5 2.6.16.46-0.14 (SLES 10),
6 2.6.18-8.1.8.EL5 (RHEL 5).
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: Fix errors in lfs documentation
18 Details : Fixes man pages
22 Description: don't use GFP_* in generic Lustre code.
23 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
27 Description: obdclass is limited by single OBD_ALLOC(idarray)
28 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
32 Description: warnings with build patchless client with vanila 2.6.19 and up
33 Details : change old ctl_table style and replace ctl_table/ctl_table_header
34 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
38 Frequency : Only for SLES
39 Description: sanity_quota fail test_1
40 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
41 file, which makes TSTID[2] inunique.
45 Frequency : Only for SLES9
46 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
47 Details : Update target/ChangeLog/which_patch .
51 Frequency : Only for test_10 in sanity-quota.sh
52 Description: a bug in quota test code
53 Details : checking a test flag in a wrong place
57 Frequency : Only for test_13 in sanity-quota.sh
58 Description: a bug in quota test script
59 Details : shouldn't check overfull allocated quota
63 Description: jbd statistics
64 Details : Port older jbd statistics patch for sles10
66 Severity : enhancement
68 Description: Add printing inode info into message about error in writepage.
73 Description: reply_lock_interpret crash due to race with it and lock cancel.
74 Details : Do not replay locks that are being cancelled. Do not reference
75 locks by their address during replay, just by their handle.
79 Frequency : When flocks are used.
80 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
81 Details : Flock locks might destroy just granted lock if it could be merged
82 with another existing flock, this is done in completion handler,
83 so teach ldlm_cli_enquque_fini that this is a valid case for
89 Description: Oops in read and write path when failing to allocate lock.
90 Details : Check if lock allocation failed and return error back.
94 Description: lstripe command fails for valid OST index
95 Details : The stripe offset is compared to lov->desc.ld_tgt_count
96 instead of lov->desc.ld_active_tgt_count.
100 Description: sanity.sh failed test 103
101 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
106 Description: df doesn't work properly if diskfs blocksize != 4K
107 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
111 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
112 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
116 Description: obd_health_check_timeout too short
117 Details : set obd_health_check_timeout as 1.5x of obd_timeout
121 Description: llapi_file_create() does not allow some changes
122 Details : add llapi_file_open() that allows specifying the mode and
123 open flags, and also returns an open file handle.
127 Description: merge and cleanup kernel patches.
128 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
132 Description: Add journal checksum support.(Kernel part)
133 Details : The journal checksum feature adds two new flags i.e
134 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
135 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
136 indicates that the commit block contains the checksum for
137 the blocks described by the descriptor blocks. Now commit
138 record can be sent to disk without waiting for descriptor
139 blocks to be written to disk. This behavior is controlled
140 using JBD2_FEATURE_ASYNC_COMMIT flag.
144 Description: OSS needs mutliple precreate threads
145 Details : Add ability to start more than one create thread per OSS.
149 Description: Sanity.sh test_65k failed: setstripe should have succeeded
150 Details : execute lfs setstripe on client
154 Description: mds_obd_create error creating tmp object
155 Details : When the user sets quota on root, llog will be affected and can't
156 create files and write files.
159 Frequency : Always on ia64 patchless client, and possibly others.
161 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
162 Details : This allows the patchless client to be loaded on architectures
167 Description: RedHat Update kernel for RHEL5
168 Details : Add the kernel config file for RHEL5.
172 Description: RedHat Update kernel for RHEL5
173 Details : Modify the kernel config file more closer RHEL5.
177 Description: Build failure against Centos5 (RHEL5)
178 Details : Define PAGE_SIZE when it isn't present.
182 Description: test_7 on sanity-quota.sh can't be used on separate machines
183 Details : add facet to handle it
187 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
188 Details : with NFS, the anon dentry's parent was set to itself in
189 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
190 corresponding dentry other than use rec->ur_name.
192 Severity : enhancement
194 Description: lfs setstripe enhancement
195 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
199 Description: testing performance impact of enabling checksumming
200 Details : enable checksum by default, allow --disable-checksum
201 configure option and "-o nochecksum" mount option
205 Description: lustre support for RHEL5
206 Details : Add support for RHEL5.
210 Description: Client eviction due to failover config
211 Details : after a connection loss, the lustre client should attempt to
212 reconnect to the last active server first before trying the
213 other potential connections.
217 Description: when mds and osts use different quota unit(32bit and 64bit),
218 quota will be released repeatly.
219 Details : void sending multiple quota reqs to mds, which will keep the
220 status between the reqs.
224 Description: osts not allocated evenly to files
225 Details : change the condition to increase offset_idx
228 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
230 Description: Data corruption for OSTs that are formatted larger than 2TB
232 Details : When generating the bio request for lustre file writes the
233 sector number would overflow a temporary variable before being
234 used for the IO. The data reads correctly from Lustre (which
235 will overflow in a similar manner) but other file data or
236 filesystem metadata may be corrupted in some cases.
240 Description: Tune the kernel for good SCSI performance.
241 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
242 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
247 Description: rewrite ext2-derived code in obdclass/uuid.c
248 Details : rewrite inherited code (uuid parsing code from ext2 utils)
249 from scratch preserving functionality.
253 Description: "lfs find -obd UUID" prints directories
254 Details : "lfs find -obd UUID" will return all directory names instead
255 of just file names. It is incorrect because the directories
256 do not reside on the OSTs.
260 Description: use i_size_read and i_size_write in 2.6 port
261 Details : replace inode->i_size access with i_size_read/write()
264 Frequency : when removing large files
266 Description: scheduling issue during removal of large Lustre files
267 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
268 It causes scheduling issues when removing large files (17TB in the
272 Frequency : only with liblustre clients on XT3
274 Description: evictions taking too long
275 Details : allow llrd to evict clients directly on OSTs
277 --------------------------------------------------------------------------------
279 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
281 * Support for kernels:
282 2.6.5-7.283 (SLES 9),
283 2.6.9-55.EL (RHEL 4),
284 2.6.16.46-0.14 (SLES 10),
285 2.6.18.8 vanilla (kernel.org)
286 * Client support for unpatched kernels:
287 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
288 2.6.16 - 2.6.22 vanilla (kernel.org)
289 * Due to recently discovered recovery problems, we do not recommend
290 using patchless RHEL 4 clients with this or any earlier release.
291 * Recommended e2fsprogs version: 1.39.cfs8
292 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
293 * Starting with this release, the ldiskfs backing filesystem required
294 by Lustre is now in its own package, lustre-ldiskfs. This package
295 should be installed. It is versioned separately from Lustre and
296 may be released separately in future.
300 Description: block reactivating mgc import until all deactivates complete
301 Details : Fix race when failing back MDT/MGS to itself (testing)
303 Severity : enhancement
305 Description: add optional extra BUILD_VERSION info
306 Details : add a new environment variable (namely LUSTRE_VERS) which allows
307 to override the lustre version.
310 Frequency : 2.6.18 servers only
312 Description: ll_kern_mount() doesn't release the module reference
313 Details : The ldiskfs module reference count never drops down to 0
314 because ll_kern_mount() doesn't release the module reference.
319 Description: server LBUG when using old ost_num_threads parameter
320 Details : Accept the old ost_num_threads parameter but warn that it
321 is deprecated, and fix an off-by-one error that caused an LBUG.
326 Description: Transient SCSI error results in persistent IO issue
327 Details : iobuf->dr_error is not reinitialized to 0 between two
331 Frequency : sometimes when underlying device returns I/O errors
333 Description: OSTs not going read-only during write failures
334 Details : OSTs are not remounted read-only when the journal commit threads
335 get I/O errors because fsfilt_ext3 calls journal_start/stop()
336 instead of the ext3 wrappers.
340 Description: poor connect scaling with increasing client count
341 Details : Don't run filter_grant_sanity_check for more than 100 exports
342 to improve scaling for large numbers of clients.
345 Frequency : SLES10 only
347 Description: sanity-quota.sh quotacheck failed: rc = -22
348 Details : Quotas cannot be enabled on SLES10.
351 Frequency : liblustre clients only
353 Description: getdirentries does not give error when run on compute nodes
354 Details : getdirentries does not fail when the size specified as an argument
355 is too small to contain at least one entry
357 Severity : enhancement
359 Description: Add LNET router traceability for debug purposes
360 Details : If a checksum failure occurs with a router as part of the
361 IO path, the NID of the last router that forwarded the bulk data
362 is printed so it can be identified.
367 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
368 Details : Due to a race condition, liblustre clients were occasionally
371 Severity : enhancement
373 Description: lfs setstripe use optional parameters instead of postional
376 Severity : enhancement
378 Description: Nanosecond timestamp support for ldiskfs
379 Details : The on-disk ldiskfs filesystem has added support for nanosecond
380 resolution timestamps. There is not yet support for this at
381 the Lustre filesystem level.
384 Frequency : during server recovery
386 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
387 Details : request with rq_no_resend flag not awake l_wait_event if they get a
391 Frequency : nfs export on patchless client
393 Description: connectathon hang when test nfs export over patchless client
394 Details : Disconnected dentry cannot be found with lookup, so we do not need
395 to unhash it or make it invalid
398 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
399 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
400 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
405 Description: open req refcounting wrong on reconnect
406 Details : If reconnect happened between getting open reply from server and
407 call to mdc_set_replay_data in ll_file_open, we will schedule
408 replay for unreferenced request that we are about to free.
409 Subsequent close will crash in variety of ways.
410 Check that request is still eligible for replay in
411 mdc_set_replay_data().
416 Description: disable writes to filesystem when reading health_check file
417 Details : the default for reading the health_check proc file has changed
418 to NOT do a journal transaction and write to disk, because this
419 can cause reads of the /proc file to hang and block HA state
420 checking on a healthy but otherwise heavily loaded system. It
421 is possible to return to the previous behaviour during configure
422 with --enable-health-write.
424 Severity : enhancement
426 Description: 64-bit inode version
427 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
428 to the inode. This will be required for version-based recovery.
433 Description: MDS fails to start if a duplicate client export is detected
434 Details : in some rare cases it was possible for a client to connect to
435 an MDS multiple times. Upon recovery the MDS would detect this
436 and fail during startup. Handle this more gracefully.
438 Severity : enhancement
440 Description: Add -o localflock option to simulate old noflock
442 Details : This will achieve local-only flock/fcntl locks
448 Description: log_commit_thread vs filter_destroy race leads to crash
449 Details : Take import reference before releasing llog record semaphore
454 Description: Wrong request locking in request set processing
455 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
456 imp_delayed_list, in this place should be used imp_lock.
459 Frequency : when reconnection
461 Description: Grant Leak when osc reconnect to OST
462 Details : When osc reconnect ost, OST(filter) should check whether it
463 should grant more space to client by comparing fed_grant and
464 cl_avail_grant, and return the granted space to client instead
465 of "new granted" space, because client will call osc_init_grant
466 to update the client grant space info.
469 Frequency : when client reconnect to OST
471 Description: Grant Leak when osc do resend and replay bulk write
472 Details : When osc reconnect to OST, OST(filter)should clear grant info of
473 bulk write request, because the grant info while be sync between
474 OSC and OST when reconnect, and we should ignore the grant info
475 these of resend/replay write req.
480 Description: Grant space more than avaiable left space sometimes.
481 Details : When then OST is about to be full, if two bulk writing from
482 different clients came to OST. Accord the avaliable space of the
483 OST, the first req should be permitted, and the second one
484 should be denied by ENOSPC. But if the seconde arrived before
485 the first one is commited. The OST might wrongly permit second
486 writing, which will cause grant space > avaiable space.
489 Frequency : when client is evicted
491 Description: Grant might be wrongly erased when osc is evicted by OST
492 Details : when the import is evicted by server, it will fork another
493 thread ptlrpc_invalidate_import_thread to invalidate the
494 import, where the grant will be set to 0. While the original
495 thread will update the grant it got when connecting. So if
496 the former happened latter, the grant will be wrongly errased
497 because of this race.
502 Description: Checking Stale with correct fid
503 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
504 to check whether it is stale, because sometimes, we want the
505 enqueue happened anyway, and op_data.fid2 will not be initialized.
507 Severity : enhancement
509 Description: update patchless client
510 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
513 Frequency : only with 2.4 kernel
515 Description: random memory corruption
516 Details : size of struct ll_inode_info is to big for union inode.u and this
517 can be cause of random memory corruption.
522 Description: Memory leak in recovery
523 Details : Lov_mds_md was not free in an error handler in mds_create_object.
524 It should also check obd_fail before fsfilt_start, otherwise if
525 fsfilt_start return -EROFS,(failover mds during mds recovery).
526 then the req will return with repmsg->transno = 0 and rc = EROFS.
527 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
528 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
529 be freed no matter whether fsfilt_commit success or not.
532 Frequency : only with huge count clients
534 Description: Prevents from taking the superblock lock in llap_from_page for
536 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
537 for llap_from_page call in ll_removepage prevents from taking the
538 superblock lock for a soon died page.
543 Description: Not check open intent error before release open handle
544 Details : in some rare cases, the open intent error is not checked before
545 release open handle, which may cause
546 ASSERTION(open_req->rq_transno != 0), because it tries to release
547 the failed open handle.
552 Description: Set cat log bitmap only after create log success.
553 Details : in some rare cases, the cat log bitmap is set too early. and it
554 should be set only after create log success.
558 Description: Accessing a block bevice can re-enable I/O when Lustre is
559 tearing down a device.
560 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
565 Description: service threads may hog cpus when there are a lot of requests
567 Details : Insert cond_resched to give other threads a chance to use some of
573 Description: the cat log was not initialized in recovery
574 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
575 unlink log on mds will not be initialized until mds post
576 recovery. And also in mds post recovery, the unlink log will
577 initialization will be done asynchronausly, so there will be race
578 between add unlink log and unlink log initialization.
582 Description: brw_stats were being printed incorrectly
583 Details : brw_stats were being printed as log2 but all of them were not
584 recorded as log2. Also remove some code duplication arising from
585 filter_tally_{read,write}.
589 Frequency : rare, only in recovery.
590 Description: ASSERTION(req->rq_type != LI_POISON) failed
591 Details : imp_lock should be held while iterating over imp_sending_list for
592 prevent destroy request after get timeout in ptlrpc_queue_wait.
596 Description: replay-single.sh test 52 fails
597 Details : A lock's skiplist need to be cleanup when it being unlinked
598 from its resource list.
602 Description: Short directio read returns full requested size rather than
604 Details : Direct I/O operations should return actual amount of bytes
605 transferred rather than requested size.
607 Severity : enhancement
609 Description: metadata RPC reduction (e.g. for rm performance)
610 Details : decrease the amount of synchronous RPC between clients and servers
611 by canceling conflicing lock before the operation on the client side
612 and packing thier handles into the main operation RPC to server.
614 Severity : enhancement
616 Description: Async OSC create to avoid the blocking unnecessarily.
617 Details : If a OST has no remain object, system will block on the creating
618 when need to create a new object on this OST. Now, ways use
619 pre-created objects when available, instead of blocking on an
620 empty osc while others are not empty. If we must block, we block
621 for the shortest possible period of time.
623 --------------------------------------------------------------------------------
625 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
630 Frequency : on some architectures
632 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
633 Details : Uninitialized flags sometimes cause configuration commands to
637 Frequency : patchless clients only
639 Description: missing __iget() symbol export
640 Details : The __iget() symbol export is missing. To avoid the need for
641 this on patchless clients the deathrow inode reaper is turned
642 off, and we depend on the VM to clean up old inodes. This
643 dependency was during via the fix for bug 12181.
645 --------------------------------------------------------------------------------
647 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
649 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
650 INTEROPERATE with older versions automatically. In many cases a
651 special upgrade step is needed. Please read the
652 user documentation before upgrading any part of a live system.
653 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
654 version will not interoperate with 1.6 betas before beta5 (1.5.95).
655 * WARNING: Lustre configuration and startup changes are required with
656 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
662 Severity : enhancement
664 Description: MountConf
665 Details : Lustre configuration is now managed via mkfs and mount
666 commands instead of lmc and lconf. New obd types (MGS, MGC)
667 are added for dynamic configuration management. See
668 https://mail.clusterfs.com/wikis/lustre/MountConf for
671 Severity : enhancement
673 Description: dynamic OST addition
674 Details : OSTs can now be added to a live filesystem
676 Severity : enhancement
678 Description: startup order invariance
679 Details : MDTs and OSTs can be started in any order. Clients only
680 require the MDT to complete startup.
682 Severity : enhancement
684 Description: parallel, asynchronous orphan cleanup
685 Details : orphan cleanup is now performed in separate threads for each
686 OST, allowing parallel non-blocking operation.
688 Severity : enhancement
690 Description: optimized stripe assignment
691 Details : stripe assignments are now made based on ost space available,
692 ost previous usage, and OSS previous usage, in order to try
693 to optimize storage space and networking resources.
695 Severity : enhancement
697 Description: Permanently set tunables
698 Details : All writable /proc/fs/lustre tunables can now be permanently
699 set on a per-server basis, at mkfs time or on a live
702 Severity : enhancement
704 Description: Lustre message v2
705 Details : Add lustre message format v2.
707 Severity : enhancement
709 Description: client OST exclusion list
710 Details : Clients can be started with a list of OSTs that should be
711 declared "inactive" for known non-responsive OSTs.
715 Description: SPEC SFS validation failure on NFS v2 over lustre.
716 Details : Changes the blocksize for regular files to be 2x RPC size,
717 and not depend on stripe size.
719 Severity : enhancement
721 Description: Multiple MD RPCs in flight.
722 Details : Further unserialise some read-only MDS RPCs - learn about intents.
723 To avoid overly-overloading MDS, introduce a limit on number of
724 MDS RPCs in flight for a single client and add /proc controls
725 to adjust this limit.
727 Severity : enhancement
729 Description: client read/write statistics
730 Details : Add client read/write call usage stats for performance
731 analysis of user processes.
732 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
733 file access. extents_stats shows chunk size distribution.
734 extents_stats_per_process show chunk size distribution per
737 Severity : enhancement
739 Description: mds statistics
740 Details : Add detailed mds operations statistics in
741 /proc/fs/lustre/mds/*/stats.
745 Description: Failure of copying files with lustre special EAs.
746 Details : Client side always return success for setxattr call for lustre
747 special xattr (currently only "trusted.lov").
752 Description: Refcount LNET uuids
753 Details : The global LNET uuid list grew linearly with every startup;
754 refcount repeated list entries instead of always adding to
757 Severity : enhancement
759 Description: Dynamic service threads
760 Details : Within a small range, start extra service threads
761 automatically when the request queue builds up.
764 Frequency : mixed-endian client/server environments
766 Description: mixed-endian crashes
767 Details : The new msg_v2 system had some failures in mixed-endian
770 Severity : enhancement
772 Description: Easy OST removal
773 Details : OSTs can be permanently deactivated with e.g. 'lctl
774 conf_param lustre-OST0001.osc.active=0'
776 Severity : enhancement
778 Description: MGS proc entries
779 Details : Added basic proc entries for the MGS showing what filesystems
782 Severity : enhancement
784 Description: provide MGS failover
785 Details : Added config lock reacquisition after MGS server failover.
787 Severity : enhancement
789 Description: add Linux 2.4 support
790 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
794 Description: a large application tries to do I/O to the same resource and dies
796 Details : Check the req->rq_arrival time after the call to
797 ost_brw_lock_get(), but before we do anything about
798 processing it & sending the BULK transfer request. This
799 should help move old stale pending locks off the queue as
800 quickly as obd_timeout.
803 Frequency : when an incorrect nid is specified during startup
805 Description: ptlrpc connect to non-existant node causes kernel crash
806 Details : LNET can't be re-entered from an event callback, which
807 happened when we expire a message after the export has been
808 cleaned up. Instead, hand the zombie cleanup off to another
811 Severity : enhancement
813 Description: plain/inodebits lock performance improvement
814 Details : Grouping plain/inodebits in granted list by their request modes
815 and bits policy, thus improving the performance of search through
819 Frequency : only if OST filesystem is corrupted
821 Description: client incorrectly hits assertion in ptlrpc_replay_req()
822 Details : for a short time RPCs with bulk IO are in the replay list,
823 but replay of bulk IOs is unimplemented. If the OST filesystem
824 is corrupted due to disk cache incoherency and then replay is
825 started it is possible to trip an assertion. Avoid putting
826 committed RPCs into the replay list at all to avoid this issue.
829 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
831 Description: System hang on startup
832 Details : This bug allowed the liblustre (e.g. catamount) client to
833 return to the app before handling all startup RPCs. This
834 could leave the node unresponsive to lustre network traffic
835 and manifested as a server ptllnd timeout.
837 Severity : enhancement
839 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
840 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
841 boolean to control whether to print peer debug info when a
842 client's RPC times out.
845 Frequency : only for kernels with patches from Lustre below 1.4.3
847 Description: Remove old rdonly API
848 Details : Remove old rdonly API which unused from at least lustre 1.4.3
851 Frequency : only for devices with external journals
853 Description: Set external device read-only also
854 Details : During a commanded failover stop, we set the disk device
855 read-only while the server shuts down. We now also set any
856 external journal device read-only at the same time.
859 Frequency : when upgrading from 1.4 while trying to change parameters
861 Description: The wrong (new) MDC name was used when setting parameters for
862 upgraded MDT's. Also allows changing of OSC (and MDC)
863 parameters if --writeconf is specified at tunefs upgrade time.
866 Frequency : when setting specific ost indicies
868 Description: QOS code breaks on skipped indicies
869 Details : Add checks for missing OST indicies in the QOS code, so OSTs
870 created with --index need not be sequential.
872 Severity : enhancement
874 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
875 Details : The uninit_groups feature works in conjunction with the kernel
876 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
877 up the pass1 processing of e2fsck. This is a read-only feature
878 in ldiskfs2 only, so older kernels and current ldiskfs cannot
879 mount filesystems that have had this feature enabled.
881 Severity : enhancement
883 Description: Improve multi-block allocation algorithm to avoid fragmentation
884 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
885 allocation locality and avoid filesystem fragmentation.
887 ------------------------------------------------------------------------------
889 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
891 * Support for kernels:
892 2.4.21-47.0.1.EL (RHEL 3)
894 2.6.9-42.0.10.EL (RHEL 4)
895 2.6.12.6 vanilla (kernel.org)
896 2.6.16.27-0.9 (SLES 10)
897 * Recommended e2fsprogs version: 1.39.cfs5
899 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
903 Frequency : occasional, depends on client load and configuration
904 Bugzilla : 12181, 12203
905 Description: data loss for recently-modified files
907 Details : In some cases it is possible that recently written or created
908 files may not be written to disk in a timely manner (this should
909 normally be within 30s unless client IO load is very high).
910 The problem appears as zero-length files or files that are a
911 multiple of 1MB in size after a client crash or client eviction
912 that are missing data at the end of the file.
914 This problem is more likely to be hit on clients where files are
915 repeatedly created and unlinked in the same directory, clients
916 have a large amount of RAM, have many CPUs, the filesystem has
917 many OSTs, the clients are rebooted frequently, and/or the files
918 are not accessed by other nodes after being written.
920 The presence of the problem can be detected by looking at
921 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
922 smaller than the second (nr_unused) then dirty files will not
923 be flushed automatically to disk. "sync; sleep 10" should be
924 run several times on the node before unmounting it to update
925 Lustre (this is also safe to run on nodes without this problem).
927 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
928 that can cause this same problem, so customers using that kernel
929 also need to update the kernel in addition to Lustre. In order
930 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
932 It is normal that files written just before a client crash (less
933 than 30s) may not yet have been flushed to disk, even for local
937 Frequency : frequent on thin XT3 nodes
939 Description: UUID collision on thin XT3 Linux nodes
940 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
941 randomly, since we relied on an insufficiently-seeded PRNG.
946 Description: OSS hangs after "All ost request buffers busy"
947 Details : A deadlock between quota and journal operations caused OSS
948 hangs after printing "All ost request buffers busy."
951 Frequency : always on liblustre builds
953 Description: Cleanup compiler warnings on liblustre
956 Frequency : always on liblustre builds on XT3
958 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
962 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
963 > 32000 subdirectories
964 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
965 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
966 subdirectory count crosses 32000. This will aid e2fsck to
967 correctly handle more than 32000 subdirectories.
970 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
972 Description: System hang on startup
973 Details : This bug allowed the liblustre (e.g. catamount) client to
974 return to the app before handling all startup RPCs. This
975 could leave the node unresponsive to lustre network traffic
976 and manifested as a server ptllnd timeout.
978 Severity : enhancement
980 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
981 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
982 boolean to control whether to print peer debug info when a
983 client's RPC times out.
988 Description: make O_SYNC working on 2.6 kernels
989 Details : 2.6 kernels use different method for mark pages for write,
990 so need add a code to lustre for O_SYNC work.
995 Description: Failure to close file and release space on NFS
996 Details : Put inode details into lock acquired in ll_intent_file_open.
997 Use mdc_intent_lock in ll_intent_open to properly
998 detect all kind of errors unhandled by mdc_enqueue.
1003 Description: proc file read during shutdown sometimes raced obd removal,
1005 Details : Add lock to prevent obd access after proc file removal.
1008 Frequency : Only for files larger than 4GB on 32-bit clients.
1010 Description: improperly doing page alignment of locks
1011 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1012 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1017 Description: RPCs being resent when they shouldn't be
1018 Details : Some RPCs that should not be resent are being resent. This
1019 can cause inconsistencies in the RPC state machine. Do not
1020 resend such requests.
1023 Frequency : rare, only with NFS export
1025 Description: Crash on NFS re-export node
1026 Details : under very unusual load conditions an assertion is hit in
1027 ll_intent_file_open()
1030 Frequency : only if OST filesystem is corrupted
1032 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1033 Details : for a short time RPCs with bulk IO are in the replay list,
1034 but replay of bulk IOs is unimplemented. If the OST filesystem
1035 is corrupted due to disk cache incoherency and then replay is
1036 started it is possible to trip an assertion. Avoid putting
1037 committed RPCs into the replay list at all to avoid this issue.
1042 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1043 Details : Large single O_DIRECT read and write calls can fail to allocate
1044 a sufficiently large buffer to process the request. In case of
1045 allocation failure the allocation is retried with a smaller
1046 buffer and broken into smaller requests.
1048 Severity : enhancement
1050 Description: Add -o localflock option to simulate old noflock behaviour.
1051 Details : This will achieve local-only flock/fcntl locks coherentness.
1056 Description: versioning check is incomplete
1057 Details : Checking the version difference of client vs. server, report
1058 error if the gap is too big.
1063 Description: add support PG_writeback bit
1064 Details : add support for PG_writeback bit for Lustre, for more carefull
1065 work with page cache in 2.6 kernel. This also fix some deadlocks
1066 and remove hack for work O_SYNC with 2.6 kernel.
1068 Severity : enhancement
1070 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1071 Details : The uninit_groups feature works in conjunction with the kernel
1072 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1073 up the pass1 processing of e2fsck. This is a read-only feature
1074 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1075 mount filesystems that have had this feature enabled.
1077 Severity : enhancement
1079 Description: Improve multi-block allocation algorithm to avoid fragmentation
1080 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1081 allocation locality and avoid filesystem fragmentation.
1083 ------------------------------------------------------------------------------
1085 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1087 * Support for kernels:
1088 2.6.9-42.0.3.EL (RHEL 4)
1089 2.6.5-7.276 (SLES 9)
1090 2.4.21-47.0.1.EL (RHEL 3)
1091 2.6.12.6 vanilla (kernel.org)
1092 2.6.16.21-0.8 (SLES10)
1093 * Recommended e2fsprogs version: 1.39.cfs2-0
1095 * The backwards-compatible /proc/sys/portals symlink has been removed
1096 in this release. Before upgrading, please ensure that you change
1097 any configuration scripts or /etc/sysctl.conf files that access
1098 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1099 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1100 in advance of the upgrade on any system running Lustre 1.4.6 or
1101 newer, since /proc/sys/lnet was added in that version.
1102 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1106 Frequency : only when quota is used
1108 Description: avoid scanning export list for quota master
1109 Details : Change the algorithms to avoid scanning export list in order
1110 to improve the efficiency.
1113 Frequency : MDS failover only, very rarely
1115 Description: "went back in time" messages on mds failover
1116 Details : The greatest transno may be lost when the current operation
1117 finishes with an error (transno==0) and the client's last_rcvd
1118 record is over-written. Save the greatest transno in the
1119 mds_last_transno for this case.
1122 Frequency : always for specific kernels and striping counts
1124 Description: client may get "Matching packet too big" without ACL support
1125 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1126 when trying to access files in certain configurations. The
1127 clients should in fact be denied when mounting because they do
1128 not understand ACLs.
1131 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1133 Description: many clients connecting with IO in progress causes connect timeouts
1134 Details : Avoid synchronous journal commits to avoid delays caused by many
1135 clients connecting/disconnecting when bulk IO is in progress.
1136 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1137 OST_IO_PORTAL to avoid delays behind potentially many pending
1141 Frequency : occasionally with multiple writers to a single file
1143 Description: shared writes to file may result in wrong size reported by stat()
1144 Details : Allow growing of kms when extent lock is cancelled
1147 Frequency : always with random mmap IO to multi-striped file
1149 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1150 Details : Only if the hole is at the end of OST object so that kms is too
1151 small. Fix is to increase kms accordingly in ll_nopage.
1154 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1156 Description: writes to a missing object would leak memory on the OST
1157 Details : If there is an inconsistency between the MDS and OST filesystems,
1158 such that the MDS references an object that doesn't exist, writes
1159 to that object will leak memory due to incorrect cleanup in the
1160 error handling path, eventually running out of memory on the OST.
1165 Description: Creating too long symlink causes lustre errors
1166 Details : Check symlink and name lengths before sending requests to MDS.
1169 Frequency : only if flock is enabled (not on by default)
1171 Description: posix locks not released on fd closure on 2.6.9+
1172 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1173 kernels, this caused such locks not to be released on fd close and
1174 then assertions on fs unmount about still used locks.
1177 Frequency : MDS failover only, very rarely
1179 Description: clients may get ASSERTION(granted_lock != NULL)
1180 Details : When request was taking a long time, and a client was resending
1181 a getattr by name lock request. The were multiple lock requests
1182 with the same client lock handle and
1183 mds_getattr_name->fixup_handle_for_resent_request found one of the
1184 lock handles but later failed with ASSERTION(granted_lock != NULL).
1189 Description: handle->h_buffer_credits > 0, assertion failure
1190 Details : h_buffer_credits is zero after truncate, causing assertion
1191 failure. This patch extends the transaction or creates a new
1195 Frequency : NFS re-export or patchless client
1196 Bugzilla : 11179, 10796
1197 Description: Crash on NFS re-export node (__d_move)
1198 Details : We do not want to hash the dentry if we don't have a lock.
1199 But if this dentry is later used in d_move, we'd hit uninitialised
1200 list head d_hash, so we just do this to init d_hash field but
1201 leave dentry unhashed.
1204 Frequency : NFS re-export or patchless client
1206 Description: NFS exports has problem with symbolic link
1207 Details : lustre client didn't properly install dentry when re-exported
1208 to NFS or running patchless client.
1211 Frequency : NFS re-export or patchless client
1213 Description: Various nfs/patchless fixes.
1214 Details : fixes reuse disconected alias for lookup process - this fixes
1215 warning "find_exported_dentry: npd != pd",
1216 fix permission error with open files at nfs.
1217 fix apply umask when do revalidate.
1220 Frequency : occasional
1222 Description: Crash on NFS re-export node
1223 Details : calling clear_page() on the wrong pointer triggered oops in
1224 generic_mapping_read().
1227 Frequency : rarely, using O_DIRECT IO
1229 Description: unaligned directio crashes client with LASSERT
1230 Details : check for unaligned buffers before trying any requests.
1233 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1235 Description: stale data returned from RAID cache
1236 Details : If only a small amount of IO is done to the RAID device before
1237 reading it again it is possible to get stale data from the RAID
1238 cache instead of reading it from disk.
1241 Frequency : always for sles10 kernel
1243 Description: sles10 support
1244 Details : ll_follow_link: compile fixes and using of nd_set_link
1245 under newer kernels.
1248 Frequency : depends on arch, kernel and compiler version, always on sles10
1251 Description: recursive or deep enough symlinks cause stack overflow
1252 Details : getting rid of large stack-allocated variable in
1256 Frequency : depends on hardware
1258 Description: lustre write performance loss in the SLES10 kernel
1259 Details : the performance loss is caused by using of write barriers in the
1260 ext3 code. The SLES10 kernel turns barrier support on by
1261 default. The fix is to undo that change for ldiskfs.
1263 ------------------------------------------------------------------------------
1265 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1267 * Support for kernels:
1268 2.6.9-42.0.3EL (RHEL 4)
1269 2.6.5-7.276 (SLES 9)
1270 2.4.21-47.0.1.EL (RHEL 3)
1271 2.6.12.6 vanilla (kernel.org)
1275 Frequency : quota enabled and large files being deleted
1277 Description: releasing more than 4GB of quota at once hangs OST
1278 Details : If a user deletes more than 4GB of files on a single OST it
1279 will cause the OST to spin in an infinite loop. Release
1280 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1285 Description: statfs data retrieved from /proc may be stale or zero
1286 Details : When reading per-device statfs data from /proc, in the
1287 {kbytes,files}_{total,free,avail} files, it may appear
1288 as zero or be out of date.
1291 Frequency : systems with MD RAID1 external journal devices
1293 Description: lconf's call to blkid is confused by RAID1 journal devices
1294 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1295 of returning all block devices that match the journal UUID.
1298 Frequency : always, for aggregate stripe size over 4GB
1300 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1301 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1302 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1305 Frequency : always if "lfs find" used on a local file/directory
1307 Description: "lfs find" segfaults if used on a local file/directory
1308 Details : The case where a directory component was not specified wasn't
1309 handled correctly. Handle this properly.
1312 Frequency : always on ppc64
1314 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1315 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1316 perform endianness conversion.
1319 Frequency : rarely (truncate to non-zero file size after write under load)
1320 Bugzilla : 10730, 10687
1321 Description: Files padded with zeros to next 4K multiple
1322 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1323 it is possible that files that are truncated to a non-zero size
1324 immediately after being written are filled with zero bytes beyond
1325 the truncated size. No file data is lost.
1327 Severity : enhancement
1329 Description: Allow recovery/failover for liblustre clients.
1330 Details : liblustre clients were unaware of failover configurations until
1333 Severity : enhancement
1335 Description: user file locks should fail when not mounting with flock option
1336 Details : Set up an error-returning stub in ll_file_operations.lock field
1337 to prevent incorrect behaviour when client is mounted without
1338 flock option. Also, set up properly f_op->flock field for
1342 Frequency : always on ia64
1344 Description: "lfs df" loops on printing out MDS statfs information
1345 Details : The obd_ioctl_data was not initialized and in some systems
1346 this caused a failure during the ioctl that did not return
1347 an error. Initialize the struct and return an error on failure.
1350 Frequency : SLES 9 only
1352 Description: Error of copying files with lustre special EAs as root
1353 Details : Client side always return success for setxattr call for lustre
1354 special xattr (currently only "trusted.lov").
1357 Frequency : rarely on clusters with both ia64+i386 clients
1359 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1360 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1361 clients are doing concurrent writes to the same file it is
1362 possible that the ia64 clients may LASSERT because the OST
1363 extent locks are not PAGE_SIZE aligned. Ensure that grown
1364 locks are always aligned on the request boundary.
1367 Frequency : specific use, occasional
1369 Description: Overwriting in use executable truncates on-disk binary image
1370 Details : If one node attempts to overwrite an executable in use by
1371 another node, we now correctly return ETXTBSY instead of
1372 truncating the file.
1374 Severity : enhancement
1376 Description: Async OSC create to avoid the blocking unnecessarily.
1377 Details : If a OST has no remain object, system will block on the creating
1378 when need to create a new object on this OST. Now, ways use
1379 pre-created objects when available, instead of blocking on an
1380 empty osc while others are not empty. If we must block, we block
1381 for the shortest possible period of time.
1386 Description: chmod on Lustre root is propagated to other clients
1387 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1388 invalid by the follow_mount time.
1391 Frequency : liblustre clients only
1393 Description: Race in 'instant cancel' lock handling could lead to such locks
1394 never to be granted in case of SMP MDS
1395 Details : Do not destroy not yet granted but cbpending locks in
1399 Frequency : replay/resend of open
1401 Description: non null lock assetion failure in mds_intent_policy
1402 Details : Trying to replay/resend lockless open requests resulted in
1403 mds_open() returning 0 with no lock. Now it sets a flag if
1404 a lock is going to be returned.
1406 Severity : enhancement
1408 Description: Checksum enhancements
1409 Details : New checksum enhancements allow for resending RPCs that failed
1412 Severity : enhancement
1414 Description: Tunables on number of dirty pages in cacche
1415 Details : Allow to set limit on number of dirty pages cached.
1420 Description: client crash on unmount - lock still has references
1421 Details : In some error handling cases it was possible to leak a lock
1422 reference on a client while accessing a file. This was not
1423 harmful to the client during operation, but would cause the
1424 client to crash when the filesystem is unmounted.
1427 Frequency : specific case, rare
1429 Description: ETXTBSY on mds though file not in use by client
1430 Details : ETXTBSY is no longer incorrectly returned when attempting to
1431 chmod or chown a directory that the user previously tried to
1432 execute or a currently-executing binary.
1435 Frequency : extremely rare except on liblustre-based clients
1437 Description: Lustre space not freed when files are deleted
1438 Details : Clean up open-unlinked files after client eviction. Previously
1439 the unlink was skipped and the files remained as orphans.
1444 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1445 Details : In some cases it was possible to send a blocking callback to
1446 a client doing a glimpse, even though that client didn't get
1447 a lock granted. When the glimpse lock is cancelled on the OST
1448 the freed lock is left on the waiting list and corrupted the list.
1451 Frequency : all core dumps
1453 Description: Broke core dumps to lustre
1454 Details : Negative dentry may be unhashed if parent does not have UPDATE
1455 lock, but some callers, e.g. do_coredump, expect dentry to be
1456 hashed after successful create, hash it in ll_create_it.
1458 ------------------------------------------------------------------------------
1460 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1462 * Support for kernels:
1463 2.6.9-42.0.2.EL (RHEL 4)
1464 2.6.5-7.276 (SLES 9)
1465 2.4.21-40.EL (RHEL 3)
1466 2.6.12.6 vanilla (kernel.org)
1470 Frequency : always on RHEL 3
1472 Description: Number of open files grows over time
1473 Details : The number of open files grows over time, whether or not
1474 Lustre is started. This was due to a filp leak introduced
1475 by one of our kernel patches.
1477 ------------------------------------------------------------------------------
1479 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1481 * Support for kernels:
1482 2.6.9-42.EL (RHEL 4)
1483 2.6.5-7.276 (SLES 9)
1484 2.4.21-40.EL (RHEL 3)
1485 2.6.12.6 vanilla (kernel.org)
1490 Bugzilla : 5719, 9635, 9792, 9684
1491 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1492 Details : If a server is under heavy load and cannot reply to new
1493 connection requests before the client resends the (re)connect,
1494 the connection handling code can behave badly if two service
1495 threads are concurrently handing separate (re)connections from
1496 the same client. Add better locking to the connection handling
1497 code, and ensure that only a single connection will be processed
1498 for a given client UUID, even if the lock is dropped.
1500 Severity : enhancement
1502 Description: add TCP zero-copy support to kernel
1503 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1504 sends if the hardware supports scatter-gather and checksumming.
1505 This allows socklnd to do client-write and server-read more
1506 efficiently and reduce CPU utilization from skbuf copying.
1509 Frequency : only if NFS exporting from client
1511 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1512 Details : If setting attributes on a file created under NFS that had
1513 never been opened it would be possible to oops the client
1514 if the file had no objects.
1517 Frequency : always for liblustre
1519 Description: liblustre client does MDS+OSTs setattr RPC for each write
1520 Details : When doing a write from a liblustre client, the client
1521 incorrectly issued an RPC to the MDS and each OST the file was
1522 striped over in order to update the timestamps. When writing
1523 with small chunks and many clients this could overwhelm the MDS
1524 with RPCs. In all cases it would slow down the write because
1525 these RPCs are unnecessary.
1527 Severity : enhancement
1529 Description: allow number of MDS service threads to be changed at module load
1530 Details : It is now possible to change the number of MDS service threads
1531 running. Adding "options mds mds_num_threads={N}" to the MDS's
1532 /etc/modprobe.conf will set the number of threads for the next
1533 time Lustre is restarted (assuming the "mds" module is also
1534 reloaded at that time). The default number of threads will
1535 stay the same, 32 for most systems.
1540 Description: OST crash if filesystem is unformatted or corrupt
1541 Details : If an OST is started on a device that has never been formatted
1542 or if the filesystem is corrupt and cannot even mount then the
1543 error handling cleanup routines would dereference a NULL pointer.
1548 Description: NULL pointer deref in llap_from_page.
1549 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1550 incorrect) mapping if the page was truncated/reclaimed while it was
1551 searched for. Check for this condition and skip such pages when
1552 doing readahead. Introduce extra check to llap_from_page() to
1553 verify page->mapping->host is non-NULL (so page is not anonymous).
1556 Frequency : Sometimes when using sys_sendfile
1558 Description: "page not covered by a lock" warnings from ll_readpage
1559 Details : sendfile called ll_readpage without right page locks present.
1560 Now we introduced ll_file_sendfile that does necessary locking
1561 around call to generic_file_sendfile() much like we do in
1565 Frequency : with certain MDS communication failures at client mount time
1567 Description: NULL pointer deref after failed client mount
1568 Details : a client connection request may delayed by the network layer
1569 and not be sent until after the PTLRPC layer has timed out the
1570 request. If the client fails the mount immediately it will try
1571 to clean up before the network times out the request. Add a
1572 reference from the request import to the obd device and delay
1573 the cleanup until the network drops the request.
1576 Frequency : occasionally during client (re)connect
1578 Description: assertion failure during client (re)connect
1579 Details : processing a client connection request may be delayed by the
1580 client or server longer than the client connect timeout. This
1581 causes the client to resend the connection request. If the
1582 original connection request is replied in this interval, the
1583 client may trip an assertion failure in ptlrpc_connect_interpret()
1584 which thought it would be the only running connect process.
1587 Frequency : only with obd_echo servers and clients that are rebooted
1589 Description: kernel BUG accessing uninitialized data structure
1590 Details : When running an obd_echo server it did not start the ping_evictor
1591 thread, and when a client was evicted an uninitialized data
1592 structure was accessed. Start the ping_evictor in the RPC
1593 service startup instead of the OBD startup.
1595 Severity : enhancement
1596 Bugzilla : 10193 (patchless)
1597 Description: Remove dependency on various unexported kernel interfaces.
1598 Details : No longer need reparent_to_init, exit_mm, exit_files,
1599 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1602 Frequency : rare (only users of deprecated and unsupported LDAP config)
1604 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1605 Details : LDAP apparently contains 'lustreName' attributes instead of
1606 'name'. A simple remapping of the name is sufficient.
1609 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1611 Description: waiting_locks_callback trips kernel BUG if client is evicted
1612 Details : Running with the dump_on_timeout debug flag turned on makes
1613 it possible that the waiting_locks_callback() can try to dump
1614 the Lustre kernel debug logs from an interrupt handler. Defer
1615 this log dumping to the expired_lock_main() thread.
1617 Severity : enhancement
1619 Description: Support NFS exporting on 2.6 kernels.
1620 Details : Implement non-rawops metadata methods for NFS server to use without
1621 changing NFS server code.
1624 Frequency : very rare (synthetic metadata workload only)
1626 Description: two racing renames might cause an MDS thread to deadlock
1627 Details : Running the "racer" program may cause one MDS thread to rename
1628 a file from being the source of a rename to being the target of
1629 a rename at exactly the same time that another thread is doing
1630 so, and the second thread has already enqueued these locks after
1631 doing a lookup of the target and is trying to relock them in
1632 order. Ensure that we don't try to re-lock the same resource.
1635 Frequency : only very large systems with liblustre clients
1637 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1638 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1639 all OSTs in parallel. This allows the eviction of stale liblustre
1640 clients to proceed much faster than if they were done in series,
1641 and also offers similar improvements for other set_info RPCs.
1646 Description: excessive CPU usage during initial read phase on client
1647 Details : During the initial read phase on a client, it would agressively
1648 retry readahead on the file, consuming too much CPU and impacting
1649 performance (since 1.4.5.8). Improve the readahead algorithm
1650 to avoid this, and also improve some other common cases (read
1651 of small files in particular, where "small" is files smaller than
1652 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1657 Description: MDS crash when receiving packet with unknown intent.
1658 Details : Do not LBUG in unknown intent case, just return -EFAULT
1660 Severity : enhancement
1661 Bugzilla : 9293, 9385
1662 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1663 Details : Do not serialize getattr (non-intent version) and statfs.
1666 Frequency : occasional, when OST network is overloaded/intermittent
1668 Description: client evicted by OST after bulk IO timeout
1669 Details : If a client sends a bulk IO request (read or write) the OST
1670 may evict the client if it is unresposive to its data GET/PUT
1671 request. This is incorrect if the network is overloaded (takes
1672 too long to transfer the RPC data) or dropped the OST GET/PUT
1673 request. There is no need to evict the client at all, since
1674 the pinger and/or lock callbacks will handle this, and the
1675 client can restart the bulk request.
1678 Frequency : Always when mmapping file with no objects
1680 Description: client crashes when mmapping file with no objects
1681 Details : Check that we actually have objects in a file before doing any
1682 operations on objects in ll_vm_open, ll_vm_close and
1688 Description: Request leak when working with deleted CWD
1689 Details : Introduce advanced request refcount tracking for requests
1690 referenced from lustre intent.
1692 Severity : Enhancement
1694 Description: Cache open file handles on client.
1695 Details : MDS now will return special lock along with openhandle, if
1696 requested and client is allowed to hold openhandle, even if unused,
1697 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1698 closing files for every read/write openration.
1700 Severity : Enhancement
1702 Description: Cache open negative dentries on client when possible.
1703 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1704 negative dentries on lock revocation.
1709 Description: Remounting a client read-only wasn't possible with a zconf mount
1710 Details : It wasn't possible to remount a client read-only with llmount.
1712 Severity : enhancement
1713 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1714 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1715 MPICH is included to add Lustre-specific ADIO interfaces.
1716 This is based closely on the UFS ADIO layer and only differs
1717 in file creation, in order to allow the OST striping to be set.
1718 This is user-contributed code and not supported by CFS.
1723 Description: extended inode attributes (immutable, append-only) work improperly
1724 when 2.4 and 2.6 kernels are used on client/server or vice versa
1725 Details : Introduce kernel-independent values for these flags.
1727 Severity : enhancement
1730 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1731 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1732 granularity. Now, Fractional MB granularity is be supported,
1733 this is very useful for low memory system.
1735 Severity : enhancement
1737 Description: Getattr by fid
1738 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1739 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1740 have locks and avoids getattr rpc requests for it on every stat.
1743 Frequency : Always, for filesystems larger than 2TB
1745 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1746 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1747 filesystems larger than 2TB to be mounted with mballoc enabled.
1750 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1752 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1753 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1754 filesystems larger than 2TB is mounted then the kernel will
1755 silently corrupt the start of the filesystem. CONFIG_LBD is
1756 enabled for all CFS-supported kernels, but the possibility of
1757 this happening with a modified kernel config exists.
1759 Severity : enhancement
1761 Description: add client O_DIRECT support for 2.6 kernels
1762 Details : It is now possible to do O_DIRECT reads and writes to files
1763 in the Lustre client mountpoint on 2.6 kernel clients.
1765 Severity : enhancement
1767 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1768 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1769 parallel, not waiting for response from every OST before sending
1770 a rpc to the next OST.
1775 Description: setattr vs write race when updating file timestamps
1776 Details : Client processes that update a file timestamp into the past
1777 right after writing to the file (e.g. tar) it is possible that
1778 the updated file modification time can be reset to the current
1779 time due to a race between processing the setattr and write RPC.
1781 Severity : enhancement
1783 Description: Bring 'lfs find' closer in line with regular Linux find.
1784 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1785 -print0 options and obtains all the needed info through the lustre
1788 Severity : enhancement
1790 Description: support up to 1024 configured devices on one node
1791 Details : change obd_dev array from statically allocated to dynamically
1792 allocated structs as they are first used to reduce memory usage
1797 Description: Flush dirty partially truncated pages during truncate
1798 Details : Immediatelly flush partially truncated pages in filter_setattr,
1799 this way we completely avoid having any pages in page cache on OST
1800 and can retire ugly workarounds during writes to flush such pages.
1805 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1806 Details : protect inode from truncation within vfs_unlink() context
1807 just take a reference before calling vfs_unlink() and release it
1808 when parent's i_sem is free.
1813 Description: last_id value checked outside lock on OST caused LASSERT failure
1814 Details : If there were multiple MDS->OST object precreate requests in
1815 flight, it was possible that the OST's last object id was checked
1816 outside a lock and incorrectly tripped an assertion. Move checks
1817 inside locks, and discard old precreate requests.
1820 Frequency : always, if extents are used on OSTs
1822 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1823 Details : OSTs using the extents format would not zero the high 16 bits of
1824 the index physical block number. This is not a problem for any
1825 OST filesystems smaller than 16TB, and no kernels support ext3
1826 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1827 new/modified files) and can be fixed for existing filesystems
1828 with e2fsprogs-1.39-cfs1.
1833 Description: import connection selection may be incorrect if timer wraps
1834 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1835 import connections to be ignored if the 32-bit jiffies counter
1836 wraps. Use a 64-bit jiffies counter.
1839 Frequency : very large clusters immediately after boot
1841 Description: LNET request buffers exhausted under heavy short-term load
1842 Details : If a large number of client requests are generated on a service
1843 that has previously never seen so many requests it is possible
1844 that the request buffer growth cannot keep up with the spike in
1845 demand. Instead of dropping incoming requests, they are held in
1846 the LND until the RPC service can accept more requests.
1849 Frequency : Sometimes during replay
1851 Description: Assertion failure in ll_local_open after replay.
1852 Details : If replay happened on an open request reply before we were able
1853 to set replay handler, reply will become not swabbed tripping the
1854 assertion in ll_local_open. Now we set the handler right after
1855 recognising of open request
1858 Frequency : very rare
1860 Description: kernel reports "badness in vsnprintf"
1861 Details : Reading from the "recovery_status" /proc file in small chunks
1862 may cause a negative length in lprocfs_obd_rd_recovery_status()
1863 call to vsnprintf() (which is otherwise harmless). Exit early
1864 if there is no more space in the output buffer.
1866 Severity : enhancement
1868 Description: clear OBD RPC statistics by writing to them
1869 Details : It is now possible to clear the OBD RPC statistics by writing
1870 to the "stats" file.
1875 Description: Client mtime is not the same on different clients after utimes
1876 Details : In some cases, the client was using the utimes() syscall on
1877 a file cached on another node. The clients now validate the
1878 ctime from the MDS + OSTs to determine which one is right.
1883 Description: Inability to activate failout mode
1884 Details : lconf script incorrectly assumed that in pythong string's numeric
1885 value is used in comparisons.
1888 Frequency : always with multiple stripes per file
1890 Description: Inefficient object allocation for mutli-stripe files
1891 Details : When selecting which OSTs to stripe files over, for files with
1892 a stripe count that divides evenly into the number of OSTs,
1893 the MDS is always picking the same starting OST for each file.
1894 Return the OST selection heuristic to the original design.
1899 Description: mount failures may take full timeout to return an error
1900 Details : Under some heavy load conditions it is possible that a
1901 failed mount can wait for the full obd_timeout interval,
1902 possibly several minutes, before reporting an error.
1903 Instead return an error as soon as the status is known.
1905 Frequency : quota enabled and large files being deleted
1907 Description: releasing more than 4GB of quota at once hangs OST
1908 Details : If a user deletes more than 4GB of files on a single OST it
1909 will cause the OST to spin in an infinite loop. Release
1910 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1915 Description: statfs data retrieved from /proc may be stale or zero
1916 Details : When reading per-device statfs data from /proc, in the
1917 {kbytes,files}_{total,free,avail} files, it may appear
1918 as zero or be out of date.
1921 Frequency : systems with MD RAID1 external journal devices
1923 Description: lconf's call to blkid is confused by RAID1 journal devices
1924 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1925 of returning all block devices that match the journal UUID.
1928 Frequency : always, for aggregate stripe size over 4GB
1930 Description: assertion fails when trying to use 4GB stripe size
1931 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1932 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1935 Frequency : always on ppc64
1937 Description: the first write on an ext3 filesystem with mballoc got stuck
1938 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1939 perform endianness conversion.
1941 ------------------------------------------------------------------------------
1943 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1945 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1946 INTEROPERATE with older versions automatically. Please read the
1947 user documentation before upgrading any part of a live system.
1948 * WARNING: Lustre networking configuration changes are required with
1949 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1952 * Support for kernels:
1953 2.6.9-22.0.2.EL (RHEL 4)
1954 2.6.5-7.244 (SLES 9)
1955 2.6.12.6 vanilla (kernel.org)
1958 Severity : enhancement
1959 Bugzilla : 7981/8208
1960 Description: Introduced Lustre Networking (LNET)
1961 Details : LNET is new networking infrastructure for Lustre, it includes
1962 a reorganized network configuration mode (see the user
1963 documentation for full details) as well as support for routing
1964 between different network fabrics. Lustre Networking Devices
1965 (LNDs) for the supported network fabrics have also been
1966 created for this new infrastructure.
1968 Severity : enhancement
1969 Description: Introduced Access control lists
1970 Details : clients can set ACLs on files and directories in order to have
1971 more fine-grained permissions than the standard Unix UGO+RWX.
1972 The MDS must be started with the "-o acl" mount option.
1974 Severity : enhancement
1975 Description: Introduced filesystem quotas
1976 Details : Administrators may now establish per-user quotas on the
1979 Severity : enhancement
1981 Description: Configuration change for the XT3
1982 The PTLLND is now used to run Lustre over Portals on the XT3
1983 The configure option(s) --with-cray-portals are no longer used.
1984 Rather --with-portals=<path-to-portals-includes> is used to
1985 enable building on the XT3. In addition to enable XT3 specific
1986 features the option --enable-cray-xt3 must be used.
1991 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1992 Details : A race between allocating a new llcd and re-getting the llcd_lock
1993 allowed another thread to grab newly-allocated llcd.
1995 Severity : enhancement
1997 Description: 2.6 OST async journal commit and locking fix to improve performance
1998 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1999 2.6 kernels are now async as they already were in 2.4 kernels so
2000 that they can commit concurrently with the network bulk transfer.
2001 For block-allocated files the filter allocation semaphore is held
2002 to avoid filesystem fragmentation during allocation. BKL lock
2003 removed for 2.6 xattr operations where it is no longer needed.
2008 Description: lconf incorrectly determined whether two IP networks could talk
2009 Details : In some more complicated routing and multiple-network
2010 configurations, lconf will avoid trying to make a network
2011 connection to a disjoint part of the IP space. It was doing the
2012 math incorrectly for one set of cases.
2017 Description: Fix for potential infinite loop processing records in an llog.
2018 Details : If an llog record is corrupted/zeroed, it is possible to loop
2019 forever in llog_process(). Validate the llog record length
2020 and skip the remainder of the block on error.
2023 Frequency : occasional (liblustre only)
2025 Description: liblustre could not open files whose last component is a symlink
2026 Details : sysio_path_walk() would incorrectly pass the open intent to
2027 intermediate path components.
2030 Frequency : rare (liblustre only with non-standard tuning)
2031 Bugzilla : 7201 (7350)
2032 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2033 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2034 after it has been released, only do this for referenced locks
2036 Severity : enhancement
2038 Description: specifying an (invalid) directory default stripe_size of -1
2039 would reset the directory default striping
2040 Details : stripe_size -1 was used internally to signal directory stripe
2041 removal, now use "all default" to signal dir stripe removal
2042 as a directory striping of "all default" is not useful
2045 Frequency : common for large clusters running liblustre clients
2047 Description: doing an ls when liblustre clients are running is slow
2048 Details : sending a glimpse AST to a liblustre client waits for every AST
2049 to time out, as liblustre clients will not respond. Since they
2050 cannot cache data we refresh the OST lock LVB from disk instead.
2052 Severity : enhancement
2054 Description: doing an ls at the same time as file IO can be slow
2055 Details : enqueue and other "small" requests can be blocked behind many
2056 large IO requests. Create a new OST IO portal for non-IO
2057 requests so they can be processed faster.
2060 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2062 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2063 created file with mode 000
2064 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2065 allow an ftruncate() as a non-root user to a file with mode 000.
2066 Lustre now respects this flag to disable mode checks when
2067 truncating a file owned by the user
2070 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2073 Description: Revoking locks from clients that went dead or catatonic might take
2075 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2076 cancellation of such locks instant on servers without waiting for
2077 any reply from clients. Clients drops these locks when cancel
2078 notification from server is received without replying.
2081 Frequency : liblustre-only, when liblustre client dies or becomes busy
2083 Description: Doing ls on Linux clients can take a long time with active
2085 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2086 granting such locks to it in the first place if possible. Locks
2087 are taken by proxy on the OST during the read or write and
2088 dropped immediately afterward. Add connect flags handling, do
2089 not grant locks to liblustre clients for glimpse ASTs.
2091 Severity : enhancement
2093 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2094 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2095 sequential reads and behaves badly with applications doing random
2096 reads. Improve it by reading ahead at least read region, and
2097 avoiding excessive large RPC for small reads.
2099 Severity : enhancement
2101 Description: Creating more than 1000 files for a single job may cause a load
2102 imbalance on the OSTs if there are also a large number of OSTs.
2103 Details : qos_prep_create() uses an OST index reseed value that is an
2104 even multiple of the number of available OSTs so that if the
2105 reseed happens in the middle of the object allocation it will
2106 still utilize the OSTs as uniformly as possible.
2111 Description: OST or MDS may oops in ping_evictor_main()
2112 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2113 but doesn't restart at beginning of obd_exports_timed list
2116 Severity : enhancement
2118 Description: improve by-nid export eviction on the MDS and OST
2119 Details : allow multiple exports with the same NID to be evicted at one
2120 time without re-searching the exports list.
2123 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2125 Description: MDS may oops in groups_free()
2126 Details : in rare race conditions a newly allocated group_info struct is
2127 freed again, and this can be NULL. The 2.4 compatibility code
2128 for groups_free() checked for a NULL pointer, but 2.6 did not.
2131 Frequency : common for liblustre clients doing little filesystem IO
2132 Bugzilla : 9352, 7313
2133 Description: server may evict liblustre clients accessing contended locks
2134 Details : if a client is granted a lock or receives a completion AST
2135 with a blocking AST already set it would not reply to the AST
2136 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2139 Frequency : lfs setstripe, only systems with more than 160 OSTs
2141 Description: unable to set striping with a starting offset beyond OST 160
2142 Details : llapi_create_file() incorrectly limited the starting stripe
2143 index to the maximum single-file stripe count.
2146 Frequency : LDAP users only
2148 Description: lconf did not handle in-kernel recovery with LDAP properly
2149 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2151 Severity : enhancement
2153 Description: bind OST threads to NUMA nodes to improve performance
2154 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2155 node and do their allocations there to localize memory access
2157 Severity : enhancement
2159 Description: llmount can determine client NID directly from Myrinet (GM)
2160 Details : the client NID code from gmnalnid was moved directly into
2161 llmount, removing the need to use this or specifying the
2162 client NID explicitly when mounting GM clients with zeroconf
2165 Frequency : if client is started with down MDS
2167 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2168 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2169 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2170 ptlrpc_import_delayed_req() because we didn't check if the
2171 request was interrupted, and we also didn't break out of the
2172 event loop if there was no timeout
2177 Description: data loss during non-page-aligned writes to a single file from
2178 both multiple nodes and multiple threads on one node at same time
2179 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2180 inconsistency led to false short-reads, that were cached and later
2181 used by ->prepare_write() to fill in partially written page,
2182 leading to data loss.
2185 Frequency : always, if lconf --abort_recovery used
2187 Description: lconf --abort_recovery fails with 'Operation not supported'
2188 Details : lconf was attempting to abort recovery on the MDT device and not
2191 Severity : enhancement
2193 Description: remove cleanup logs
2194 Details : replace lconf-generated cleanup logs with lustre internal
2195 cleanup routines. Eliminates the need for client-cleanup and
2198 Severity : enhancement
2200 Description: add support for EAs (user and system) on lustre filesystems
2201 Details : it is now possible to store extended attributes in the Lustre
2202 client filesystem, and with the user_xattr mount option it
2203 is possible to allow users to store EAs on their files also
2205 Severity : enhancement
2207 Description: Add possibility (config option) to show minimal available OST free
2209 Details : When compiled with --enable-mindf configure option, statfs(2)
2210 (and so, df) will return least minimal free space available from
2211 all OSTs as amount of free space on FS, instead of summary of
2212 free spaces of all OSTs.
2214 Severity : enhancement
2216 Description: do not expand extent locks acquired on OST-side
2217 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2218 by server: they are not cached anyway.
2221 Frequency : when mmap is used/binaries executed from Lustre
2223 Description: Unmmap pages before throwing them away from read cache.
2224 Details : llap_shrink cache now attempts to unmap pages before discarding
2225 them (if unmapping failed - do not discard). SLES9 kernel has
2226 extra checks that trigger if this unmapping is not done first.
2231 Description: lconf didn't resolve symlinks before checking to see whether a
2232 given mountpoint was already in use
2235 Frequency : when migrating failover services
2236 Bugzilla : 6395, 9514
2237 Description: When migrating a subset of services from a node (e.g. failback
2238 from a failover service node) the remaining services would
2239 time out and evict clients.
2240 Details : lconf --force (implied by --failover) sets the global obd_timeout
2241 to 5 seconds in order to quickly disconnect, but this caused
2242 other RPCs to time out too quickly. Do not change the global
2243 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2245 Severity : enhancement
2246 Frequency : if MDS is started with down OST
2247 Bugzilla : 9439,5706
2248 Description: Allow startup/shutdown of an MDS without depending on the
2249 availability of the OSTs.
2250 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2251 Add appropriate locking and lov-osc refcounts for safe
2252 cleaning. Add osc abort_inflight calls in case the
2253 synchronize never started.
2256 Frequency : occasional (Cray XT3 only)
2258 Description: root not authorized to access files in CRAY_PORTALS environment
2259 Details : The client process capabilities were not honoured on the MDS in
2260 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2261 been accessed by an authorized user then root was able to access
2262 the file on the local client also. The root user capabilities
2263 are now allowed on the MDS, as this environment has secure UID.
2266 Frequency : occasional
2268 Description: ldiskfs "too long searching" message happens too often
2269 Details : A debugging message (otherwise harmless) prints too often on
2270 the OST console. This has been reduced to only happen when
2271 there are fragmentation problems on the filesystem.
2276 Description: Division by zero in statfs when all OSCs are inactive
2277 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2278 lov_statfs divides by value returned by lov_get_stripecnt().
2282 Bugzilla : 9489, 3273
2283 Description: First write from each client to each OST was only 4kB in size,
2284 to initialize client writeback cache, which caused sub-optimal
2285 RPCs and poor layout on disk for the first writen file.
2286 Details : Clients now request an initial cache grant at (re)connect time
2287 and so that they can start streaming writes to the cache right
2288 away and always do full-sized RPCs if there is enough data.
2289 If the OST is rebooted the client also re-establishes its grant
2290 so that client cached writes will be honoured under the grant.
2295 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2296 Details : Now I/O RPCs go to different portal number and (presumably) fast
2297 lock requests (and glimses) and other RPCs get their own service
2298 threads pool that should be able to service those RPCs
2301 Severity : enhancement
2303 Description: Ability to exchange lustre version between client and servers and
2304 issue warnings at client side if client is too old. Also for
2305 liblustre clients there is ability to refuse connection of too old
2307 Details : New 'version' field is added to connect data structure that is
2308 filled with version info. That info is later checked by server and
2312 Frequency : rare, liblustre only.
2313 Bugzilla : 9296, 9581
2314 Description: Two simultaneous writes from liblustre at offset within same page
2315 might proceed at the same time overwriting eachother with stale
2317 Details : I/O lock withing llu_file_prwv was released too early, before data
2318 actually was hitting the wire. Extended lock-holding time until
2319 server acknowledges receiving data.
2322 Frequency : extremely rare. Never observed in practice.
2324 Description: avoid generating lustre_handle cookie of 0.
2325 Details : class_handle_hash() generates handle cookies by incrementing
2326 global counter, and can hit 0 occasionaly (this is unlikely, but
2327 not impossible, because initial value of cookie counter is
2328 selected randonly). Value of 0 is used as a sentinel meaning
2329 "unassigned handle" --- avoid it. Also coalesce two critical
2330 sections in this function into one.
2332 Severity : enhancement
2334 Description: allow liblustre clients to delegate truncate locking to OST
2335 Details : To avoid overhead of locking, liblustre client instructs OST to
2336 take extent lock in ost_punch() on client's behalf. New connection
2337 flag is added to handle backward compatibility.
2339 Severity : enhancement
2340 Bugzilla : 4928, 7341, 9758
2341 Description: allow number of OST service threads to be specified
2342 Details : a module parameter allows the number of OST service threads
2343 to be specified via "options ost ost_num_threads={N}" in the
2344 OSS's /etc/modules.conf or /etc/modprobe.conf.
2348 Bugzilla : 6146, 9635, 9895
2349 Description: servers crash with bad pointer in target_handle_connect()
2350 Details : In rare cases when a client is reconnecting it was possible that
2351 the connection request was the last reference for that export.
2352 We would temporarily drop the export reference and get a new
2353 one, but this may have been the last reference and the export
2354 was just destroyed. Get new reference before dropping old one.
2356 Severity : enhancement
2357 Frequency : if client is started with failover MDS
2359 Description: Allow multiple MDS hostnames in the mount command
2360 Details : Try to read the configuration from all specified MDS
2361 hostnames during a client mount in case the "primary"
2364 Severity : enhancement
2366 Description: Stop sending data to evicted clients as soon as possible.
2367 Details : Check if the client we are about to send or are sending data to
2368 was evicted already. (Check is done every second of waiting,
2369 for which l_wait_event interface was extended to allow checking
2370 of exit condition at specified intervals).
2373 Frequency : rare, normally only when NFS exporting is done from client
2375 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2377 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2380 Frequency : always, if srand() is called before liblustre initialization
2382 Description: Liblustre uses system PRNG disturbing its usage by user application
2383 Details : Introduce internal to lustre fast and high-quality PRNG for
2384 lustre usage and make liblustre and some other places in generic
2385 lustre code to use it.
2387 Severity : enhancement
2388 Bugzilla : 9477, 9557, 9870
2389 Description: Verify that the MDS configuration logs are updated when xml is
2390 Details : Check if the .xml configuration logs are newer than the config
2391 logs stored on the MDS and report an error if this is the case.
2392 Request --write-conf, or allow starting with --old_conf.
2394 Severity : enhancement
2396 Description: Handle symlinks in the path when checking if Lustre is mounted.
2397 Details : Resolve intermediate symlinks when checking if a client has
2398 mounted a filesystem to avoid duplicate client mounts.
2403 Description: lconf can hit an error exception but still return success.
2404 Details : The lconf command catches the Command error exception at the top
2405 level script context and will exit with the associated exit
2406 status, but doesn't ensure that this exit status is non-zero.
2411 Description: failure of ptlrpc thread startup can cause oops
2412 Details : Starting a ptlrpc service thread can fail if there are a large
2413 number of threads or the server memory is very fragmented.
2414 Handle this without oopsing.
2417 Frequency : always, only if liblustre and non-default acceptor port was used
2419 Description: liblustre cannot connect to servers with non-default acceptor port
2420 Details : tcpnal_set_default_params() was not called and was therefore
2421 ignoring the environment varaible TCPNAL_PORT, as well as other
2422 TCPNAL_ environment variables
2427 Description: two objects could be created on the same OST for a single file
2428 Details : If an OST is down, in some cases it was possible to create two
2429 objects on a single OST for a single file. No problems other
2430 than potential performance impact and spurious error messages.
2434 Bugzilla : 5681, 9562
2435 Description: Client may oops in ll_unhash_aliases
2436 Details : Client dcache may become inconsistent in race condition.
2437 In some cases "getcwd" can fail if the current directory is
2443 Description: Inode refcounting problems in NFS export code
2444 Details : link_raw functions used to call d_instantiate without obtaining
2445 extra inode reference first.
2449 Bugzilla : 9942, 9903
2450 Description: Referencing freed requests leading to crash, memleaks with NFS.
2451 Details : We used to require that call to ll_revalidate_it was always
2452 followed by ll_lookup_it. Also with revalidate_special() it is
2453 possible to call ll_revalidate_it() twice for the same dentry
2454 even if first occurence returned success. This fix changes semantic
2455 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2456 reference on a request referred from the intent.
2457 ll_intent_release() then releases such a request.
2460 Frequency : rare, normally benchmark loads only
2462 Description: unlinked inodes were kept in memory on the client
2463 Details : If a client is repeatedly creating and unlinking files it
2464 can accumulate a lot of stale inodes in the inode slab cache.
2465 If there is no other client load running this can cause the
2466 client node to run out of memory. Instead flush old inodes
2467 from client cache that have the same inode number as a new inode.
2470 Frequency : SLES9 2.6.5 kernel and long filenames only
2471 Bugzilla : 9969, 10379
2472 Description: utime reports stale NFS file handle
2473 Details : SLES9 uses out-of-dentry names in some cases, which confused
2474 the lustre dentry revalidation. Change it to always use the
2478 Frequency : rare, unless heavy write-truncate concurrency is continuous
2479 Bugzilla : 4180, 6984, 7171, 9963, 9331
2480 Description: OST becomes very slow and/or deadlocked during object unlink
2481 Details : filter_destroy() was holding onto the parent directory lock
2482 while truncating+unlinking objects. For very large objects this
2483 may block other threads for a long time and slow overall OST
2484 responsiveness. It may also be possible to get a lock ordering
2485 deadlock in this case, or run out of journal credits because of
2486 the combined truncate+unlink. Solution is to do object truncate
2487 first in one transaction without parent lock, and then do the
2488 final unlink in a new transaction with the parent lock. This
2489 reduces the lock hold time dramatically.
2492 Frequency : rare, 2.4 kernels only
2494 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2495 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2496 before unmounting their filesystem in order to avoid deadlock.
2497 The kernel_locked() function in 2.4 kernels only checks whether
2498 the kernel lock is held, not whether it is this process that is
2499 holding it as 2.6 kernels do.
2504 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2505 Details : The client ptlrpc code may be trying to reconnect to a down
2506 server before a previous connection attempt has timed out.
2507 Increase the reconnect interval to be longer than the connection
2508 timeout interval to avoid sending duplicate connections to
2512 Frequency : echo_client brw_test command
2514 Description: fix echo_client to work with OST preallocated code
2515 Details : OST preallocation code (5137) didn't take echo_client IO path
2516 into account: echo_client calls filter methods outside of any
2517 OST thread and, hence, there is no per-thread preallocated
2518 pages and buffers to use. Solution: hijack pga pages for IO. As
2519 a byproduct, this avoids unnecessary data copying.
2523 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2524 Description: Client can oops in mdc_commit_close() after open replay
2525 Details : It was possible for the MDS to return an open request with no
2526 transaction number in mds_finish_transno() if the client was
2527 evicted, but without actually returning an error. Clients
2528 would later try to replay that open and may trip an assertion
2529 Simplify the client close codepath, and always return an error
2530 from the MDS in case the open is not successful.
2533 Frequency : rare, 2.6 OSTs only
2535 Description: OST may deadlock under high load on fragmented files
2536 Details : If there was a heavy load and highly-fragmented OST filesystems
2537 it was possible to have all the OST threads deadlock waiting on
2538 allocation of biovecs, because the biovecs were not released
2539 until the entire RPC IO was completed. Instead, release biovecs
2540 as soon as they are complete to ensure forward IO progress.
2542 Severity : enhancement
2544 Description: Support for specifying external journal device at mount
2545 Details : If an OST or MDS device is formatted with an external journal
2546 device, this device major/minor is stored in the ext3 superblock
2547 and may not be valid for failover. Allow detecting and
2548 specifying the external journal at mount time.
2553 Description: Mounting an MDS with pending unlinked files may cause oops
2554 Details : target_finish_recovery() calls mds_postrecov() which returned
2555 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2556 considers this an error and immediately begins cleaning up the
2557 lov, just after starting the mds_lov process
2559 Severity : enhancement
2561 Description: Implement 'lfs df' to report actual free space on per-OST basis
2562 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2563 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2564 usage of objects; '-h' to report in human readable format.
2566 ------------------------------------------------------------------------------
2568 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2575 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2576 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2577 at startup during orphan unlinking. Instead, skip buddy update
2578 before setup, as it will be regenerated after recovery is complete.
2583 Description: If an OST is inactive, its locks might reference stale inodes.
2584 Details : lov_change_cbdata() must iterate over all namespaces, even if
2585 they are inactive to clear inode references from the lock.
2587 Severity : enhancement
2588 Frequency : occasional, if non-standard max_dirty_mb used
2590 Description: Client will block write RPCs if not enough grant
2591 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2592 then the client will block writes while waiting for another RPC
2593 to complete instead of consuming its dirty limit. With change
2594 we get improved performance when max_dirty_mb is small.
2596 Severity : enhancement
2597 Bugzilla : 3389, 6253
2598 Description: Add support for supplementary groups on the MDS.
2599 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2600 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2601 lookups for user supplementary groups into a cache.
2605 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2606 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2607 for liblustre O_EXCL is enforced by MDS.
2611 Description: readdir never returns NULL in liblustre.
2612 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2613 dentry in struct dirent.
2617 Frequency : liblustre only, depends on application IO pattern
2618 Description: liblustre clients evicted if not contacting servers
2619 Details : Don't put liblustre clients into the ping_evictor list, so
2620 they will not be evicted by the pinger ever.
2622 Severity : enhancement
2624 Description: Add ability to evict clients by NID from MDS.
2625 Details : By echoing "nid:$NID" string into
2626 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2627 $NID would be instantly evicted from this MDS and from all active
2628 OSTs connected to it.
2632 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2633 Details : lookup_it_finish() used to query file size from OSTs that was not
2638 Description: service threads change working directory to that of init
2639 Details : Starting lustre service threads may pin the working directory
2640 of the parent thread, making that filesystem busy. Threads
2641 now change to the working directory of init to avoid this.
2645 Frequency : during shutdown only
2646 Description: shutdown with a failed MDS or OST can cause unmount to hang
2647 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2652 Frequency : 2.6 only
2653 Description: chmod/chown may include an extra supplementary group
2654 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2655 supplementary group and if none is specified this is used.
2658 Bugzilla : 5479 (6816)
2660 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2661 Details : If lookup is in progress on a file that is unlinked we might try
2662 to revalidate the inode and fail in revalidate after lookup is
2663 complete and ll_file_open() enqueues the open again but
2664 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2667 Frequency : always, if lconf --abort_recovery used
2669 Description: lconf --abort_recovery fails with 'Operation not supported'
2670 Details : lconf was attempting to abort recovery on the MDT device and not
2673 ------------------------------------------------------------------------------
2675 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2680 Frequency : rare (only unsupported configurations with a node running as an
2682 Bugzilla : 6514, 5137
2683 Description: Mounting a Lustre file system on a node running as an OST could
2685 Details : OSTs now preallocates memory needed to write out data at
2686 startup, instead of when needed, to avoid having to
2687 allocate memory in possibly low memory situations.
2688 Specifically, if the file system is mounted on on OST,
2689 memory pressure could force it to try to write out data,
2690 which it needed to allocate memory to do. Due to the low
2691 memory, it would be unable to do so and the node would
2692 become unresponsive.
2694 Severity : enhancement
2696 Description: Addition of lconf --service command line option
2697 Details : lconf now accepts a '--service <arg>' option, which is
2698 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2700 Severity : enhancement
2702 Description: Failover mode is now the default for OSTs.
2703 Details : By default, OSTs will now run in failover mode. To return to
2704 the old behaviour, add '--failout' to the lmc line for OSTs.
2706 Severity : enhancement
2708 Description: Health checks are now provided for MDS and OSTs
2709 Details : Additional detailed health check information on MSD and OSTs
2710 is now provided through the procfs health_check value.
2713 Frequency : occasional, depends on IO load
2715 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2716 after numerous create/delete cycles
2717 Details : The ext3 inode allocation policy would not allocate new inodes
2718 very well on the OSTs because there are no new directories
2719 being created. Instead we look for groups with free space if
2720 the parent directories are nearly full.
2725 Description: Network or server problems during mount may cause partially
2726 mounted clients instead of returning an error.
2727 Details : The config llog parsing code may overwrite the error return
2728 code during mount error handling, returning success instead
2734 Description: MDS can fail to allocate large reply buffers
2735 Details : After long uptimes the MDS can fail to allocate large reply
2736 buffers (e.g. zconf client mount config records) due to memory
2737 fragmentation or consumption by the buffer cache. Preallocate
2738 some large reply buffers so that these replies can be sent even
2739 under memory pressure.
2743 Frequency : rare (liblustre)
2744 Description: fsx running with liblustre complained that using truncate() to
2745 extend the file doesn't work. This patch corrects that issue.
2746 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2747 ATTR_SIZE and lsm use in llu_setattr_raw.
2751 Frequency : rare, only 2.6 kernels
2752 Description: Unusual file access patterns on the MDS may result in inode
2753 data being lost in very rare circumstances.
2754 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2755 code in the 2.6 kernel caused the inode and/or EA data not to
2756 be read from disk, causing single-file corruption.
2760 Frequency : rare, only 2.6 filesystems using extents
2761 Description: Heavy concurrent write and delete load may cause data corruption.
2762 Details : It was possible under high-load situations to have an extent
2763 metadata block in the block device cache from a just-unlinked
2764 file overwrite a newly-allocated data block. We now unmap any
2765 metadata buffers that alias just-allocated data blocks.
2769 Frequency : filesystems with default stripe_count larger than 77
2770 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2771 Details : lconf specifies an inode size of 4096 bytes when the default
2772 stripe_count is larger than 77. This conflicts with the default
2773 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2774 this case to avoid pinning too much memory for large EAs.
2776 ------------------------------------------------------------------------------
2778 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2783 Frequency : rare (extremely heavy IO load with hundreds of clients)
2785 Description: Client is evicted, gets IO error writing to file
2786 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2787 caused clients to be evicted for AST timeouts. The fixes in
2788 bug 5192 mean we no longer need to have such short AST timeouts
2789 so ldlm_timeout has been increased.
2792 Frequency : occasional during --force or --failover shutdown under load
2793 Bugzilla : 5949, 4834
2794 Description: Server oops/LBUG if stopped with --force or --failover under load
2795 Details : a collection of import/export refcount and cleanup ordering
2796 issues fixed for safer force cleanup
2799 Frequency : only filesystems larger than 120 OSTs
2800 Bugzilla : 5990, 6223
2801 Description: lfs getstripe would oops on a very large filesystem
2802 Details : lov_getconfig used kfree on vmalloc'd memory
2805 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2806 Bugzilla : 6242, 6243
2807 Description: reading from files that had been truncated to a non-zero size
2808 but never opened returned no data
2809 Details : ll_file_read() reads zeros from no-object files to EOF
2814 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2815 Details : There's little harm in aborting MDS/OSS recovery and letting it
2816 try again, so I removed the LASSERT and return an error instead.
2818 Severity : enhancement
2820 Description: New debugging infrastructure for tracking down data corruption
2821 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2822 (b) cover more of the client-side code path, and (c) try to narrow
2823 down where problems occurred
2827 Bugzilla : 3819, 4364, 4397, 6313
2828 Description: Racing close and eviction MDS could cause assertion in mds_close
2829 Details : It was possible to get multiple mfd references during close and
2830 client eviction, leading to one thread referencing a freed mfd.
2832 Severity: : enhancement
2833 Bugzilla : 3262, 6359
2834 Description: Attempts to reconnect to servers are now more aggressive.
2835 Details : This builds on the enhanced upcall-less recovery that was added
2836 in 1.4.2. When trying to reconnect to servers, clients will
2837 now try each server in the failover group every 10 seconds. By
2838 default, clients would previously try one server every 25 seconds.
2843 Description: After recovery, certain operations trigger a failed
2844 assertion on a client.
2845 Details : Failing over an mds, using lconf -d --failover, while a
2846 client was doing a readdir() call would cause the client to
2847 LBUG after recovery completed and the readdir() was resent.
2849 Severity : enhancement
2851 Description: Default groups are now added by lconf
2852 Details : You can now run lconf --group <servicename> without having to
2853 manually add groups with lmc.
2856 Frequency : occasional
2858 Description: Nodes with an elan id of 0 trigger a failed assertion
2861 Frequency : always when accessing e.g. tty/console device nodes
2863 Description: tty and some other devices nodes cannot be used on lustre
2864 Details : file's private_data field is used by device data and lustre
2865 values in there got lost. New field was added to struct file to
2866 store fs-specific private data.
2869 Frequency : when exporting Lustre via NFS
2871 Description: NFSD failed occasionally when looking up a path component
2872 Details : NFSD is looking up ".." which was broken in ext3 directories
2873 that had grown large enough to become hashed.
2876 Frequency : Clusters with multiple interfaces not on the same subnet
2878 Description: Nodes will repeatedly try to reconnect to an interface which it
2879 cannot reach and report an error to the log.
2880 Details : Extra peer list entries will be created by lconf with some peers
2881 unreachable. lconf now validates the peer before adding it.
2884 Frequency : Only if a default stripe is set on the filesystem root.
2886 Description: Setting a default stripe on the filesystem root prevented the
2887 filesystem from being remounted.
2888 Details : The client was sending extra request flags in the root getattr
2889 request and did not allocate a reply buffer for the dir EA.
2892 Frequency : occasional, higher if lots of files are accessed by one client
2893 Bugzilla : 6159, 6097
2894 Description: Client trips assertion regarding lsm mismatch/magic
2895 Details : While revalidating inodes the VFS looks up inodes with ifind()
2896 and in rare cases can find an inode that is being freed.
2897 The ll_test_inode() code will free the lsm during ifind()
2898 when it finds an existing inode and then the VFS later attaches
2899 this free lsm to a new inode.
2903 Bugzilla : 6422, 7030
2904 Description: MDS deadlock between mkdir and client eviction
2905 Details : Creating a new file via mkdir or mknod (starting a transaction
2906 and getting the ns lock) can deadlock with client eviction
2907 (gets ns lock and trying to finish a synchronous transaction).
2910 Frequency : occasional
2911 Description: While starting a server, the fsfilt_ext3 module could not be
2913 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2914 kernels. Previously, lconf would still use the ext3 name
2915 when trying to load modules. Now, it will correctly use
2916 ext3 on 2.4 and ldiskfs on 2.6.
2918 Severity : enhancement
2919 Description: The default stripe count has been changed to 1
2920 Details : The interpretation of the default stripe count (0, to lfs
2921 or lmc) has been changed to mean striping across a single
2922 OST, rather than all available. For general usage we have
2923 found a stripe count of 1 or 2 works best.
2925 Severity : enhancement
2926 Description: Add support for compiling against Cray portals.
2927 Details : Conditional compiling for some areas that are different
2931 Frequency : occasional
2932 Bugzilla : 6409, 6834
2933 Description: Creating files with an explicit stripe count may lead to
2934 a failed assertion on the MDS
2935 Details : If some OSTs are full or unavailable, creating files may
2936 trigger a failed assertion on the MDS. Now, Lustre will
2937 try to use other servers or return an error to the
2941 Frequency : occasional
2943 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2944 caused lock timeout during readahead (since 1.4.2).
2945 Details : Processes doing readahead might match a lock that hasn't been
2946 granted yet if there are overlapping and conflicting lock
2947 requests. The readahead process waits on ungranted lock
2948 (original lock is CBPENDING), while OST waits for that process
2949 to cancel CBPENDING read lock and eventually evicts client.
2951 Severity : enhancement
2953 Description: Initial enabling of flock support for clients
2954 Details : Implements fcntl advisory locking and file status functions.
2955 This feature is provided as an optional mount flag (default
2956 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2957 locking are implemented yet, and those that are are not guaranteed
2958 to be completely correct in production environments.
2959 mount -t lustre -o [flock|noflock] ...
2962 Frequency : occasional
2964 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2965 assertion in the ext3 JBD (journaling) layer.
2966 Details : The b_committed_data struct is protected by the big kernel lock
2967 in 2.4 kernels, serializing journal_commit_transaction() and
2968 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2969 access to this struct. In 2.6 kernels there is finer grained
2970 locking to improve SMP performance of the JBD layer.
2974 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2979 Description: merge and cleanup kernel patches.
2982 -----------------------------------------------------------------------------
2984 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2986 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2987 versions of Lustre. Please update all servers and clients to
2988 version 1.4.2 or later at the same time. You must also run
2989 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2990 to update the configuration logs.
2992 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2993 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2994 - swab llog records of type '0' so we get proper header size/idx (5861)
2995 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2996 - fix rename of one directory over another leaking an inode (5953)
2997 - avoid SetPageDirty on 2.6 (5981)
2998 - don't re-add just-being-destroyed locks to the waiting list (5653)
2999 - when creating new directories, inherit the parent's custom
3000 striping settings if present parent (3048)
3001 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3002 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3003 - don't hold client locks on temporary worklist from l_lru (5666)
3004 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3005 - automatically evict dead clients (5921)
3006 - Update file size properly in create+truncate+fstat case (6196)
3007 - Do not unhash mountpoint dentries, do not allow removal of
3009 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3010 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3011 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3012 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3013 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3014 - Correctly handle reads of files with no objects (6243)
3015 - lctl recover will also mark a device active if deactivate used (5933)
3017 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3018 - allow --write-conf on an MDS with different nettype than client (5619)
3019 - don't write config llogs to MDS for mounts not from that MDS (5617)
3020 - lconf should create multiple TCP connections from a client (5201)
3021 - init scripts are now turned off by default; run chkconfig --on
3022 lustre and chkconfig --on lustrefs to use them
3023 - upcalls are no longer needed for clients to recover to failover
3025 - add --abort-recovery option to lconf to abort recovery on device
3027 - add support for an arbitrary number of OSTs (3026)
3028 - Quota support protocol changes.
3029 - forward compatibility changes to wire structs (6007)
3030 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3031 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3032 - improved "lustre status" script
3033 - initialize blocksize for non-regular files (6062)
3034 - added --disable-server and --disable-client configure options (5782)
3035 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3036 - Vanilla 2.4.29 support
3037 - increase maximum number of obd devices to 520 (6242)
3038 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3039 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3040 SLES 9 2.6.5 kernels
3041 - limit stripes per file to 160 (the maximum EA size) (6093)
3043 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3046 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3047 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3048 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3049 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3050 - small ext3 extents cleanups and fixes (5733)
3051 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3052 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3053 - avoid needless client->OST connect, fix handle mismatch (5317)
3054 - fix DLM error path that led to out-of-sync client, long delays (5779)
3055 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3056 - fix several locking issues related to i_size (5492,5624,5654,5672)
3057 - don't move pending lock onto export if it is already evicted (5683)
3058 - fix kernel oops when creating .foo in unlinked directory (5548)
3059 - fix deadlock in obdfilter statistics vs. object create (5811)
3060 - use time_{before,after} to avoid timer jiffies wrap (5882)
3061 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3062 - Do not leak request if server was not able to process it (5154)
3063 - If mds_open unable to find parent dir, make that negative lookup(5154)
3064 - don't create new directories with extent-mapping (5909, 5936)
3066 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3067 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3068 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3069 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3070 - support for RHEL4 kernel on i686 (5773)
3071 - provide error messages when incompatible logs are encountered (5898)
3073 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3074 * version 1.4.0.10 (1.4.1 release candidate 1)
3076 - don't keep a lock reference when lock is not granted (4238)
3077 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3078 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3079 - drop import inflight refcount on signal_completed_replay error (5255)
3080 - unlock page after async write error during send (3677)
3081 - handle missing objects in filter_preprw_read properly (5265)
3082 - no transno return for symlink open, don't save no-trasno open (3440)
3083 - don't try to complete elan receive that already failed (4012)
3084 - free RPC server reply state on error (5406)
3085 - clean up thread from ptlrpc_start_thread() on error (5160)
3086 - readahead could read extra page into cache that wasn't ejected (5388)
3087 - prevent races in class_attach/setup/cleanup/detach (5260)
3088 - don't dereference de->d_inode after l_dput of de (5458)
3089 - use "int" for stripe value returned from lock_to_stripe (5544)
3090 - mballoc allocation and error-checking fixes in 2.6 (5504)
3091 - block device patches to fix I/O request sizes in 2.6 (5482)
3092 - look up hostnames for IB nals (5602)
3093 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3094 - don't start multiple acceptors for the same port (5277)
3095 - fix incorrect LASSERT in mds_getattr_name (5635)
3096 - export a proc file for general "ping" checking (5628)
3097 - fix "lfs check" to not block when the MDS is down (5628)
3099 - service request history (4965)
3100 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3101 - create an "evict_client" /proc entry on OSTs, like the MDS has
3102 - fix mount usage message, return errors per mount(8) (5168)
3103 - change grep [] to grep "[]" in tests so they work in more UMLs
3104 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3105 - remove ancient LOV_MAGIC_V0 EA support (5047)
3106 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3107 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3108 - allow readahead allocations to fail when low on memory (5383)
3109 - mmap locking landed again, after considerable improvement (2828)
3110 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3112 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3115 - send OST transaction number in read/write reply to free req (4966)
3116 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3117 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3118 - the watchdog thread now runs as interruptible (5246)
3119 - flock/lockf fixes (but it's still disabled, pending 5135)
3120 - don't use EXT3 constants in llite code (5094)
3121 - memory shortage at startup could cause assertion (5176)
3123 - reorganization of lov code
3124 - single portals codebase
3126 - add extents/mballoc support (5025)
3127 - direct I/O reads in the obdfilter (4048)
3128 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3130 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3133 - send OST transaction number in read/write reply to free req (4966)
3134 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3135 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3136 - the watchdog thread now runs as interruptible (5246)
3137 - handle missing objects in filter_preprw_read properly (5265)
3138 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3139 - drop import inflight refcount on signal_completed_replay error (5255)
3140 - unlock page after async write error during send (3677)
3141 - return original error code on reconstructed replies (3761)
3142 - no transno return for symlink open, don't save no-trasno open (3440)
3144 - add pid to ldlm debugging output (4922)
3145 - bump the watchdog timeouts -- we can't handle 30sec yet
3146 - extra debugging for orphan dentry/inode bug (5259)
3148 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3151 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3152 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3153 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3154 - fix NULL dereference in /proc/sys/portals/routes (4827)
3155 - allow failed mdc_close() operations to be interrupted (4561)
3156 - stop precreate on OST before MDS would time out on it (4778)
3157 - don't send partial-page writes before EOF from client (4410)
3158 - discard client grant for sub-page writes on large-page clients (4520)
3159 - don't free dentries not owned by NFS code, check generation (4806)
3160 - fix lsm leak if mds_create_objects() fails (4801)
3161 - limit debug_daemon file size, always print CERROR messages (4789)
3162 - use transno after validating reply (3892)
3163 - process timed out requests if import state changes (3754)
3164 - update mtime on OST during writes, return in glimpse (4829)
3165 - add mkfsoptions to LDAP (4679)
3166 - use ->max_readahead method instead of zapping global ra (5039)
3167 - don't interrupt __l_wait_event() during strace
3169 - add software watchdogs to catch hung threads quickly (4941)
3170 - make lustrefs init script start after nfs is mounted
3171 - fix CWARN/ERROR duplication (4930)
3172 - return async write errors to application if possible (2248)
3173 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3174 - print NAL number in %x format (4645)
3175 - update barely-supported suse-2.4.21-171 series (4842)
3176 - support for sles 9 %post scripts
3177 - support for building 2.6 kernel-source packages
3178 - support for sles km_* packages
3180 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3183 - ignore -ENOENT errors in osc_destroy (3639)
3184 - notify osc create thread that OSC is being cleaned up (4600)
3185 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3186 - reconstruct ost_handle() like mds_handle() (4657)
3187 - create a new thread to do import eviction to avoid deadlock (3969)
3188 - let lconf resolve symlinked-to devices (4629)
3189 - don't unlink "objects" from directory with default EA (4554)
3190 - hold socknal file ref over connect in case target is down (4394)
3191 - allow more than 32000 subdirectories in a single directory (3244)
3192 - fix blocks count for O_DIRECT writes (3751)
3193 - OST returns ENOSPC from object create when no space left (4539)
3194 - don't send truncate RPC if file size isn't changing (4410)
3195 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3196 - bind to privileged port in socknal and tcpnal (3689)
3198 - rate limit CERROR/CWARN console message to avoid overload (4519)
3199 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3200 - basic mmap support (3918)
3201 - kernel patch series update from b1_4 (4711)
3203 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3206 - avoid crash during MDS cleanup with OST shut down (2775)
3207 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3208 - don't use bad inodes on the MDS (3744)
3209 - dynamic object preallocation to improve recovery speed (4236)
3210 - don't hold spinlock over lock dumping or change debug flags (4401)
3211 - don't zero obd_dev when it is force cleaned (3651)
3212 - print grants to console if they go negative (4431)
3213 - "lctl deactivate" will stop automatic recovery attempts (3406)
3214 - look for existing locks in ldlm_handle_enqueue() (3764)
3215 - don't resolve lock handle twice in recovery avoiding race (4401)
3216 - revalidate should check working dir is a directory (4134)
3218 - don't always mark "slow" obdfilter messages as errors (4418)
3220 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3223 - don't close LustreDB during write_conf until it is done (3860)
3224 - fix typo in lconf for_each_profile (3821)
3225 - allow dumping logs from multiple threads at one time (3820)
3226 - don't allow multiple threads in OSC recovery (3812)
3227 - fix debug_size parameters (3864)
3228 - fix mds_postrecov to initialize import for llog ctxt (3121)
3229 - replace config semaphore with spinlock (3306)
3230 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3231 - don't allow enqueue to complete on a destroyed export (3822)
3232 - down write_lock before checking llog header bitmap (3825)
3233 - recover from lock replay timeout (3764)
3234 - up llog sem before sending rpc (3652)
3235 - reduce ns lock hold times when setting kms (3267)
3236 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3237 - fix NULL deref and obd_dev leak on setup error (3312)
3238 - replace some LBUG about llog ops with error handling (3841)
3239 - don't match INVALID dentries from d_lookup and spin (3784)
3240 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3241 - fix invalid assertion in ptlrpc_set_wait (3880)
3243 - add libwrap support for the TCP acceptor (3996)
3244 - add /proc/sys/portals/routes for non-root route listing (3994)
3245 - allow setting MDS UUID in .xml (2580)
3246 - print the stack of a process that LBUGs (4228)
3248 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3251 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3252 - make sure to unset replay flag from failed open requests (3440)
3253 - if default stripe count is 0, use OST count for inode size (3636)
3254 - update parent mtime/ctime on client for create/unlink (2611)
3255 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3256 - free recovery state on server during a forced cleanup (3571)
3257 - unregister_reply for resent reqs (3063)
3258 - loop back devices mounting and status check on 2.6 (3563)
3259 - fix resource-creation race that can provoke i_size == 0 (3513)
3260 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3261 - more debugging for page-accounting assertion (3746)
3262 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3263 - avoid deadlock after precreation failure (3758)
3264 - fix race and lock order deadlock in orphan handling (3450, 3750)
3265 - add validity checks when grabbing inodes from l_ast_data (3599)
3267 - add /proc/.../recovery_status to obdfilter (3428)
3268 - lightweight CDEBUG infrastructure, debug daemon (3668)
3269 - change default OSC RPC parameters to be better on small clusters
3270 - turn off OST read cache for files smaller than 32MB
3271 - install man pages and include them in rpms (3100)
3272 - add new init script for (un)mounting lustre filesystems (2593)
3273 - run chkconfig in %post for init scripts (3701)
3274 - drop scimac NAL (unmaintained)
3276 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3279 - clean kiobufs before and after use (3485)
3280 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3281 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3282 - change init script to fail more clearly if not run as root (1528)
3283 - allow clients to reconnect during replay (1742)
3284 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3285 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3286 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3287 - fix cond_resched (3554)
3288 - only evict unfinished clients after recovery (3515)
3289 - allow bulk resend, prevent data loss (3570)
3290 - dynamic ptlrpc request buffer allocation (2102)
3291 - don't allow unlinking open directory if it isn't empty (2904)
3292 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3293 - remove extraneous obd dereference causing LASSERT failure (3334)
3294 - don't use get_cycles() when creating temp. files on the mds (3156)
3295 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3296 - handle EEXIST for set-stripe, set proper directory name (3336)
3298 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3299 - fix ksocknal_fmb_callback() error messages (2918)
3301 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3304 - don't copy lvb into (possibly NULL) reply on error (2983)
3305 - don't deref dentry after dput, don't free lvb on error (2922)
3306 - use the kms to determine writeback rpc length (2947)
3307 - increment oti_logcookies when osc is inactive (2948)
3308 - update client's i_blocks count via lvb messages (2543)
3309 - handle intent open/close of special files properly (1557)
3310 - mount MDS with errors=remount-ro, like obdfilter (2009)
3311 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3312 - don't use cancelling-locks' kms values (2947)
3313 - use highest lock extent for kms, not last one (2925)
3314 - don't dereference ERR_PTR() dentry in error handling path (3107)
3315 - fix thread race in portals_debug_dumplog() (3122)
3316 - create lprocfs device entries at setup instead of at attach (1519)
3317 - common AST error handler, don't evict client on completion race (3145)
3318 - zero nameidata in detach_mnt in 2.6 (3118)
3319 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3320 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3321 - initialize RPC timeout timer earlier for 2.6 (3219)
3322 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3323 - print nal/nid for unknown nid (3258)
3324 - additional checks for oscc recovery before doing precreate (3284)
3325 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3326 - don't crash in mdc_close for bad permissions on open (3285)
3327 - zero i_rdev for non-device files (3147)
3328 - clear page->private before handing to FS, better assertion (3119)
3329 - tune the read pipeline (3236)
3330 - fix incorrect decref of invalidated dentry (2350)
3331 - provide read-ahead stats and refine rpc in flight stats (3328)
3332 - don't hold journal transaction open across create RPC (3313)
3333 - update atime on MDS at close time (3265)
3334 - close LDAP connection when recovering to avoid server load (3315)
3335 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3336 - don't leak open file on MDS after open resend (3325)
3337 - serialize filter_precreate and filter_destroy_precreated (3329)
3338 - loop device shouldn't call sync_dev() for nul device (3092)
3339 - clear page cache after eviction (2766)
3340 - resynchronize MDS->OST in background (2824)
3341 - refuse to mount the same filesystem twice on same mountpoint (3394)
3342 - allow llmount to create routes for mounting behind routers (3320)
3343 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3344 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3345 - fix rare problem with rename on htree directories (3417)
3347 - allow default OST striping configuration per directory (1414)
3348 - fix compilation for qswnal for 2.6 kernels (3125)
3349 - increase maximum number of MDS request buffers for large systems
3350 - change liblustreapi to be useful for external progs like lfsck (3098)
3351 - increase local configuration timeout for slow disks (3353)
3352 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3354 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3357 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3358 - don't overwrite extent policy data in reply if lock was blocked (2901)
3359 - drop filter export grants atomically with removal from device (2663)
3360 - del obd_self_export from work_list in class_disconnect_exports (2908)
3361 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3362 - swab reply message in mdc_close, other PPC fixes (2464)
3363 - fix destroying of named logs (2325)
3364 - overwrite old logs when running lconf --write_conf (2264)
3365 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3366 - fix race in target_handle_connect (2898)
3367 - mds_reint_create() should take same inode create lock (2926)
3368 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3369 - don't close files for self_export to avoid uninitialized obd (2936)
3370 - allow MDS with the same name as client node (2939)
3371 - hold dentry reference for closed log files for unlink (2325)
3372 - reserve space for all logs during transactions (2059)
3373 - don't evict page beyond end of stripe extent (2925)
3374 - don't oops on a deleted current working directory (2399)
3375 - handle hard links to targets without a parent properly (2517)
3376 - don't dereference NULL lock when racing during eviction (2867)
3377 - don't grow lock extents when lots of conflicting locks (2919)
3379 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3382 - account for cache space usage on clients to avoid data loss (974)
3383 - lfsck support in lustre kernel code (2349)
3384 - reduce journal credits needed for BRW writes (2370)
3385 - orphan handling to avoid losing space on client/server crashes
3386 - ptlrpcd can be blocked, stopping ALL progress (2477)
3387 - use lock value blocks to assist in proper KMS, faster stat (1021)
3388 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3389 - recovery for initial connections (2355)
3390 - fixes for mds_cleanup_orphans (1934)
3391 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3392 - block all file creations until orphan recovery completes (1901)
3393 - client remove rq_connection from request struct (2423)
3394 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3395 - recovery timer race (2670)
3396 - mdc_close recovey bug (2532)
3397 - ptlrpc cleanup bug (2710)
3398 - mds timeout on local locks (2588)
3399 - namespace lock held during RPCs (2431)
3400 - handle interrupted sync write properly (2503)
3401 - don't try to handle a message that hasn't been replied to (2699)
3402 - client assert failure during cleanup after abort recovery (2701)
3403 - leak mdc device after failed mount (2712)
3404 - ptlrpc_check_set allows timedout requests to complete (2714)
3405 - wait for inflight reqs when ptlrpcd finishes (2710)
3406 - make sure unregistered services are removed from the srv_list
3407 - reset bulk XID's when resending them (caught by 1138 test)
3408 - unregister_bulk after timeout
3409 - fix lconf error (2694)
3410 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3411 - readahead locks pages, leaves pending causing memory pressure (2673)
3412 - increase OST request buffers to 4096 on large machines (2729)
3413 - fix up permission of existing directories in simple_mkdir (2661)
3414 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3415 - don't assign transno to errored transactions (2742)
3416 - don't delete objects on OST if given a bogus objid from MDS (2751)
3417 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3418 - if rq_no_resend, then timeout request after recovery (2432)
3419 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3420 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3421 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3422 - MDS recovery shouldn't delete objects that it has given out (2730)
3423 - if enqueue arrives after completion, don't clobber LVB (2819)
3424 - don't unlock pages twice when trigger_group_io returns error (2814)
3425 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3426 - don't write pages to disk if there was an error (1450)
3427 - don't ping imports that have recovery disabled (2676)
3428 - take buffered bytes into account when balancing socknal conn (2817)
3429 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3430 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3431 - remove little-endian swabbing of llog records (1987)
3432 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3433 - retry reposting request buffers if they fail (1191)
3434 - grow extent at grant time to avoid granting a revoked lock (2809)
3435 - lock revoke doesn't evict page if covered by a second lock (2765)
3436 - disable VM readahead to avoid reading outside lock extents (2805)
3438 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3439 - updated kernel patches for hp-2.4.20 kernel (2681)
3441 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3444 - fix truncated write corruption (2366)
3445 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3447 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3448 - ensure len(uuid) < 37 in lmc (1171)
3449 - fix ia64 OOPS in llog_test (2255)
3450 - zero end of page at obdfilter for partial page writes (2648)
3451 - don't leave stale dentries around after renames (bug 2428)
3452 - fix timeouts when evicting a client with a single lock held (2642)
3453 - set deadline for the initial HELLO message to drain (2634)
3454 - print out dotted-quad IP addresses in the socknal (2302)
3456 - additional debugging for MDS client eviction problem (2443)
3457 - fix mkfsoptions support for osts (2603, 2604)
3459 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3462 - add series for the vanilla 2.6.0 kernel
3463 - add series for the vanilla 2.4.24 kernel
3464 - add series for a cray x86/64 UL kernel drop
3465 - fix xattr patches for the vanilla 2.4.19 series
3467 - generate true UUIDs in lmc (1171)
3468 - have portals stack dumping break in UML (2466)
3469 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3470 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3471 - wake callback waiting threads on client eviction (2460)
3472 - Add --ptldebug and --subsystem to lmc (1719)
3473 - update assertion to allow safe interrupt allocation
3474 - set rq_no_resend for cancel requests (2432)
3475 - recalculate ptlrpcd timeout after resend (2494)
3476 - call vfs_rmdir when removing pending directories (2368)
3477 - fix renaming a file to itself (2429)
3478 - lmc creates a default one-stripe lov (2454)
3479 - expand procfs space to handle large clusters (2326)
3480 - increase UML stack to avoid overflow
3481 - update lconf's list of debug and subsystem masks
3482 - fix lfs find --obd (2510)
3483 - /proc tunable for disabling filter read caching (2591)
3484 - stop rpm packages from altering slapd.conf (2301)
3485 - disable nagle in the socknal under 0conf (2578)
3486 - choose mds inode size based on stripe count (2572)
3487 - fix kernel-source rpm problems (2516)
3489 - add --disable-doc to avoid pdf generation (2421)
3490 - update documentation, tests, type-os, comments
3491 - avoid format warnings on ia64
3492 - remove the TOE NAL
3493 - tiny code cleanups by removing unused fields
3495 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3498 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3499 - remove the most blatant lies from BUILDING file (2371)
3500 - change default debug level to reasonable production setting
3501 - reduce client side cache size to reduce cache flush time
3502 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3503 - make TCP zerocopy and pinger support enabled by default (2476)
3504 - sync writes completed after process exits caused crashes (2319)
3505 - maintain correct mount count on the MDS (2356)
3506 - backout 1557, because 2316 wasn't really fixed
3507 - better file I/O statistics gathering in /proc
3508 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3509 - another kernel patch to fix zero-copy TCP function export
3510 - don't take duplicate lock when processing re-sent getattr (2420)
3511 - lctl uses obd_self_export instead of creating new conn (2353)
3512 - MDS/OST recovery case which requires object creation asserted (2425)
3513 - move lfs from /usr/sbin to /usr/bin in packages
3514 - fix race between mds_client_add and mds_client_free (2417)
3515 - use kmalloc instead of slabs in portals (2430)
3516 - don't create duplicate records when a failover MDS is present (2442)
3517 - remove unnecessary mount age check (2332)
3518 - don't remove directory inodes from locks prematurely (2451)
3519 - don't break if MDS service name is the same as hostname (2103)
3520 - fix races in client write RPC generation when cache full (2482)
3522 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3525 - remove now-unused request->rq_obd (278)
3526 - if an allocation fails, print out how much memory we've used (1933)
3527 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3528 - add the "configurable stack size" patch to most series files (1256)
3529 - ability to write large log records, for 100+ OST configs (2306)
3530 - fix NULL deref when filter_prep fails (2314)
3531 - fix operator precedence error in filter_sync
3532 - dynamic allocation of socknal TX descriptors (2315)
3533 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3534 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3535 - ensure that CWARN messages in Portals always get to the syslog
3536 - __init/__exit are not for prototype decls (ldlm_init/exit)
3537 - x86-64 compile warning fixes
3538 - fix gateway LMC keyword conflict (2318)
3539 - fix MDS lock inversions in getattr/reint paths (1844)
3540 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3541 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3542 - DLM race condition prevented some lock evictions (2328)
3543 - ENOMEM detection and retry on socknal sends (2230)
3544 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3545 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3546 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3547 - better error messages when a client is rejected during recovery (1505)
3548 - avoid cancelling locks which were never granted, after failure (2330)
3549 - fix i_sem/journal inversion in mds_client_add (2333)
3550 - fix truncate/getattr lock cycle deadlock (2334)
3551 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3552 - fix two rare exit paths which could leak an l_lock() ref (2321)
3553 - fencepost error in MDS/OST orphan recovery (2226)
3554 - make log record alignment 8 bytes (1988)
3555 - lstripe now fails when requested offset > ost_count (2237)
3556 - ensure that all kernel series have a complete list.h (1607)
3557 - fix crashes in special-file operations (2316)
3558 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3560 - allow configurable automake binary, for testing new versions
3561 - small update to the lfs documentation
3563 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3565 * fix negative export reference count in fsfilt_sync (2312)
3567 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3568 * release candidate 0.9.1
3570 - orphans are moved into the PENDING directory for possible recovery
3571 - replayed opens now open by fid for orphan/rename safety (1042)
3572 - last close of an orphan inode generates a transno (683)
3573 - chdir() and mount() now pin the directory entry (1020)
3574 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3575 - discard very old requests without processing them (1502)
3576 - remove some common, well-understood CERRORs (1505)
3577 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3578 - clear "grant" flags in OST replies until OST grant code lands (1644)
3579 - fix read performance by not clobbering i_blksize on client (1598)
3580 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3581 - make LRU size a /proc tunable, clears locks when reduced (707)
3582 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3583 - server locks take references on exports now (1558)
3584 - build fixes for 2.4.20-rh trees (1663)
3585 - return an error from lov_create if all OSCs are inactive (1751)
3586 - fix import levels when a reconnect happens without a timeout (1597)
3587 - exit early from mds_open if we get a lookup error (1749)
3588 - partial page read at EOF wouldn't wait for disk before sending (1642)
3589 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3590 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3591 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3592 - write/truncate lock inversion (1639)
3593 - Don't auto-load obdclass, portals modules during cleanup (1495)
3594 - fix timestamps from jumping to "now" (1763)
3595 - extra journal assertions (1648)
3596 - add an extra multiunlink test (1771)
3597 - fix read_record/write_record API (1776)
3598 - fix leak of offset_extent, possible incorrect i_size later (1772)
3599 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3600 - Debugging for the kqswnal_get_idle_tx problems (1820)
3601 - Allow recovery to be attempted multiple times (1536)
3602 - Write out MDS last_rcvd file after it is first created (1600)
3603 - Fix tx_descriptor leak in failed transmit situations (1827)
3604 - ext3 journaling fixes for assertion failure after IO error (1871)
3605 - class_export_put() on freed export after completion AST error (1896)
3606 - Fix revalidate looping in VFS (1322)
3607 - Don't access a freed export during MDS_REINT timeout (1521)
3608 - Add open-unlink recovery support on the MDS (1673,1764)
3609 - Return an error if no MDS data was read from last_rcvd (1946)
3610 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3611 - Don't setup a disk device that doesn't match exported UUID (317)
3612 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3613 - avoid committing NULL handle in force close
3614 - local.sh is now a one-stripe LOV configuration
3615 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3616 - fix ext3 htree duplicate directory entry corruption (1516)
3617 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3618 - update to new LOV EA format (2097)
3619 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3621 2003-06-15 Phil Schwan <phil@clusterfs.com>
3624 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3625 - per-import/export recovery handling (958, 931, 959)
3626 - multiple last-rcvd slots, for serving multiple FSes (949)
3627 - connections are again shared between multiple imp/exports (963, 964)
3628 - "umount -f" would hang if any requests needed to be sent (393, 978)
3629 - avoid pinning large req buffer by copying for queued messages (989)
3630 - add "uuid" to "lctl device" command to help upcalls (991)
3631 - "open" RPCs with transnos would confuse recovery counters (1037)
3632 - do proper endian conversion of all wire messages (288, 340, 891)
3633 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3634 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3635 - fix ptlrpc_connection leak in target_handle_connect (1174)
3636 - fix import refcounting bug in OST and MDS cleanup (1134)
3637 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3638 - fix very unlikely obd_types race condition (501)
3639 - remove osc_open hack for echo_client (1187)
3640 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3641 - a failure in read_inode2 leads to deadlock (1139)
3642 - cancel ack-locks as soon as transaction is committed (1072)
3643 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3644 - make sure to commitrw after any preprw to avoid deadlock (1162)
3645 - failing to execute a file in a lustre FS would lock inode (1203)
3646 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3647 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3648 - obd_uuid2tgt would walk off the end of the list (1255)
3649 - on IA64 the osc would give portals incorrect bulk size (1258)
3650 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3651 - fix lock inversion caused by new llite matching code (1282)
3652 - limit the number of dirty pages on a client to 10MB (1286)
3653 - timed out locks were not being corrected cancelled (1289)
3654 - fix O_DIRECT above 4GB on IA-32 (1292)
3655 * major user-visible changes
3656 - fail out/fail over policy now controlled by the upcall (993)
3658 - add OBD_PING to check server availability and failure (954)
3659 - lustre messages are now sent in sending host order (288, 340, 891)
3660 - add eadatalen to MDS getattr reply (340)
3661 - OST read replies may contain second buffer, with per-page status (593)
3663 2003-03-11 Phil Schwan <phil@clusterfs.com>
3666 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3667 - failed open()s could cause deadlock; fixed (867, 869)
3668 - stop cancelling OST locks when files are closed (481)
3669 - overlapping XID spaces caused network corruption (851, 853)
3670 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3671 - setattr_raw added, to do single-RPC, server-side setattrs
3672 - lmc/lconf syntax change for OST UUIDs
3673 - fix crashy race condition between ptlrpc_free_req and osc_close
3674 - don't use request in mdc_enqueue if we hit a timeout (889)
3675 - don't set the inode i_size for regular files from the MDS (896)
3676 - handle out of order completion AST (842)
3677 - don't LBUG if a lock request times out after receiving AST (913)
3678 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3679 - if a bad lock AST arrives, send an error instead of dropping entirely
3680 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3681 - fix leak in bulk IO when only partially completed (899, 900, 926)
3682 - fix O_DIRECT for ia64 (55)
3683 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3684 - C-z after timeout could hang a process forever; fixed (977)
3686 - client-side I/O cache (678, 924, 929, 941, 970)
3688 - READPAGE and SETATTRs which don't take server-side locks get
3691 2003-02-11 Phil Schwan <phil@clusterfs.com>
3694 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3695 - Fix fsfilt_extN_readpage() to read a full page of directory
3696 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3697 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3698 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3699 - Most intent operations were removed, in favour of a new RPC mode
3700 that does a single RPC to the server and bypasses most of the VFS
3701 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3702 - Aggressively cancel local locks on DLM servers
3703 - mds_reint_unlink sends EA to the client if it's the last nlink.
3704 client uses that EA to unlink OST objects.
3705 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3706 - recursive symlinks were fixed (439)
3707 - fixed NULL deref in DEBUG_REQ
3708 - filter_update_lastobjid no longer calls sync, which annoyed extN
3709 - fixed multi-client small-writes to a single file problem (445)
3710 - fixed mtime updates during file writes (607)
3711 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3712 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3713 - fixed Linux OST size reporting problem (444, 656)
3714 - OST now updates object mtime with writes or setattr (607, 619)
3715 - client verifies file size before zeroing page past EOF (445)
3716 - OST now writes last allocated objid to disk with allocation (108)
3717 - LOV on echo now works (409)
3719 - mds_reint_unlink sends a new buffer, with the EA included. this
3720 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3721 set if a regular file was being unlinked, and it was the last link
3722 - use PtlGet from the target for bulk writes (315)
3723 - OST now updates object mtime with writes or setattr (607, 619)
3724 - LDLM now has a grant-time callback to revalidate locked items, if
3726 - Many MDS operations were reorganized to combat race conditions
3728 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3729 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3731 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3734 - Fully reactivate OST imports after reconnection (512, others)
3735 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3736 - More graceful error handling for truncating on dead OST (515)
3737 - Don't error out unless we're actually accessing dead stripes (474)
3738 - Fix garbage sizes when stripes are missing (410)
3739 - LRU counters were broken, causing constant lock purge (433, 432)
3740 - garbage on read from stripes with failed OSTs (441)
3741 - mark OSCs as active before reconnecting during recovery (438)
3742 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3743 - lfind did not preserve OST order in output (443)
3744 - symlinks cause hung clients, incorrect data (439)
3745 - stop dereferencing request after dropping refcount (457)
3746 - don't LASSERT(spin_is_locked) on non-SMP (455)
3747 - fixes for many rename() bugs
3748 - fstat didn't correctly synchronize attributes (399)
3749 - server must handle lock cancellation during blocking AST prep (487)
3750 - bulk descriptors were free()d too soon (511)
3751 - fix paths in lconf, which would load incorrect modules (451, 507)
3752 - fix confusing lconf 'host not found' error message (386)
3753 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3754 - fix race condition in mdc_blocking_ast() for inode access (526)
3755 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3756 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3757 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3758 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3759 - fix lstripe in several ways (kernel side) (527)
3760 - fix request leak in ldlm_cli_enqueue (262)
3761 - incorrect OSC was marked inactive after OST failure
3762 - call mds_fs_cleanup before unmounting filesystem (524)
3763 - fix races between taking ns_lock and ldlm_lock_change_resource
3764 - fix races updating LOV export open file list
3765 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3766 - fix recovery NULL deref in ldlm_cli_cancel_unused
3767 - fix some DLM races by using new hash table for lock handles (419)
3768 - permit the client to specify desired inodes, at replay
3769 - duplicate requests when we queue them for replay reintegration
3770 - fix last_rcvd offset calculation
3771 - sync after each recovered transaction, so we always make progress
3772 - never, not always, ERESTART requests without transnos
3773 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3775 - skip replay if the MDS says that the client is already connected
3776 - don't check for a recovery-enabled export to match lctl's UUID
3777 - don't INC_USE_COUNT for phantom exports
3778 - don't crash when cleaning up phantom exports (567)
3779 - don't double-finish or set replay data for errored mdc_open requests
3780 - abort requests when they time out, so we don't get old replies
3781 - send/receive replies for AST messages again
3782 - if the client says that it doesn't have the lock, cancel it on the
3784 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3785 mark it as destroyed, it will all work out when decref is called
3786 - fix module use counts (22, 581)
3788 - ASTs now expect a reply (server cancels lock on error reply)
3790 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3793 - fix many simultaneous client startup (392)
3794 - fix dentry->d_it clobbering
3795 - credentials weren't being shipped for readdir/getattr operations
3796 - remove invalid assertions triggered during some concurrent MD
3798 - proper Lustre versions added (336, 389)
3799 - fix memory leak for create error case (398)
3800 - fix LOV locking bug that would get cli/srv out of sync
3801 - fix echo client over LOV (409)
3802 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3803 - fix double-O_EXCL intent crash (424)
3804 - avoid sending multiple lock CANCELs (352)
3806 - MDS can do multi-client recovery (modulo bugs in new code)
3808 - many updates, edits, cleanups
3810 2002-11-18 Phil Schwan <phil@clusterfs.com>
3813 - fix null d_it dereference (346)
3814 - fix full OST/dbench hang (333)
3815 - fix permission problem with file removal (286)
3816 - fix removal of OSCs from LOV when they fail
3817 - fix NULL deref during bulk timeout (214)
3818 - fix problems related to multiple filesystems on one MDS (241)
3819 - fixed serious subtle metadata locking bugs
3820 - free locks on clients when inodes are removed due to memory
3822 - fix inode pointer in lock data (285)
3823 - partial support for multiple MDS on a single host (241)
3824 - data locks weren't cancelled at clear_inode time (290, 311)
3825 - intent locks could lead to unbounded lock growth (205)
3826 - added a maximum lock count, an LRU list, and a flusher
3827 - fix multiple rename (365)
3828 - properly abstracted the echo client
3829 - OSC locked 1 byte too many; fixed
3830 - rewrote brw callback code:
3831 - fixed recovery bugs related to LOVs (306)
3832 - fixed too-many-pages-in-one-write crash (191)
3833 - fixed (again) crash in sync_io_timeout (214)
3834 - probably fixed callback-related race (385)
3836 - Add capability to MDS protocol
3837 - LDLM cancellations and callbacks on different portals
3839 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3842 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3843 - defer open object destruction to close time (601981, 138)
3844 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3845 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3846 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3847 - don't call obd_setattr in ll_file_release for destroyed objects
3848 * protocol change to lustre_msg: move |version| and add |flags|
3849 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3850 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3851 * added replay of create, unlink, link and rename operations during
3852 MDS failover; recovery should be much more robust now
3853 * remove failed OSCs from LOVs (only lov_create uses this so far)
3854 * the lustre-HOWTO was brought (more) up to date (582544)
3856 2002-10-23 Phil Schwan <phil@clusterfs.com>
3859 - in-use dentries weren't being reused properly (617851)
3860 - prevent multiple LDLM setup (599178)
3861 - fix LOV size calculations for truncate (617853)
3862 - fix client handling of MDS intent errors (POSIX)
3863 - fix permission bug in lovstripe.c test (624321)
3864 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3865 - truncate past end of file could corrupt data
3866 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3867 - a race in recovery could return ETIMEDOUT to apps (623947)
3868 - building outside the source directory was fixed
3869 * the lustre-HOWTO was brought (more) up to date (582544)
3870 * major progress was made on recovery functionality
3872 2002-10-10 Phil Schwan <phil@clusterfs.com>
3875 - recovery deadlock fix
3876 - rm -rf causes LBUG fix (617817)
3877 - file open by multiple tasks fix (618962)
3878 - directory permissions bugs (602707 and 620007)
3879 - journal_stop fixed with locking (611313)
3880 - O_APPEND failures resolved (618273, perhaps 614459)
3881 - lconf PATH fix (619770)
3882 - IA64 build fix (621450)
3883 - RPC buffer sizes scale with amount of memory
3885 2002-10-01 Phil Schwan <phil@clusterfs.com>
3888 - locks would be cancelled without throwing away data pages,
3889 resulting in inconsistent data (605627)
3890 - inode attributes were not always being refreshed (605627, 612449)
3891 - lconf now continues to cleanup after lctl reports an error
3892 - MDS now enforces user permissions (602707)
3893 - lprocfs cleanup fixed, but not yet enabled (614157)
3894 - fixed infinite server hang, should a client not respond to an AST
3895 - avoid going into recovery if user calls readlink() with a buffer
3896 that's too small (613941)
3897 - AST RPCs no longer require replies (614867) -- this may be changed
3898 - don't crash server if client sends an IOV that's too big (611336)
3899 - fixed lock conversion deadlock (611892)
3900 - fixed the following of symlinks (614622)
3901 * recovery: the server can remove locks from a client that dies, other
3902 clients can make progress
3903 * more extN patch fixes
3904 * compile-time configurable ptlrpc buffer allocations
3906 - collaborative read cache document
3907 - Lustre Lite Performance CDR document-in-progress
3909 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3912 - fix typo in patch-2.4.18
3914 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3917 - clear ptlrpc request each time in handle_incoming_request()
3918 - unlink of files now destroys the object on the OST
3920 2002-09-19 Peter Braam <braam@clusterfs.com>
3922 * add hard link support
3923 * change obdfile creation method
3924 * kernel patch changed
3926 2002-09-19 Peter Braam <braam@clusterfs.com>
3929 - stack overflow bug in extN fixed
3931 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3933 * documentation updates
3934 - add man pages for config tools
3935 - update tests/README to describe testing with new config tools
3936 - finish metadata API descriptions
3937 * bug fixes and cleanups
3938 - statfs workaround for 16TB limit
3939 - LOV stripe allocation improved, can stripe on subset of OSTs
3940 - LOV file size/IO offset was wrong for files > 4GB in size
3941 - object EA data was being dropped, caused files to be unreadable
3942 - memory overflow with non-LOV OST caused memory corruption
3943 - fixed regression tests to work with new config tools, obdfilter
3944 - fixed bug when directory size became larger than 1 block
3945 - fixed bug (for single client case) when PWD was deleted
3946 - invalidate local directory pages when doing intent-based ops
3947 - avoid LDLM oops when lock callback contained bad data
3949 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3951 * documentation updates
3952 * bug fixes and cleanups
3953 - configuration tools
3956 - 64-bit compile warnings
3957 - 64-bit internal statfs data
3959 * test_brw on persistent OST devices
3961 * lprocfs (disabled)
3963 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3965 * documentation updates
3966 * bug fixes and cleanups
3967 * configuration tools
3969 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3972 * many small fixes to 0_5_4
3973 * io/network handling
3974 * thinkos in MDS operations
3976 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3979 * crucial basic fixes to 0.5.3
3980 * IOR, Iozone work over Elan
3983 2002-08-07 Phil Schwan <phil@clusterfs.com>
3984 * version 0_5_3, our first alpha
3985 * we use the new Portals iovs
3986 * documentation updates
3987 * bug fixes and cleanups
3988 * small changes in the DLM wire protocol
3990 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3991 * version 0_5_1 with some initial stability,
3992 * locking on MD and file I/O.
3993 * documentation updates
3994 * several bug fixes since 0.5.0
3995 * small changes in wire protocol
3997 2002-07-18 Phil Schwan <phil@clusterfs.com>
3999 * delivered as Lustre Light Alpha
4000 * fixed a crash after handling invalid MDS requests
4001 * fixed directory pages for architectures with non-4k pages sizes
4003 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4004 * release version v0_4_4
4005 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4007 * Fixes a number of interruption problems with OST operations.
4008 * Update documentation for portals header changes
4009 * Move all wire protocol structs/defines to lustre_idl.h
4010 * Fixes symlink length bug.
4011 * Add tcpdump to repository.
4013 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4014 * release version v0_4_3
4015 * Fixes statfs for inodes on extN.
4016 * Fixes bug in runtests which would delete /etc/hosts.
4017 * Use 64-bit object IDs wherever possible (not into VFS though)
4018 Remove ost_get_info, which is unused by lustre, and out of date.
4020 2002-07-03 Peter Braam <braam@clusterfs.com>
4021 * release version v0_4_2 Fixes a lookup error (type not passed)
4022 * move forward to head of Portals
4023 * move forward to latest Lustre kernel
4025 2002-06-25 Peter Braam <braam@clusterfs.com>
4026 * release version v0_4_1. Hopefully stable on single node use.