows socklnd to do client-write and server-read more
1440 efficiently and reduce CPU utilization from skbuf copying.
1443 Frequency : only if NFS exporting from client
1445 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1446 Details : If setting attributes on a file created under NFS that had
1447 never been opened it would be possible to oops the client
1448 if the file had no objects.
1451 Frequency : always for liblustre
1453 Description: liblustre client does MDS+OSTs setattr RPC for each write
1454 Details : When doing a write from a liblustre client, the client
1455 incorrectly issued an RPC to the MDS and each OST the file was
1456 striped over in order to update the timestamps. When writing
1457 with small chunks and many clients this could overwhelm the MDS
1458 with RPCs. In all cases it would slow down the write because
1459 these RPCs are unnecessary.
1461 Severity : enhancement
1463 Description: allow number of MDS service threads to be changed at module load
1464 Details : It is now possible to change the number of MDS service threads
1465 running. Adding "options mds mds_num_threads={N}" to the MDS's
1466 /etc/modprobe.conf will set the number of threads for the next
1467 time Lustre is restarted (assuming the "mds" module is also
1468 reloaded at that time). The default number of threads will
1469 stay the same, 32 for most systems.
1474 Description: OST crash if filesystem is unformatted or corrupt
1475 Details : If an OST is started on a device that has never been formatted
1476 or if the filesystem is corrupt and cannot even mount then the
1477 error handling cleanup routines would dereference a NULL pointer.
1482 Description: NULL pointer deref in llap_from_page.
1483 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1484 incorrect) mapping if the page was truncated/reclaimed while it was
1485 searched for. Check for this condition and skip such pages when
1486 doing readahead. Introduce extra check to llap_from_page() to
1487 verify page->mapping->host is non-NULL (so page is not anonymous).
1490 Frequency : Sometimes when using sys_sendfile
1492 Description: "page not covered by a lock" warnings from ll_readpage
1493 Details : sendfile called ll_readpage without right page locks present.
1494 Now we introduced ll_file_sendfile that does necessary locking
1495 around call to generic_file_sendfile() much like we do in
1499 Frequency : with certain MDS communication failures at client mount time
1501 Description: NULL pointer deref after failed client mount
1502 Details : a client connection request may delayed by the network layer
1503 and not be sent until after the PTLRPC layer has timed out the
1504 request. If the client fails the mount immediately it will try
1505 to clean up before the network times out the request. Add a
1506 reference from the request import to the obd device and delay
1507 the cleanup until the network drops the request.
1510 Frequency : occasionally during client (re)connect
1512 Description: assertion failure during client (re)connect
1513 Details : processing a client connection request may be delayed by the
1514 client or server longer than the client connect timeout. This
1515 causes the client to resend the connection request. If the
1516 original connection request is replied in this interval, the
1517 client may trip an assertion failure in ptlrpc_connect_interpret()
1518 which thought it would be the only running connect process.
1521 Frequency : only with obd_echo servers and clients that are rebooted
1523 Description: kernel BUG accessing uninitialized data structure
1524 Details : When running an obd_echo server it did not start the ping_evictor
1525 thread, and when a client was evicted an uninitialized data
1526 structure was accessed. Start the ping_evictor in the RPC
1527 service startup instead of the OBD startup.
1529 Severity : enhancement
1530 Bugzilla : 10193 (patchless)
1531 Description: Remove dependency on various unexported kernel interfaces.
1532 Details : No longer need reparent_to_init, exit_mm, exit_files,
1533 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1536 Frequency : rare (only users of deprecated and unsupported LDAP config)
1538 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1539 Details : LDAP apparently contains 'lustreName' attributes instead of
1540 'name'. A simple remapping of the name is sufficient.
1543 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1545 Description: waiting_locks_callback trips kernel BUG if client is evicted
1546 Details : Running with the dump_on_timeout debug flag turned on makes
1547 it possible that the waiting_locks_callback() can try to dump
1548 the Lustre kernel debug logs from an interrupt handler. Defer
1549 this log dumping to the expired_lock_main() thread.
1551 Severity : enhancement
1553 Description: Support NFS exporting on 2.6 kernels.
1554 Details : Implement non-rawops metadata methods for NFS server to use without
1555 changing NFS server code.
1558 Frequency : very rare (synthetic metadata workload only)
1560 Description: two racing renames might cause an MDS thread to deadlock
1561 Details : Running the "racer" program may cause one MDS thread to rename
1562 a file from being the source of a rename to being the target of
1563 a rename at exactly the same time that another thread is doing
1564 so, and the second thread has already enqueued these locks after
1565 doing a lookup of the target and is trying to relock them in
1566 order. Ensure that we don't try to re-lock the same resource.
1569 Frequency : only very large systems with liblustre clients
1571 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1572 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1573 all OSTs in parallel. This allows the eviction of stale liblustre
1574 clients to proceed much faster than if they were done in series,
1575 and also offers similar improvements for other set_info RPCs.
1580 Description: excessive CPU usage during initial read phase on client
1581 Details : During the initial read phase on a client, it would agressively
1582 retry readahead on the file, consuming too much CPU and impacting
1583 performance (since 1.4.5.8). Improve the readahead algorithm
1584 to avoid this, and also improve some other common cases (read
1585 of small files in particular, where "small" is files smaller than
1586 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1591 Description: MDS crash when receiving packet with unknown intent.
1592 Details : Do not LBUG in unknown intent case, just return -EFAULT
1594 Severity : enhancement
1595 Bugzilla : 9293, 9385
1596 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1597 Details : Do not serialize getattr (non-intent version) and statfs.
1600 Frequency : occasional, when OST network is overloaded/intermittent
1602 Description: client evicted by OST after bulk IO timeout
1603 Details : If a client sends a bulk IO request (read or write) the OST
1604 may evict the client if it is unresposive to its data GET/PUT
1605 request. This is incorrect if the network is overloaded (takes
1606 too long to transfer the RPC data) or dropped the OST GET/PUT
1607 request. There is no need to evict the client at all, since
1608 the pinger and/or lock callbacks will handle this, and the
1609 client can restart the bulk request.
1612 Frequency : Always when mmapping file with no objects
1614 Description: client crashes when mmapping file with no objects
1615 Details : Check that we actually have objects in a file before doing any
1616 operations on objects in ll_vm_open, ll_vm_close and
1622 Description: Request leak when working with deleted CWD
1623 Details : Introduce advanced request refcount tracking for requests
1624 referenced from lustre intent.
1626 Severity : Enhancement
1628 Description: Cache open file handles on client.
1629 Details : MDS now will return special lock along with openhandle, if
1630 requested and client is allowed to hold openhandle, even if unused,
1631 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1632 closing files for every read/write openration.
1634 Severity : Enhancement
1636 Description: Cache open negative dentries on client when possible.
1637 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1638 negative dentries on lock revocation.
1643 Description: Remounting a client read-only wasn't possible with a zconf mount
1644 Details : It wasn't possible to remount a client read-only with llmount.
1646 Severity : enhancement
1647 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1648 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1649 MPICH is included to add Lustre-specific ADIO interfaces.
1650 This is based closely on the UFS ADIO layer and only differs
1651 in file creation, in order to allow the OST striping to be set.
1652 This is user-contributed code and not supported by CFS.
1657 Description: extended inode attributes (immutable, append-only) work improperly
1658 when 2.4 and 2.6 kernels are used on client/server or vice versa
1659 Details : Introduce kernel-independent values for these flags.
1661 Severity : enhancement
1664 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1665 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1666 granularity. Now, Fractional MB granularity is be supported,
1667 this is very useful for low memory system.
1669 Severity : enhancement
1671 Description: Getattr by fid
1672 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1673 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1674 have locks and avoids getattr rpc requests for it on every stat.
1677 Frequency : Always, for filesystems larger than 2TB
1679 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1680 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1681 filesystems larger than 2TB to be mounted with mballoc enabled.
1684 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1686 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1687 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1688 filesystems larger than 2TB is mounted then the kernel will
1689 silently corrupt the start of the filesystem. CONFIG_LBD is
1690 enabled for all CFS-supported kernels, but the possibility of
1691 this happening with a modified kernel config exists.
1693 Severity : enhancement
1695 Description: add client O_DIRECT support for 2.6 kernels
1696 Details : It is now possible to do O_DIRECT reads and writes to files
1697 in the Lustre client mountpoint on 2.6 kernel clients.
1699 Severity : enhancement
1701 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1702 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1703 parallel, not waiting for response from every OST before sending
1704 a rpc to the next OST.
1709 Description: setattr vs write race when updating file timestamps
1710 Details : Client processes that update a file timestamp into the past
1711 right after writing to the file (e.g. tar) it is possible that
1712 the updated file modification time can be reset to the current
1713 time due to a race between processing the setattr and write RPC.
1715 Severity : enhancement
1717 Description: Bring 'lfs find' closer in line with regular Linux find.
1718 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1719 -print0 options and obtains all the needed info through the lustre
1722 Severity : enhancement
1724 Description: support up to 1024 configured devices on one node
1725 Details : change obd_dev array from statically allocated to dynamically
1726 allocated structs as they are first used to reduce memory usage
1731 Description: Flush dirty partially truncated pages during truncate
1732 Details : Immediatelly flush partially truncated pages in filter_setattr,
1733 this way we completely avoid having any pages in page cache on OST
1734 and can retire ugly workarounds during writes to flush such pages.
1739 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1740 Details : protect inode from truncation within vfs_unlink() context
1741 just take a reference before calling vfs_unlink() and release it
1742 when parent's i_sem is free.
1747 Description: last_id value checked outside lock on OST caused LASSERT failure
1748 Details : If there were multiple MDS->OST object precreate requests in
1749 flight, it was possible that the OST's last object id was checked
1750 outside a lock and incorrectly tripped an assertion. Move checks
1751 inside locks, and discard old precreate requests.
1754 Frequency : always, if extents are used on OSTs
1756 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1757 Details : OSTs using the extents format would not zero the high 16 bits of
1758 the index physical block number. This is not a problem for any
1759 OST filesystems smaller than 16TB, and no kernels support ext3
1760 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1761 new/modified files) and can be fixed for existing filesystems
1762 with e2fsprogs-1.39-cfs1.
1767 Description: import connection selection may be incorrect if timer wraps
1768 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1769 import connections to be ignored if the 32-bit jiffies counter
1770 wraps. Use a 64-bit jiffies counter.
1773 Frequency : very large clusters immediately after boot
1775 Description: LNET request buffers exhausted under heavy short-term load
1776 Details : If a large number of client requests are generated on a service
1777 that has previously never seen so many requests it is possible
1778 that the request buffer growth cannot keep up with the spike in
1779 demand. Instead of dropping incoming requests, they are held in
1780 the LND until the RPC service can accept more requests.
1783 Frequency : Sometimes during replay
1785 Description: Assertion failure in ll_local_open after replay.
1786 Details : If replay happened on an open request reply before we were able
1787 to set replay handler, reply will become not swabbed tripping the
1788 assertion in ll_local_open. Now we set the handler right after
1789 recognising of open request
1792 Frequency : very rare
1794 Description: kernel reports "badness in vsnprintf"
1795 Details : Reading from the "recovery_status" /proc file in small chunks
1796 may cause a negative length in lprocfs_obd_rd_recovery_status()
1797 call to vsnprintf() (which is otherwise harmless). Exit early
1798 if there is no more space in the output buffer.
1800 Severity : enhancement
1802 Description: clear OBD RPC statistics by writing to them
1803 Details : It is now possible to clear the OBD RPC statistics by writing
1804 to the "stats" file.
1809 Description: Client mtime is not the same on different clients after utimes
1810 Details : In some cases, the client was using the utimes() syscall on
1811 a file cached on another node. The clients now validate the
1812 ctime from the MDS + OSTs to determine which one is right.
1817 Description: Inability to activate failout mode
1818 Details : lconf script incorrectly assumed that in pythong string's numeric
1819 value is used in comparisons.
1822 Frequency : always with multiple stripes per file
1824 Description: Inefficient object allocation for mutli-stripe files
1825 Details : When selecting which OSTs to stripe files over, for files with
1826 a stripe count that divides evenly into the number of OSTs,
1827 the MDS is always picking the same starting OST for each file.
1828 Return the OST selection heuristic to the original design.
1833 Description: mount failures may take full timeout to return an error
1834 Details : Under some heavy load conditions it is possible that a
1835 failed mount can wait for the full obd_timeout interval,
1836 possibly several minutes, before reporting an error.
1837 Instead return an error as soon as the status is known.
1839 Frequency : quota enabled and large files being deleted
1841 Description: releasing more than 4GB of quota at once hangs OST
1842 Details : If a user deletes more than 4GB of files on a single OST it
1843 will cause the OST to spin in an infinite loop. Release
1844 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1849 Description: statfs data retrieved from /proc may be stale or zero
1850 Details : When reading per-device statfs data from /proc, in the
1851 {kbytes,files}_{total,free,avail} files, it may appear
1852 as zero or be out of date.
1855 Frequency : systems with MD RAID1 external journal devices
1857 Description: lconf's call to blkid is confused by RAID1 journal devices
1858 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1859 of returning all block devices that match the journal UUID.
1862 Frequency : always, for aggregate stripe size over 4GB
1864 Description: assertion fails when trying to use 4GB stripe size
1865 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1866 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1869 Frequency : always on ppc64
1871 Description: the first write on an ext3 filesystem with mballoc got stuck
1872 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1873 perform endianness conversion.
1875 ------------------------------------------------------------------------------
1877 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1879 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1880 INTEROPERATE with older versions automatically. Please read the
1881 user documentation before upgrading any part of a live system.
1882 * WARNING: Lustre networking configuration changes are required with
1883 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1886 * Support for kernels:
1887 2.6.9-22.0.2.EL (RHEL 4)
1888 2.6.5-7.244 (SLES 9)
1889 2.6.12.6 vanilla (kernel.org)
1892 Severity : enhancement
1893 Bugzilla : 7981/8208
1894 Description: Introduced Lustre Networking (LNET)
1895 Details : LNET is new networking infrastructure for Lustre, it includes
1896 a reorganized network configuration mode (see the user
1897 documentation for full details) as well as support for routing
1898 between different network fabrics. Lustre Networking Devices
1899 (LNDs) for the supported network fabrics have also been
1900 created for this new infrastructure.
1902 Severity : enhancement
1903 Description: Introduced Access control lists
1904 Details : clients can set ACLs on files and directories in order to have
1905 more fine-grained permissions than the standard Unix UGO+RWX.
1906 The MDS must be started with the "-o acl" mount option.
1908 Severity : enhancement
1909 Description: Introduced filesystem quotas
1910 Details : Administrators may now establish per-user quotas on the
1913 Severity : enhancement
1915 Description: Configuration change for the XT3
1916 The PTLLND is now used to run Lustre over Portals on the XT3
1917 The configure option(s) --with-cray-portals are no longer used.
1918 Rather --with-portals=<path-to-portals-includes> is used to
1919 enable building on the XT3. In addition to enable XT3 specific
1920 features the option --enable-cray-xt3 must be used.
1925 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1926 Details : A race between allocating a new llcd and re-getting the llcd_lock
1927 allowed another thread to grab newly-allocated llcd.
1929 Severity : enhancement
1931 Description: 2.6 OST async journal commit and locking fix to improve performance
1932 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1933 2.6 kernels are now async as they already were in 2.4 kernels so
1934 that they can commit concurrently with the network bulk transfer.
1935 For block-allocated files the filter allocation semaphore is held
1936 to avoid filesystem fragmentation during allocation. BKL lock
1937 removed for 2.6 xattr operations where it is no longer needed.
1942 Description: lconf incorrectly determined whether two IP networks could talk
1943 Details : In some more complicated routing and multiple-network
1944 configurations, lconf will avoid trying to make a network
1945 connection to a disjoint part of the IP space. It was doing the
1946 math incorrectly for one set of cases.
1951 Description: Fix for potential infinite loop processing records in an llog.
1952 Details : If an llog record is corrupted/zeroed, it is possible to loop
1953 forever in llog_process(). Validate the llog record length
1954 and skip the remainder of the block on error.
1957 Frequency : occasional (liblustre only)
1959 Description: liblustre could not open files whose last component is a symlink
1960 Details : sysio_path_walk() would incorrectly pass the open intent to
1961 intermediate path components.
1964 Frequency : rare (liblustre only with non-standard tuning)
1965 Bugzilla : 7201 (7350)
1966 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1967 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1968 after it has been released, only do this for referenced locks
1970 Severity : enhancement
1972 Description: specifying an (invalid) directory default stripe_size of -1
1973 would reset the directory default striping
1974 Details : stripe_size -1 was used internally to signal directory stripe
1975 removal, now use "all default" to signal dir stripe removal
1976 as a directory striping of "all default" is not useful
1979 Frequency : common for large clusters running liblustre clients
1981 Description: doing an ls when liblustre clients are running is slow
1982 Details : sending a glimpse AST to a liblustre client waits for every AST
1983 to time out, as liblustre clients will not respond. Since they
1984 cannot cache data we refresh the OST lock LVB from disk instead.
1986 Severity : enhancement
1988 Description: doing an ls at the same time as file IO can be slow
1989 Details : enqueue and other "small" requests can be blocked behind many
1990 large IO requests. Create a new OST IO portal for non-IO
1991 requests so they can be processed faster.
1994 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1996 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1997 created file with mode 000
1998 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1999 allow an ftruncate() as a non-root user to a file with mode 000.
2000 Lustre now respects this flag to disable mode checks when
2001 truncating a file owned by the user
2004 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2007 Description: Revoking locks from clients that went dead or catatonic might take
2009 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2010 cancellation of such locks instant on servers without waiting for
2011 any reply from clients. Clients drops these locks when cancel
2012 notification from server is received without replying.
2015 Frequency : liblustre-only, when liblustre client dies or becomes busy
2017 Description: Doing ls on Linux clients can take a long time with active
2019 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2020 granting such locks to it in the first place if possible. Locks
2021 are taken by proxy on the OST during the read or write and
2022 dropped immediately afterward. Add connect flags handling, do
2023 not grant locks to liblustre clients for glimpse ASTs.
2025 Severity : enhancement
2027 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2028 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2029 sequential reads and behaves badly with applications doing random
2030 reads. Improve it by reading ahead at least read region, and
2031 avoiding excessive large RPC for small reads.
2033 Severity : enhancement
2035 Description: Creating more than 1000 files for a single job may cause a load
2036 imbalance on the OSTs if there are also a large number of OSTs.
2037 Details : qos_prep_create() uses an OST index reseed value that is an
2038 even multiple of the number of available OSTs so that if the
2039 reseed happens in the middle of the object allocation it will
2040 still utilize the OSTs as uniformly as possible.
2045 Description: OST or MDS may oops in ping_evictor_main()
2046 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2047 but doesn't restart at beginning of obd_exports_timed list
2050 Severity : enhancement
2052 Description: improve by-nid export eviction on the MDS and OST
2053 Details : allow multiple exports with the same NID to be evicted at one
2054 time without re-searching the exports list.
2057 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2059 Description: MDS may oops in groups_free()
2060 Details : in rare race conditions a newly allocated group_info struct is
2061 freed again, and this can be NULL. The 2.4 compatibility code
2062 for groups_free() checked for a NULL pointer, but 2.6 did not.
2065 Frequency : common for liblustre clients doing little filesystem IO
2066 Bugzilla : 9352, 7313
2067 Description: server may evict liblustre clients accessing contended locks
2068 Details : if a client is granted a lock or receives a completion AST
2069 with a blocking AST already set it would not reply to the AST
2070 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2073 Frequency : lfs setstripe, only systems with more than 160 OSTs
2075 Description: unable to set striping with a starting offset beyond OST 160
2076 Details : llapi_create_file() incorrectly limited the starting stripe
2077 index to the maximum single-file stripe count.
2080 Frequency : LDAP users only
2082 Description: lconf did not handle in-kernel recovery with LDAP properly
2083 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2085 Severity : enhancement
2087 Description: bind OST threads to NUMA nodes to improve performance
2088 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2089 node and do their allocations there to localize memory access
2091 Severity : enhancement
2093 Description: llmount can determine client NID directly from Myrinet (GM)
2094 Details : the client NID code from gmnalnid was moved directly into
2095 llmount, removing the need to use this or specifying the
2096 client NID explicitly when mounting GM clients with zeroconf
2099 Frequency : if client is started with down MDS
2101 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2102 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2103 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2104 ptlrpc_import_delayed_req() because we didn't check if the
2105 request was interrupted, and we also didn't break out of the
2106 event loop if there was no timeout
2111 Description: data loss during non-page-aligned writes to a single file from
2112 both multiple nodes and multiple threads on one node at same time
2113 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2114 inconsistency led to false short-reads, that were cached and later
2115 used by ->prepare_write() to fill in partially written page,
2116 leading to data loss.
2119 Frequency : always, if lconf --abort_recovery used
2121 Description: lconf --abort_recovery fails with 'Operation not supported'
2122 Details : lconf was attempting to abort recovery on the MDT device and not
2125 Severity : enhancement
2127 Description: remove cleanup logs
2128 Details : replace lconf-generated cleanup logs with lustre internal
2129 cleanup routines. Eliminates the need for client-cleanup and
2132 Severity : enhancement
2134 Description: add support for EAs (user and system) on lustre filesystems
2135 Details : it is now possible to store extended attributes in the Lustre
2136 client filesystem, and with the user_xattr mount option it
2137 is possible to allow users to store EAs on their files also
2139 Severity : enhancement
2141 Description: Add possibility (config option) to show minimal available OST free
2143 Details : When compiled with --enable-mindf configure option, statfs(2)
2144 (and so, df) will return least minimal free space available from
2145 all OSTs as amount of free space on FS, instead of summary of
2146 free spaces of all OSTs.
2148 Severity : enhancement
2150 Description: do not expand extent locks acquired on OST-side
2151 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2152 by server: they are not cached anyway.
2155 Frequency : when mmap is used/binaries executed from Lustre
2157 Description: Unmmap pages before throwing them away from read cache.
2158 Details : llap_shrink cache now attempts to unmap pages before discarding
2159 them (if unmapping failed - do not discard). SLES9 kernel has
2160 extra checks that trigger if this unmapping is not done first.
2165 Description: lconf didn't resolve symlinks before checking to see whether a
2166 given mountpoint was already in use
2169 Frequency : when migrating failover services
2170 Bugzilla : 6395, 9514
2171 Description: When migrating a subset of services from a node (e.g. failback
2172 from a failover service node) the remaining services would
2173 time out and evict clients.
2174 Details : lconf --force (implied by --failover) sets the global obd_timeout
2175 to 5 seconds in order to quickly disconnect, but this caused
2176 other RPCs to time out too quickly. Do not change the global
2177 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2179 Severity : enhancement
2180 Frequency : if MDS is started with down OST
2181 Bugzilla : 9439,5706
2182 Description: Allow startup/shutdown of an MDS without depending on the
2183 availability of the OSTs.
2184 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2185 Add appropriate locking and lov-osc refcounts for safe
2186 cleaning. Add osc abort_inflight calls in case the
2187 synchronize never started.
2190 Frequency : occasional (Cray XT3 only)
2192 Description: root not authorized to access files in CRAY_PORTALS environment
2193 Details : The client process capabilities were not honoured on the MDS in
2194 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2195 been accessed by an authorized user then root was able to access
2196 the file on the local client also. The root user capabilities
2197 are now allowed on the MDS, as this environment has secure UID.
2200 Frequency : occasional
2202 Description: ldiskfs "too long searching" message happens too often
2203 Details : A debugging message (otherwise harmless) prints too often on
2204 the OST console. This has been reduced to only happen when
2205 there are fragmentation problems on the filesystem.
2210 Description: Division by zero in statfs when all OSCs are inactive
2211 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2212 lov_statfs divides by value returned by lov_get_stripecnt().
2216 Bugzilla : 9489, 3273
2217 Description: First write from each client to each OST was only 4kB in size,
2218 to initialize client writeback cache, which caused sub-optimal
2219 RPCs and poor layout on disk for the first writen file.
2220 Details : Clients now request an initial cache grant at (re)connect time
2221 and so that they can start streaming writes to the cache right
2222 away and always do full-sized RPCs if there is enough data.
2223 If the OST is rebooted the client also re-establishes its grant
2224 so that client cached writes will be honoured under the grant.
2229 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2230 Details : Now I/O RPCs go to different portal number and (presumably) fast
2231 lock requests (and glimses) and other RPCs get their own service
2232 threads pool that should be able to service those RPCs
2235 Severity : enhancement
2237 Description: Ability to exchange lustre version between client and servers and
2238 issue warnings at client side if client is too old. Also for
2239 liblustre clients there is ability to refuse connection of too old
2241 Details : New 'version' field is added to connect data structure that is
2242 filled with version info. That info is later checked by server and
2246 Frequency : rare, liblustre only.
2247 Bugzilla : 9296, 9581
2248 Description: Two simultaneous writes from liblustre at offset within same page
2249 might proceed at the same time overwriting eachother with stale
2251 Details : I/O lock withing llu_file_prwv was released too early, before data
2252 actually was hitting the wire. Extended lock-holding time until
2253 server acknowledges receiving data.
2256 Frequency : extremely rare. Never observed in practice.
2258 Description: avoid generating lustre_handle cookie of 0.
2259 Details : class_handle_hash() generates handle cookies by incrementing
2260 global counter, and can hit 0 occasionaly (this is unlikely, but
2261 not impossible, because initial value of cookie counter is
2262 selected randonly). Value of 0 is used as a sentinel meaning
2263 "unassigned handle" --- avoid it. Also coalesce two critical
2264 sections in this function into one.
2266 Severity : enhancement
2268 Description: allow liblustre clients to delegate truncate locking to OST
2269 Details : To avoid overhead of locking, liblustre client instructs OST to
2270 take extent lock in ost_punch() on client's behalf. New connection
2271 flag is added to handle backward compatibility.
2273 Severity : enhancement
2274 Bugzilla : 4928, 7341, 9758
2275 Description: allow number of OST service threads to be specified
2276 Details : a module parameter allows the number of OST service threads
2277 to be specified via "options ost ost_num_threads={N}" in the
2278 OSS's /etc/modules.conf or /etc/modprobe.conf.
2282 Bugzilla : 6146, 9635, 9895
2283 Description: servers crash with bad pointer in target_handle_connect()
2284 Details : In rare cases when a client is reconnecting it was possible that
2285 the connection request was the last reference for that export.
2286 We would temporarily drop the export reference and get a new
2287 one, but this may have been the last reference and the export
2288 was just destroyed. Get new reference before dropping old one.
2290 Severity : enhancement
2291 Frequency : if client is started with failover MDS
2293 Description: Allow multiple MDS hostnames in the mount command
2294 Details : Try to read the configuration from all specified MDS
2295 hostnames during a client mount in case the "primary"
2298 Severity : enhancement
2300 Description: Stop sending data to evicted clients as soon as possible.
2301 Details : Check if the client we are about to send or are sending data to
2302 was evicted already. (Check is done every second of waiting,
2303 for which l_wait_event interface was extended to allow checking
2304 of exit condition at specified intervals).
2307 Frequency : rare, normally only when NFS exporting is done from client
2309 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2311 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2314 Frequency : always, if srand() is called before liblustre initialization
2316 Description: Liblustre uses system PRNG disturbing its usage by user application
2317 Details : Introduce internal to lustre fast and high-quality PRNG for
2318 lustre usage and make liblustre and some other places in generic
2319 lustre code to use it.
2321 Severity : enhancement
2322 Bugzilla : 9477, 9557, 9870
2323 Description: Verify that the MDS configuration logs are updated when xml is
2324 Details : Check if the .xml configuration logs are newer than the config
2325 logs stored on the MDS and report an error if this is the case.
2326 Request --write-conf, or allow starting with --old_conf.
2328 Severity : enhancement
2330 Description: Handle symlinks in the path when checking if Lustre is mounted.
2331 Details : Resolve intermediate symlinks when checking if a client has
2332 mounted a filesystem to avoid duplicate client mounts.
2337 Description: lconf can hit an error exception but still return success.
2338 Details : The lconf command catches the Command error exception at the top
2339 level script context and will exit with the associated exit
2340 status, but doesn't ensure that this exit status is non-zero.
2345 Description: failure of ptlrpc thread startup can cause oops
2346 Details : Starting a ptlrpc service thread can fail if there are a large
2347 number of threads or the server memory is very fragmented.
2348 Handle this without oopsing.
2351 Frequency : always, only if liblustre and non-default acceptor port was used
2353 Description: liblustre cannot connect to servers with non-default acceptor port
2354 Details : tcpnal_set_default_params() was not called and was therefore
2355 ignoring the environment varaible TCPNAL_PORT, as well as other
2356 TCPNAL_ environment variables
2361 Description: two objects could be created on the same OST for a single file
2362 Details : If an OST is down, in some cases it was possible to create two
2363 objects on a single OST for a single file. No problems other
2364 than potential performance impact and spurious error messages.
2368 Bugzilla : 5681, 9562
2369 Description: Client may oops in ll_unhash_aliases
2370 Details : Client dcache may become inconsistent in race condition.
2371 In some cases "getcwd" can fail if the current directory is
2377 Description: Inode refcounting problems in NFS export code
2378 Details : link_raw functions used to call d_instantiate without obtaining
2379 extra inode reference first.
2383 Bugzilla : 9942, 9903
2384 Description: Referencing freed requests leading to crash, memleaks with NFS.
2385 Details : We used to require that call to ll_revalidate_it was always
2386 followed by ll_lookup_it. Also with revalidate_special() it is
2387 possible to call ll_revalidate_it() twice for the same dentry
2388 even if first occurence returned success. This fix changes semantic
2389 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2390 reference on a request referred from the intent.
2391 ll_intent_release() then releases such a request.
2394 Frequency : rare, normally benchmark loads only
2396 Description: unlinked inodes were kept in memory on the client
2397 Details : If a client is repeatedly creating and unlinking files it
2398 can accumulate a lot of stale inodes in the inode slab cache.
2399 If there is no other client load running this can cause the
2400 client node to run out of memory. Instead flush old inodes
2401 from client cache that have the same inode number as a new inode.
2404 Frequency : SLES9 2.6.5 kernel and long filenames only
2405 Bugzilla : 9969, 10379
2406 Description: utime reports stale NFS file handle
2407 Details : SLES9 uses out-of-dentry names in some cases, which confused
2408 the lustre dentry revalidation. Change it to always use the
2412 Frequency : rare, unless heavy write-truncate concurrency is continuous
2413 Bugzilla : 4180, 6984, 7171, 9963, 9331
2414 Description: OST becomes very slow and/or deadlocked during object unlink
2415 Details : filter_destroy() was holding onto the parent directory lock
2416 while truncating+unlinking objects. For very large objects this
2417 may block other threads for a long time and slow overall OST
2418 responsiveness. It may also be possible to get a lock ordering
2419 deadlock in this case, or run out of journal credits because of
2420 the combined truncate+unlink. Solution is to do object truncate
2421 first in one transaction without parent lock, and then do the
2422 final unlink in a new transaction with the parent lock. This
2423 reduces the lock hold time dramatically.
2426 Frequency : rare, 2.4 kernels only
2428 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2429 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2430 before unmounting their filesystem in order to avoid deadlock.
2431 The kernel_locked() function in 2.4 kernels only checks whether
2432 the kernel lock is held, not whether it is this process that is
2433 holding it as 2.6 kernels do.
2438 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2439 Details : The client ptlrpc code may be trying to reconnect to a down
2440 server before a previous connection attempt has timed out.
2441 Increase the reconnect interval to be longer than the connection
2442 timeout interval to avoid sending duplicate connections to
2446 Frequency : echo_client brw_test command
2448 Description: fix echo_client to work with OST preallocated code
2449 Details : OST preallocation code (5137) didn't take echo_client IO path
2450 into account: echo_client calls filter methods outside of any
2451 OST thread and, hence, there is no per-thread preallocated
2452 pages and buffers to use. Solution: hijack pga pages for IO. As
2453 a byproduct, this avoids unnecessary data copying.
2457 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2458 Description: Client can oops in mdc_commit_close() after open replay
2459 Details : It was possible for the MDS to return an open request with no
2460 transaction number in mds_finish_transno() if the client was
2461 evicted, but without actually returning an error. Clients
2462 would later try to replay that open and may trip an assertion
2463 Simplify the client close codepath, and always return an error
2464 from the MDS in case the open is not successful.
2467 Frequency : rare, 2.6 OSTs only
2469 Description: OST may deadlock under high load on fragmented files
2470 Details : If there was a heavy load and highly-fragmented OST filesystems
2471 it was possible to have all the OST threads deadlock waiting on
2472 allocation of biovecs, because the biovecs were not released
2473 until the entire RPC IO was completed. Instead, release biovecs
2474 as soon as they are complete to ensure forward IO progress.
2476 Severity : enhancement
2478 Description: Support for specifying external journal device at mount
2479 Details : If an OST or MDS device is formatted with an external journal
2480 device, this device major/minor is stored in the ext3 superblock
2481 and may not be valid for failover. Allow detecting and
2482 specifying the external journal at mount time.
2487 Description: Mounting an MDS with pending unlinked files may cause oops
2488 Details : target_finish_recovery() calls mds_postrecov() which returned
2489 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2490 considers this an error and immediately begins cleaning up the
2491 lov, just after starting the mds_lov process
2493 Severity : enhancement
2495 Description: Implement 'lfs df' to report actual free space on per-OST basis
2496 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2497 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2498 usage of objects; '-h' to report in human readable format.
2500 ------------------------------------------------------------------------------
2502 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2509 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2510 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2511 at startup during orphan unlinking. Instead, skip buddy update
2512 before setup, as it will be regenerated after recovery is complete.
2517 Description: If an OST is inactive, its locks might reference stale inodes.
2518 Details : lov_change_cbdata() must iterate over all namespaces, even if
2519 they are inactive to clear inode references from the lock.
2521 Severity : enhancement
2522 Frequency : occasional, if non-standard max_dirty_mb used
2524 Description: Client will block write RPCs if not enough grant
2525 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2526 then the client will block writes while waiting for another RPC
2527 to complete instead of consuming its dirty limit. With change
2528 we get improved performance when max_dirty_mb is small.
2530 Severity : enhancement
2531 Bugzilla : 3389, 6253
2532 Description: Add support for supplementary groups on the MDS.
2533 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2534 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2535 lookups for user supplementary groups into a cache.
2539 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2540 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2541 for liblustre O_EXCL is enforced by MDS.
2545 Description: readdir never returns NULL in liblustre.
2546 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2547 dentry in struct dirent.
2551 Frequency : liblustre only, depends on application IO pattern
2552 Description: liblustre clients evicted if not contacting servers
2553 Details : Don't put liblustre clients into the ping_evictor list, so
2554 they will not be evicted by the pinger ever.
2556 Severity : enhancement
2558 Description: Add ability to evict clients by NID from MDS.
2559 Details : By echoing "nid:$NID" string into
2560 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2561 $NID would be instantly evicted from this MDS and from all active
2562 OSTs connected to it.
2566 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2567 Details : lookup_it_finish() used to query file size from OSTs that was not
2572 Description: service threads change working directory to that of init
2573 Details : Starting lustre service threads may pin the working directory
2574 of the parent thread, making that filesystem busy. Threads
2575 now change to the working directory of init to avoid this.
2579 Frequency : during shutdown only
2580 Description: shutdown with a failed MDS or OST can cause unmount to hang
2581 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2586 Frequency : 2.6 only
2587 Description: chmod/chown may include an extra supplementary group
2588 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2589 supplementary group and if none is specified this is used.
2592 Bugzilla : 5479 (6816)
2594 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2595 Details : If lookup is in progress on a file that is unlinked we might try
2596 to revalidate the inode and fail in revalidate after lookup is
2597 complete and ll_file_open() enqueues the open again but
2598 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2601 Frequency : always, if lconf --abort_recovery used
2603 Description: lconf --abort_recovery fails with 'Operation not supported'
2604 Details : lconf was attempting to abort recovery on the MDT device and not
2607 ------------------------------------------------------------------------------
2609 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2614 Frequency : rare (only unsupported configurations with a node running as an
2616 Bugzilla : 6514, 5137
2617 Description: Mounting a Lustre file system on a node running as an OST could
2619 Details : OSTs now preallocates memory needed to write out data at
2620 startup, instead of when needed, to avoid having to
2621 allocate memory in possibly low memory situations.
2622 Specifically, if the file system is mounted on on OST,
2623 memory pressure could force it to try to write out data,
2624 which it needed to allocate memory to do. Due to the low
2625 memory, it would be unable to do so and the node would
2626 become unresponsive.
2628 Severity : enhancement
2630 Description: Addition of lconf --service command line option
2631 Details : lconf now accepts a '--service <arg>' option, which is
2632 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2634 Severity : enhancement
2636 Description: Failover mode is now the default for OSTs.
2637 Details : By default, OSTs will now run in failover mode. To return to
2638 the old behaviour, add '--failout' to the lmc line for OSTs.
2640 Severity : enhancement
2642 Description: Health checks are now provided for MDS and OSTs
2643 Details : Additional detailed health check information on MSD and OSTs
2644 is now provided through the procfs health_check value.
2647 Frequency : occasional, depends on IO load
2649 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2650 after numerous create/delete cycles
2651 Details : The ext3 inode allocation policy would not allocate new inodes
2652 very well on the OSTs because there are no new directories
2653 being created. Instead we look for groups with free space if
2654 the parent directories are nearly full.
2659 Description: Network or server problems during mount may cause partially
2660 mounted clients instead of returning an error.
2661 Details : The config llog parsing code may overwrite the error return
2662 code during mount error handling, returning success instead
2668 Description: MDS can fail to allocate large reply buffers
2669 Details : After long uptimes the MDS can fail to allocate large reply
2670 buffers (e.g. zconf client mount config records) due to memory
2671 fragmentation or consumption by the buffer cache. Preallocate
2672 some large reply buffers so that these replies can be sent even
2673 under memory pressure.
2677 Frequency : rare (liblustre)
2678 Description: fsx running with liblustre complained that using truncate() to
2679 extend the file doesn't work. This patch corrects that issue.
2680 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2681 ATTR_SIZE and lsm use in llu_setattr_raw.
2685 Frequency : rare, only 2.6 kernels
2686 Description: Unusual file access patterns on the MDS may result in inode
2687 data being lost in very rare circumstances.
2688 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2689 code in the 2.6 kernel caused the inode and/or EA data not to
2690 be read from disk, causing single-file corruption.
2694 Frequency : rare, only 2.6 filesystems using extents
2695 Description: Heavy concurrent write and delete load may cause data corruption.
2696 Details : It was possible under high-load situations to have an extent
2697 metadata block in the block device cache from a just-unlinked
2698 file overwrite a newly-allocated data block. We now unmap any
2699 metadata buffers that alias just-allocated data blocks.
2703 Frequency : filesystems with default stripe_count larger than 77
2704 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2705 Details : lconf specifies an inode size of 4096 bytes when the default
2706 stripe_count is larger than 77. This conflicts with the default
2707 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2708 this case to avoid pinning too much memory for large EAs.
2710 ------------------------------------------------------------------------------
2712 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2717 Frequency : rare (extremely heavy IO load with hundreds of clients)
2719 Description: Client is evicted, gets IO error writing to file
2720 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2721 caused clients to be evicted for AST timeouts. The fixes in
2722 bug 5192 mean we no longer need to have such short AST timeouts
2723 so ldlm_timeout has been increased.
2726 Frequency : occasional during --force or --failover shutdown under load
2727 Bugzilla : 5949, 4834
2728 Description: Server oops/LBUG if stopped with --force or --failover under load
2729 Details : a collection of import/export refcount and cleanup ordering
2730 issues fixed for safer force cleanup
2733 Frequency : only filesystems larger than 120 OSTs
2734 Bugzilla : 5990, 6223
2735 Description: lfs getstripe would oops on a very large filesystem
2736 Details : lov_getconfig used kfree on vmalloc'd memory
2739 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2740 Bugzilla : 6242, 6243
2741 Description: reading from files that had been truncated to a non-zero size
2742 but never opened returned no data
2743 Details : ll_file_read() reads zeros from no-object files to EOF
2748 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2749 Details : There's little harm in aborting MDS/OSS recovery and letting it
2750 try again, so I removed the LASSERT and return an error instead.
2752 Severity : enhancement
2754 Description: New debugging infrastructure for tracking down data corruption
2755 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2756 (b) cover more of the client-side code path, and (c) try to narrow
2757 down where problems occurred
2761 Bugzilla : 3819, 4364, 4397, 6313
2762 Description: Racing close and eviction MDS could cause assertion in mds_close
2763 Details : It was possible to get multiple mfd references during close and
2764 client eviction, leading to one thread referencing a freed mfd.
2766 Severity: : enhancement
2767 Bugzilla : 3262, 6359
2768 Description: Attempts to reconnect to servers are now more aggressive.
2769 Details : This builds on the enhanced upcall-less recovery that was added
2770 in 1.4.2. When trying to reconnect to servers, clients will
2771 now try each server in the failover group every 10 seconds. By
2772 default, clients would previously try one server every 25 seconds.
2777 Description: After recovery, certain operations trigger a failed
2778 assertion on a client.
2779 Details : Failing over an mds, using lconf -d --failover, while a
2780 client was doing a readdir() call would cause the client to
2781 LBUG after recovery completed and the readdir() was resent.
2783 Severity : enhancement
2785 Description: Default groups are now added by lconf
2786 Details : You can now run lconf --group <servicename> without having to
2787 manually add groups with lmc.
2790 Frequency : occasional
2792 Description: Nodes with an elan id of 0 trigger a failed assertion
2795 Frequency : always when accessing e.g. tty/console device nodes
2797 Description: tty and some other devices nodes cannot be used on lustre
2798 Details : file's private_data field is used by device data and lustre
2799 values in there got lost. New field was added to struct file to
2800 store fs-specific private data.
2803 Frequency : when exporting Lustre via NFS
2805 Description: NFSD failed occasionally when looking up a path component
2806 Details : NFSD is looking up ".." which was broken in ext3 directories
2807 that had grown large enough to become hashed.
2810 Frequency : Clusters with multiple interfaces not on the same subnet
2812 Description: Nodes will repeatedly try to reconnect to an interface which it
2813 cannot reach and report an error to the log.
2814 Details : Extra peer list entries will be created by lconf with some peers
2815 unreachable. lconf now validates the peer before adding it.
2818 Frequency : Only if a default stripe is set on the filesystem root.
2820 Description: Setting a default stripe on the filesystem root prevented the
2821 filesystem from being remounted.
2822 Details : The client was sending extra request flags in the root getattr
2823 request and did not allocate a reply buffer for the dir EA.
2826 Frequency : occasional, higher if lots of files are accessed by one client
2827 Bugzilla : 6159, 6097
2828 Description: Client trips assertion regarding lsm mismatch/magic
2829 Details : While revalidating inodes the VFS looks up inodes with ifind()
2830 and in rare cases can find an inode that is being freed.
2831 The ll_test_inode() code will free the lsm during ifind()
2832 when it finds an existing inode and then the VFS later attaches
2833 this free lsm to a new inode.
2837 Bugzilla : 6422, 7030
2838 Description: MDS deadlock between mkdir and client eviction
2839 Details : Creating a new file via mkdir or mknod (starting a transaction
2840 and getting the ns lock) can deadlock with client eviction
2841 (gets ns lock and trying to finish a synchronous transaction).
2844 Frequency : occasional
2845 Description: While starting a server, the fsfilt_ext3 module could not be
2847 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2848 kernels. Previously, lconf would still use the ext3 name
2849 when trying to load modules. Now, it will correctly use
2850 ext3 on 2.4 and ldiskfs on 2.6.
2852 Severity : enhancement
2853 Description: The default stripe count has been changed to 1
2854 Details : The interpretation of the default stripe count (0, to lfs
2855 or lmc) has been changed to mean striping across a single
2856 OST, rather than all available. For general usage we have
2857 found a stripe count of 1 or 2 works best.
2859 Severity : enhancement
2860 Description: Add support for compiling against Cray portals.
2861 Details : Conditional compiling for some areas that are different
2865 Frequency : occasional
2866 Bugzilla : 6409, 6834
2867 Description: Creating files with an explicit stripe count may lead to
2868 a failed assertion on the MDS
2869 Details : If some OSTs are full or unavailable, creating files may
2870 trigger a failed assertion on the MDS. Now, Lustre will
2871 try to use other servers or return an error to the
2875 Frequency : occasional
2877 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2878 caused lock timeout during readahead (since 1.4.2).
2879 Details : Processes doing readahead might match a lock that hasn't been
2880 granted yet if there are overlapping and conflicting lock
2881 requests. The readahead process waits on ungranted lock
2882 (original lock is CBPENDING), while OST waits for that process
2883 to cancel CBPENDING read lock and eventually evicts client.
2885 Severity : enhancement
2887 Description: Initial enabling of flock support for clients
2888 Details : Implements fcntl advisory locking and file status functions.
2889 This feature is provided as an optional mount flag (default
2890 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2891 locking are implemented yet, and those that are are not guaranteed
2892 to be completely correct in production environments.
2893 mount -t lustre -o [flock|noflock] ...
2896 Frequency : occasional
2898 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2899 assertion in the ext3 JBD (journaling) layer.
2900 Details : The b_committed_data struct is protected by the big kernel lock
2901 in 2.4 kernels, serializing journal_commit_transaction() and
2902 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2903 access to this struct. In 2.6 kernels there is finer grained
2904 locking to improve SMP performance of the JBD layer.
2908 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2913 Description: merge and cleanup kernel patches.
2916 -----------------------------------------------------------------------------
2918 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2920 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2921 versions of Lustre. Please update all servers and clients to
2922 version 1.4.2 or later at the same time. You must also run
2923 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2924 to update the configuration logs.
2926 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2927 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2928 - swab llog records of type '0' so we get proper header size/idx (5861)
2929 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2930 - fix rename of one directory over another leaking an inode (5953)
2931 - avoid SetPageDirty on 2.6 (5981)
2932 - don't re-add just-being-destroyed locks to the waiting list (5653)
2933 - when creating new directories, inherit the parent's custom
2934 striping settings if present parent (3048)
2935 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2936 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2937 - don't hold client locks on temporary worklist from l_lru (5666)
2938 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2939 - automatically evict dead clients (5921)
2940 - Update file size properly in create+truncate+fstat case (6196)
2941 - Do not unhash mountpoint dentries, do not allow removal of
2943 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2944 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2945 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2946 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2947 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2948 - Correctly handle reads of files with no objects (6243)
2949 - lctl recover will also mark a device active if deactivate used (5933)
2951 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2952 - allow --write-conf on an MDS with different nettype than client (5619)
2953 - don't write config llogs to MDS for mounts not from that MDS (5617)
2954 - lconf should create multiple TCP connections from a client (5201)
2955 - init scripts are now turned off by default; run chkconfig --on
2956 lustre and chkconfig --on lustrefs to use them
2957 - upcalls are no longer needed for clients to recover to failover
2959 - add --abort-recovery option to lconf to abort recovery on device
2961 - add support for an arbitrary number of OSTs (3026)
2962 - Quota support protocol changes.
2963 - forward compatibility changes to wire structs (6007)
2964 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2965 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2966 - improved "lustre status" script
2967 - initialize blocksize for non-regular files (6062)
2968 - added --disable-server and --disable-client configure options (5782)
2969 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2970 - Vanilla 2.4.29 support
2971 - increase maximum number of obd devices to 520 (6242)
2972 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2973 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2974 SLES 9 2.6.5 kernels
2975 - limit stripes per file to 160 (the maximum EA size) (6093)
2977 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2980 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2981 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2982 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2983 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2984 - small ext3 extents cleanups and fixes (5733)
2985 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2986 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2987 - avoid needless client->OST connect, fix handle mismatch (5317)
2988 - fix DLM error path that led to out-of-sync client, long delays (5779)
2989 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2990 - fix several locking issues related to i_size (5492,5624,5654,5672)
2991 - don't move pending lock onto export if it is already evicted (5683)
2992 - fix kernel oops when creating .foo in unlinked directory (5548)
2993 - fix deadlock in obdfilter statistics vs. object create (5811)
2994 - use time_{before,after} to avoid timer jiffies wrap (5882)
2995 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2996 - Do not leak request if server was not able to process it (5154)
2997 - If mds_open unable to find parent dir, make that negative lookup(5154)
2998 - don't create new directories with extent-mapping (5909, 5936)
3000 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3001 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3002 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3003 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3004 - support for RHEL4 kernel on i686 (5773)
3005 - provide error messages when incompatible logs are encountered (5898)
3007 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3008 * version 1.4.0.10 (1.4.1 release candidate 1)
3010 - don't keep a lock reference when lock is not granted (4238)
3011 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3012 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3013 - drop import inflight refcount on signal_completed_replay error (5255)
3014 - unlock page after async write error during send (3677)
3015 - handle missing objects in filter_preprw_read properly (5265)
3016 - no transno return for symlink open, don't save no-trasno open (3440)
3017 - don't try to complete elan receive that already failed (4012)
3018 - free RPC server reply state on error (5406)
3019 - clean up thread from ptlrpc_start_thread() on error (5160)
3020 - readahead could read extra page into cache that wasn't ejected (5388)
3021 - prevent races in class_attach/setup/cleanup/detach (5260)
3022 - don't dereference de->d_inode after l_dput of de (5458)
3023 - use "int" for stripe value returned from lock_to_stripe (5544)
3024 - mballoc allocation and error-checking fixes in 2.6 (5504)
3025 - block device patches to fix I/O request sizes in 2.6 (5482)
3026 - look up hostnames for IB nals (5602)
3027 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3028 - don't start multiple acceptors for the same port (5277)
3029 - fix incorrect LASSERT in mds_getattr_name (5635)
3030 - export a proc file for general "ping" checking (5628)
3031 - fix "lfs check" to not block when the MDS is down (5628)
3033 - service request history (4965)
3034 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3035 - create an "evict_client" /proc entry on OSTs, like the MDS has
3036 - fix mount usage message, return errors per mount(8) (5168)
3037 - change grep [] to grep "[]" in tests so they work in more UMLs
3038 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3039 - remove ancient LOV_MAGIC_V0 EA support (5047)
3040 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3041 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3042 - allow readahead allocations to fail when low on memory (5383)
3043 - mmap locking landed again, after considerable improvement (2828)
3044 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3046 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3049 - send OST transaction number in read/write reply to free req (4966)
3050 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3051 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3052 - the watchdog thread now runs as interruptible (5246)
3053 - flock/lockf fixes (but it's still disabled, pending 5135)
3054 - don't use EXT3 constants in llite code (5094)
3055 - memory shortage at startup could cause assertion (5176)
3057 - reorganization of lov code
3058 - single portals codebase
3060 - add extents/mballoc support (5025)
3061 - direct I/O reads in the obdfilter (4048)
3062 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3064 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3067 - send OST transaction number in read/write reply to free req (4966)
3068 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3069 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3070 - the watchdog thread now runs as interruptible (5246)
3071 - handle missing objects in filter_preprw_read properly (5265)
3072 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3073 - drop import inflight refcount on signal_completed_replay error (5255)
3074 - unlock page after async write error during send (3677)
3075 - return original error code on reconstructed replies (3761)
3076 - no transno return for symlink open, don't save no-trasno open (3440)
3078 - add pid to ldlm debugging output (4922)
3079 - bump the watchdog timeouts -- we can't handle 30sec yet
3080 - extra debugging for orphan dentry/inode bug (5259)
3082 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3085 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3086 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3087 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3088 - fix NULL dereference in /proc/sys/portals/routes (4827)
3089 - allow failed mdc_close() operations to be interrupted (4561)
3090 - stop precreate on OST before MDS would time out on it (4778)
3091 - don't send partial-page writes before EOF from client (4410)
3092 - discard client grant for sub-page writes on large-page clients (4520)
3093 - don't free dentries not owned by NFS code, check generation (4806)
3094 - fix lsm leak if mds_create_objects() fails (4801)
3095 - limit debug_daemon file size, always print CERROR messages (4789)
3096 - use transno after validating reply (3892)
3097 - process timed out requests if import state changes (3754)
3098 - update mtime on OST during writes, return in glimpse (4829)
3099 - add mkfsoptions to LDAP (4679)
3100 - use ->max_readahead method instead of zapping global ra (5039)
3101 - don't interrupt __l_wait_event() during strace
3103 - add software watchdogs to catch hung threads quickly (4941)
3104 - make lustrefs init script start after nfs is mounted
3105 - fix CWARN/ERROR duplication (4930)
3106 - return async write errors to application if possible (2248)
3107 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3108 - print NAL number in %x format (4645)
3109 - update barely-supported suse-2.4.21-171 series (4842)
3110 - support for sles 9 %post scripts
3111 - support for building 2.6 kernel-source packages
3112 - support for sles km_* packages
3114 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3117 - ignore -ENOENT errors in osc_destroy (3639)
3118 - notify osc create thread that OSC is being cleaned up (4600)
3119 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3120 - reconstruct ost_handle() like mds_handle() (4657)
3121 - create a new thread to do import eviction to avoid deadlock (3969)
3122 - let lconf resolve symlinked-to devices (4629)
3123 - don't unlink "objects" from directory with default EA (4554)
3124 - hold socknal file ref over connect in case target is down (4394)
3125 - allow more than 32000 subdirectories in a single directory (3244)
3126 - fix blocks count for O_DIRECT writes (3751)
3127 - OST returns ENOSPC from object create when no space left (4539)
3128 - don't send truncate RPC if file size isn't changing (4410)
3129 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3130 - bind to privileged port in socknal and tcpnal (3689)
3132 - rate limit CERROR/CWARN console message to avoid overload (4519)
3133 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3134 - basic mmap support (3918)
3135 - kernel patch series update from b1_4 (4711)
3137 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3140 - avoid crash during MDS cleanup with OST shut down (2775)
3141 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3142 - don't use bad inodes on the MDS (3744)
3143 - dynamic object preallocation to improve recovery speed (4236)
3144 - don't hold spinlock over lock dumping or change debug flags (4401)
3145 - don't zero obd_dev when it is force cleaned (3651)
3146 - print grants to console if they go negative (4431)
3147 - "lctl deactivate" will stop automatic recovery attempts (3406)
3148 - look for existing locks in ldlm_handle_enqueue() (3764)
3149 - don't resolve lock handle twice in recovery avoiding race (4401)
3150 - revalidate should check working dir is a directory (4134)
3152 - don't always mark "slow" obdfilter messages as errors (4418)
3154 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3157 - don't close LustreDB during write_conf until it is done (3860)
3158 - fix typo in lconf for_each_profile (3821)
3159 - allow dumping logs from multiple threads at one time (3820)
3160 - don't allow multiple threads in OSC recovery (3812)
3161 - fix debug_size parameters (3864)
3162 - fix mds_postrecov to initialize import for llog ctxt (3121)
3163 - replace config semaphore with spinlock (3306)
3164 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3165 - don't allow enqueue to complete on a destroyed export (3822)
3166 - down write_lock before checking llog header bitmap (3825)
3167 - recover from lock replay timeout (3764)
3168 - up llog sem before sending rpc (3652)
3169 - reduce ns lock hold times when setting kms (3267)
3170 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3171 - fix NULL deref and obd_dev leak on setup error (3312)
3172 - replace some LBUG about llog ops with error handling (3841)
3173 - don't match INVALID dentries from d_lookup and spin (3784)
3174 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3175 - fix invalid assertion in ptlrpc_set_wait (3880)
3177 - add libwrap support for the TCP acceptor (3996)
3178 - add /proc/sys/portals/routes for non-root route listing (3994)
3179 - allow setting MDS UUID in .xml (2580)
3180 - print the stack of a process that LBUGs (4228)
3182 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3185 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3186 - make sure to unset replay flag from failed open requests (3440)
3187 - if default stripe count is 0, use OST count for inode size (3636)
3188 - update parent mtime/ctime on client for create/unlink (2611)
3189 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3190 - free recovery state on server during a forced cleanup (3571)
3191 - unregister_reply for resent reqs (3063)
3192 - loop back devices mounting and status check on 2.6 (3563)
3193 - fix resource-creation race that can provoke i_size == 0 (3513)
3194 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3195 - more debugging for page-accounting assertion (3746)
3196 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3197 - avoid deadlock after precreation failure (3758)
3198 - fix race and lock order deadlock in orphan handling (3450, 3750)
3199 - add validity checks when grabbing inodes from l_ast_data (3599)
3201 - add /proc/.../recovery_status to obdfilter (3428)
3202 - lightweight CDEBUG infrastructure, debug daemon (3668)
3203 - change default OSC RPC parameters to be better on small clusters
3204 - turn off OST read cache for files smaller than 32MB
3205 - install man pages and include them in rpms (3100)
3206 - add new init script for (un)mounting lustre filesystems (2593)
3207 - run chkconfig in %post for init scripts (3701)
3208 - drop scimac NAL (unmaintained)
3210 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3213 - clean kiobufs before and after use (3485)
3214 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3215 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3216 - change init script to fail more clearly if not run as root (1528)
3217 - allow clients to reconnect during replay (1742)
3218 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3219 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3220 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3221 - fix cond_resched (3554)
3222 - only evict unfinished clients after recovery (3515)
3223 - allow bulk resend, prevent data loss (3570)
3224 - dynamic ptlrpc request buffer allocation (2102)
3225 - don't allow unlinking open directory if it isn't empty (2904)
3226 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3227 - remove extraneous obd dereference causing LASSERT failure (3334)
3228 - don't use get_cycles() when creating temp. files on the mds (3156)
3229 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3230 - handle EEXIST for set-stripe, set proper directory name (3336)
3232 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3233 - fix ksocknal_fmb_callback() error messages (2918)
3235 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3238 - don't copy lvb into (possibly NULL) reply on error (2983)
3239 - don't deref dentry after dput, don't free lvb on error (2922)
3240 - use the kms to determine writeback rpc length (2947)
3241 - increment oti_logcookies when osc is inactive (2948)
3242 - update client's i_blocks count via lvb messages (2543)
3243 - handle intent open/close of special files properly (1557)
3244 - mount MDS with errors=remount-ro, like obdfilter (2009)
3245 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3246 - don't use cancelling-locks' kms values (2947)
3247 - use highest lock extent for kms, not last one (2925)
3248 - don't dereference ERR_PTR() dentry in error handling path (3107)
3249 - fix thread race in portals_debug_dumplog() (3122)
3250 - create lprocfs device entries at setup instead of at attach (1519)
3251 - common AST error handler, don't evict client on completion race (3145)
3252 - zero nameidata in detach_mnt in 2.6 (3118)
3253 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3254 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3255 - initialize RPC timeout timer earlier for 2.6 (3219)
3256 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3257 - print nal/nid for unknown nid (3258)
3258 - additional checks for oscc recovery before doing precreate (3284)
3259 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3260 - don't crash in mdc_close for bad permissions on open (3285)
3261 - zero i_rdev for non-device files (3147)
3262 - clear page->private before handing to FS, better assertion (3119)
3263 - tune the read pipeline (3236)
3264 - fix incorrect decref of invalidated dentry (2350)
3265 - provide read-ahead stats and refine rpc in flight stats (3328)
3266 - don't hold journal transaction open across create RPC (3313)
3267 - update atime on MDS at close time (3265)
3268 - close LDAP connection when recovering to avoid server load (3315)
3269 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3270 - don't leak open file on MDS after open resend (3325)
3271 - serialize filter_precreate and filter_destroy_precreated (3329)
3272 - loop device shouldn't call sync_dev() for nul device (3092)
3273 - clear page cache after eviction (2766)
3274 - resynchronize MDS->OST in background (2824)
3275 - refuse to mount the same filesystem twice on same mountpoint (3394)
3276 - allow llmount to create routes for mounting behind routers (3320)
3277 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3278 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3279 - fix rare problem with rename on htree directories (3417)
3281 - allow default OST striping configuration per directory (1414)
3282 - fix compilation for qswnal for 2.6 kernels (3125)
3283 - increase maximum number of MDS request buffers for large systems
3284 - change liblustreapi to be useful for external progs like lfsck (3098)
3285 - increase local configuration timeout for slow disks (3353)
3286 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3288 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3291 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3292 - don't overwrite extent policy data in reply if lock was blocked (2901)
3293 - drop filter export grants atomically with removal from device (2663)
3294 - del obd_self_export from work_list in class_disconnect_exports (2908)
3295 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3296 - swab reply message in mdc_close, other PPC fixes (2464)
3297 - fix destroying of named logs (2325)
3298 - overwrite old logs when running lconf --write_conf (2264)
3299 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3300 - fix race in target_handle_connect (2898)
3301 - mds_reint_create() should take same inode create lock (2926)
3302 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3303 - don't close files for self_export to avoid uninitialized obd (2936)
3304 - allow MDS with the same name as client node (2939)
3305 - hold dentry reference for closed log files for unlink (2325)
3306 - reserve space for all logs during transactions (2059)
3307 - don't evict page beyond end of stripe extent (2925)
3308 - don't oops on a deleted current working directory (2399)
3309 - handle hard links to targets without a parent properly (2517)
3310 - don't dereference NULL lock when racing during eviction (2867)
3311 - don't grow lock extents when lots of conflicting locks (2919)
3313 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3316 - account for cache space usage on clients to avoid data loss (974)
3317 - lfsck support in lustre kernel code (2349)
3318 - reduce journal credits needed for BRW writes (2370)
3319 - orphan handling to avoid losing space on client/server crashes
3320 - ptlrpcd can be blocked, stopping ALL progress (2477)
3321 - use lock value blocks to assist in proper KMS, faster stat (1021)
3322 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3323 - recovery for initial connections (2355)
3324 - fixes for mds_cleanup_orphans (1934)
3325 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3326 - block all file creations until orphan recovery completes (1901)
3327 - client remove rq_connection from request struct (2423)
3328 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3329 - recovery timer race (2670)
3330 - mdc_close recovey bug (2532)
3331 - ptlrpc cleanup bug (2710)
3332 - mds timeout on local locks (2588)
3333 - namespace lock held during RPCs (2431)
3334 - handle interrupted sync write properly (2503)
3335 - don't try to handle a message that hasn't been replied to (2699)
3336 - client assert failure during cleanup after abort recovery (2701)
3337 - leak mdc device after failed mount (2712)
3338 - ptlrpc_check_set allows timedout requests to complete (2714)
3339 - wait for inflight reqs when ptlrpcd finishes (2710)
3340 - make sure unregistered services are removed from the srv_list
3341 - reset bulk XID's when resending them (caught by 1138 test)
3342 - unregister_bulk after timeout
3343 - fix lconf error (2694)
3344 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3345 - readahead locks pages, leaves pending causing memory pressure (2673)
3346 - increase OST request buffers to 4096 on large machines (2729)
3347 - fix up permission of existing directories in simple_mkdir (2661)
3348 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3349 - don't assign transno to errored transactions (2742)
3350 - don't delete objects on OST if given a bogus objid from MDS (2751)
3351 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3352 - if rq_no_resend, then timeout request after recovery (2432)
3353 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3354 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3355 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3356 - MDS recovery shouldn't delete objects that it has given out (2730)
3357 - if enqueue arrives after completion, don't clobber LVB (2819)
3358 - don't unlock pages twice when trigger_group_io returns error (2814)
3359 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3360 - don't write pages to disk if there was an error (1450)
3361 - don't ping imports that have recovery disabled (2676)
3362 - take buffered bytes into account when balancing socknal conn (2817)
3363 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3364 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3365 - remove little-endian swabbing of llog records (1987)
3366 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3367 - retry reposting request buffers if they fail (1191)
3368 - grow extent at grant time to avoid granting a revoked lock (2809)
3369 - lock revoke doesn't evict page if covered by a second lock (2765)
3370 - disable VM readahead to avoid reading outside lock extents (2805)
3372 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3373 - updated kernel patches for hp-2.4.20 kernel (2681)
3375 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3378 - fix truncated write corruption (2366)
3379 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3381 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3382 - ensure len(uuid) < 37 in lmc (1171)
3383 - fix ia64 OOPS in llog_test (2255)
3384 - zero end of page at obdfilter for partial page writes (2648)
3385 - don't leave stale dentries around after renames (bug 2428)
3386 - fix timeouts when evicting a client with a single lock held (2642)
3387 - set deadline for the initial HELLO message to drain (2634)
3388 - print out dotted-quad IP addresses in the socknal (2302)
3390 - additional debugging for MDS client eviction problem (2443)
3391 - fix mkfsoptions support for osts (2603, 2604)
3393 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3396 - add series for the vanilla 2.6.0 kernel
3397 - add series for the vanilla 2.4.24 kernel
3398 - add series for a cray x86/64 UL kernel drop
3399 - fix xattr patches for the vanilla 2.4.19 series
3401 - generate true UUIDs in lmc (1171)
3402 - have portals stack dumping break in UML (2466)
3403 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3404 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3405 - wake callback waiting threads on client eviction (2460)
3406 - Add --ptldebug and --subsystem to lmc (1719)
3407 - update assertion to allow safe interrupt allocation
3408 - set rq_no_resend for cancel requests (2432)
3409 - recalculate ptlrpcd timeout after resend (2494)
3410 - call vfs_rmdir when removing pending directories (2368)
3411 - fix renaming a file to itself (2429)
3412 - lmc creates a default one-stripe lov (2454)
3413 - expand procfs space to handle large clusters (2326)
3414 - increase UML stack to avoid overflow
3415 - update lconf's list of debug and subsystem masks
3416 - fix lfs find --obd (2510)
3417 - /proc tunable for disabling filter read caching (2591)
3418 - stop rpm packages from altering slapd.conf (2301)
3419 - disable nagle in the socknal under 0conf (2578)
3420 - choose mds inode size based on stripe count (2572)
3421 - fix kernel-source rpm problems (2516)
3423 - add --disable-doc to avoid pdf generation (2421)
3424 - update documentation, tests, type-os, comments
3425 - avoid format warnings on ia64
3426 - remove the TOE NAL
3427 - tiny code cleanups by removing unused fields
3429 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3432 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3433 - remove the most blatant lies from BUILDING file (2371)
3434 - change default debug level to reasonable production setting
3435 - reduce client side cache size to reduce cache flush time
3436 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3437 - make TCP zerocopy and pinger support enabled by default (2476)
3438 - sync writes completed after process exits caused crashes (2319)
3439 - maintain correct mount count on the MDS (2356)
3440 - backout 1557, because 2316 wasn't really fixed
3441 - better file I/O statistics gathering in /proc
3442 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3443 - another kernel patch to fix zero-copy TCP function export
3444 - don't take duplicate lock when processing re-sent getattr (2420)
3445 - lctl uses obd_self_export instead of creating new conn (2353)
3446 - MDS/OST recovery case which requires object creation asserted (2425)
3447 - move lfs from /usr/sbin to /usr/bin in packages
3448 - fix race between mds_client_add and mds_client_free (2417)
3449 - use kmalloc instead of slabs in portals (2430)
3450 - don't create duplicate records when a failover MDS is present (2442)
3451 - remove unnecessary mount age check (2332)
3452 - don't remove directory inodes from locks prematurely (2451)
3453 - don't break if MDS service name is the same as hostname (2103)
3454 - fix races in client write RPC generation when cache full (2482)
3456 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3459 - remove now-unused request->rq_obd (278)
3460 - if an allocation fails, print out how much memory we've used (1933)
3461 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3462 - add the "configurable stack size" patch to most series files (1256)
3463 - ability to write large log records, for 100+ OST configs (2306)
3464 - fix NULL deref when filter_prep fails (2314)
3465 - fix operator precedence error in filter_sync
3466 - dynamic allocation of socknal TX descriptors (2315)
3467 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3468 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3469 - ensure that CWARN messages in Portals always get to the syslog
3470 - __init/__exit are not for prototype decls (ldlm_init/exit)
3471 - x86-64 compile warning fixes
3472 - fix gateway LMC keyword conflict (2318)
3473 - fix MDS lock inversions in getattr/reint paths (1844)
3474 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3475 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3476 - DLM race condition prevented some lock evictions (2328)
3477 - ENOMEM detection and retry on socknal sends (2230)
3478 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3479 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3480 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3481 - better error messages when a client is rejected during recovery (1505)
3482 - avoid cancelling locks which were never granted, after failure (2330)
3483 - fix i_sem/journal inversion in mds_client_add (2333)
3484 - fix truncate/getattr lock cycle deadlock (2334)
3485 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3486 - fix two rare exit paths which could leak an l_lock() ref (2321)
3487 - fencepost error in MDS/OST orphan recovery (2226)
3488 - make log record alignment 8 bytes (1988)
3489 - lstripe now fails when requested offset > ost_count (2237)
3490 - ensure that all kernel series have a complete list.h (1607)
3491 - fix crashes in special-file operations (2316)
3492 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3494 - allow configurable automake binary, for testing new versions
3495 - small update to the lfs documentation
3497 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3499 * fix negative export reference count in fsfilt_sync (2312)
3501 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3502 * release candidate 0.9.1
3504 - orphans are moved into the PENDING directory for possible recovery
3505 - replayed opens now open by fid for orphan/rename safety (1042)
3506 - last close of an orphan inode generates a transno (683)
3507 - chdir() and mount() now pin the directory entry (1020)
3508 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3509 - discard very old requests without processing them (1502)
3510 - remove some common, well-understood CERRORs (1505)
3511 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3512 - clear "grant" flags in OST replies until OST grant code lands (1644)
3513 - fix read performance by not clobbering i_blksize on client (1598)
3514 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3515 - make LRU size a /proc tunable, clears locks when reduced (707)
3516 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3517 - server locks take references on exports now (1558)
3518 - build fixes for 2.4.20-rh trees (1663)
3519 - return an error from lov_create if all OSCs are inactive (1751)
3520 - fix import levels when a reconnect happens without a timeout (1597)
3521 - exit early from mds_open if we get a lookup error (1749)
3522 - partial page read at EOF wouldn't wait for disk before sending (1642)
3523 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3524 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3525 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3526 - write/truncate lock inversion (1639)
3527 - Don't auto-load obdclass, portals modules during cleanup (1495)
3528 - fix timestamps from jumping to "now" (1763)
3529 - extra journal assertions (1648)
3530 - add an extra multiunlink test (1771)
3531 - fix read_record/write_record API (1776)
3532 - fix leak of offset_extent, possible incorrect i_size later (1772)
3533 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3534 - Debugging for the kqswnal_get_idle_tx problems (1820)
3535 - Allow recovery to be attempted multiple times (1536)
3536 - Write out MDS last_rcvd file after it is first created (1600)
3537 - Fix tx_descriptor leak in failed transmit situations (1827)
3538 - ext3 journaling fixes for assertion failure after IO error (1871)
3539 - class_export_put() on freed export after completion AST error (1896)
3540 - Fix revalidate looping in VFS (1322)
3541 - Don't access a freed export during MDS_REINT timeout (1521)
3542 - Add open-unlink recovery support on the MDS (1673,1764)
3543 - Return an error if no MDS data was read from last_rcvd (1946)
3544 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3545 - Don't setup a disk device that doesn't match exported UUID (317)
3546 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3547 - avoid committing NULL handle in force close
3548 - local.sh is now a one-stripe LOV configuration
3549 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3550 - fix ext3 htree duplicate directory entry corruption (1516)
3551 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3552 - update to new LOV EA format (2097)
3553 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3555 2003-06-15 Phil Schwan <phil@clusterfs.com>
3558 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3559 - per-import/export recovery handling (958, 931, 959)
3560 - multiple last-rcvd slots, for serving multiple FSes (949)
3561 - connections are again shared between multiple imp/exports (963, 964)
3562 - "umount -f" would hang if any requests needed to be sent (393, 978)
3563 - avoid pinning large req buffer by copying for queued messages (989)
3564 - add "uuid" to "lctl device" command to help upcalls (991)
3565 - "open" RPCs with transnos would confuse recovery counters (1037)
3566 - do proper endian conversion of all wire messages (288, 340, 891)
3567 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3568 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3569 - fix ptlrpc_connection leak in target_handle_connect (1174)
3570 - fix import refcounting bug in OST and MDS cleanup (1134)
3571 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3572 - fix very unlikely obd_types race condition (501)
3573 - remove osc_open hack for echo_client (1187)
3574 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3575 - a failure in read_inode2 leads to deadlock (1139)
3576 - cancel ack-locks as soon as transaction is committed (1072)
3577 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3578 - make sure to commitrw after any preprw to avoid deadlock (1162)
3579 - failing to execute a file in a lustre FS would lock inode (1203)
3580 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3581 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3582 - obd_uuid2tgt would walk off the end of the list (1255)
3583 - on IA64 the osc would give portals incorrect bulk size (1258)
3584 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3585 - fix lock inversion caused by new llite matching code (1282)
3586 - limit the number of dirty pages on a client to 10MB (1286)
3587 - timed out locks were not being corrected cancelled (1289)
3588 - fix O_DIRECT above 4GB on IA-32 (1292)
3589 * major user-visible changes
3590 - fail out/fail over policy now controlled by the upcall (993)
3592 - add OBD_PING to check server availability and failure (954)
3593 - lustre messages are now sent in sending host order (288, 340, 891)
3594 - add eadatalen to MDS getattr reply (340)
3595 - OST read replies may contain second buffer, with per-page status (593)
3597 2003-03-11 Phil Schwan <phil@clusterfs.com>
3600 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3601 - failed open()s could cause deadlock; fixed (867, 869)
3602 - stop cancelling OST locks when files are closed (481)
3603 - overlapping XID spaces caused network corruption (851, 853)
3604 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3605 - setattr_raw added, to do single-RPC, server-side setattrs
3606 - lmc/lconf syntax change for OST UUIDs
3607 - fix crashy race condition between ptlrpc_free_req and osc_close
3608 - don't use request in mdc_enqueue if we hit a timeout (889)
3609 - don't set the inode i_size for regular files from the MDS (896)
3610 - handle out of order completion AST (842)
3611 - don't LBUG if a lock request times out after receiving AST (913)
3612 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3613 - if a bad lock AST arrives, send an error instead of dropping entirely
3614 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3615 - fix leak in bulk IO when only partially completed (899, 900, 926)
3616 - fix O_DIRECT for ia64 (55)
3617 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3618 - C-z after timeout could hang a process forever; fixed (977)
3620 - client-side I/O cache (678, 924, 929, 941, 970)
3622 - READPAGE and SETATTRs which don't take server-side locks get
3625 2003-02-11 Phil Schwan <phil@clusterfs.com>
3628 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3629 - Fix fsfilt_extN_readpage() to read a full page of directory
3630 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3631 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3632 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3633 - Most intent operations were removed, in favour of a new RPC mode
3634 that does a single RPC to the server and bypasses most of the VFS
3635 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3636 - Aggressively cancel local locks on DLM servers
3637 - mds_reint_unlink sends EA to the client if it's the last nlink.
3638 client uses that EA to unlink OST objects.
3639 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3640 - recursive symlinks were fixed (439)
3641 - fixed NULL deref in DEBUG_REQ
3642 - filter_update_lastobjid no longer calls sync, which annoyed extN
3643 - fixed multi-client small-writes to a single file problem (445)
3644 - fixed mtime updates during file writes (607)
3645 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3646 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3647 - fixed Linux OST size reporting problem (444, 656)
3648 - OST now updates object mtime with writes or setattr (607, 619)
3649 - client verifies file size before zeroing page past EOF (445)
3650 - OST now writes last allocated objid to disk with allocation (108)
3651 - LOV on echo now works (409)
3653 - mds_reint_unlink sends a new buffer, with the EA included. this
3654 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3655 set if a regular file was being unlinked, and it was the last link
3656 - use PtlGet from the target for bulk writes (315)
3657 - OST now updates object mtime with writes or setattr (607, 619)
3658 - LDLM now has a grant-time callback to revalidate locked items, if
3660 - Many MDS operations were reorganized to combat race conditions
3662 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3663 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3665 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3668 - Fully reactivate OST imports after reconnection (512, others)
3669 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3670 - More graceful error handling for truncating on dead OST (515)
3671 - Don't error out unless we're actually accessing dead stripes (474)
3672 - Fix garbage sizes when stripes are missing (410)
3673 - LRU counters were broken, causing constant lock purge (433, 432)
3674 - garbage on read from stripes with failed OSTs (441)
3675 - mark OSCs as active before reconnecting during recovery (438)
3676 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3677 - lfind did not preserve OST order in output (443)
3678 - symlinks cause hung clients, incorrect data (439)
3679 - stop dereferencing request after dropping refcount (457)
3680 - don't LASSERT(spin_is_locked) on non-SMP (455)
3681 - fixes for many rename() bugs
3682 - fstat didn't correctly synchronize attributes (399)
3683 - server must handle lock cancellation during blocking AST prep (487)
3684 - bulk descriptors were free()d too soon (511)
3685 - fix paths in lconf, which would load incorrect modules (451, 507)
3686 - fix confusing lconf 'host not found' error message (386)
3687 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3688 - fix race condition in mdc_blocking_ast() for inode access (526)
3689 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3690 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3691 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3692 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3693 - fix lstripe in several ways (kernel side) (527)
3694 - fix request leak in ldlm_cli_enqueue (262)
3695 - incorrect OSC was marked inactive after OST failure
3696 - call mds_fs_cleanup before unmounting filesystem (524)
3697 - fix races between taking ns_lock and ldlm_lock_change_resource
3698 - fix races updating LOV export open file list
3699 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3700 - fix recovery NULL deref in ldlm_cli_cancel_unused
3701 - fix some DLM races by using new hash table for lock handles (419)
3702 - permit the client to specify desired inodes, at replay
3703 - duplicate requests when we queue them for replay reintegration
3704 - fix last_rcvd offset calculation
3705 - sync after each recovered transaction, so we always make progress
3706 - never, not always, ERESTART requests without transnos
3707 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3709 - skip replay if the MDS says that the client is already connected
3710 - don't check for a recovery-enabled export to match lctl's UUID
3711 - don't INC_USE_COUNT for phantom exports
3712 - don't crash when cleaning up phantom exports (567)
3713 - don't double-finish or set replay data for errored mdc_open requests
3714 - abort requests when they time out, so we don't get old replies
3715 - send/receive replies for AST messages again
3716 - if the client says that it doesn't have the lock, cancel it on the
3718 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3719 mark it as destroyed, it will all work out when decref is called
3720 - fix module use counts (22, 581)
3722 - ASTs now expect a reply (server cancels lock on error reply)
3724 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3727 - fix many simultaneous client startup (392)
3728 - fix dentry->d_it clobbering
3729 - credentials weren't being shipped for readdir/getattr operations
3730 - remove invalid assertions triggered during some concurrent MD
3732 - proper Lustre versions added (336, 389)
3733 - fix memory leak for create error case (398)
3734 - fix LOV locking bug that would get cli/srv out of sync
3735 - fix echo client over LOV (409)
3736 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3737 - fix double-O_EXCL intent crash (424)
3738 - avoid sending multiple lock CANCELs (352)
3740 - MDS can do multi-client recovery (modulo bugs in new code)
3742 - many updates, edits, cleanups
3744 2002-11-18 Phil Schwan <phil@clusterfs.com>
3747 - fix null d_it dereference (346)
3748 - fix full OST/dbench hang (333)
3749 - fix permission problem with file removal (286)
3750 - fix removal of OSCs from LOV when they fail
3751 - fix NULL deref during bulk timeout (214)
3752 - fix problems related to multiple filesystems on one MDS (241)
3753 - fixed serious subtle metadata locking bugs
3754 - free locks on clients when inodes are removed due to memory
3756 - fix inode pointer in lock data (285)
3757 - partial support for multiple MDS on a single host (241)
3758 - data locks weren't cancelled at clear_inode time (290, 311)
3759 - intent locks could lead to unbounded lock growth (205)
3760 - added a maximum lock count, an LRU list, and a flusher
3761 - fix multiple rename (365)
3762 - properly abstracted the echo client
3763 - OSC locked 1 byte too many; fixed
3764 - rewrote brw callback code:
3765 - fixed recovery bugs related to LOVs (306)
3766 - fixed too-many-pages-in-one-write crash (191)
3767 - fixed (again) crash in sync_io_timeout (214)
3768 - probably fixed callback-related race (385)
3770 - Add capability to MDS protocol
3771 - LDLM cancellations and callbacks on different portals
3773 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3776 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3777 - defer open object destruction to close time (601981, 138)
3778 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3779 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3780 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3781 - don't call obd_setattr in ll_file_release for destroyed objects
3782 * protocol change to lustre_msg: move |version| and add |flags|
3783 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3784 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3785 * added replay of create, unlink, link and rename operations during
3786 MDS failover; recovery should be much more robust now
3787 * remove failed OSCs from LOVs (only lov_create uses this so far)
3788 * the lustre-HOWTO was brought (more) up to date (582544)
3790 2002-10-23 Phil Schwan <phil@clusterfs.com>
3793 - in-use dentries weren't being reused properly (617851)
3794 - prevent multiple LDLM setup (599178)
3795 - fix LOV size calculations for truncate (617853)
3796 - fix client handling of MDS intent errors (POSIX)
3797 - fix permission bug in lovstripe.c test (624321)
3798 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3799 - truncate past end of file could corrupt data
3800 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3801 - a race in recovery could return ETIMEDOUT to apps (623947)
3802 - building outside the source directory was fixed
3803 * the lustre-HOWTO was brought (more) up to date (582544)
3804 * major progress was made on recovery functionality
3806 2002-10-10 Phil Schwan <phil@clusterfs.com>
3809 - recovery deadlock fix
3810 - rm -rf causes LBUG fix (617817)
3811 - file open by multiple tasks fix (618962)
3812 - directory permissions bugs (602707 and 620007)
3813 - journal_stop fixed with locking (611313)
3814 - O_APPEND failures resolved (618273, perhaps 614459)
3815 - lconf PATH fix (619770)
3816 - IA64 build fix (621450)
3817 - RPC buffer sizes scale with amount of memory
3819 2002-10-01 Phil Schwan <phil@clusterfs.com>
3822 - locks would be cancelled without throwing away data pages,
3823 resulting in inconsistent data (605627)
3824 - inode attributes were not always being refreshed (605627, 612449)
3825 - lconf now continues to cleanup after lctl reports an error
3826 - MDS now enforces user permissions (602707)
3827 - lprocfs cleanup fixed, but not yet enabled (614157)
3828 - fixed infinite server hang, should a client not respond to an AST
3829 - avoid going into recovery if user calls readlink() with a buffer
3830 that's too small (613941)
3831 - AST RPCs no longer require replies (614867) -- this may be changed
3832 - don't crash server if client sends an IOV that's too big (611336)
3833 - fixed lock conversion deadlock (611892)
3834 - fixed the following of symlinks (614622)
3835 * recovery: the server can remove locks from a client that dies, other
3836 clients can make progress
3837 * more extN patch fixes
3838 * compile-time configurable ptlrpc buffer allocations
3840 - collaborative read cache document
3841 - Lustre Lite Performance CDR document-in-progress
3843 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3846 - fix typo in patch-2.4.18
3848 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3851 - clear ptlrpc request each time in handle_incoming_request()
3852 - unlink of files now destroys the object on the OST
3854 2002-09-19 Peter Braam <braam@clusterfs.com>
3856 * add hard link support
3857 * change obdfile creation method
3858 * kernel patch changed
3860 2002-09-19 Peter Braam <braam@clusterfs.com>
3863 - stack overflow bug in extN fixed
3865 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3867 * documentation updates
3868 - add man pages for config tools
3869 - update tests/README to describe testing with new config tools
3870 - finish metadata API descriptions
3871 * bug fixes and cleanups
3872 - statfs workaround for 16TB limit
3873 - LOV stripe allocation improved, can stripe on subset of OSTs
3874 - LOV file size/IO offset was wrong for files > 4GB in size
3875 - object EA data was being dropped, caused files to be unreadable
3876 - memory overflow with non-LOV OST caused memory corruption
3877 - fixed regression tests to work with new config tools, obdfilter
3878 - fixed bug when directory size became larger than 1 block
3879 - fixed bug (for single client case) when PWD was deleted
3880 - invalidate local directory pages when doing intent-based ops
3881 - avoid LDLM oops when lock callback contained bad data
3883 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3885 * documentation updates
3886 * bug fixes and cleanups
3887 - configuration tools
3890 - 64-bit compile warnings
3891 - 64-bit internal statfs data
3893 * test_brw on persistent OST devices
3895 * lprocfs (disabled)
3897 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3899 * documentation updates
3900 * bug fixes and cleanups
3901 * configuration tools
3903 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3906 * many small fixes to 0_5_4
3907 * io/network handling
3908 * thinkos in MDS operations
3910 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3913 * crucial basic fixes to 0.5.3
3914 * IOR, Iozone work over Elan
3917 2002-08-07 Phil Schwan <phil@clusterfs.com>
3918 * version 0_5_3, our first alpha
3919 * we use the new Portals iovs
3920 * documentation updates
3921 * bug fixes and cleanups
3922 * small changes in the DLM wire protocol
3924 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3925 * version 0_5_1 with some initial stability,
3926 * locking on MD and file I/O.
3927 * documentation updates
3928 * several bug fixes since 0.5.0
3929 * small changes in wire protocol
3931 2002-07-18 Phil Schwan <phil@clusterfs.com>
3933 * delivered as Lustre Light Alpha
3934 * fixed a crash after handling invalid MDS requests
3935 * fixed directory pages for architectures with non-4k pages sizes
3937 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3938 * release version v0_4_4
3939 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3941 * Fixes a number of interruption problems with OST operations.
3942 * Update documentation for portals header changes
3943 * Move all wire protocol structs/defines to lustre_idl.h
3944 * Fixes symlink length bug.
3945 * Add tcpdump to repository.
3947 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3948 * release version v0_4_3
3949 * Fixes statfs for inodes on extN.
3950 * Fixes bug in runtests which would delete /etc/hosts.
3951 * Use 64-bit object IDs wherever possible (not into VFS though)
3952 Remove ost_get_info, which is unused by lustre, and out of date.
3954 2002-07-03 Peter Braam <braam@clusterfs.com>
3955 * release version v0_4_2 Fixes a lookup error (type not passed)
3956 * move forward to head of Portals
3957 * move forward to latest Lustre kernel
3959 2002-06-25 Peter Braam <braam@clusterfs.com>
3960 * release version v0_4_1. Hopefully stable on single node use.