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
271 --------------------------------------------------------------------------------
273 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
275 * Support for kernels:
276 2.6.5-7.283 (SLES 9),
277 2.6.9-55.EL (RHEL 4),
278 2.6.16.46-0.14 (SLES 10),
279 2.6.18.8 vanilla (kernel.org)
280 * Client support for unpatched kernels:
281 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
282 2.6.16 - 2.6.22 vanilla (kernel.org)
283 * Due to recently discovered recovery problems, we do not recommend
284 using patchless RHEL 4 clients with this or any earlier release.
285 * Recommended e2fsprogs version: 1.39.cfs8
286 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
287 * Starting with this release, the ldiskfs backing filesystem required
288 by Lustre is now in its own package, lustre-ldiskfs. This package
289 should be installed. It is versioned separately from Lustre and
290 may be released separately in future.
294 Description: block reactivating mgc import until all deactivates complete
295 Details : Fix race when failing back MDT/MGS to itself (testing)
297 Severity : enhancement
299 Description: add optional extra BUILD_VERSION info
300 Details : add a new environment variable (namely LUSTRE_VERS) which allows
301 to override the lustre version.
304 Frequency : 2.6.18 servers only
306 Description: ll_kern_mount() doesn't release the module reference
307 Details : The ldiskfs module reference count never drops down to 0
308 because ll_kern_mount() doesn't release the module reference.
313 Description: server LBUG when using old ost_num_threads parameter
314 Details : Accept the old ost_num_threads parameter but warn that it
315 is deprecated, and fix an off-by-one error that caused an LBUG.
320 Description: Transient SCSI error results in persistent IO issue
321 Details : iobuf->dr_error is not reinitialized to 0 between two
325 Frequency : sometimes when underlying device returns I/O errors
327 Description: OSTs not going read-only during write failures
328 Details : OSTs are not remounted read-only when the journal commit threads
329 get I/O errors because fsfilt_ext3 calls journal_start/stop()
330 instead of the ext3 wrappers.
334 Description: poor connect scaling with increasing client count
335 Details : Don't run filter_grant_sanity_check for more than 100 exports
336 to improve scaling for large numbers of clients.
339 Frequency : SLES10 only
341 Description: sanity-quota.sh quotacheck failed: rc = -22
342 Details : Quotas cannot be enabled on SLES10.
345 Frequency : liblustre clients only
347 Description: getdirentries does not give error when run on compute nodes
348 Details : getdirentries does not fail when the size specified as an argument
349 is too small to contain at least one entry
351 Severity : enhancement
353 Description: Add LNET router traceability for debug purposes
354 Details : If a checksum failure occurs with a router as part of the
355 IO path, the NID of the last router that forwarded the bulk data
356 is printed so it can be identified.
361 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
362 Details : Due to a race condition, liblustre clients were occasionally
365 Severity : enhancement
367 Description: lfs setstripe use optional parameters instead of postional
370 Severity : enhancement
372 Description: Nanosecond timestamp support for ldiskfs
373 Details : The on-disk ldiskfs filesystem has added support for nanosecond
374 resolution timestamps. There is not yet support for this at
375 the Lustre filesystem level.
378 Frequency : during server recovery
380 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
381 Details : request with rq_no_resend flag not awake l_wait_event if they get a
385 Frequency : nfs export on patchless client
387 Description: connectathon hang when test nfs export over patchless client
388 Details : Disconnected dentry cannot be found with lookup, so we do not need
389 to unhash it or make it invalid
392 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
393 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
394 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
399 Description: open req refcounting wrong on reconnect
400 Details : If reconnect happened between getting open reply from server and
401 call to mdc_set_replay_data in ll_file_open, we will schedule
402 replay for unreferenced request that we are about to free.
403 Subsequent close will crash in variety of ways.
404 Check that request is still eligible for replay in
405 mdc_set_replay_data().
410 Description: disable writes to filesystem when reading health_check file
411 Details : the default for reading the health_check proc file has changed
412 to NOT do a journal transaction and write to disk, because this
413 can cause reads of the /proc file to hang and block HA state
414 checking on a healthy but otherwise heavily loaded system. It
415 is possible to return to the previous behaviour during configure
416 with --enable-health-write.
418 Severity : enhancement
420 Description: 64-bit inode version
421 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
422 to the inode. This will be required for version-based recovery.
427 Description: MDS fails to start if a duplicate client export is detected
428 Details : in some rare cases it was possible for a client to connect to
429 an MDS multiple times. Upon recovery the MDS would detect this
430 and fail during startup. Handle this more gracefully.
432 Severity : enhancement
434 Description: Add -o localflock option to simulate old noflock
436 Details : This will achieve local-only flock/fcntl locks
442 Description: log_commit_thread vs filter_destroy race leads to crash
443 Details : Take import reference before releasing llog record semaphore
448 Description: Wrong request locking in request set processing
449 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
450 imp_delayed_list, in this place should be used imp_lock.
453 Frequency : when reconnection
455 Description: Grant Leak when osc reconnect to OST
456 Details : When osc reconnect ost, OST(filter) should check whether it
457 should grant more space to client by comparing fed_grant and
458 cl_avail_grant, and return the granted space to client instead
459 of "new granted" space, because client will call osc_init_grant
460 to update the client grant space info.
463 Frequency : when client reconnect to OST
465 Description: Grant Leak when osc do resend and replay bulk write
466 Details : When osc reconnect to OST, OST(filter)should clear grant info of
467 bulk write request, because the grant info while be sync between
468 OSC and OST when reconnect, and we should ignore the grant info
469 these of resend/replay write req.
474 Description: Grant space more than avaiable left space sometimes.
475 Details : When then OST is about to be full, if two bulk writing from
476 different clients came to OST. Accord the avaliable space of the
477 OST, the first req should be permitted, and the second one
478 should be denied by ENOSPC. But if the seconde arrived before
479 the first one is commited. The OST might wrongly permit second
480 writing, which will cause grant space > avaiable space.
483 Frequency : when client is evicted
485 Description: Grant might be wrongly erased when osc is evicted by OST
486 Details : when the import is evicted by server, it will fork another
487 thread ptlrpc_invalidate_import_thread to invalidate the
488 import, where the grant will be set to 0. While the original
489 thread will update the grant it got when connecting. So if
490 the former happened latter, the grant will be wrongly errased
491 because of this race.
496 Description: Checking Stale with correct fid
497 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
498 to check whether it is stale, because sometimes, we want the
499 enqueue happened anyway, and op_data.fid2 will not be initialized.
501 Severity : enhancement
503 Description: update patchless client
504 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
507 Frequency : only with 2.4 kernel
509 Description: random memory corruption
510 Details : size of struct ll_inode_info is to big for union inode.u and this
511 can be cause of random memory corruption.
516 Description: Memory leak in recovery
517 Details : Lov_mds_md was not free in an error handler in mds_create_object.
518 It should also check obd_fail before fsfilt_start, otherwise if
519 fsfilt_start return -EROFS,(failover mds during mds recovery).
520 then the req will return with repmsg->transno = 0 and rc = EROFS.
521 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
522 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
523 be freed no matter whether fsfilt_commit success or not.
526 Frequency : only with huge count clients
528 Description: Prevents from taking the superblock lock in llap_from_page for
530 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
531 for llap_from_page call in ll_removepage prevents from taking the
532 superblock lock for a soon died page.
537 Description: Not check open intent error before release open handle
538 Details : in some rare cases, the open intent error is not checked before
539 release open handle, which may cause
540 ASSERTION(open_req->rq_transno != 0), because it tries to release
541 the failed open handle.
546 Description: Set cat log bitmap only after create log success.
547 Details : in some rare cases, the cat log bitmap is set too early. and it
548 should be set only after create log success.
552 Description: Accessing a block bevice can re-enable I/O when Lustre is
553 tearing down a device.
554 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
559 Description: service threads may hog cpus when there are a lot of requests
561 Details : Insert cond_resched to give other threads a chance to use some of
567 Description: the cat log was not initialized in recovery
568 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
569 unlink log on mds will not be initialized until mds post
570 recovery. And also in mds post recovery, the unlink log will
571 initialization will be done asynchronausly, so there will be race
572 between add unlink log and unlink log initialization.
576 Description: brw_stats were being printed incorrectly
577 Details : brw_stats were being printed as log2 but all of them were not
578 recorded as log2. Also remove some code duplication arising from
579 filter_tally_{read,write}.
583 Frequency : rare, only in recovery.
584 Description: ASSERTION(req->rq_type != LI_POISON) failed
585 Details : imp_lock should be held while iterating over imp_sending_list for
586 prevent destroy request after get timeout in ptlrpc_queue_wait.
590 Description: replay-single.sh test 52 fails
591 Details : A lock's skiplist need to be cleanup when it being unlinked
592 from its resource list.
596 Description: Short directio read returns full requested size rather than
598 Details : Direct I/O operations should return actual amount of bytes
599 transferred rather than requested size.
601 Severity : enhancement
603 Description: metadata RPC reduction (e.g. for rm performance)
604 Details : decrease the amount of synchronous RPC between clients and servers
605 by canceling conflicing lock before the operation on the client side
606 and packing thier handles into the main operation RPC to server.
608 Severity : enhancement
610 Description: Async OSC create to avoid the blocking unnecessarily.
611 Details : If a OST has no remain object, system will block on the creating
612 when need to create a new object on this OST. Now, ways use
613 pre-created objects when available, instead of blocking on an
614 empty osc while others are not empty. If we must block, we block
615 for the shortest possible period of time.
617 --------------------------------------------------------------------------------
619 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
624 Frequency : on some architectures
626 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
627 Details : Uninitialized flags sometimes cause configuration commands to
631 Frequency : patchless clients only
633 Description: missing __iget() symbol export
634 Details : The __iget() symbol export is missing. To avoid the need for
635 this on patchless clients the deathrow inode reaper is turned
636 off, and we depend on the VM to clean up old inodes. This
637 dependency was during via the fix for bug 12181.
639 --------------------------------------------------------------------------------
641 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
643 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
644 INTEROPERATE with older versions automatically. In many cases a
645 special upgrade step is needed. Please read the
646 user documentation before upgrading any part of a live system.
647 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
648 version will not interoperate with 1.6 betas before beta5 (1.5.95).
649 * WARNING: Lustre configuration and startup changes are required with
650 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
656 Severity : enhancement
658 Description: MountConf
659 Details : Lustre configuration is now managed via mkfs and mount
660 commands instead of lmc and lconf. New obd types (MGS, MGC)
661 are added for dynamic configuration management. See
662 https://mail.clusterfs.com/wikis/lustre/MountConf for
665 Severity : enhancement
667 Description: dynamic OST addition
668 Details : OSTs can now be added to a live filesystem
670 Severity : enhancement
672 Description: startup order invariance
673 Details : MDTs and OSTs can be started in any order. Clients only
674 require the MDT to complete startup.
676 Severity : enhancement
678 Description: parallel, asynchronous orphan cleanup
679 Details : orphan cleanup is now performed in separate threads for each
680 OST, allowing parallel non-blocking operation.
682 Severity : enhancement
684 Description: optimized stripe assignment
685 Details : stripe assignments are now made based on ost space available,
686 ost previous usage, and OSS previous usage, in order to try
687 to optimize storage space and networking resources.
689 Severity : enhancement
691 Description: Permanently set tunables
692 Details : All writable /proc/fs/lustre tunables can now be permanently
693 set on a per-server basis, at mkfs time or on a live
696 Severity : enhancement
698 Description: Lustre message v2
699 Details : Add lustre message format v2.
701 Severity : enhancement
703 Description: client OST exclusion list
704 Details : Clients can be started with a list of OSTs that should be
705 declared "inactive" for known non-responsive OSTs.
709 Description: SPEC SFS validation failure on NFS v2 over lustre.
710 Details : Changes the blocksize for regular files to be 2x RPC size,
711 and not depend on stripe size.
713 Severity : enhancement
715 Description: Multiple MD RPCs in flight.
716 Details : Further unserialise some read-only MDS RPCs - learn about intents.
717 To avoid overly-overloading MDS, introduce a limit on number of
718 MDS RPCs in flight for a single client and add /proc controls
719 to adjust this limit.
721 Severity : enhancement
723 Description: client read/write statistics
724 Details : Add client read/write call usage stats for performance
725 analysis of user processes.
726 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
727 file access. extents_stats shows chunk size distribution.
728 extents_stats_per_process show chunk size distribution per
731 Severity : enhancement
733 Description: mds statistics
734 Details : Add detailed mds operations statistics in
735 /proc/fs/lustre/mds/*/stats.
739 Description: Failure of copying files with lustre special EAs.
740 Details : Client side always return success for setxattr call for lustre
741 special xattr (currently only "trusted.lov").
746 Description: Refcount LNET uuids
747 Details : The global LNET uuid list grew linearly with every startup;
748 refcount repeated list entries instead of always adding to
751 Severity : enhancement
753 Description: Dynamic service threads
754 Details : Within a small range, start extra service threads
755 automatically when the request queue builds up.
758 Frequency : mixed-endian client/server environments
760 Description: mixed-endian crashes
761 Details : The new msg_v2 system had some failures in mixed-endian
764 Severity : enhancement
766 Description: Easy OST removal
767 Details : OSTs can be permanently deactivated with e.g. 'lctl
768 conf_param lustre-OST0001.osc.active=0'
770 Severity : enhancement
772 Description: MGS proc entries
773 Details : Added basic proc entries for the MGS showing what filesystems
776 Severity : enhancement
778 Description: provide MGS failover
779 Details : Added config lock reacquisition after MGS server failover.
781 Severity : enhancement
783 Description: add Linux 2.4 support
784 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
788 Description: a large application tries to do I/O to the same resource and dies
790 Details : Check the req->rq_arrival time after the call to
791 ost_brw_lock_get(), but before we do anything about
792 processing it & sending the BULK transfer request. This
793 should help move old stale pending locks off the queue as
794 quickly as obd_timeout.
797 Frequency : when an incorrect nid is specified during startup
799 Description: ptlrpc connect to non-existant node causes kernel crash
800 Details : LNET can't be re-entered from an event callback, which
801 happened when we expire a message after the export has been
802 cleaned up. Instead, hand the zombie cleanup off to another
805 Severity : enhancement
807 Description: plain/inodebits lock performance improvement
808 Details : Grouping plain/inodebits in granted list by their request modes
809 and bits policy, thus improving the performance of search through
813 Frequency : only if OST filesystem is corrupted
815 Description: client incorrectly hits assertion in ptlrpc_replay_req()
816 Details : for a short time RPCs with bulk IO are in the replay list,
817 but replay of bulk IOs is unimplemented. If the OST filesystem
818 is corrupted due to disk cache incoherency and then replay is
819 started it is possible to trip an assertion. Avoid putting
820 committed RPCs into the replay list at all to avoid this issue.
823 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
825 Description: System hang on startup
826 Details : This bug allowed the liblustre (e.g. catamount) client to
827 return to the app before handling all startup RPCs. This
828 could leave the node unresponsive to lustre network traffic
829 and manifested as a server ptllnd timeout.
831 Severity : enhancement
833 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
834 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
835 boolean to control whether to print peer debug info when a
836 client's RPC times out.
839 Frequency : only for kernels with patches from Lustre below 1.4.3
841 Description: Remove old rdonly API
842 Details : Remove old rdonly API which unused from at least lustre 1.4.3
845 Frequency : only for devices with external journals
847 Description: Set external device read-only also
848 Details : During a commanded failover stop, we set the disk device
849 read-only while the server shuts down. We now also set any
850 external journal device read-only at the same time.
853 Frequency : when upgrading from 1.4 while trying to change parameters
855 Description: The wrong (new) MDC name was used when setting parameters for
856 upgraded MDT's. Also allows changing of OSC (and MDC)
857 parameters if --writeconf is specified at tunefs upgrade time.
860 Frequency : when setting specific ost indicies
862 Description: QOS code breaks on skipped indicies
863 Details : Add checks for missing OST indicies in the QOS code, so OSTs
864 created with --index need not be sequential.
866 Severity : enhancement
868 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
869 Details : The uninit_groups feature works in conjunction with the kernel
870 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
871 up the pass1 processing of e2fsck. This is a read-only feature
872 in ldiskfs2 only, so older kernels and current ldiskfs cannot
873 mount filesystems that have had this feature enabled.
875 Severity : enhancement
877 Description: Improve multi-block allocation algorithm to avoid fragmentation
878 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
879 allocation locality and avoid filesystem fragmentation.
881 ------------------------------------------------------------------------------
883 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
885 * Support for kernels:
886 2.4.21-47.0.1.EL (RHEL 3)
888 2.6.9-42.0.10.EL (RHEL 4)
889 2.6.12.6 vanilla (kernel.org)
890 2.6.16.27-0.9 (SLES 10)
891 * Recommended e2fsprogs version: 1.39.cfs5
893 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
897 Frequency : occasional, depends on client load and configuration
898 Bugzilla : 12181, 12203
899 Description: data loss for recently-modified files
901 Details : In some cases it is possible that recently written or created
902 files may not be written to disk in a timely manner (this should
903 normally be within 30s unless client IO load is very high).
904 The problem appears as zero-length files or files that are a
905 multiple of 1MB in size after a client crash or client eviction
906 that are missing data at the end of the file.
908 This problem is more likely to be hit on clients where files are
909 repeatedly created and unlinked in the same directory, clients
910 have a large amount of RAM, have many CPUs, the filesystem has
911 many OSTs, the clients are rebooted frequently, and/or the files
912 are not accessed by other nodes after being written.
914 The presence of the problem can be detected by looking at
915 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
916 smaller than the second (nr_unused) then dirty files will not
917 be flushed automatically to disk. "sync; sleep 10" should be
918 run several times on the node before unmounting it to update
919 Lustre (this is also safe to run on nodes without this problem).
921 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
922 that can cause this same problem, so customers using that kernel
923 also need to update the kernel in addition to Lustre. In order
924 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
926 It is normal that files written just before a client crash (less
927 than 30s) may not yet have been flushed to disk, even for local
931 Frequency : frequent on thin XT3 nodes
933 Description: UUID collision on thin XT3 Linux nodes
934 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
935 randomly, since we relied on an insufficiently-seeded PRNG.
940 Description: OSS hangs after "All ost request buffers busy"
941 Details : A deadlock between quota and journal operations caused OSS
942 hangs after printing "All ost request buffers busy."
945 Frequency : always on liblustre builds
947 Description: Cleanup compiler warnings on liblustre
950 Frequency : always on liblustre builds on XT3
952 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
956 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
957 > 32000 subdirectories
958 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
959 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
960 subdirectory count crosses 32000. This will aid e2fsck to
961 correctly handle more than 32000 subdirectories.
964 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
966 Description: System hang on startup
967 Details : This bug allowed the liblustre (e.g. catamount) client to
968 return to the app before handling all startup RPCs. This
969 could leave the node unresponsive to lustre network traffic
970 and manifested as a server ptllnd timeout.
972 Severity : enhancement
974 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
975 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
976 boolean to control whether to print peer debug info when a
977 client's RPC times out.
982 Description: make O_SYNC working on 2.6 kernels
983 Details : 2.6 kernels use different method for mark pages for write,
984 so need add a code to lustre for O_SYNC work.
989 Description: Failure to close file and release space on NFS
990 Details : Put inode details into lock acquired in ll_intent_file_open.
991 Use mdc_intent_lock in ll_intent_open to properly
992 detect all kind of errors unhandled by mdc_enqueue.
997 Description: proc file read during shutdown sometimes raced obd removal,
999 Details : Add lock to prevent obd access after proc file removal.
1002 Frequency : Only for files larger than 4GB on 32-bit clients.
1004 Description: improperly doing page alignment of locks
1005 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1006 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1011 Description: RPCs being resent when they shouldn't be
1012 Details : Some RPCs that should not be resent are being resent. This
1013 can cause inconsistencies in the RPC state machine. Do not
1014 resend such requests.
1017 Frequency : rare, only with NFS export
1019 Description: Crash on NFS re-export node
1020 Details : under very unusual load conditions an assertion is hit in
1021 ll_intent_file_open()
1024 Frequency : only if OST filesystem is corrupted
1026 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1027 Details : for a short time RPCs with bulk IO are in the replay list,
1028 but replay of bulk IOs is unimplemented. If the OST filesystem
1029 is corrupted due to disk cache incoherency and then replay is
1030 started it is possible to trip an assertion. Avoid putting
1031 committed RPCs into the replay list at all to avoid this issue.
1036 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1037 Details : Large single O_DIRECT read and write calls can fail to allocate
1038 a sufficiently large buffer to process the request. In case of
1039 allocation failure the allocation is retried with a smaller
1040 buffer and broken into smaller requests.
1042 Severity : enhancement
1044 Description: Add -o localflock option to simulate old noflock behaviour.
1045 Details : This will achieve local-only flock/fcntl locks coherentness.
1050 Description: versioning check is incomplete
1051 Details : Checking the version difference of client vs. server, report
1052 error if the gap is too big.
1057 Description: add support PG_writeback bit
1058 Details : add support for PG_writeback bit for Lustre, for more carefull
1059 work with page cache in 2.6 kernel. This also fix some deadlocks
1060 and remove hack for work O_SYNC with 2.6 kernel.
1062 Severity : enhancement
1064 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1065 Details : The uninit_groups feature works in conjunction with the kernel
1066 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1067 up the pass1 processing of e2fsck. This is a read-only feature
1068 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1069 mount filesystems that have had this feature enabled.
1071 Severity : enhancement
1073 Description: Improve multi-block allocation algorithm to avoid fragmentation
1074 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1075 allocation locality and avoid filesystem fragmentation.
1077 ------------------------------------------------------------------------------
1079 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1081 * Support for kernels:
1082 2.6.9-42.0.3.EL (RHEL 4)
1083 2.6.5-7.276 (SLES 9)
1084 2.4.21-47.0.1.EL (RHEL 3)
1085 2.6.12.6 vanilla (kernel.org)
1086 2.6.16.21-0.8 (SLES10)
1087 * Recommended e2fsprogs version: 1.39.cfs2-0
1089 * The backwards-compatible /proc/sys/portals symlink has been removed
1090 in this release. Before upgrading, please ensure that you change
1091 any configuration scripts or /etc/sysctl.conf files that access
1092 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1093 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1094 in advance of the upgrade on any system running Lustre 1.4.6 or
1095 newer, since /proc/sys/lnet was added in that version.
1096 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1100 Frequency : only when quota is used
1102 Description: avoid scanning export list for quota master
1103 Details : Change the algorithms to avoid scanning export list in order
1104 to improve the efficiency.
1107 Frequency : MDS failover only, very rarely
1109 Description: "went back in time" messages on mds failover
1110 Details : The greatest transno may be lost when the current operation
1111 finishes with an error (transno==0) and the client's last_rcvd
1112 record is over-written. Save the greatest transno in the
1113 mds_last_transno for this case.
1116 Frequency : always for specific kernels and striping counts
1118 Description: client may get "Matching packet too big" without ACL support
1119 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1120 when trying to access files in certain configurations. The
1121 clients should in fact be denied when mounting because they do
1122 not understand ACLs.
1125 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1127 Description: many clients connecting with IO in progress causes connect timeouts
1128 Details : Avoid synchronous journal commits to avoid delays caused by many
1129 clients connecting/disconnecting when bulk IO is in progress.
1130 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1131 OST_IO_PORTAL to avoid delays behind potentially many pending
1135 Frequency : occasionally with multiple writers to a single file
1137 Description: shared writes to file may result in wrong size reported by stat()
1138 Details : Allow growing of kms when extent lock is cancelled
1141 Frequency : always with random mmap IO to multi-striped file
1143 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1144 Details : Only if the hole is at the end of OST object so that kms is too
1145 small. Fix is to increase kms accordingly in ll_nopage.
1148 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1150 Description: writes to a missing object would leak memory on the OST
1151 Details : If there is an inconsistency between the MDS and OST filesystems,
1152 such that the MDS references an object that doesn't exist, writes
1153 to that object will leak memory due to incorrect cleanup in the
1154 error handling path, eventually running out of memory on the OST.
1159 Description: Creating too long symlink causes lustre errors
1160 Details : Check symlink and name lengths before sending requests to MDS.
1163 Frequency : only if flock is enabled (not on by default)
1165 Description: posix locks not released on fd closure on 2.6.9+
1166 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1167 kernels, this caused such locks not to be released on fd close and
1168 then assertions on fs unmount about still used locks.
1171 Frequency : MDS failover only, very rarely
1173 Description: clients may get ASSERTION(granted_lock != NULL)
1174 Details : When request was taking a long time, and a client was resending
1175 a getattr by name lock request. The were multiple lock requests
1176 with the same client lock handle and
1177 mds_getattr_name->fixup_handle_for_resent_request found one of the
1178 lock handles but later failed with ASSERTION(granted_lock != NULL).
1183 Description: handle->h_buffer_credits > 0, assertion failure
1184 Details : h_buffer_credits is zero after truncate, causing assertion
1185 failure. This patch extends the transaction or creates a new
1189 Frequency : NFS re-export or patchless client
1190 Bugzilla : 11179, 10796
1191 Description: Crash on NFS re-export node (__d_move)
1192 Details : We do not want to hash the dentry if we don't have a lock.
1193 But if this dentry is later used in d_move, we'd hit uninitialised
1194 list head d_hash, so we just do this to init d_hash field but
1195 leave dentry unhashed.
1198 Frequency : NFS re-export or patchless client
1200 Description: NFS exports has problem with symbolic link
1201 Details : lustre client didn't properly install dentry when re-exported
1202 to NFS or running patchless client.
1205 Frequency : NFS re-export or patchless client
1207 Description: Various nfs/patchless fixes.
1208 Details : fixes reuse disconected alias for lookup process - this fixes
1209 warning "find_exported_dentry: npd != pd",
1210 fix permission error with open files at nfs.
1211 fix apply umask when do revalidate.
1214 Frequency : occasional
1216 Description: Crash on NFS re-export node
1217 Details : calling clear_page() on the wrong pointer triggered oops in
1218 generic_mapping_read().
1221 Frequency : rarely, using O_DIRECT IO
1223 Description: unaligned directio crashes client with LASSERT
1224 Details : check for unaligned buffers before trying any requests.
1227 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1229 Description: stale data returned from RAID cache
1230 Details : If only a small amount of IO is done to the RAID device before
1231 reading it again it is possible to get stale data from the RAID
1232 cache instead of reading it from disk.
1235 Frequency : always for sles10 kernel
1237 Description: sles10 support
1238 Details : ll_follow_link: compile fixes and using of nd_set_link
1239 under newer kernels.
1242 Frequency : depends on arch, kernel and compiler version, always on sles10
1245 Description: recursive or deep enough symlinks cause stack overflow
1246 Details : getting rid of large stack-allocated variable in
1250 Frequency : depends on hardware
1252 Description: lustre write performance loss in the SLES10 kernel
1253 Details : the performance loss is caused by using of write barriers in the
1254 ext3 code. The SLES10 kernel turns barrier support on by
1255 default. The fix is to undo that change for ldiskfs.
1257 ------------------------------------------------------------------------------
1259 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1261 * Support for kernels:
1262 2.6.9-42.0.3EL (RHEL 4)
1263 2.6.5-7.276 (SLES 9)
1264 2.4.21-47.0.1.EL (RHEL 3)
1265 2.6.12.6 vanilla (kernel.org)
1269 Frequency : quota enabled and large files being deleted
1271 Description: releasing more than 4GB of quota at once hangs OST
1272 Details : If a user deletes more than 4GB of files on a single OST it
1273 will cause the OST to spin in an infinite loop. Release
1274 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1279 Description: statfs data retrieved from /proc may be stale or zero
1280 Details : When reading per-device statfs data from /proc, in the
1281 {kbytes,files}_{total,free,avail} files, it may appear
1282 as zero or be out of date.
1285 Frequency : systems with MD RAID1 external journal devices
1287 Description: lconf's call to blkid is confused by RAID1 journal devices
1288 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1289 of returning all block devices that match the journal UUID.
1292 Frequency : always, for aggregate stripe size over 4GB
1294 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1295 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1296 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1299 Frequency : always if "lfs find" used on a local file/directory
1301 Description: "lfs find" segfaults if used on a local file/directory
1302 Details : The case where a directory component was not specified wasn't
1303 handled correctly. Handle this properly.
1306 Frequency : always on ppc64
1308 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1309 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1310 perform endianness conversion.
1313 Frequency : rarely (truncate to non-zero file size after write under load)
1314 Bugzilla : 10730, 10687
1315 Description: Files padded with zeros to next 4K multiple
1316 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1317 it is possible that files that are truncated to a non-zero size
1318 immediately after being written are filled with zero bytes beyond
1319 the truncated size. No file data is lost.
1321 Severity : enhancement
1323 Description: Allow recovery/failover for liblustre clients.
1324 Details : liblustre clients were unaware of failover configurations until
1327 Severity : enhancement
1329 Description: user file locks should fail when not mounting with flock option
1330 Details : Set up an error-returning stub in ll_file_operations.lock field
1331 to prevent incorrect behaviour when client is mounted without
1332 flock option. Also, set up properly f_op->flock field for
1336 Frequency : always on ia64
1338 Description: "lfs df" loops on printing out MDS statfs information
1339 Details : The obd_ioctl_data was not initialized and in some systems
1340 this caused a failure during the ioctl that did not return
1341 an error. Initialize the struct and return an error on failure.
1344 Frequency : SLES 9 only
1346 Description: Error of copying files with lustre special EAs as root
1347 Details : Client side always return success for setxattr call for lustre
1348 special xattr (currently only "trusted.lov").
1351 Frequency : rarely on clusters with both ia64+i386 clients
1353 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1354 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1355 clients are doing concurrent writes to the same file it is
1356 possible that the ia64 clients may LASSERT because the OST
1357 extent locks are not PAGE_SIZE aligned. Ensure that grown
1358 locks are always aligned on the request boundary.
1361 Frequency : specific use, occasional
1363 Description: Overwriting in use executable truncates on-disk binary image
1364 Details : If one node attempts to overwrite an executable in use by
1365 another node, we now correctly return ETXTBSY instead of
1366 truncating the file.
1368 Severity : enhancement
1370 Description: Async OSC create to avoid the blocking unnecessarily.
1371 Details : If a OST has no remain object, system will block on the creating
1372 when need to create a new object on this OST. Now, ways use
1373 pre-created objects when available, instead of blocking on an
1374 empty osc while others are not empty. If we must block, we block
1375 for the shortest possible period of time.
1380 Description: chmod on Lustre root is propagated to other clients
1381 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1382 invalid by the follow_mount time.
1385 Frequency : liblustre clients only
1387 Description: Race in 'instant cancel' lock handling could lead to such locks
1388 never to be granted in case of SMP MDS
1389 Details : Do not destroy not yet granted but cbpending locks in
1393 Frequency : replay/resend of open
1395 Description: non null lock assetion failure in mds_intent_policy
1396 Details : Trying to replay/resend lockless open requests resulted in
1397 mds_open() returning 0 with no lock. Now it sets a flag if
1398 a lock is going to be returned.
1400 Severity : enhancement
1402 Description: Checksum enhancements
1403 Details : New checksum enhancements allow for resending RPCs that failed
1406 Severity : enhancement
1408 Description: Tunables on number of dirty pages in cacche
1409 Details : Allow to set limit on number of dirty pages cached.
1414 Description: client crash on unmount - lock still has references
1415 Details : In some error handling cases it was possible to leak a lock
1416 reference on a client while accessing a file. This was not
1417 harmful to the client during operation, but would cause the
1418 client to crash when the filesystem is unmounted.
1421 Frequency : specific case, rare
1423 Description: ETXTBSY on mds though file not in use by client
1424 Details : ETXTBSY is no longer incorrectly returned when attempting to
1425 chmod or chown a directory that the user previously tried to
1426 execute or a currently-executing binary.
1429 Frequency : extremely rare except on liblustre-based clients
1431 Description: Lustre space not freed when files are deleted
1432 Details : Clean up open-unlinked files after client eviction. Previously
1433 the unlink was skipped and the files remained as orphans.
1438 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1439 Details : In some cases it was possible to send a blocking callback to
1440 a client doing a glimpse, even though that client didn't get
1441 a lock granted. When the glimpse lock is cancelled on the OST
1442 the freed lock is left on the waiting list and corrupted the list.
1445 Frequency : all core dumps
1447 Description: Broke core dumps to lustre
1448 Details : Negative dentry may be unhashed if parent does not have UPDATE
1449 lock, but some callers, e.g. do_coredump, expect dentry to be
1450 hashed after successful create, hash it in ll_create_it.
1452 ------------------------------------------------------------------------------
1454 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1456 * Support for kernels:
1457 2.6.9-42.0.2.EL (RHEL 4)
1458 2.6.5-7.276 (SLES 9)
1459 2.4.21-40.EL (RHEL 3)
1460 2.6.12.6 vanilla (kernel.org)
1464 Frequency : always on RHEL 3
1466 Description: Number of open files grows over time
1467 Details : The number of open files grows over time, whether or not
1468 Lustre is started. This was due to a filp leak introduced
1469 by one of our kernel patches.
1471 ------------------------------------------------------------------------------
1473 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1475 * Support for kernels:
1476 2.6.9-42.EL (RHEL 4)
1477 2.6.5-7.276 (SLES 9)
1478 2.4.21-40.EL (RHEL 3)
1479 2.6.12.6 vanilla (kernel.org)
1484 Bugzilla : 5719, 9635, 9792, 9684
1485 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1486 Details : If a server is under heavy load and cannot reply to new
1487 connection requests before the client resends the (re)connect,
1488 the connection handling code can behave badly if two service
1489 threads are concurrently handing separate (re)connections from
1490 the same client. Add better locking to the connection handling
1491 code, and ensure that only a single connection will be processed
1492 for a given client UUID, even if the lock is dropped.
1494 Severity : enhancement
1496 Description: add TCP zero-copy support to kernel
1497 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1498 sends if the hardware supports scatter-gather and checksumming.
1499 This allows socklnd to do client-write and server-read more
1500 efficiently and reduce CPU utilization from skbuf copying.
1503 Frequency : only if NFS exporting from client
1505 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1506 Details : If setting attributes on a file created under NFS that had
1507 never been opened it would be possible to oops the client
1508 if the file had no objects.
1511 Frequency : always for liblustre
1513 Description: liblustre client does MDS+OSTs setattr RPC for each write
1514 Details : When doing a write from a liblustre client, the client
1515 incorrectly issued an RPC to the MDS and each OST the file was
1516 striped over in order to update the timestamps. When writing
1517 with small chunks and many clients this could overwhelm the MDS
1518 with RPCs. In all cases it would slow down the write because
1519 these RPCs are unnecessary.
1521 Severity : enhancement
1523 Description: allow number of MDS service threads to be changed at module load
1524 Details : It is now possible to change the number of MDS service threads
1525 running. Adding "options mds mds_num_threads={N}" to the MDS's
1526 /etc/modprobe.conf will set the number of threads for the next
1527 time Lustre is restarted (assuming the "mds" module is also
1528 reloaded at that time). The default number of threads will
1529 stay the same, 32 for most systems.
1534 Description: OST crash if filesystem is unformatted or corrupt
1535 Details : If an OST is started on a device that has never been formatted
1536 or if the filesystem is corrupt and cannot even mount then the
1537 error handling cleanup routines would dereference a NULL pointer.
1542 Description: NULL pointer deref in llap_from_page.
1543 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1544 incorrect) mapping if the page was truncated/reclaimed while it was
1545 searched for. Check for this condition and skip such pages when
1546 doing readahead. Introduce extra check to llap_from_page() to
1547 verify page->mapping->host is non-NULL (so page is not anonymous).
1550 Frequency : Sometimes when using sys_sendfile
1552 Description: "page not covered by a lock" warnings from ll_readpage
1553 Details : sendfile called ll_readpage without right page locks present.
1554 Now we introduced ll_file_sendfile that does necessary locking
1555 around call to generic_file_sendfile() much like we do in
1559 Frequency : with certain MDS communication failures at client mount time
1561 Description: NULL pointer deref after failed client mount
1562 Details : a client connection request may delayed by the network layer
1563 and not be sent until after the PTLRPC layer has timed out the
1564 request. If the client fails the mount immediately it will try
1565 to clean up before the network times out the request. Add a
1566 reference from the request import to the obd device and delay
1567 the cleanup until the network drops the request.
1570 Frequency : occasionally during client (re)connect
1572 Description: assertion failure during client (re)connect
1573 Details : processing a client connection request may be delayed by the
1574 client or server longer than the client connect timeout. This
1575 causes the client to resend the connection request. If the
1576 original connection request is replied in this interval, the
1577 client may trip an assertion failure in ptlrpc_connect_interpret()
1578 which thought it would be the only running connect process.
1581 Frequency : only with obd_echo servers and clients that are rebooted
1583 Description: kernel BUG accessing uninitialized data structure
1584 Details : When running an obd_echo server it did not start the ping_evictor
1585 thread, and when a client was evicted an uninitialized data
1586 structure was accessed. Start the ping_evictor in the RPC
1587 service startup instead of the OBD startup.
1589 Severity : enhancement
1590 Bugzilla : 10193 (patchless)
1591 Description: Remove dependency on various unexported kernel interfaces.
1592 Details : No longer need reparent_to_init, exit_mm, exit_files,
1593 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1596 Frequency : rare (only users of deprecated and unsupported LDAP config)
1598 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1599 Details : LDAP apparently contains 'lustreName' attributes instead of
1600 'name'. A simple remapping of the name is sufficient.
1603 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1605 Description: waiting_locks_callback trips kernel BUG if client is evicted
1606 Details : Running with the dump_on_timeout debug flag turned on makes
1607 it possible that the waiting_locks_callback() can try to dump
1608 the Lustre kernel debug logs from an interrupt handler. Defer
1609 this log dumping to the expired_lock_main() thread.
1611 Severity : enhancement
1613 Description: Support NFS exporting on 2.6 kernels.
1614 Details : Implement non-rawops metadata methods for NFS server to use without
1615 changing NFS server code.
1618 Frequency : very rare (synthetic metadata workload only)
1620 Description: two racing renames might cause an MDS thread to deadlock
1621 Details : Running the "racer" program may cause one MDS thread to rename
1622 a file from being the source of a rename to being the target of
1623 a rename at exactly the same time that another thread is doing
1624 so, and the second thread has already enqueued these locks after
1625 doing a lookup of the target and is trying to relock them in
1626 order. Ensure that we don't try to re-lock the same resource.
1629 Frequency : only very large systems with liblustre clients
1631 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1632 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1633 all OSTs in parallel. This allows the eviction of stale liblustre
1634 clients to proceed much faster than if they were done in series,
1635 and also offers similar improvements for other set_info RPCs.
1640 Description: excessive CPU usage during initial read phase on client
1641 Details : During the initial read phase on a client, it would agressively
1642 retry readahead on the file, consuming too much CPU and impacting
1643 performance (since 1.4.5.8). Improve the readahead algorithm
1644 to avoid this, and also improve some other common cases (read
1645 of small files in particular, where "small" is files smaller than
1646 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1651 Description: MDS crash when receiving packet with unknown intent.
1652 Details : Do not LBUG in unknown intent case, just return -EFAULT
1654 Severity : enhancement
1655 Bugzilla : 9293, 9385
1656 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1657 Details : Do not serialize getattr (non-intent version) and statfs.
1660 Frequency : occasional, when OST network is overloaded/intermittent
1662 Description: client evicted by OST after bulk IO timeout
1663 Details : If a client sends a bulk IO request (read or write) the OST
1664 may evict the client if it is unresposive to its data GET/PUT
1665 request. This is incorrect if the network is overloaded (takes
1666 too long to transfer the RPC data) or dropped the OST GET/PUT
1667 request. There is no need to evict the client at all, since
1668 the pinger and/or lock callbacks will handle this, and the
1669 client can restart the bulk request.
1672 Frequency : Always when mmapping file with no objects
1674 Description: client crashes when mmapping file with no objects
1675 Details : Check that we actually have objects in a file before doing any
1676 operations on objects in ll_vm_open, ll_vm_close and
1682 Description: Request leak when working with deleted CWD
1683 Details : Introduce advanced request refcount tracking for requests
1684 referenced from lustre intent.
1686 Severity : Enhancement
1688 Description: Cache open file handles on client.
1689 Details : MDS now will return special lock along with openhandle, if
1690 requested and client is allowed to hold openhandle, even if unused,
1691 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1692 closing files for every read/write openration.
1694 Severity : Enhancement
1696 Description: Cache open negative dentries on client when possible.
1697 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1698 negative dentries on lock revocation.
1703 Description: Remounting a client read-only wasn't possible with a zconf mount
1704 Details : It wasn't possible to remount a client read-only with llmount.
1706 Severity : enhancement
1707 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1708 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1709 MPICH is included to add Lustre-specific ADIO interfaces.
1710 This is based closely on the UFS ADIO layer and only differs
1711 in file creation, in order to allow the OST striping to be set.
1712 This is user-contributed code and not supported by CFS.
1717 Description: extended inode attributes (immutable, append-only) work improperly
1718 when 2.4 and 2.6 kernels are used on client/server or vice versa
1719 Details : Introduce kernel-independent values for these flags.
1721 Severity : enhancement
1724 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1725 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1726 granularity. Now, Fractional MB granularity is be supported,
1727 this is very useful for low memory system.
1729 Severity : enhancement
1731 Description: Getattr by fid
1732 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1733 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1734 have locks and avoids getattr rpc requests for it on every stat.
1737 Frequency : Always, for filesystems larger than 2TB
1739 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1740 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1741 filesystems larger than 2TB to be mounted with mballoc enabled.
1744 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1746 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1747 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1748 filesystems larger than 2TB is mounted then the kernel will
1749 silently corrupt the start of the filesystem. CONFIG_LBD is
1750 enabled for all CFS-supported kernels, but the possibility of
1751 this happening with a modified kernel config exists.
1753 Severity : enhancement
1755 Description: add client O_DIRECT support for 2.6 kernels
1756 Details : It is now possible to do O_DIRECT reads and writes to files
1757 in the Lustre client mountpoint on 2.6 kernel clients.
1759 Severity : enhancement
1761 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1762 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1763 parallel, not waiting for response from every OST before sending
1764 a rpc to the next OST.
1769 Description: setattr vs write race when updating file timestamps
1770 Details : Client processes that update a file timestamp into the past
1771 right after writing to the file (e.g. tar) it is possible that
1772 the updated file modification time can be reset to the current
1773 time due to a race between processing the setattr and write RPC.
1775 Severity : enhancement
1777 Description: Bring 'lfs find' closer in line with regular Linux find.
1778 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1779 -print0 options and obtains all the needed info through the lustre
1782 Severity : enhancement
1784 Description: support up to 1024 configured devices on one node
1785 Details : change obd_dev array from statically allocated to dynamically
1786 allocated structs as they are first used to reduce memory usage
1791 Description: Flush dirty partially truncated pages during truncate
1792 Details : Immediatelly flush partially truncated pages in filter_setattr,
1793 this way we completely avoid having any pages in page cache on OST
1794 and can retire ugly workarounds during writes to flush such pages.
1799 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1800 Details : protect inode from truncation within vfs_unlink() context
1801 just take a reference before calling vfs_unlink() and release it
1802 when parent's i_sem is free.
1807 Description: last_id value checked outside lock on OST caused LASSERT failure
1808 Details : If there were multiple MDS->OST object precreate requests in
1809 flight, it was possible that the OST's last object id was checked
1810 outside a lock and incorrectly tripped an assertion. Move checks
1811 inside locks, and discard old precreate requests.
1814 Frequency : always, if extents are used on OSTs
1816 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1817 Details : OSTs using the extents format would not zero the high 16 bits of
1818 the index physical block number. This is not a problem for any
1819 OST filesystems smaller than 16TB, and no kernels support ext3
1820 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1821 new/modified files) and can be fixed for existing filesystems
1822 with e2fsprogs-1.39-cfs1.
1827 Description: import connection selection may be incorrect if timer wraps
1828 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1829 import connections to be ignored if the 32-bit jiffies counter
1830 wraps. Use a 64-bit jiffies counter.
1833 Frequency : very large clusters immediately after boot
1835 Description: LNET request buffers exhausted under heavy short-term load
1836 Details : If a large number of client requests are generated on a service
1837 that has previously never seen so many requests it is possible
1838 that the request buffer growth cannot keep up with the spike in
1839 demand. Instead of dropping incoming requests, they are held in
1840 the LND until the RPC service can accept more requests.
1843 Frequency : Sometimes during replay
1845 Description: Assertion failure in ll_local_open after replay.
1846 Details : If replay happened on an open request reply before we were able
1847 to set replay handler, reply will become not swabbed tripping the
1848 assertion in ll_local_open. Now we set the handler right after
1849 recognising of open request
1852 Frequency : very rare
1854 Description: kernel reports "badness in vsnprintf"
1855 Details : Reading from the "recovery_status" /proc file in small chunks
1856 may cause a negative length in lprocfs_obd_rd_recovery_status()
1857 call to vsnprintf() (which is otherwise harmless). Exit early
1858 if there is no more space in the output buffer.
1860 Severity : enhancement
1862 Description: clear OBD RPC statistics by writing to them
1863 Details : It is now possible to clear the OBD RPC statistics by writing
1864 to the "stats" file.
1869 Description: Client mtime is not the same on different clients after utimes
1870 Details : In some cases, the client was using the utimes() syscall on
1871 a file cached on another node. The clients now validate the
1872 ctime from the MDS + OSTs to determine which one is right.
1877 Description: Inability to activate failout mode
1878 Details : lconf script incorrectly assumed that in pythong string's numeric
1879 value is used in comparisons.
1882 Frequency : always with multiple stripes per file
1884 Description: Inefficient object allocation for mutli-stripe files
1885 Details : When selecting which OSTs to stripe files over, for files with
1886 a stripe count that divides evenly into the number of OSTs,
1887 the MDS is always picking the same starting OST for each file.
1888 Return the OST selection heuristic to the original design.
1893 Description: mount failures may take full timeout to return an error
1894 Details : Under some heavy load conditions it is possible that a
1895 failed mount can wait for the full obd_timeout interval,
1896 possibly several minutes, before reporting an error.
1897 Instead return an error as soon as the status is known.
1899 Frequency : quota enabled and large files being deleted
1901 Description: releasing more than 4GB of quota at once hangs OST
1902 Details : If a user deletes more than 4GB of files on a single OST it
1903 will cause the OST to spin in an infinite loop. Release
1904 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1909 Description: statfs data retrieved from /proc may be stale or zero
1910 Details : When reading per-device statfs data from /proc, in the
1911 {kbytes,files}_{total,free,avail} files, it may appear
1912 as zero or be out of date.
1915 Frequency : systems with MD RAID1 external journal devices
1917 Description: lconf's call to blkid is confused by RAID1 journal devices
1918 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1919 of returning all block devices that match the journal UUID.
1922 Frequency : always, for aggregate stripe size over 4GB
1924 Description: assertion fails when trying to use 4GB stripe size
1925 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1926 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1929 Frequency : always on ppc64
1931 Description: the first write on an ext3 filesystem with mballoc got stuck
1932 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1933 perform endianness conversion.
1935 ------------------------------------------------------------------------------
1937 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1939 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1940 INTEROPERATE with older versions automatically. Please read the
1941 user documentation before upgrading any part of a live system.
1942 * WARNING: Lustre networking configuration changes are required with
1943 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1946 * Support for kernels:
1947 2.6.9-22.0.2.EL (RHEL 4)
1948 2.6.5-7.244 (SLES 9)
1949 2.6.12.6 vanilla (kernel.org)
1952 Severity : enhancement
1953 Bugzilla : 7981/8208
1954 Description: Introduced Lustre Networking (LNET)
1955 Details : LNET is new networking infrastructure for Lustre, it includes
1956 a reorganized network configuration mode (see the user
1957 documentation for full details) as well as support for routing
1958 between different network fabrics. Lustre Networking Devices
1959 (LNDs) for the supported network fabrics have also been
1960 created for this new infrastructure.
1962 Severity : enhancement
1963 Description: Introduced Access control lists
1964 Details : clients can set ACLs on files and directories in order to have
1965 more fine-grained permissions than the standard Unix UGO+RWX.
1966 The MDS must be started with the "-o acl" mount option.
1968 Severity : enhancement
1969 Description: Introduced filesystem quotas
1970 Details : Administrators may now establish per-user quotas on the
1973 Severity : enhancement
1975 Description: Configuration change for the XT3
1976 The PTLLND is now used to run Lustre over Portals on the XT3
1977 The configure option(s) --with-cray-portals are no longer used.
1978 Rather --with-portals=<path-to-portals-includes> is used to
1979 enable building on the XT3. In addition to enable XT3 specific
1980 features the option --enable-cray-xt3 must be used.
1985 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1986 Details : A race between allocating a new llcd and re-getting the llcd_lock
1987 allowed another thread to grab newly-allocated llcd.
1989 Severity : enhancement
1991 Description: 2.6 OST async journal commit and locking fix to improve performance
1992 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1993 2.6 kernels are now async as they already were in 2.4 kernels so
1994 that they can commit concurrently with the network bulk transfer.
1995 For block-allocated files the filter allocation semaphore is held
1996 to avoid filesystem fragmentation during allocation. BKL lock
1997 removed for 2.6 xattr operations where it is no longer needed.
2002 Description: lconf incorrectly determined whether two IP networks could talk
2003 Details : In some more complicated routing and multiple-network
2004 configurations, lconf will avoid trying to make a network
2005 connection to a disjoint part of the IP space. It was doing the
2006 math incorrectly for one set of cases.
2011 Description: Fix for potential infinite loop processing records in an llog.
2012 Details : If an llog record is corrupted/zeroed, it is possible to loop
2013 forever in llog_process(). Validate the llog record length
2014 and skip the remainder of the block on error.
2017 Frequency : occasional (liblustre only)
2019 Description: liblustre could not open files whose last component is a symlink
2020 Details : sysio_path_walk() would incorrectly pass the open intent to
2021 intermediate path components.
2024 Frequency : rare (liblustre only with non-standard tuning)
2025 Bugzilla : 7201 (7350)
2026 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2027 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2028 after it has been released, only do this for referenced locks
2030 Severity : enhancement
2032 Description: specifying an (invalid) directory default stripe_size of -1
2033 would reset the directory default striping
2034 Details : stripe_size -1 was used internally to signal directory stripe
2035 removal, now use "all default" to signal dir stripe removal
2036 as a directory striping of "all default" is not useful
2039 Frequency : common for large clusters running liblustre clients
2041 Description: doing an ls when liblustre clients are running is slow
2042 Details : sending a glimpse AST to a liblustre client waits for every AST
2043 to time out, as liblustre clients will not respond. Since they
2044 cannot cache data we refresh the OST lock LVB from disk instead.
2046 Severity : enhancement
2048 Description: doing an ls at the same time as file IO can be slow
2049 Details : enqueue and other "small" requests can be blocked behind many
2050 large IO requests. Create a new OST IO portal for non-IO
2051 requests so they can be processed faster.
2054 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2056 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2057 created file with mode 000
2058 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2059 allow an ftruncate() as a non-root user to a file with mode 000.
2060 Lustre now respects this flag to disable mode checks when
2061 truncating a file owned by the user
2064 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2067 Description: Revoking locks from clients that went dead or catatonic might take
2069 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2070 cancellation of such locks instant on servers without waiting for
2071 any reply from clients. Clients drops these locks when cancel
2072 notification from server is received without replying.
2075 Frequency : liblustre-only, when liblustre client dies or becomes busy
2077 Description: Doing ls on Linux clients can take a long time with active
2079 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2080 granting such locks to it in the first place if possible. Locks
2081 are taken by proxy on the OST during the read or write and
2082 dropped immediately afterward. Add connect flags handling, do
2083 not grant locks to liblustre clients for glimpse ASTs.
2085 Severity : enhancement
2087 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2088 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2089 sequential reads and behaves badly with applications doing random
2090 reads. Improve it by reading ahead at least read region, and
2091 avoiding excessive large RPC for small reads.
2093 Severity : enhancement
2095 Description: Creating more than 1000 files for a single job may cause a load
2096 imbalance on the OSTs if there are also a large number of OSTs.
2097 Details : qos_prep_create() uses an OST index reseed value that is an
2098 even multiple of the number of available OSTs so that if the
2099 reseed happens in the middle of the object allocation it will
2100 still utilize the OSTs as uniformly as possible.
2105 Description: OST or MDS may oops in ping_evictor_main()
2106 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2107 but doesn't restart at beginning of obd_exports_timed list
2110 Severity : enhancement
2112 Description: improve by-nid export eviction on the MDS and OST
2113 Details : allow multiple exports with the same NID to be evicted at one
2114 time without re-searching the exports list.
2117 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2119 Description: MDS may oops in groups_free()
2120 Details : in rare race conditions a newly allocated group_info struct is
2121 freed again, and this can be NULL. The 2.4 compatibility code
2122 for groups_free() checked for a NULL pointer, but 2.6 did not.
2125 Frequency : common for liblustre clients doing little filesystem IO
2126 Bugzilla : 9352, 7313
2127 Description: server may evict liblustre clients accessing contended locks
2128 Details : if a client is granted a lock or receives a completion AST
2129 with a blocking AST already set it would not reply to the AST
2130 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2133 Frequency : lfs setstripe, only systems with more than 160 OSTs
2135 Description: unable to set striping with a starting offset beyond OST 160
2136 Details : llapi_create_file() incorrectly limited the starting stripe
2137 index to the maximum single-file stripe count.
2140 Frequency : LDAP users only
2142 Description: lconf did not handle in-kernel recovery with LDAP properly
2143 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2145 Severity : enhancement
2147 Description: bind OST threads to NUMA nodes to improve performance
2148 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2149 node and do their allocations there to localize memory access
2151 Severity : enhancement
2153 Description: llmount can determine client NID directly from Myrinet (GM)
2154 Details : the client NID code from gmnalnid was moved directly into
2155 llmount, removing the need to use this or specifying the
2156 client NID explicitly when mounting GM clients with zeroconf
2159 Frequency : if client is started with down MDS
2161 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2162 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2163 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2164 ptlrpc_import_delayed_req() because we didn't check if the
2165 request was interrupted, and we also didn't break out of the
2166 event loop if there was no timeout
2171 Description: data loss during non-page-aligned writes to a single file from
2172 both multiple nodes and multiple threads on one node at same time
2173 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2174 inconsistency led to false short-reads, that were cached and later
2175 used by ->prepare_write() to fill in partially written page,
2176 leading to data loss.
2179 Frequency : always, if lconf --abort_recovery used
2181 Description: lconf --abort_recovery fails with 'Operation not supported'
2182 Details : lconf was attempting to abort recovery on the MDT device and not
2185 Severity : enhancement
2187 Description: remove cleanup logs
2188 Details : replace lconf-generated cleanup logs with lustre internal
2189 cleanup routines. Eliminates the need for client-cleanup and
2192 Severity : enhancement
2194 Description: add support for EAs (user and system) on lustre filesystems
2195 Details : it is now possible to store extended attributes in the Lustre
2196 client filesystem, and with the user_xattr mount option it
2197 is possible to allow users to store EAs on their files also
2199 Severity : enhancement
2201 Description: Add possibility (config option) to show minimal available OST free
2203 Details : When compiled with --enable-mindf configure option, statfs(2)
2204 (and so, df) will return least minimal free space available from
2205 all OSTs as amount of free space on FS, instead of summary of
2206 free spaces of all OSTs.
2208 Severity : enhancement
2210 Description: do not expand extent locks acquired on OST-side
2211 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2212 by server: they are not cached anyway.
2215 Frequency : when mmap is used/binaries executed from Lustre
2217 Description: Unmmap pages before throwing them away from read cache.
2218 Details : llap_shrink cache now attempts to unmap pages before discarding
2219 them (if unmapping failed - do not discard). SLES9 kernel has
2220 extra checks that trigger if this unmapping is not done first.
2225 Description: lconf didn't resolve symlinks before checking to see whether a
2226 given mountpoint was already in use
2229 Frequency : when migrating failover services
2230 Bugzilla : 6395, 9514
2231 Description: When migrating a subset of services from a node (e.g. failback
2232 from a failover service node) the remaining services would
2233 time out and evict clients.
2234 Details : lconf --force (implied by --failover) sets the global obd_timeout
2235 to 5 seconds in order to quickly disconnect, but this caused
2236 other RPCs to time out too quickly. Do not change the global
2237 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2239 Severity : enhancement
2240 Frequency : if MDS is started with down OST
2241 Bugzilla : 9439,5706
2242 Description: Allow startup/shutdown of an MDS without depending on the
2243 availability of the OSTs.
2244 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2245 Add appropriate locking and lov-osc refcounts for safe
2246 cleaning. Add osc abort_inflight calls in case the
2247 synchronize never started.
2250 Frequency : occasional (Cray XT3 only)
2252 Description: root not authorized to access files in CRAY_PORTALS environment
2253 Details : The client process capabilities were not honoured on the MDS in
2254 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2255 been accessed by an authorized user then root was able to access
2256 the file on the local client also. The root user capabilities
2257 are now allowed on the MDS, as this environment has secure UID.
2260 Frequency : occasional
2262 Description: ldiskfs "too long searching" message happens too often
2263 Details : A debugging message (otherwise harmless) prints too often on
2264 the OST console. This has been reduced to only happen when
2265 there are fragmentation problems on the filesystem.
2270 Description: Division by zero in statfs when all OSCs are inactive
2271 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2272 lov_statfs divides by value returned by lov_get_stripecnt().
2276 Bugzilla : 9489, 3273
2277 Description: First write from each client to each OST was only 4kB in size,
2278 to initialize client writeback cache, which caused sub-optimal
2279 RPCs and poor layout on disk for the first writen file.
2280 Details : Clients now request an initial cache grant at (re)connect time
2281 and so that they can start streaming writes to the cache right
2282 away and always do full-sized RPCs if there is enough data.
2283 If the OST is rebooted the client also re-establishes its grant
2284 so that client cached writes will be honoured under the grant.
2289 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2290 Details : Now I/O RPCs go to different portal number and (presumably) fast
2291 lock requests (and glimses) and other RPCs get their own service
2292 threads pool that should be able to service those RPCs
2295 Severity : enhancement
2297 Description: Ability to exchange lustre version between client and servers and
2298 issue warnings at client side if client is too old. Also for
2299 liblustre clients there is ability to refuse connection of too old
2301 Details : New 'version' field is added to connect data structure that is
2302 filled with version info. That info is later checked by server and
2306 Frequency : rare, liblustre only.
2307 Bugzilla : 9296, 9581
2308 Description: Two simultaneous writes from liblustre at offset within same page
2309 might proceed at the same time overwriting eachother with stale
2311 Details : I/O lock withing llu_file_prwv was released too early, before data
2312 actually was hitting the wire. Extended lock-holding time until
2313 server acknowledges receiving data.
2316 Frequency : extremely rare. Never observed in practice.
2318 Description: avoid generating lustre_handle cookie of 0.
2319 Details : class_handle_hash() generates handle cookies by incrementing
2320 global counter, and can hit 0 occasionaly (this is unlikely, but
2321 not impossible, because initial value of cookie counter is
2322 selected randonly). Value of 0 is used as a sentinel meaning
2323 "unassigned handle" --- avoid it. Also coalesce two critical
2324 sections in this function into one.
2326 Severity : enhancement
2328 Description: allow liblustre clients to delegate truncate locking to OST
2329 Details : To avoid overhead of locking, liblustre client instructs OST to
2330 take extent lock in ost_punch() on client's behalf. New connection
2331 flag is added to handle backward compatibility.
2333 Severity : enhancement
2334 Bugzilla : 4928, 7341, 9758
2335 Description: allow number of OST service threads to be specified
2336 Details : a module parameter allows the number of OST service threads
2337 to be specified via "options ost ost_num_threads={N}" in the
2338 OSS's /etc/modules.conf or /etc/modprobe.conf.
2342 Bugzilla : 6146, 9635, 9895
2343 Description: servers crash with bad pointer in target_handle_connect()
2344 Details : In rare cases when a client is reconnecting it was possible that
2345 the connection request was the last reference for that export.
2346 We would temporarily drop the export reference and get a new
2347 one, but this may have been the last reference and the export
2348 was just destroyed. Get new reference before dropping old one.
2350 Severity : enhancement
2351 Frequency : if client is started with failover MDS
2353 Description: Allow multiple MDS hostnames in the mount command
2354 Details : Try to read the configuration from all specified MDS
2355 hostnames during a client mount in case the "primary"
2358 Severity : enhancement
2360 Description: Stop sending data to evicted clients as soon as possible.
2361 Details : Check if the client we are about to send or are sending data to
2362 was evicted already. (Check is done every second of waiting,
2363 for which l_wait_event interface was extended to allow checking
2364 of exit condition at specified intervals).
2367 Frequency : rare, normally only when NFS exporting is done from client
2369 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2371 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2374 Frequency : always, if srand() is called before liblustre initialization
2376 Description: Liblustre uses system PRNG disturbing its usage by user application
2377 Details : Introduce internal to lustre fast and high-quality PRNG for
2378 lustre usage and make liblustre and some other places in generic
2379 lustre code to use it.
2381 Severity : enhancement
2382 Bugzilla : 9477, 9557, 9870
2383 Description: Verify that the MDS configuration logs are updated when xml is
2384 Details : Check if the .xml configuration logs are newer than the config
2385 logs stored on the MDS and report an error if this is the case.
2386 Request --write-conf, or allow starting with --old_conf.
2388 Severity : enhancement
2390 Description: Handle symlinks in the path when checking if Lustre is mounted.
2391 Details : Resolve intermediate symlinks when checking if a client has
2392 mounted a filesystem to avoid duplicate client mounts.
2397 Description: lconf can hit an error exception but still return success.
2398 Details : The lconf command catches the Command error exception at the top
2399 level script context and will exit with the associated exit
2400 status, but doesn't ensure that this exit status is non-zero.
2405 Description: failure of ptlrpc thread startup can cause oops
2406 Details : Starting a ptlrpc service thread can fail if there are a large
2407 number of threads or the server memory is very fragmented.
2408 Handle this without oopsing.
2411 Frequency : always, only if liblustre and non-default acceptor port was used
2413 Description: liblustre cannot connect to servers with non-default acceptor port
2414 Details : tcpnal_set_default_params() was not called and was therefore
2415 ignoring the environment varaible TCPNAL_PORT, as well as other
2416 TCPNAL_ environment variables
2421 Description: two objects could be created on the same OST for a single file
2422 Details : If an OST is down, in some cases it was possible to create two
2423 objects on a single OST for a single file. No problems other
2424 than potential performance impact and spurious error messages.
2428 Bugzilla : 5681, 9562
2429 Description: Client may oops in ll_unhash_aliases
2430 Details : Client dcache may become inconsistent in race condition.
2431 In some cases "getcwd" can fail if the current directory is
2437 Description: Inode refcounting problems in NFS export code
2438 Details : link_raw functions used to call d_instantiate without obtaining
2439 extra inode reference first.
2443 Bugzilla : 9942, 9903
2444 Description: Referencing freed requests leading to crash, memleaks with NFS.
2445 Details : We used to require that call to ll_revalidate_it was always
2446 followed by ll_lookup_it. Also with revalidate_special() it is
2447 possible to call ll_revalidate_it() twice for the same dentry
2448 even if first occurence returned success. This fix changes semantic
2449 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2450 reference on a request referred from the intent.
2451 ll_intent_release() then releases such a request.
2454 Frequency : rare, normally benchmark loads only
2456 Description: unlinked inodes were kept in memory on the client
2457 Details : If a client is repeatedly creating and unlinking files it
2458 can accumulate a lot of stale inodes in the inode slab cache.
2459 If there is no other client load running this can cause the
2460 client node to run out of memory. Instead flush old inodes
2461 from client cache that have the same inode number as a new inode.
2464 Frequency : SLES9 2.6.5 kernel and long filenames only
2465 Bugzilla : 9969, 10379
2466 Description: utime reports stale NFS file handle
2467 Details : SLES9 uses out-of-dentry names in some cases, which confused
2468 the lustre dentry revalidation. Change it to always use the
2472 Frequency : rare, unless heavy write-truncate concurrency is continuous
2473 Bugzilla : 4180, 6984, 7171, 9963, 9331
2474 Description: OST becomes very slow and/or deadlocked during object unlink
2475 Details : filter_destroy() was holding onto the parent directory lock
2476 while truncating+unlinking objects. For very large objects this
2477 may block other threads for a long time and slow overall OST
2478 responsiveness. It may also be possible to get a lock ordering
2479 deadlock in this case, or run out of journal credits because of
2480 the combined truncate+unlink. Solution is to do object truncate
2481 first in one transaction without parent lock, and then do the
2482 final unlink in a new transaction with the parent lock. This
2483 reduces the lock hold time dramatically.
2486 Frequency : rare, 2.4 kernels only
2488 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2489 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2490 before unmounting their filesystem in order to avoid deadlock.
2491 The kernel_locked() function in 2.4 kernels only checks whether
2492 the kernel lock is held, not whether it is this process that is
2493 holding it as 2.6 kernels do.
2498 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2499 Details : The client ptlrpc code may be trying to reconnect to a down
2500 server before a previous connection attempt has timed out.
2501 Increase the reconnect interval to be longer than the connection
2502 timeout interval to avoid sending duplicate connections to
2506 Frequency : echo_client brw_test command
2508 Description: fix echo_client to work with OST preallocated code
2509 Details : OST preallocation code (5137) didn't take echo_client IO path
2510 into account: echo_client calls filter methods outside of any
2511 OST thread and, hence, there is no per-thread preallocated
2512 pages and buffers to use. Solution: hijack pga pages for IO. As
2513 a byproduct, this avoids unnecessary data copying.
2517 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2518 Description: Client can oops in mdc_commit_close() after open replay
2519 Details : It was possible for the MDS to return an open request with no
2520 transaction number in mds_finish_transno() if the client was
2521 evicted, but without actually returning an error. Clients
2522 would later try to replay that open and may trip an assertion
2523 Simplify the client close codepath, and always return an error
2524 from the MDS in case the open is not successful.
2527 Frequency : rare, 2.6 OSTs only
2529 Description: OST may deadlock under high load on fragmented files
2530 Details : If there was a heavy load and highly-fragmented OST filesystems
2531 it was possible to have all the OST threads deadlock waiting on
2532 allocation of biovecs, because the biovecs were not released
2533 until the entire RPC IO was completed. Instead, release biovecs
2534 as soon as they are complete to ensure forward IO progress.
2536 Severity : enhancement
2538 Description: Support for specifying external journal device at mount
2539 Details : If an OST or MDS device is formatted with an external journal
2540 device, this device major/minor is stored in the ext3 superblock
2541 and may not be valid for failover. Allow detecting and
2542 specifying the external journal at mount time.
2547 Description: Mounting an MDS with pending unlinked files may cause oops
2548 Details : target_finish_recovery() calls mds_postrecov() which returned
2549 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2550 considers this an error and immediately begins cleaning up the
2551 lov, just after starting the mds_lov process
2553 Severity : enhancement
2555 Description: Implement 'lfs df' to report actual free space on per-OST basis
2556 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2557 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2558 usage of objects; '-h' to report in human readable format.
2560 ------------------------------------------------------------------------------
2562 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2569 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2570 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2571 at startup during orphan unlinking. Instead, skip buddy update
2572 before setup, as it will be regenerated after recovery is complete.
2577 Description: If an OST is inactive, its locks might reference stale inodes.
2578 Details : lov_change_cbdata() must iterate over all namespaces, even if
2579 they are inactive to clear inode references from the lock.
2581 Severity : enhancement
2582 Frequency : occasional, if non-standard max_dirty_mb used
2584 Description: Client will block write RPCs if not enough grant
2585 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2586 then the client will block writes while waiting for another RPC
2587 to complete instead of consuming its dirty limit. With change
2588 we get improved performance when max_dirty_mb is small.
2590 Severity : enhancement
2591 Bugzilla : 3389, 6253
2592 Description: Add support for supplementary groups on the MDS.
2593 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2594 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2595 lookups for user supplementary groups into a cache.
2599 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2600 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2601 for liblustre O_EXCL is enforced by MDS.
2605 Description: readdir never returns NULL in liblustre.
2606 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2607 dentry in struct dirent.
2611 Frequency : liblustre only, depends on application IO pattern
2612 Description: liblustre clients evicted if not contacting servers
2613 Details : Don't put liblustre clients into the ping_evictor list, so
2614 they will not be evicted by the pinger ever.
2616 Severity : enhancement
2618 Description: Add ability to evict clients by NID from MDS.
2619 Details : By echoing "nid:$NID" string into
2620 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2621 $NID would be instantly evicted from this MDS and from all active
2622 OSTs connected to it.
2626 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2627 Details : lookup_it_finish() used to query file size from OSTs that was not
2632 Description: service threads change working directory to that of init
2633 Details : Starting lustre service threads may pin the working directory
2634 of the parent thread, making that filesystem busy. Threads
2635 now change to the working directory of init to avoid this.
2639 Frequency : during shutdown only
2640 Description: shutdown with a failed MDS or OST can cause unmount to hang
2641 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2646 Frequency : 2.6 only
2647 Description: chmod/chown may include an extra supplementary group
2648 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2649 supplementary group and if none is specified this is used.
2652 Bugzilla : 5479 (6816)
2654 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2655 Details : If lookup is in progress on a file that is unlinked we might try
2656 to revalidate the inode and fail in revalidate after lookup is
2657 complete and ll_file_open() enqueues the open again but
2658 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2661 Frequency : always, if lconf --abort_recovery used
2663 Description: lconf --abort_recovery fails with 'Operation not supported'
2664 Details : lconf was attempting to abort recovery on the MDT device and not
2667 ------------------------------------------------------------------------------
2669 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2674 Frequency : rare (only unsupported configurations with a node running as an
2676 Bugzilla : 6514, 5137
2677 Description: Mounting a Lustre file system on a node running as an OST could
2679 Details : OSTs now preallocates memory needed to write out data at
2680 startup, instead of when needed, to avoid having to
2681 allocate memory in possibly low memory situations.
2682 Specifically, if the file system is mounted on on OST,
2683 memory pressure could force it to try to write out data,
2684 which it needed to allocate memory to do. Due to the low
2685 memory, it would be unable to do so and the node would
2686 become unresponsive.
2688 Severity : enhancement
2690 Description: Addition of lconf --service command line option
2691 Details : lconf now accepts a '--service <arg>' option, which is
2692 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2694 Severity : enhancement
2696 Description: Failover mode is now the default for OSTs.
2697 Details : By default, OSTs will now run in failover mode. To return to
2698 the old behaviour, add '--failout' to the lmc line for OSTs.
2700 Severity : enhancement
2702 Description: Health checks are now provided for MDS and OSTs
2703 Details : Additional detailed health check information on MSD and OSTs
2704 is now provided through the procfs health_check value.
2707 Frequency : occasional, depends on IO load
2709 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2710 after numerous create/delete cycles
2711 Details : The ext3 inode allocation policy would not allocate new inodes
2712 very well on the OSTs because there are no new directories
2713 being created. Instead we look for groups with free space if
2714 the parent directories are nearly full.
2719 Description: Network or server problems during mount may cause partially
2720 mounted clients instead of returning an error.
2721 Details : The config llog parsing code may overwrite the error return
2722 code during mount error handling, returning success instead
2728 Description: MDS can fail to allocate large reply buffers
2729 Details : After long uptimes the MDS can fail to allocate large reply
2730 buffers (e.g. zconf client mount config records) due to memory
2731 fragmentation or consumption by the buffer cache. Preallocate
2732 some large reply buffers so that these replies can be sent even
2733 under memory pressure.
2737 Frequency : rare (liblustre)
2738 Description: fsx running with liblustre complained that using truncate() to
2739 extend the file doesn't work. This patch corrects that issue.
2740 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2741 ATTR_SIZE and lsm use in llu_setattr_raw.
2745 Frequency : rare, only 2.6 kernels
2746 Description: Unusual file access patterns on the MDS may result in inode
2747 data being lost in very rare circumstances.
2748 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2749 code in the 2.6 kernel caused the inode and/or EA data not to
2750 be read from disk, causing single-file corruption.
2754 Frequency : rare, only 2.6 filesystems using extents
2755 Description: Heavy concurrent write and delete load may cause data corruption.
2756 Details : It was possible under high-load situations to have an extent
2757 metadata block in the block device cache from a just-unlinked
2758 file overwrite a newly-allocated data block. We now unmap any
2759 metadata buffers that alias just-allocated data blocks.
2763 Frequency : filesystems with default stripe_count larger than 77
2764 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2765 Details : lconf specifies an inode size of 4096 bytes when the default
2766 stripe_count is larger than 77. This conflicts with the default
2767 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2768 this case to avoid pinning too much memory for large EAs.
2770 ------------------------------------------------------------------------------
2772 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2777 Frequency : rare (extremely heavy IO load with hundreds of clients)
2779 Description: Client is evicted, gets IO error writing to file
2780 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2781 caused clients to be evicted for AST timeouts. The fixes in
2782 bug 5192 mean we no longer need to have such short AST timeouts
2783 so ldlm_timeout has been increased.
2786 Frequency : occasional during --force or --failover shutdown under load
2787 Bugzilla : 5949, 4834
2788 Description: Server oops/LBUG if stopped with --force or --failover under load
2789 Details : a collection of import/export refcount and cleanup ordering
2790 issues fixed for safer force cleanup
2793 Frequency : only filesystems larger than 120 OSTs
2794 Bugzilla : 5990, 6223
2795 Description: lfs getstripe would oops on a very large filesystem
2796 Details : lov_getconfig used kfree on vmalloc'd memory
2799 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2800 Bugzilla : 6242, 6243
2801 Description: reading from files that had been truncated to a non-zero size
2802 but never opened returned no data
2803 Details : ll_file_read() reads zeros from no-object files to EOF
2808 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2809 Details : There's little harm in aborting MDS/OSS recovery and letting it
2810 try again, so I removed the LASSERT and return an error instead.
2812 Severity : enhancement
2814 Description: New debugging infrastructure for tracking down data corruption
2815 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2816 (b) cover more of the client-side code path, and (c) try to narrow
2817 down where problems occurred
2821 Bugzilla : 3819, 4364, 4397, 6313
2822 Description: Racing close and eviction MDS could cause assertion in mds_close
2823 Details : It was possible to get multiple mfd references during close and
2824 client eviction, leading to one thread referencing a freed mfd.
2826 Severity: : enhancement
2827 Bugzilla : 3262, 6359
2828 Description: Attempts to reconnect to servers are now more aggressive.
2829 Details : This builds on the enhanced upcall-less recovery that was added
2830 in 1.4.2. When trying to reconnect to servers, clients will
2831 now try each server in the failover group every 10 seconds. By
2832 default, clients would previously try one server every 25 seconds.
2837 Description: After recovery, certain operations trigger a failed
2838 assertion on a client.
2839 Details : Failing over an mds, using lconf -d --failover, while a
2840 client was doing a readdir() call would cause the client to
2841 LBUG after recovery completed and the readdir() was resent.
2843 Severity : enhancement
2845 Description: Default groups are now added by lconf
2846 Details : You can now run lconf --group <servicename> without having to
2847 manually add groups with lmc.
2850 Frequency : occasional
2852 Description: Nodes with an elan id of 0 trigger a failed assertion
2855 Frequency : always when accessing e.g. tty/console device nodes
2857 Description: tty and some other devices nodes cannot be used on lustre
2858 Details : file's private_data field is used by device data and lustre
2859 values in there got lost. New field was added to struct file to
2860 store fs-specific private data.
2863 Frequency : when exporting Lustre via NFS
2865 Description: NFSD failed occasionally when looking up a path component
2866 Details : NFSD is looking up ".." which was broken in ext3 directories
2867 that had grown large enough to become hashed.
2870 Frequency : Clusters with multiple interfaces not on the same subnet
2872 Description: Nodes will repeatedly try to reconnect to an interface which it
2873 cannot reach and report an error to the log.
2874 Details : Extra peer list entries will be created by lconf with some peers
2875 unreachable. lconf now validates the peer before adding it.
2878 Frequency : Only if a default stripe is set on the filesystem root.
2880 Description: Setting a default stripe on the filesystem root prevented the
2881 filesystem from being remounted.
2882 Details : The client was sending extra request flags in the root getattr
2883 request and did not allocate a reply buffer for the dir EA.
2886 Frequency : occasional, higher if lots of files are accessed by one client
2887 Bugzilla : 6159, 6097
2888 Description: Client trips assertion regarding lsm mismatch/magic
2889 Details : While revalidating inodes the VFS looks up inodes with ifind()
2890 and in rare cases can find an inode that is being freed.
2891 The ll_test_inode() code will free the lsm during ifind()
2892 when it finds an existing inode and then the VFS later attaches
2893 this free lsm to a new inode.
2897 Bugzilla : 6422, 7030
2898 Description: MDS deadlock between mkdir and client eviction
2899 Details : Creating a new file via mkdir or mknod (starting a transaction
2900 and getting the ns lock) can deadlock with client eviction
2901 (gets ns lock and trying to finish a synchronous transaction).
2904 Frequency : occasional
2905 Description: While starting a server, the fsfilt_ext3 module could not be
2907 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2908 kernels. Previously, lconf would still use the ext3 name
2909 when trying to load modules. Now, it will correctly use
2910 ext3 on 2.4 and ldiskfs on 2.6.
2912 Severity : enhancement
2913 Description: The default stripe count has been changed to 1
2914 Details : The interpretation of the default stripe count (0, to lfs
2915 or lmc) has been changed to mean striping across a single
2916 OST, rather than all available. For general usage we have
2917 found a stripe count of 1 or 2 works best.
2919 Severity : enhancement
2920 Description: Add support for compiling against Cray portals.
2921 Details : Conditional compiling for some areas that are different
2925 Frequency : occasional
2926 Bugzilla : 6409, 6834
2927 Description: Creating files with an explicit stripe count may lead to
2928 a failed assertion on the MDS
2929 Details : If some OSTs are full or unavailable, creating files may
2930 trigger a failed assertion on the MDS. Now, Lustre will
2931 try to use other servers or return an error to the
2935 Frequency : occasional
2937 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2938 caused lock timeout during readahead (since 1.4.2).
2939 Details : Processes doing readahead might match a lock that hasn't been
2940 granted yet if there are overlapping and conflicting lock
2941 requests. The readahead process waits on ungranted lock
2942 (original lock is CBPENDING), while OST waits for that process
2943 to cancel CBPENDING read lock and eventually evicts client.
2945 Severity : enhancement
2947 Description: Initial enabling of flock support for clients
2948 Details : Implements fcntl advisory locking and file status functions.
2949 This feature is provided as an optional mount flag (default
2950 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2951 locking are implemented yet, and those that are are not guaranteed
2952 to be completely correct in production environments.
2953 mount -t lustre -o [flock|noflock] ...
2956 Frequency : occasional
2958 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2959 assertion in the ext3 JBD (journaling) layer.
2960 Details : The b_committed_data struct is protected by the big kernel lock
2961 in 2.4 kernels, serializing journal_commit_transaction() and
2962 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2963 access to this struct. In 2.6 kernels there is finer grained
2964 locking to improve SMP performance of the JBD layer.
2968 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2973 Description: merge and cleanup kernel patches.
2976 -----------------------------------------------------------------------------
2978 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2980 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2981 versions of Lustre. Please update all servers and clients to
2982 version 1.4.2 or later at the same time. You must also run
2983 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2984 to update the configuration logs.
2986 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2987 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2988 - swab llog records of type '0' so we get proper header size/idx (5861)
2989 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2990 - fix rename of one directory over another leaking an inode (5953)
2991 - avoid SetPageDirty on 2.6 (5981)
2992 - don't re-add just-being-destroyed locks to the waiting list (5653)
2993 - when creating new directories, inherit the parent's custom
2994 striping settings if present parent (3048)
2995 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2996 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2997 - don't hold client locks on temporary worklist from l_lru (5666)
2998 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2999 - automatically evict dead clients (5921)
3000 - Update file size properly in create+truncate+fstat case (6196)
3001 - Do not unhash mountpoint dentries, do not allow removal of
3003 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3004 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3005 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3006 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3007 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3008 - Correctly handle reads of files with no objects (6243)
3009 - lctl recover will also mark a device active if deactivate used (5933)
3011 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3012 - allow --write-conf on an MDS with different nettype than client (5619)
3013 - don't write config llogs to MDS for mounts not from that MDS (5617)
3014 - lconf should create multiple TCP connections from a client (5201)
3015 - init scripts are now turned off by default; run chkconfig --on
3016 lustre and chkconfig --on lustrefs to use them
3017 - upcalls are no longer needed for clients to recover to failover
3019 - add --abort-recovery option to lconf to abort recovery on device
3021 - add support for an arbitrary number of OSTs (3026)
3022 - Quota support protocol changes.
3023 - forward compatibility changes to wire structs (6007)
3024 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3025 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3026 - improved "lustre status" script
3027 - initialize blocksize for non-regular files (6062)
3028 - added --disable-server and --disable-client configure options (5782)
3029 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3030 - Vanilla 2.4.29 support
3031 - increase maximum number of obd devices to 520 (6242)
3032 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3033 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3034 SLES 9 2.6.5 kernels
3035 - limit stripes per file to 160 (the maximum EA size) (6093)
3037 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3040 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3041 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3042 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3043 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3044 - small ext3 extents cleanups and fixes (5733)
3045 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3046 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3047 - avoid needless client->OST connect, fix handle mismatch (5317)
3048 - fix DLM error path that led to out-of-sync client, long delays (5779)
3049 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3050 - fix several locking issues related to i_size (5492,5624,5654,5672)
3051 - don't move pending lock onto export if it is already evicted (5683)
3052 - fix kernel oops when creating .foo in unlinked directory (5548)
3053 - fix deadlock in obdfilter statistics vs. object create (5811)
3054 - use time_{before,after} to avoid timer jiffies wrap (5882)
3055 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3056 - Do not leak request if server was not able to process it (5154)
3057 - If mds_open unable to find parent dir, make that negative lookup(5154)
3058 - don't create new directories with extent-mapping (5909, 5936)
3060 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3061 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3062 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3063 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3064 - support for RHEL4 kernel on i686 (5773)
3065 - provide error messages when incompatible logs are encountered (5898)
3067 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3068 * version 1.4.0.10 (1.4.1 release candidate 1)
3070 - don't keep a lock reference when lock is not granted (4238)
3071 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3072 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3073 - drop import inflight refcount on signal_completed_replay error (5255)
3074 - unlock page after async write error during send (3677)
3075 - handle missing objects in filter_preprw_read properly (5265)
3076 - no transno return for symlink open, don't save no-trasno open (3440)
3077 - don't try to complete elan receive that already failed (4012)
3078 - free RPC server reply state on error (5406)
3079 - clean up thread from ptlrpc_start_thread() on error (5160)
3080 - readahead could read extra page into cache that wasn't ejected (5388)
3081 - prevent races in class_attach/setup/cleanup/detach (5260)
3082 - don't dereference de->d_inode after l_dput of de (5458)
3083 - use "int" for stripe value returned from lock_to_stripe (5544)
3084 - mballoc allocation and error-checking fixes in 2.6 (5504)
3085 - block device patches to fix I/O request sizes in 2.6 (5482)
3086 - look up hostnames for IB nals (5602)
3087 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3088 - don't start multiple acceptors for the same port (5277)
3089 - fix incorrect LASSERT in mds_getattr_name (5635)
3090 - export a proc file for general "ping" checking (5628)
3091 - fix "lfs check" to not block when the MDS is down (5628)
3093 - service request history (4965)
3094 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3095 - create an "evict_client" /proc entry on OSTs, like the MDS has
3096 - fix mount usage message, return errors per mount(8) (5168)
3097 - change grep [] to grep "[]" in tests so they work in more UMLs
3098 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3099 - remove ancient LOV_MAGIC_V0 EA support (5047)
3100 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3101 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3102 - allow readahead allocations to fail when low on memory (5383)
3103 - mmap locking landed again, after considerable improvement (2828)
3104 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3106 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3109 - send OST transaction number in read/write reply to free req (4966)
3110 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3111 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3112 - the watchdog thread now runs as interruptible (5246)
3113 - flock/lockf fixes (but it's still disabled, pending 5135)
3114 - don't use EXT3 constants in llite code (5094)
3115 - memory shortage at startup could cause assertion (5176)
3117 - reorganization of lov code
3118 - single portals codebase
3120 - add extents/mballoc support (5025)
3121 - direct I/O reads in the obdfilter (4048)
3122 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3124 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3127 - send OST transaction number in read/write reply to free req (4966)
3128 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3129 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3130 - the watchdog thread now runs as interruptible (5246)
3131 - handle missing objects in filter_preprw_read properly (5265)
3132 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3133 - drop import inflight refcount on signal_completed_replay error (5255)
3134 - unlock page after async write error during send (3677)
3135 - return original error code on reconstructed replies (3761)
3136 - no transno return for symlink open, don't save no-trasno open (3440)
3138 - add pid to ldlm debugging output (4922)
3139 - bump the watchdog timeouts -- we can't handle 30sec yet
3140 - extra debugging for orphan dentry/inode bug (5259)
3142 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3145 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3146 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3147 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3148 - fix NULL dereference in /proc/sys/portals/routes (4827)
3149 - allow failed mdc_close() operations to be interrupted (4561)
3150 - stop precreate on OST before MDS would time out on it (4778)
3151 - don't send partial-page writes before EOF from client (4410)
3152 - discard client grant for sub-page writes on large-page clients (4520)
3153 - don't free dentries not owned by NFS code, check generation (4806)
3154 - fix lsm leak if mds_create_objects() fails (4801)
3155 - limit debug_daemon file size, always print CERROR messages (4789)
3156 - use transno after validating reply (3892)
3157 - process timed out requests if import state changes (3754)
3158 - update mtime on OST during writes, return in glimpse (4829)
3159 - add mkfsoptions to LDAP (4679)
3160 - use ->max_readahead method instead of zapping global ra (5039)
3161 - don't interrupt __l_wait_event() during strace
3163 - add software watchdogs to catch hung threads quickly (4941)
3164 - make lustrefs init script start after nfs is mounted
3165 - fix CWARN/ERROR duplication (4930)
3166 - return async write errors to application if possible (2248)
3167 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3168 - print NAL number in %x format (4645)
3169 - update barely-supported suse-2.4.21-171 series (4842)
3170 - support for sles 9 %post scripts
3171 - support for building 2.6 kernel-source packages
3172 - support for sles km_* packages
3174 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3177 - ignore -ENOENT errors in osc_destroy (3639)
3178 - notify osc create thread that OSC is being cleaned up (4600)
3179 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3180 - reconstruct ost_handle() like mds_handle() (4657)
3181 - create a new thread to do import eviction to avoid deadlock (3969)
3182 - let lconf resolve symlinked-to devices (4629)
3183 - don't unlink "objects" from directory with default EA (4554)
3184 - hold socknal file ref over connect in case target is down (4394)
3185 - allow more than 32000 subdirectories in a single directory (3244)
3186 - fix blocks count for O_DIRECT writes (3751)
3187 - OST returns ENOSPC from object create when no space left (4539)
3188 - don't send truncate RPC if file size isn't changing (4410)
3189 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3190 - bind to privileged port in socknal and tcpnal (3689)
3192 - rate limit CERROR/CWARN console message to avoid overload (4519)
3193 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3194 - basic mmap support (3918)
3195 - kernel patch series update from b1_4 (4711)
3197 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3200 - avoid crash during MDS cleanup with OST shut down (2775)
3201 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3202 - don't use bad inodes on the MDS (3744)
3203 - dynamic object preallocation to improve recovery speed (4236)
3204 - don't hold spinlock over lock dumping or change debug flags (4401)
3205 - don't zero obd_dev when it is force cleaned (3651)
3206 - print grants to console if they go negative (4431)
3207 - "lctl deactivate" will stop automatic recovery attempts (3406)
3208 - look for existing locks in ldlm_handle_enqueue() (3764)
3209 - don't resolve lock handle twice in recovery avoiding race (4401)
3210 - revalidate should check working dir is a directory (4134)
3212 - don't always mark "slow" obdfilter messages as errors (4418)
3214 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3217 - don't close LustreDB during write_conf until it is done (3860)
3218 - fix typo in lconf for_each_profile (3821)
3219 - allow dumping logs from multiple threads at one time (3820)
3220 - don't allow multiple threads in OSC recovery (3812)
3221 - fix debug_size parameters (3864)
3222 - fix mds_postrecov to initialize import for llog ctxt (3121)
3223 - replace config semaphore with spinlock (3306)
3224 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3225 - don't allow enqueue to complete on a destroyed export (3822)
3226 - down write_lock before checking llog header bitmap (3825)
3227 - recover from lock replay timeout (3764)
3228 - up llog sem before sending rpc (3652)
3229 - reduce ns lock hold times when setting kms (3267)
3230 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3231 - fix NULL deref and obd_dev leak on setup error (3312)
3232 - replace some LBUG about llog ops with error handling (3841)
3233 - don't match INVALID dentries from d_lookup and spin (3784)
3234 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3235 - fix invalid assertion in ptlrpc_set_wait (3880)
3237 - add libwrap support for the TCP acceptor (3996)
3238 - add /proc/sys/portals/routes for non-root route listing (3994)
3239 - allow setting MDS UUID in .xml (2580)
3240 - print the stack of a process that LBUGs (4228)
3242 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3245 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3246 - make sure to unset replay flag from failed open requests (3440)
3247 - if default stripe count is 0, use OST count for inode size (3636)
3248 - update parent mtime/ctime on client for create/unlink (2611)
3249 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3250 - free recovery state on server during a forced cleanup (3571)
3251 - unregister_reply for resent reqs (3063)
3252 - loop back devices mounting and status check on 2.6 (3563)
3253 - fix resource-creation race that can provoke i_size == 0 (3513)
3254 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3255 - more debugging for page-accounting assertion (3746)
3256 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3257 - avoid deadlock after precreation failure (3758)
3258 - fix race and lock order deadlock in orphan handling (3450, 3750)
3259 - add validity checks when grabbing inodes from l_ast_data (3599)
3261 - add /proc/.../recovery_status to obdfilter (3428)
3262 - lightweight CDEBUG infrastructure, debug daemon (3668)
3263 - change default OSC RPC parameters to be better on small clusters
3264 - turn off OST read cache for files smaller than 32MB
3265 - install man pages and include them in rpms (3100)
3266 - add new init script for (un)mounting lustre filesystems (2593)
3267 - run chkconfig in %post for init scripts (3701)
3268 - drop scimac NAL (unmaintained)
3270 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3273 - clean kiobufs before and after use (3485)
3274 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3275 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3276 - change init script to fail more clearly if not run as root (1528)
3277 - allow clients to reconnect during replay (1742)
3278 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3279 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3280 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3281 - fix cond_resched (3554)
3282 - only evict unfinished clients after recovery (3515)
3283 - allow bulk resend, prevent data loss (3570)
3284 - dynamic ptlrpc request buffer allocation (2102)
3285 - don't allow unlinking open directory if it isn't empty (2904)
3286 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3287 - remove extraneous obd dereference causing LASSERT failure (3334)
3288 - don't use get_cycles() when creating temp. files on the mds (3156)
3289 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3290 - handle EEXIST for set-stripe, set proper directory name (3336)
3292 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3293 - fix ksocknal_fmb_callback() error messages (2918)
3295 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3298 - don't copy lvb into (possibly NULL) reply on error (2983)
3299 - don't deref dentry after dput, don't free lvb on error (2922)
3300 - use the kms to determine writeback rpc length (2947)
3301 - increment oti_logcookies when osc is inactive (2948)
3302 - update client's i_blocks count via lvb messages (2543)
3303 - handle intent open/close of special files properly (1557)
3304 - mount MDS with errors=remount-ro, like obdfilter (2009)
3305 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3306 - don't use cancelling-locks' kms values (2947)
3307 - use highest lock extent for kms, not last one (2925)
3308 - don't dereference ERR_PTR() dentry in error handling path (3107)
3309 - fix thread race in portals_debug_dumplog() (3122)
3310 - create lprocfs device entries at setup instead of at attach (1519)
3311 - common AST error handler, don't evict client on completion race (3145)
3312 - zero nameidata in detach_mnt in 2.6 (3118)
3313 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3314 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3315 - initialize RPC timeout timer earlier for 2.6 (3219)
3316 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3317 - print nal/nid for unknown nid (3258)
3318 - additional checks for oscc recovery before doing precreate (3284)
3319 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3320 - don't crash in mdc_close for bad permissions on open (3285)
3321 - zero i_rdev for non-device files (3147)
3322 - clear page->private before handing to FS, better assertion (3119)
3323 - tune the read pipeline (3236)
3324 - fix incorrect decref of invalidated dentry (2350)
3325 - provide read-ahead stats and refine rpc in flight stats (3328)
3326 - don't hold journal transaction open across create RPC (3313)
3327 - update atime on MDS at close time (3265)
3328 - close LDAP connection when recovering to avoid server load (3315)
3329 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3330 - don't leak open file on MDS after open resend (3325)
3331 - serialize filter_precreate and filter_destroy_precreated (3329)
3332 - loop device shouldn't call sync_dev() for nul device (3092)
3333 - clear page cache after eviction (2766)
3334 - resynchronize MDS->OST in background (2824)
3335 - refuse to mount the same filesystem twice on same mountpoint (3394)
3336 - allow llmount to create routes for mounting behind routers (3320)
3337 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3338 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3339 - fix rare problem with rename on htree directories (3417)
3341 - allow default OST striping configuration per directory (1414)
3342 - fix compilation for qswnal for 2.6 kernels (3125)
3343 - increase maximum number of MDS request buffers for large systems
3344 - change liblustreapi to be useful for external progs like lfsck (3098)
3345 - increase local configuration timeout for slow disks (3353)
3346 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3348 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3351 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3352 - don't overwrite extent policy data in reply if lock was blocked (2901)
3353 - drop filter export grants atomically with removal from device (2663)
3354 - del obd_self_export from work_list in class_disconnect_exports (2908)
3355 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3356 - swab reply message in mdc_close, other PPC fixes (2464)
3357 - fix destroying of named logs (2325)
3358 - overwrite old logs when running lconf --write_conf (2264)
3359 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3360 - fix race in target_handle_connect (2898)
3361 - mds_reint_create() should take same inode create lock (2926)
3362 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3363 - don't close files for self_export to avoid uninitialized obd (2936)
3364 - allow MDS with the same name as client node (2939)
3365 - hold dentry reference for closed log files for unlink (2325)
3366 - reserve space for all logs during transactions (2059)
3367 - don't evict page beyond end of stripe extent (2925)
3368 - don't oops on a deleted current working directory (2399)
3369 - handle hard links to targets without a parent properly (2517)
3370 - don't dereference NULL lock when racing during eviction (2867)
3371 - don't grow lock extents when lots of conflicting locks (2919)
3373 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3376 - account for cache space usage on clients to avoid data loss (974)
3377 - lfsck support in lustre kernel code (2349)
3378 - reduce journal credits needed for BRW writes (2370)
3379 - orphan handling to avoid losing space on client/server crashes
3380 - ptlrpcd can be blocked, stopping ALL progress (2477)
3381 - use lock value blocks to assist in proper KMS, faster stat (1021)
3382 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3383 - recovery for initial connections (2355)
3384 - fixes for mds_cleanup_orphans (1934)
3385 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3386 - block all file creations until orphan recovery completes (1901)
3387 - client remove rq_connection from request struct (2423)
3388 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3389 - recovery timer race (2670)
3390 - mdc_close recovey bug (2532)
3391 - ptlrpc cleanup bug (2710)
3392 - mds timeout on local locks (2588)
3393 - namespace lock held during RPCs (2431)
3394 - handle interrupted sync write properly (2503)
3395 - don't try to handle a message that hasn't been replied to (2699)
3396 - client assert failure during cleanup after abort recovery (2701)
3397 - leak mdc device after failed mount (2712)
3398 - ptlrpc_check_set allows timedout requests to complete (2714)
3399 - wait for inflight reqs when ptlrpcd finishes (2710)
3400 - make sure unregistered services are removed from the srv_list
3401 - reset bulk XID's when resending them (caught by 1138 test)
3402 - unregister_bulk after timeout
3403 - fix lconf error (2694)
3404 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3405 - readahead locks pages, leaves pending causing memory pressure (2673)
3406 - increase OST request buffers to 4096 on large machines (2729)
3407 - fix up permission of existing directories in simple_mkdir (2661)
3408 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3409 - don't assign transno to errored transactions (2742)
3410 - don't delete objects on OST if given a bogus objid from MDS (2751)
3411 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3412 - if rq_no_resend, then timeout request after recovery (2432)
3413 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3414 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3415 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3416 - MDS recovery shouldn't delete objects that it has given out (2730)
3417 - if enqueue arrives after completion, don't clobber LVB (2819)
3418 - don't unlock pages twice when trigger_group_io returns error (2814)
3419 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3420 - don't write pages to disk if there was an error (1450)
3421 - don't ping imports that have recovery disabled (2676)
3422 - take buffered bytes into account when balancing socknal conn (2817)
3423 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3424 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3425 - remove little-endian swabbing of llog records (1987)
3426 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3427 - retry reposting request buffers if they fail (1191)
3428 - grow extent at grant time to avoid granting a revoked lock (2809)
3429 - lock revoke doesn't evict page if covered by a second lock (2765)
3430 - disable VM readahead to avoid reading outside lock extents (2805)
3432 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3433 - updated kernel patches for hp-2.4.20 kernel (2681)
3435 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3438 - fix truncated write corruption (2366)
3439 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3441 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3442 - ensure len(uuid) < 37 in lmc (1171)
3443 - fix ia64 OOPS in llog_test (2255)
3444 - zero end of page at obdfilter for partial page writes (2648)
3445 - don't leave stale dentries around after renames (bug 2428)
3446 - fix timeouts when evicting a client with a single lock held (2642)
3447 - set deadline for the initial HELLO message to drain (2634)
3448 - print out dotted-quad IP addresses in the socknal (2302)
3450 - additional debugging for MDS client eviction problem (2443)
3451 - fix mkfsoptions support for osts (2603, 2604)
3453 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3456 - add series for the vanilla 2.6.0 kernel
3457 - add series for the vanilla 2.4.24 kernel
3458 - add series for a cray x86/64 UL kernel drop
3459 - fix xattr patches for the vanilla 2.4.19 series
3461 - generate true UUIDs in lmc (1171)
3462 - have portals stack dumping break in UML (2466)
3463 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3464 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3465 - wake callback waiting threads on client eviction (2460)
3466 - Add --ptldebug and --subsystem to lmc (1719)
3467 - update assertion to allow safe interrupt allocation
3468 - set rq_no_resend for cancel requests (2432)
3469 - recalculate ptlrpcd timeout after resend (2494)
3470 - call vfs_rmdir when removing pending directories (2368)
3471 - fix renaming a file to itself (2429)
3472 - lmc creates a default one-stripe lov (2454)
3473 - expand procfs space to handle large clusters (2326)
3474 - increase UML stack to avoid overflow
3475 - update lconf's list of debug and subsystem masks
3476 - fix lfs find --obd (2510)
3477 - /proc tunable for disabling filter read caching (2591)
3478 - stop rpm packages from altering slapd.conf (2301)
3479 - disable nagle in the socknal under 0conf (2578)
3480 - choose mds inode size based on stripe count (2572)
3481 - fix kernel-source rpm problems (2516)
3483 - add --disable-doc to avoid pdf generation (2421)
3484 - update documentation, tests, type-os, comments
3485 - avoid format warnings on ia64
3486 - remove the TOE NAL
3487 - tiny code cleanups by removing unused fields
3489 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3492 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3493 - remove the most blatant lies from BUILDING file (2371)
3494 - change default debug level to reasonable production setting
3495 - reduce client side cache size to reduce cache flush time
3496 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3497 - make TCP zerocopy and pinger support enabled by default (2476)
3498 - sync writes completed after process exits caused crashes (2319)
3499 - maintain correct mount count on the MDS (2356)
3500 - backout 1557, because 2316 wasn't really fixed
3501 - better file I/O statistics gathering in /proc
3502 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3503 - another kernel patch to fix zero-copy TCP function export
3504 - don't take duplicate lock when processing re-sent getattr (2420)
3505 - lctl uses obd_self_export instead of creating new conn (2353)
3506 - MDS/OST recovery case which requires object creation asserted (2425)
3507 - move lfs from /usr/sbin to /usr/bin in packages
3508 - fix race between mds_client_add and mds_client_free (2417)
3509 - use kmalloc instead of slabs in portals (2430)
3510 - don't create duplicate records when a failover MDS is present (2442)
3511 - remove unnecessary mount age check (2332)
3512 - don't remove directory inodes from locks prematurely (2451)
3513 - don't break if MDS service name is the same as hostname (2103)
3514 - fix races in client write RPC generation when cache full (2482)
3516 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3519 - remove now-unused request->rq_obd (278)
3520 - if an allocation fails, print out how much memory we've used (1933)
3521 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3522 - add the "configurable stack size" patch to most series files (1256)
3523 - ability to write large log records, for 100+ OST configs (2306)
3524 - fix NULL deref when filter_prep fails (2314)
3525 - fix operator precedence error in filter_sync
3526 - dynamic allocation of socknal TX descriptors (2315)
3527 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3528 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3529 - ensure that CWARN messages in Portals always get to the syslog
3530 - __init/__exit are not for prototype decls (ldlm_init/exit)
3531 - x86-64 compile warning fixes
3532 - fix gateway LMC keyword conflict (2318)
3533 - fix MDS lock inversions in getattr/reint paths (1844)
3534 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3535 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3536 - DLM race condition prevented some lock evictions (2328)
3537 - ENOMEM detection and retry on socknal sends (2230)
3538 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3539 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3540 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3541 - better error messages when a client is rejected during recovery (1505)
3542 - avoid cancelling locks which were never granted, after failure (2330)
3543 - fix i_sem/journal inversion in mds_client_add (2333)
3544 - fix truncate/getattr lock cycle deadlock (2334)
3545 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3546 - fix two rare exit paths which could leak an l_lock() ref (2321)
3547 - fencepost error in MDS/OST orphan recovery (2226)
3548 - make log record alignment 8 bytes (1988)
3549 - lstripe now fails when requested offset > ost_count (2237)
3550 - ensure that all kernel series have a complete list.h (1607)
3551 - fix crashes in special-file operations (2316)
3552 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3554 - allow configurable automake binary, for testing new versions
3555 - small update to the lfs documentation
3557 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3559 * fix negative export reference count in fsfilt_sync (2312)
3561 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3562 * release candidate 0.9.1
3564 - orphans are moved into the PENDING directory for possible recovery
3565 - replayed opens now open by fid for orphan/rename safety (1042)
3566 - last close of an orphan inode generates a transno (683)
3567 - chdir() and mount() now pin the directory entry (1020)
3568 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3569 - discard very old requests without processing them (1502)
3570 - remove some common, well-understood CERRORs (1505)
3571 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3572 - clear "grant" flags in OST replies until OST grant code lands (1644)
3573 - fix read performance by not clobbering i_blksize on client (1598)
3574 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3575 - make LRU size a /proc tunable, clears locks when reduced (707)
3576 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3577 - server locks take references on exports now (1558)
3578 - build fixes for 2.4.20-rh trees (1663)
3579 - return an error from lov_create if all OSCs are inactive (1751)
3580 - fix import levels when a reconnect happens without a timeout (1597)
3581 - exit early from mds_open if we get a lookup error (1749)
3582 - partial page read at EOF wouldn't wait for disk before sending (1642)
3583 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3584 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3585 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3586 - write/truncate lock inversion (1639)
3587 - Don't auto-load obdclass, portals modules during cleanup (1495)
3588 - fix timestamps from jumping to "now" (1763)
3589 - extra journal assertions (1648)
3590 - add an extra multiunlink test (1771)
3591 - fix read_record/write_record API (1776)
3592 - fix leak of offset_extent, possible incorrect i_size later (1772)
3593 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3594 - Debugging for the kqswnal_get_idle_tx problems (1820)
3595 - Allow recovery to be attempted multiple times (1536)
3596 - Write out MDS last_rcvd file after it is first created (1600)
3597 - Fix tx_descriptor leak in failed transmit situations (1827)
3598 - ext3 journaling fixes for assertion failure after IO error (1871)
3599 - class_export_put() on freed export after completion AST error (1896)
3600 - Fix revalidate looping in VFS (1322)
3601 - Don't access a freed export during MDS_REINT timeout (1521)
3602 - Add open-unlink recovery support on the MDS (1673,1764)
3603 - Return an error if no MDS data was read from last_rcvd (1946)
3604 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3605 - Don't setup a disk device that doesn't match exported UUID (317)
3606 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3607 - avoid committing NULL handle in force close
3608 - local.sh is now a one-stripe LOV configuration
3609 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3610 - fix ext3 htree duplicate directory entry corruption (1516)
3611 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3612 - update to new LOV EA format (2097)
3613 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3615 2003-06-15 Phil Schwan <phil@clusterfs.com>
3618 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3619 - per-import/export recovery handling (958, 931, 959)
3620 - multiple last-rcvd slots, for serving multiple FSes (949)
3621 - connections are again shared between multiple imp/exports (963, 964)
3622 - "umount -f" would hang if any requests needed to be sent (393, 978)
3623 - avoid pinning large req buffer by copying for queued messages (989)
3624 - add "uuid" to "lctl device" command to help upcalls (991)
3625 - "open" RPCs with transnos would confuse recovery counters (1037)
3626 - do proper endian conversion of all wire messages (288, 340, 891)
3627 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3628 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3629 - fix ptlrpc_connection leak in target_handle_connect (1174)
3630 - fix import refcounting bug in OST and MDS cleanup (1134)
3631 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3632 - fix very unlikely obd_types race condition (501)
3633 - remove osc_open hack for echo_client (1187)
3634 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3635 - a failure in read_inode2 leads to deadlock (1139)
3636 - cancel ack-locks as soon as transaction is committed (1072)
3637 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3638 - make sure to commitrw after any preprw to avoid deadlock (1162)
3639 - failing to execute a file in a lustre FS would lock inode (1203)
3640 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3641 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3642 - obd_uuid2tgt would walk off the end of the list (1255)
3643 - on IA64 the osc would give portals incorrect bulk size (1258)
3644 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3645 - fix lock inversion caused by new llite matching code (1282)
3646 - limit the number of dirty pages on a client to 10MB (1286)
3647 - timed out locks were not being corrected cancelled (1289)
3648 - fix O_DIRECT above 4GB on IA-32 (1292)
3649 * major user-visible changes
3650 - fail out/fail over policy now controlled by the upcall (993)
3652 - add OBD_PING to check server availability and failure (954)
3653 - lustre messages are now sent in sending host order (288, 340, 891)
3654 - add eadatalen to MDS getattr reply (340)
3655 - OST read replies may contain second buffer, with per-page status (593)
3657 2003-03-11 Phil Schwan <phil@clusterfs.com>
3660 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3661 - failed open()s could cause deadlock; fixed (867, 869)
3662 - stop cancelling OST locks when files are closed (481)
3663 - overlapping XID spaces caused network corruption (851, 853)
3664 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3665 - setattr_raw added, to do single-RPC, server-side setattrs
3666 - lmc/lconf syntax change for OST UUIDs
3667 - fix crashy race condition between ptlrpc_free_req and osc_close
3668 - don't use request in mdc_enqueue if we hit a timeout (889)
3669 - don't set the inode i_size for regular files from the MDS (896)
3670 - handle out of order completion AST (842)
3671 - don't LBUG if a lock request times out after receiving AST (913)
3672 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3673 - if a bad lock AST arrives, send an error instead of dropping entirely
3674 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3675 - fix leak in bulk IO when only partially completed (899, 900, 926)
3676 - fix O_DIRECT for ia64 (55)
3677 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3678 - C-z after timeout could hang a process forever; fixed (977)
3680 - client-side I/O cache (678, 924, 929, 941, 970)
3682 - READPAGE and SETATTRs which don't take server-side locks get
3685 2003-02-11 Phil Schwan <phil@clusterfs.com>
3688 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3689 - Fix fsfilt_extN_readpage() to read a full page of directory
3690 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3691 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3692 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3693 - Most intent operations were removed, in favour of a new RPC mode
3694 that does a single RPC to the server and bypasses most of the VFS
3695 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3696 - Aggressively cancel local locks on DLM servers
3697 - mds_reint_unlink sends EA to the client if it's the last nlink.
3698 client uses that EA to unlink OST objects.
3699 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3700 - recursive symlinks were fixed (439)
3701 - fixed NULL deref in DEBUG_REQ
3702 - filter_update_lastobjid no longer calls sync, which annoyed extN
3703 - fixed multi-client small-writes to a single file problem (445)
3704 - fixed mtime updates during file writes (607)
3705 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3706 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3707 - fixed Linux OST size reporting problem (444, 656)
3708 - OST now updates object mtime with writes or setattr (607, 619)
3709 - client verifies file size before zeroing page past EOF (445)
3710 - OST now writes last allocated objid to disk with allocation (108)
3711 - LOV on echo now works (409)
3713 - mds_reint_unlink sends a new buffer, with the EA included. this
3714 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3715 set if a regular file was being unlinked, and it was the last link
3716 - use PtlGet from the target for bulk writes (315)
3717 - OST now updates object mtime with writes or setattr (607, 619)
3718 - LDLM now has a grant-time callback to revalidate locked items, if
3720 - Many MDS operations were reorganized to combat race conditions
3722 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3723 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3725 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3728 - Fully reactivate OST imports after reconnection (512, others)
3729 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3730 - More graceful error handling for truncating on dead OST (515)
3731 - Don't error out unless we're actually accessing dead stripes (474)
3732 - Fix garbage sizes when stripes are missing (410)
3733 - LRU counters were broken, causing constant lock purge (433, 432)
3734 - garbage on read from stripes with failed OSTs (441)
3735 - mark OSCs as active before reconnecting during recovery (438)
3736 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3737 - lfind did not preserve OST order in output (443)
3738 - symlinks cause hung clients, incorrect data (439)
3739 - stop dereferencing request after dropping refcount (457)
3740 - don't LASSERT(spin_is_locked) on non-SMP (455)
3741 - fixes for many rename() bugs
3742 - fstat didn't correctly synchronize attributes (399)
3743 - server must handle lock cancellation during blocking AST prep (487)
3744 - bulk descriptors were free()d too soon (511)
3745 - fix paths in lconf, which would load incorrect modules (451, 507)
3746 - fix confusing lconf 'host not found' error message (386)
3747 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3748 - fix race condition in mdc_blocking_ast() for inode access (526)
3749 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3750 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3751 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3752 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3753 - fix lstripe in several ways (kernel side) (527)
3754 - fix request leak in ldlm_cli_enqueue (262)
3755 - incorrect OSC was marked inactive after OST failure
3756 - call mds_fs_cleanup before unmounting filesystem (524)
3757 - fix races between taking ns_lock and ldlm_lock_change_resource
3758 - fix races updating LOV export open file list
3759 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3760 - fix recovery NULL deref in ldlm_cli_cancel_unused
3761 - fix some DLM races by using new hash table for lock handles (419)
3762 - permit the client to specify desired inodes, at replay
3763 - duplicate requests when we queue them for replay reintegration
3764 - fix last_rcvd offset calculation
3765 - sync after each recovered transaction, so we always make progress
3766 - never, not always, ERESTART requests without transnos
3767 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3769 - skip replay if the MDS says that the client is already connected
3770 - don't check for a recovery-enabled export to match lctl's UUID
3771 - don't INC_USE_COUNT for phantom exports
3772 - don't crash when cleaning up phantom exports (567)
3773 - don't double-finish or set replay data for errored mdc_open requests
3774 - abort requests when they time out, so we don't get old replies
3775 - send/receive replies for AST messages again
3776 - if the client says that it doesn't have the lock, cancel it on the
3778 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3779 mark it as destroyed, it will all work out when decref is called
3780 - fix module use counts (22, 581)
3782 - ASTs now expect a reply (server cancels lock on error reply)
3784 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3787 - fix many simultaneous client startup (392)
3788 - fix dentry->d_it clobbering
3789 - credentials weren't being shipped for readdir/getattr operations
3790 - remove invalid assertions triggered during some concurrent MD
3792 - proper Lustre versions added (336, 389)
3793 - fix memory leak for create error case (398)
3794 - fix LOV locking bug that would get cli/srv out of sync
3795 - fix echo client over LOV (409)
3796 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3797 - fix double-O_EXCL intent crash (424)
3798 - avoid sending multiple lock CANCELs (352)
3800 - MDS can do multi-client recovery (modulo bugs in new code)
3802 - many updates, edits, cleanups
3804 2002-11-18 Phil Schwan <phil@clusterfs.com>
3807 - fix null d_it dereference (346)
3808 - fix full OST/dbench hang (333)
3809 - fix permission problem with file removal (286)
3810 - fix removal of OSCs from LOV when they fail
3811 - fix NULL deref during bulk timeout (214)
3812 - fix problems related to multiple filesystems on one MDS (241)
3813 - fixed serious subtle metadata locking bugs
3814 - free locks on clients when inodes are removed due to memory
3816 - fix inode pointer in lock data (285)
3817 - partial support for multiple MDS on a single host (241)
3818 - data locks weren't cancelled at clear_inode time (290, 311)
3819 - intent locks could lead to unbounded lock growth (205)
3820 - added a maximum lock count, an LRU list, and a flusher
3821 - fix multiple rename (365)
3822 - properly abstracted the echo client
3823 - OSC locked 1 byte too many; fixed
3824 - rewrote brw callback code:
3825 - fixed recovery bugs related to LOVs (306)
3826 - fixed too-many-pages-in-one-write crash (191)
3827 - fixed (again) crash in sync_io_timeout (214)
3828 - probably fixed callback-related race (385)
3830 - Add capability to MDS protocol
3831 - LDLM cancellations and callbacks on different portals
3833 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3836 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3837 - defer open object destruction to close time (601981, 138)
3838 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3839 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3840 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3841 - don't call obd_setattr in ll_file_release for destroyed objects
3842 * protocol change to lustre_msg: move |version| and add |flags|
3843 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3844 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3845 * added replay of create, unlink, link and rename operations during
3846 MDS failover; recovery should be much more robust now
3847 * remove failed OSCs from LOVs (only lov_create uses this so far)
3848 * the lustre-HOWTO was brought (more) up to date (582544)
3850 2002-10-23 Phil Schwan <phil@clusterfs.com>
3853 - in-use dentries weren't being reused properly (617851)
3854 - prevent multiple LDLM setup (599178)
3855 - fix LOV size calculations for truncate (617853)
3856 - fix client handling of MDS intent errors (POSIX)
3857 - fix permission bug in lovstripe.c test (624321)
3858 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3859 - truncate past end of file could corrupt data
3860 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3861 - a race in recovery could return ETIMEDOUT to apps (623947)
3862 - building outside the source directory was fixed
3863 * the lustre-HOWTO was brought (more) up to date (582544)
3864 * major progress was made on recovery functionality
3866 2002-10-10 Phil Schwan <phil@clusterfs.com>
3869 - recovery deadlock fix
3870 - rm -rf causes LBUG fix (617817)
3871 - file open by multiple tasks fix (618962)
3872 - directory permissions bugs (602707 and 620007)
3873 - journal_stop fixed with locking (611313)
3874 - O_APPEND failures resolved (618273, perhaps 614459)
3875 - lconf PATH fix (619770)
3876 - IA64 build fix (621450)
3877 - RPC buffer sizes scale with amount of memory
3879 2002-10-01 Phil Schwan <phil@clusterfs.com>
3882 - locks would be cancelled without throwing away data pages,
3883 resulting in inconsistent data (605627)
3884 - inode attributes were not always being refreshed (605627, 612449)
3885 - lconf now continues to cleanup after lctl reports an error
3886 - MDS now enforces user permissions (602707)
3887 - lprocfs cleanup fixed, but not yet enabled (614157)
3888 - fixed infinite server hang, should a client not respond to an AST
3889 - avoid going into recovery if user calls readlink() with a buffer
3890 that's too small (613941)
3891 - AST RPCs no longer require replies (614867) -- this may be changed
3892 - don't crash server if client sends an IOV that's too big (611336)
3893 - fixed lock conversion deadlock (611892)
3894 - fixed the following of symlinks (614622)
3895 * recovery: the server can remove locks from a client that dies, other
3896 clients can make progress
3897 * more extN patch fixes
3898 * compile-time configurable ptlrpc buffer allocations
3900 - collaborative read cache document
3901 - Lustre Lite Performance CDR document-in-progress
3903 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3906 - fix typo in patch-2.4.18
3908 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3911 - clear ptlrpc request each time in handle_incoming_request()
3912 - unlink of files now destroys the object on the OST
3914 2002-09-19 Peter Braam <braam@clusterfs.com>
3916 * add hard link support
3917 * change obdfile creation method
3918 * kernel patch changed
3920 2002-09-19 Peter Braam <braam@clusterfs.com>
3923 - stack overflow bug in extN fixed
3925 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3927 * documentation updates
3928 - add man pages for config tools
3929 - update tests/README to describe testing with new config tools
3930 - finish metadata API descriptions
3931 * bug fixes and cleanups
3932 - statfs workaround for 16TB limit
3933 - LOV stripe allocation improved, can stripe on subset of OSTs
3934 - LOV file size/IO offset was wrong for files > 4GB in size
3935 - object EA data was being dropped, caused files to be unreadable
3936 - memory overflow with non-LOV OST caused memory corruption
3937 - fixed regression tests to work with new config tools, obdfilter
3938 - fixed bug when directory size became larger than 1 block
3939 - fixed bug (for single client case) when PWD was deleted
3940 - invalidate local directory pages when doing intent-based ops
3941 - avoid LDLM oops when lock callback contained bad data
3943 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3945 * documentation updates
3946 * bug fixes and cleanups
3947 - configuration tools
3950 - 64-bit compile warnings
3951 - 64-bit internal statfs data
3953 * test_brw on persistent OST devices
3955 * lprocfs (disabled)
3957 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3959 * documentation updates
3960 * bug fixes and cleanups
3961 * configuration tools
3963 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3966 * many small fixes to 0_5_4
3967 * io/network handling
3968 * thinkos in MDS operations
3970 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3973 * crucial basic fixes to 0.5.3
3974 * IOR, Iozone work over Elan
3977 2002-08-07 Phil Schwan <phil@clusterfs.com>
3978 * version 0_5_3, our first alpha
3979 * we use the new Portals iovs
3980 * documentation updates
3981 * bug fixes and cleanups
3982 * small changes in the DLM wire protocol
3984 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3985 * version 0_5_1 with some initial stability,
3986 * locking on MD and file I/O.
3987 * documentation updates
3988 * several bug fixes since 0.5.0
3989 * small changes in wire protocol
3991 2002-07-18 Phil Schwan <phil@clusterfs.com>
3993 * delivered as Lustre Light Alpha
3994 * fixed a crash after handling invalid MDS requests
3995 * fixed directory pages for architectures with non-4k pages sizes
3997 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3998 * release version v0_4_4
3999 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4001 * Fixes a number of interruption problems with OST operations.
4002 * Update documentation for portals header changes
4003 * Move all wire protocol structs/defines to lustre_idl.h
4004 * Fixes symlink length bug.
4005 * Add tcpdump to repository.
4007 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4008 * release version v0_4_3
4009 * Fixes statfs for inodes on extN.
4010 * Fixes bug in runtests which would delete /etc/hosts.
4011 * Use 64-bit object IDs wherever possible (not into VFS though)
4012 Remove ost_get_info, which is unused by lustre, and out of date.
4014 2002-07-03 Peter Braam <braam@clusterfs.com>
4015 * release version v0_4_2 Fixes a lookup error (type not passed)
4016 * move forward to head of Portals
4017 * move forward to latest Lustre kernel
4019 2002-06-25 Peter Braam <braam@clusterfs.com>
4020 * release version v0_4_1. Hopefully stable on single node use.