4 2.6.9-55.0.6.EL (RHEL 4),
5 2.6.16.46-0.14 (SLES 10),
6 2.6.18-8.1.10.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.
15 Severity : enhancement
17 Description: self-adjustable client's lru lists
18 Details : use adaptive algorithm for managing client cached locks lru
19 lists according to current server load, other client's work
20 pattern, memory activities, etc. Both, server and client
21 side namespaces provide number of proc tunables for controlling
24 Severity : enhancement
26 Description: light-weight GSS support
27 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
28 and reduce performance overhead.
30 Severity : enhancement
32 Description: Linux keyring support
33 Details : Support using service of Linux keyring for Lustre GSS internal
34 context refresh/cache mechanism.
38 Description: Fix errors in lfs documentation
39 Details : Fixes man pages
43 Description: don't use GFP_* in generic Lustre code.
44 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
48 Description: obdclass is limited by single OBD_ALLOC(idarray)
49 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
53 Description: warnings with build patchless client with vanila 2.6.19 and up
54 Details : change old ctl_table style and replace ctl_table/ctl_table_header
55 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
59 Frequency : Only for SLES
60 Description: sanity_quota fail test_1
61 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
62 file, which makes TSTID[2] inunique.
66 Frequency : Only for SLES9
67 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
68 Details : Update target/ChangeLog/which_patch .
72 Frequency : Only for test_10 in sanity-quota.sh
73 Description: a bug in quota test code
74 Details : checking a test flag in a wrong place
78 Frequency : Only for test_13 in sanity-quota.sh
79 Description: a bug in quota test script
80 Details : shouldn't check overfull allocated quota
84 Description: jbd statistics
85 Details : Port older jbd statistics patch for sles10
87 Severity : enhancement
89 Description: Add printing inode info into message about error in writepage.
94 Description: reply_lock_interpret crash due to race with it and lock cancel.
95 Details : Do not replay locks that are being cancelled. Do not reference
96 locks by their address during replay, just by their handle.
100 Frequency : When flocks are used.
101 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
102 Details : Flock locks might destroy just granted lock if it could be merged
103 with another existing flock, this is done in completion handler,
104 so teach ldlm_cli_enquque_fini that this is a valid case for
110 Description: Oops in read and write path when failing to allocate lock.
111 Details : Check if lock allocation failed and return error back.
115 Description: lstripe command fails for valid OST index
116 Details : The stripe offset is compared to lov->desc.ld_tgt_count
117 instead of lov->desc.ld_active_tgt_count.
121 Description: sanity.sh failed test 103
122 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
127 Description: df doesn't work properly if diskfs blocksize != 4K
128 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
132 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
133 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
137 Description: obd_health_check_timeout too short
138 Details : set obd_health_check_timeout as 1.5x of obd_timeout
142 Description: llapi_file_create() does not allow some changes
143 Details : add llapi_file_open() that allows specifying the mode and
144 open flags, and also returns an open file handle.
148 Description: merge and cleanup kernel patches.
149 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
153 Description: Add journal checksum support.(Kernel part)
154 Details : The journal checksum feature adds two new flags i.e
155 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
156 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
157 indicates that the commit block contains the checksum for
158 the blocks described by the descriptor blocks. Now commit
159 record can be sent to disk without waiting for descriptor
160 blocks to be written to disk. This behavior is controlled
161 using JBD2_FEATURE_ASYNC_COMMIT flag.
165 Description: OSS needs mutliple precreate threads
166 Details : Add ability to start more than one create thread per OSS.
170 Description: Sanity.sh test_65k failed: setstripe should have succeeded
171 Details : execute lfs setstripe on client
175 Description: mds_obd_create error creating tmp object
176 Details : When the user sets quota on root, llog will be affected and can't
177 create files and write files.
180 Frequency : Always on ia64 patchless client, and possibly others.
182 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
183 Details : This allows the patchless client to be loaded on architectures
188 Description: RedHat Update kernel for RHEL5
189 Details : Add the kernel config file for RHEL5.
193 Description: RedHat Update kernel for RHEL5
194 Details : Modify the kernel config file more closer RHEL5.
198 Description: Build failure against Centos5 (RHEL5)
199 Details : Define PAGE_SIZE when it isn't present.
203 Description: test_7 on sanity-quota.sh can't be used on separate machines
204 Details : add facet to handle it
208 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
209 Details : with NFS, the anon dentry's parent was set to itself in
210 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
211 corresponding dentry other than use rec->ur_name.
213 Severity : enhancement
215 Description: lfs setstripe enhancement
216 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
220 Description: testing performance impact of enabling checksumming
221 Details : enable checksum by default, allow --disable-checksum
222 configure option and "-o nochecksum" mount option
226 Description: lustre support for RHEL5
227 Details : Add support for RHEL5.
231 Description: Client eviction due to failover config
232 Details : after a connection loss, the lustre client should attempt to
233 reconnect to the last active server first before trying the
234 other potential connections.
238 Description: when mds and osts use different quota unit(32bit and 64bit),
239 quota will be released repeatly.
240 Details : void sending multiple quota reqs to mds, which will keep the
241 status between the reqs.
245 Description: osts not allocated evenly to files
246 Details : change the condition to increase offset_idx
249 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
251 Description: Data corruption for OSTs that are formatted larger than 2TB
253 Details : When generating the bio request for lustre file writes the
254 sector number would overflow a temporary variable before being
255 used for the IO. The data reads correctly from Lustre (which
256 will overflow in a similar manner) but other file data or
257 filesystem metadata may be corrupted in some cases.
261 Description: Tune the kernel for good SCSI performance.
262 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
263 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
268 Description: rewrite ext2-derived code in obdclass/uuid.c
269 Details : rewrite inherited code (uuid parsing code from ext2 utils)
270 from scratch preserving functionality.
274 Description: "lfs find -obd UUID" prints directories
275 Details : "lfs find -obd UUID" will return all directory names instead
276 of just file names. It is incorrect because the directories
277 do not reside on the OSTs.
281 Description: use i_size_read and i_size_write in 2.6 port
282 Details : replace inode->i_size access with i_size_read/write()
285 Frequency : when removing large files
287 Description: scheduling issue during removal of large Lustre files
288 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
289 It causes scheduling issues when removing large files (17TB in the
293 Frequency : only with liblustre clients on XT3
295 Description: evictions taking too long
296 Details : allow llrd to evict clients directly on OSTs
299 Frequency : only on ppc
301 Description: /proc/fs/lustre/devices broken on ppc
302 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
303 We should make sure the type of 'index' is loff_t and then cast
304 explicitly as needed below. Do not assign an explicitly cast
308 Frequency : only for rhel5
310 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
311 Details : Modify the target file & which_kernel.
313 Severity : enhancement
315 Description: omit set fsid for export NFS
316 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
317 as FS_REQUIRES_DEV for avoid problems with generate fsid.
321 Frequency : Always, for kernels after 2.6.16
322 Description: Fix warning idr_remove called for id=.. which is not
324 Details : Last kernels save old s_dev before kill super and not allow
325 to restore from callback - restore it before call kill_anon_super.
329 Description: buffer overruns could theoretically occur
330 Details : llapi_semantic_traverse() modifies the "path" argument by
331 appending values to the end of the origin string, and a
332 overrun may occur. Adding buffer overrun check in liblustreapi.
336 Description: client sometimes tried revalidate root dentry.
337 Details : not revalidate dentry if it root dentry.
341 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
342 Details : Modify vm-tunables-rhel4.patch.
344 --------------------------------------------------------------------------------
346 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
348 * Support for kernels:
349 2.6.5-7.283 (SLES 9),
350 2.6.9-55.EL (RHEL 4),
351 2.6.16.46-0.14 (SLES 10),
352 2.6.18.8 vanilla (kernel.org)
353 * Client support for unpatched kernels:
354 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
355 2.6.16 - 2.6.22 vanilla (kernel.org)
356 * Due to recently discovered recovery problems, we do not recommend
357 using patchless RHEL 4 clients with this or any earlier release.
358 * Recommended e2fsprogs version: 1.39.cfs8
359 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
360 * Starting with this release, the ldiskfs backing filesystem required
361 by Lustre is now in its own package, lustre-ldiskfs. This package
362 should be installed. It is versioned separately from Lustre and
363 may be released separately in future.
367 Description: block reactivating mgc import until all deactivates complete
368 Details : Fix race when failing back MDT/MGS to itself (testing)
370 Severity : enhancement
372 Description: add optional extra BUILD_VERSION info
373 Details : add a new environment variable (namely LUSTRE_VERS) which allows
374 to override the lustre version.
377 Frequency : 2.6.18 servers only
379 Description: ll_kern_mount() doesn't release the module reference
380 Details : The ldiskfs module reference count never drops down to 0
381 because ll_kern_mount() doesn't release the module reference.
386 Description: server LBUG when using old ost_num_threads parameter
387 Details : Accept the old ost_num_threads parameter but warn that it
388 is deprecated, and fix an off-by-one error that caused an LBUG.
393 Description: Transient SCSI error results in persistent IO issue
394 Details : iobuf->dr_error is not reinitialized to 0 between two
398 Frequency : sometimes when underlying device returns I/O errors
400 Description: OSTs not going read-only during write failures
401 Details : OSTs are not remounted read-only when the journal commit threads
402 get I/O errors because fsfilt_ext3 calls journal_start/stop()
403 instead of the ext3 wrappers.
407 Description: poor connect scaling with increasing client count
408 Details : Don't run filter_grant_sanity_check for more than 100 exports
409 to improve scaling for large numbers of clients.
412 Frequency : SLES10 only
414 Description: sanity-quota.sh quotacheck failed: rc = -22
415 Details : Quotas cannot be enabled on SLES10.
418 Frequency : liblustre clients only
420 Description: getdirentries does not give error when run on compute nodes
421 Details : getdirentries does not fail when the size specified as an argument
422 is too small to contain at least one entry
424 Severity : enhancement
426 Description: Add LNET router traceability for debug purposes
427 Details : If a checksum failure occurs with a router as part of the
428 IO path, the NID of the last router that forwarded the bulk data
429 is printed so it can be identified.
434 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
435 Details : Due to a race condition, liblustre clients were occasionally
438 Severity : enhancement
440 Description: lfs setstripe use optional parameters instead of postional
443 Severity : enhancement
445 Description: Nanosecond timestamp support for ldiskfs
446 Details : The on-disk ldiskfs filesystem has added support for nanosecond
447 resolution timestamps. There is not yet support for this at
448 the Lustre filesystem level.
451 Frequency : during server recovery
453 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
454 Details : request with rq_no_resend flag not awake l_wait_event if they get a
458 Frequency : nfs export on patchless client
460 Description: connectathon hang when test nfs export over patchless client
461 Details : Disconnected dentry cannot be found with lookup, so we do not need
462 to unhash it or make it invalid
465 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
466 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
467 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
472 Description: open req refcounting wrong on reconnect
473 Details : If reconnect happened between getting open reply from server and
474 call to mdc_set_replay_data in ll_file_open, we will schedule
475 replay for unreferenced request that we are about to free.
476 Subsequent close will crash in variety of ways.
477 Check that request is still eligible for replay in
478 mdc_set_replay_data().
483 Description: disable writes to filesystem when reading health_check file
484 Details : the default for reading the health_check proc file has changed
485 to NOT do a journal transaction and write to disk, because this
486 can cause reads of the /proc file to hang and block HA state
487 checking on a healthy but otherwise heavily loaded system. It
488 is possible to return to the previous behaviour during configure
489 with --enable-health-write.
491 Severity : enhancement
493 Description: 64-bit inode version
494 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
495 to the inode. This will be required for version-based recovery.
500 Description: MDS fails to start if a duplicate client export is detected
501 Details : in some rare cases it was possible for a client to connect to
502 an MDS multiple times. Upon recovery the MDS would detect this
503 and fail during startup. Handle this more gracefully.
505 Severity : enhancement
507 Description: Add -o localflock option to simulate old noflock
509 Details : This will achieve local-only flock/fcntl locks
515 Description: log_commit_thread vs filter_destroy race leads to crash
516 Details : Take import reference before releasing llog record semaphore
521 Description: Wrong request locking in request set processing
522 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
523 imp_delayed_list, in this place should be used imp_lock.
526 Frequency : when reconnection
528 Description: Grant Leak when osc reconnect to OST
529 Details : When osc reconnect ost, OST(filter) should check whether it
530 should grant more space to client by comparing fed_grant and
531 cl_avail_grant, and return the granted space to client instead
532 of "new granted" space, because client will call osc_init_grant
533 to update the client grant space info.
536 Frequency : when client reconnect to OST
538 Description: Grant Leak when osc do resend and replay bulk write
539 Details : When osc reconnect to OST, OST(filter)should clear grant info of
540 bulk write request, because the grant info while be sync between
541 OSC and OST when reconnect, and we should ignore the grant info
542 these of resend/replay write req.
547 Description: Grant space more than avaiable left space sometimes.
548 Details : When then OST is about to be full, if two bulk writing from
549 different clients came to OST. Accord the avaliable space of the
550 OST, the first req should be permitted, and the second one
551 should be denied by ENOSPC. But if the seconde arrived before
552 the first one is commited. The OST might wrongly permit second
553 writing, which will cause grant space > avaiable space.
556 Frequency : when client is evicted
558 Description: Grant might be wrongly erased when osc is evicted by OST
559 Details : when the import is evicted by server, it will fork another
560 thread ptlrpc_invalidate_import_thread to invalidate the
561 import, where the grant will be set to 0. While the original
562 thread will update the grant it got when connecting. So if
563 the former happened latter, the grant will be wrongly errased
564 because of this race.
569 Description: Checking Stale with correct fid
570 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
571 to check whether it is stale, because sometimes, we want the
572 enqueue happened anyway, and op_data.fid2 will not be initialized.
574 Severity : enhancement
576 Description: update patchless client
577 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
580 Frequency : only with 2.4 kernel
582 Description: random memory corruption
583 Details : size of struct ll_inode_info is to big for union inode.u and this
584 can be cause of random memory corruption.
589 Description: Memory leak in recovery
590 Details : Lov_mds_md was not free in an error handler in mds_create_object.
591 It should also check obd_fail before fsfilt_start, otherwise if
592 fsfilt_start return -EROFS,(failover mds during mds recovery).
593 then the req will return with repmsg->transno = 0 and rc = EROFS.
594 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
595 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
596 be freed no matter whether fsfilt_commit success or not.
599 Frequency : only with huge count clients
601 Description: Prevents from taking the superblock lock in llap_from_page for
603 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
604 for llap_from_page call in ll_removepage prevents from taking the
605 superblock lock for a soon died page.
610 Description: Not check open intent error before release open handle
611 Details : in some rare cases, the open intent error is not checked before
612 release open handle, which may cause
613 ASSERTION(open_req->rq_transno != 0), because it tries to release
614 the failed open handle.
619 Description: Set cat log bitmap only after create log success.
620 Details : in some rare cases, the cat log bitmap is set too early. and it
621 should be set only after create log success.
625 Description: Accessing a block bevice can re-enable I/O when Lustre is
626 tearing down a device.
627 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
632 Description: service threads may hog cpus when there are a lot of requests
634 Details : Insert cond_resched to give other threads a chance to use some of
640 Description: the cat log was not initialized in recovery
641 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
642 unlink log on mds will not be initialized until mds post
643 recovery. And also in mds post recovery, the unlink log will
644 initialization will be done asynchronausly, so there will be race
645 between add unlink log and unlink log initialization.
649 Description: brw_stats were being printed incorrectly
650 Details : brw_stats were being printed as log2 but all of them were not
651 recorded as log2. Also remove some code duplication arising from
652 filter_tally_{read,write}.
656 Frequency : rare, only in recovery.
657 Description: ASSERTION(req->rq_type != LI_POISON) failed
658 Details : imp_lock should be held while iterating over imp_sending_list for
659 prevent destroy request after get timeout in ptlrpc_queue_wait.
663 Description: replay-single.sh test 52 fails
664 Details : A lock's skiplist need to be cleanup when it being unlinked
665 from its resource list.
669 Description: Short directio read returns full requested size rather than
671 Details : Direct I/O operations should return actual amount of bytes
672 transferred rather than requested size.
674 Severity : enhancement
676 Description: metadata RPC reduction (e.g. for rm performance)
677 Details : decrease the amount of synchronous RPC between clients and servers
678 by canceling conflicing lock before the operation on the client side
679 and packing thier handles into the main operation RPC to server.
681 Severity : enhancement
683 Description: Async OSC create to avoid the blocking unnecessarily.
684 Details : If a OST has no remain object, system will block on the creating
685 when need to create a new object on this OST. Now, ways use
686 pre-created objects when available, instead of blocking on an
687 empty osc while others are not empty. If we must block, we block
688 for the shortest possible period of time.
690 --------------------------------------------------------------------------------
692 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
697 Frequency : on some architectures
699 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
700 Details : Uninitialized flags sometimes cause configuration commands to
704 Frequency : patchless clients only
706 Description: missing __iget() symbol export
707 Details : The __iget() symbol export is missing. To avoid the need for
708 this on patchless clients the deathrow inode reaper is turned
709 off, and we depend on the VM to clean up old inodes. This
710 dependency was during via the fix for bug 12181.
712 --------------------------------------------------------------------------------
714 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
716 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
717 INTEROPERATE with older versions automatically. In many cases a
718 special upgrade step is needed. Please read the
719 user documentation before upgrading any part of a live system.
720 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
721 version will not interoperate with 1.6 betas before beta5 (1.5.95).
722 * WARNING: Lustre configuration and startup changes are required with
723 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
729 Severity : enhancement
731 Description: MountConf
732 Details : Lustre configuration is now managed via mkfs and mount
733 commands instead of lmc and lconf. New obd types (MGS, MGC)
734 are added for dynamic configuration management. See
735 https://mail.clusterfs.com/wikis/lustre/MountConf for
738 Severity : enhancement
740 Description: dynamic OST addition
741 Details : OSTs can now be added to a live filesystem
743 Severity : enhancement
745 Description: startup order invariance
746 Details : MDTs and OSTs can be started in any order. Clients only
747 require the MDT to complete startup.
749 Severity : enhancement
751 Description: parallel, asynchronous orphan cleanup
752 Details : orphan cleanup is now performed in separate threads for each
753 OST, allowing parallel non-blocking operation.
755 Severity : enhancement
757 Description: optimized stripe assignment
758 Details : stripe assignments are now made based on ost space available,
759 ost previous usage, and OSS previous usage, in order to try
760 to optimize storage space and networking resources.
762 Severity : enhancement
764 Description: Permanently set tunables
765 Details : All writable /proc/fs/lustre tunables can now be permanently
766 set on a per-server basis, at mkfs time or on a live
769 Severity : enhancement
771 Description: Lustre message v2
772 Details : Add lustre message format v2.
774 Severity : enhancement
776 Description: client OST exclusion list
777 Details : Clients can be started with a list of OSTs that should be
778 declared "inactive" for known non-responsive OSTs.
782 Description: SPEC SFS validation failure on NFS v2 over lustre.
783 Details : Changes the blocksize for regular files to be 2x RPC size,
784 and not depend on stripe size.
786 Severity : enhancement
788 Description: Multiple MD RPCs in flight.
789 Details : Further unserialise some read-only MDS RPCs - learn about intents.
790 To avoid overly-overloading MDS, introduce a limit on number of
791 MDS RPCs in flight for a single client and add /proc controls
792 to adjust this limit.
794 Severity : enhancement
796 Description: client read/write statistics
797 Details : Add client read/write call usage stats for performance
798 analysis of user processes.
799 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
800 file access. extents_stats shows chunk size distribution.
801 extents_stats_per_process show chunk size distribution per
804 Severity : enhancement
806 Description: mds statistics
807 Details : Add detailed mds operations statistics in
808 /proc/fs/lustre/mds/*/stats.
812 Description: Failure of copying files with lustre special EAs.
813 Details : Client side always return success for setxattr call for lustre
814 special xattr (currently only "trusted.lov").
819 Description: Refcount LNET uuids
820 Details : The global LNET uuid list grew linearly with every startup;
821 refcount repeated list entries instead of always adding to
824 Severity : enhancement
826 Description: Dynamic service threads
827 Details : Within a small range, start extra service threads
828 automatically when the request queue builds up.
831 Frequency : mixed-endian client/server environments
833 Description: mixed-endian crashes
834 Details : The new msg_v2 system had some failures in mixed-endian
837 Severity : enhancement
839 Description: Easy OST removal
840 Details : OSTs can be permanently deactivated with e.g. 'lctl
841 conf_param lustre-OST0001.osc.active=0'
843 Severity : enhancement
845 Description: MGS proc entries
846 Details : Added basic proc entries for the MGS showing what filesystems
849 Severity : enhancement
851 Description: provide MGS failover
852 Details : Added config lock reacquisition after MGS server failover.
854 Severity : enhancement
856 Description: add Linux 2.4 support
857 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
861 Description: a large application tries to do I/O to the same resource and dies
863 Details : Check the req->rq_arrival time after the call to
864 ost_brw_lock_get(), but before we do anything about
865 processing it & sending the BULK transfer request. This
866 should help move old stale pending locks off the queue as
867 quickly as obd_timeout.
870 Frequency : when an incorrect nid is specified during startup
872 Description: ptlrpc connect to non-existant node causes kernel crash
873 Details : LNET can't be re-entered from an event callback, which
874 happened when we expire a message after the export has been
875 cleaned up. Instead, hand the zombie cleanup off to another
878 Severity : enhancement
880 Description: plain/inodebits lock performance improvement
881 Details : Grouping plain/inodebits in granted list by their request modes
882 and bits policy, thus improving the performance of search through
886 Frequency : only if OST filesystem is corrupted
888 Description: client incorrectly hits assertion in ptlrpc_replay_req()
889 Details : for a short time RPCs with bulk IO are in the replay list,
890 but replay of bulk IOs is unimplemented. If the OST filesystem
891 is corrupted due to disk cache incoherency and then replay is
892 started it is possible to trip an assertion. Avoid putting
893 committed RPCs into the replay list at all to avoid this issue.
896 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
898 Description: System hang on startup
899 Details : This bug allowed the liblustre (e.g. catamount) client to
900 return to the app before handling all startup RPCs. This
901 could leave the node unresponsive to lustre network traffic
902 and manifested as a server ptllnd timeout.
904 Severity : enhancement
906 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
907 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
908 boolean to control whether to print peer debug info when a
909 client's RPC times out.
912 Frequency : only for kernels with patches from Lustre below 1.4.3
914 Description: Remove old rdonly API
915 Details : Remove old rdonly API which unused from at least lustre 1.4.3
918 Frequency : only for devices with external journals
920 Description: Set external device read-only also
921 Details : During a commanded failover stop, we set the disk device
922 read-only while the server shuts down. We now also set any
923 external journal device read-only at the same time.
926 Frequency : when upgrading from 1.4 while trying to change parameters
928 Description: The wrong (new) MDC name was used when setting parameters for
929 upgraded MDT's. Also allows changing of OSC (and MDC)
930 parameters if --writeconf is specified at tunefs upgrade time.
933 Frequency : when setting specific ost indicies
935 Description: QOS code breaks on skipped indicies
936 Details : Add checks for missing OST indicies in the QOS code, so OSTs
937 created with --index need not be sequential.
939 Severity : enhancement
941 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
942 Details : The uninit_groups feature works in conjunction with the kernel
943 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
944 up the pass1 processing of e2fsck. This is a read-only feature
945 in ldiskfs2 only, so older kernels and current ldiskfs cannot
946 mount filesystems that have had this feature enabled.
948 Severity : enhancement
950 Description: Improve multi-block allocation algorithm to avoid fragmentation
951 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
952 allocation locality and avoid filesystem fragmentation.
954 ------------------------------------------------------------------------------
956 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
958 * Support for kernels:
959 2.4.21-47.0.1.EL (RHEL 3)
961 2.6.9-42.0.10.EL (RHEL 4)
962 2.6.12.6 vanilla (kernel.org)
963 2.6.16.27-0.9 (SLES 10)
964 * Recommended e2fsprogs version: 1.39.cfs5
966 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
970 Frequency : occasional, depends on client load and configuration
971 Bugzilla : 12181, 12203
972 Description: data loss for recently-modified files
974 Details : In some cases it is possible that recently written or created
975 files may not be written to disk in a timely manner (this should
976 normally be within 30s unless client IO load is very high).
977 The problem appears as zero-length files or files that are a
978 multiple of 1MB in size after a client crash or client eviction
979 that are missing data at the end of the file.
981 This problem is more likely to be hit on clients where files are
982 repeatedly created and unlinked in the same directory, clients
983 have a large amount of RAM, have many CPUs, the filesystem has
984 many OSTs, the clients are rebooted frequently, and/or the files
985 are not accessed by other nodes after being written.
987 The presence of the problem can be detected by looking at
988 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
989 smaller than the second (nr_unused) then dirty files will not
990 be flushed automatically to disk. "sync; sleep 10" should be
991 run several times on the node before unmounting it to update
992 Lustre (this is also safe to run on nodes without this problem).
994 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
995 that can cause this same problem, so customers using that kernel
996 also need to update the kernel in addition to Lustre. In order
997 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
999 It is normal that files written just before a client crash (less
1000 than 30s) may not yet have been flushed to disk, even for local
1004 Frequency : frequent on thin XT3 nodes
1006 Description: UUID collision on thin XT3 Linux nodes
1007 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1008 randomly, since we relied on an insufficiently-seeded PRNG.
1013 Description: OSS hangs after "All ost request buffers busy"
1014 Details : A deadlock between quota and journal operations caused OSS
1015 hangs after printing "All ost request buffers busy."
1018 Frequency : always on liblustre builds
1020 Description: Cleanup compiler warnings on liblustre
1023 Frequency : always on liblustre builds on XT3
1025 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1029 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1030 > 32000 subdirectories
1031 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1032 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1033 subdirectory count crosses 32000. This will aid e2fsck to
1034 correctly handle more than 32000 subdirectories.
1037 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1039 Description: System hang on startup
1040 Details : This bug allowed the liblustre (e.g. catamount) client to
1041 return to the app before handling all startup RPCs. This
1042 could leave the node unresponsive to lustre network traffic
1043 and manifested as a server ptllnd timeout.
1045 Severity : enhancement
1047 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1048 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1049 boolean to control whether to print peer debug info when a
1050 client's RPC times out.
1055 Description: make O_SYNC working on 2.6 kernels
1056 Details : 2.6 kernels use different method for mark pages for write,
1057 so need add a code to lustre for O_SYNC work.
1062 Description: Failure to close file and release space on NFS
1063 Details : Put inode details into lock acquired in ll_intent_file_open.
1064 Use mdc_intent_lock in ll_intent_open to properly
1065 detect all kind of errors unhandled by mdc_enqueue.
1070 Description: proc file read during shutdown sometimes raced obd removal,
1072 Details : Add lock to prevent obd access after proc file removal.
1075 Frequency : Only for files larger than 4GB on 32-bit clients.
1077 Description: improperly doing page alignment of locks
1078 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1079 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1084 Description: RPCs being resent when they shouldn't be
1085 Details : Some RPCs that should not be resent are being resent. This
1086 can cause inconsistencies in the RPC state machine. Do not
1087 resend such requests.
1090 Frequency : rare, only with NFS export
1092 Description: Crash on NFS re-export node
1093 Details : under very unusual load conditions an assertion is hit in
1094 ll_intent_file_open()
1097 Frequency : only if OST filesystem is corrupted
1099 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1100 Details : for a short time RPCs with bulk IO are in the replay list,
1101 but replay of bulk IOs is unimplemented. If the OST filesystem
1102 is corrupted due to disk cache incoherency and then replay is
1103 started it is possible to trip an assertion. Avoid putting
1104 committed RPCs into the replay list at all to avoid this issue.
1109 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1110 Details : Large single O_DIRECT read and write calls can fail to allocate
1111 a sufficiently large buffer to process the request. In case of
1112 allocation failure the allocation is retried with a smaller
1113 buffer and broken into smaller requests.
1115 Severity : enhancement
1117 Description: Add -o localflock option to simulate old noflock behaviour.
1118 Details : This will achieve local-only flock/fcntl locks coherentness.
1123 Description: versioning check is incomplete
1124 Details : Checking the version difference of client vs. server, report
1125 error if the gap is too big.
1130 Description: add support PG_writeback bit
1131 Details : add support for PG_writeback bit for Lustre, for more carefull
1132 work with page cache in 2.6 kernel. This also fix some deadlocks
1133 and remove hack for work O_SYNC with 2.6 kernel.
1135 Severity : enhancement
1137 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1138 Details : The uninit_groups feature works in conjunction with the kernel
1139 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1140 up the pass1 processing of e2fsck. This is a read-only feature
1141 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1142 mount filesystems that have had this feature enabled.
1144 Severity : enhancement
1146 Description: Improve multi-block allocation algorithm to avoid fragmentation
1147 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1148 allocation locality and avoid filesystem fragmentation.
1150 ------------------------------------------------------------------------------
1152 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1154 * Support for kernels:
1155 2.6.9-42.0.3.EL (RHEL 4)
1156 2.6.5-7.276 (SLES 9)
1157 2.4.21-47.0.1.EL (RHEL 3)
1158 2.6.12.6 vanilla (kernel.org)
1159 2.6.16.21-0.8 (SLES10)
1160 * Recommended e2fsprogs version: 1.39.cfs2-0
1162 * The backwards-compatible /proc/sys/portals symlink has been removed
1163 in this release. Before upgrading, please ensure that you change
1164 any configuration scripts or /etc/sysctl.conf files that access
1165 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1166 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1167 in advance of the upgrade on any system running Lustre 1.4.6 or
1168 newer, since /proc/sys/lnet was added in that version.
1169 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1173 Frequency : only when quota is used
1175 Description: avoid scanning export list for quota master
1176 Details : Change the algorithms to avoid scanning export list in order
1177 to improve the efficiency.
1180 Frequency : MDS failover only, very rarely
1182 Description: "went back in time" messages on mds failover
1183 Details : The greatest transno may be lost when the current operation
1184 finishes with an error (transno==0) and the client's last_rcvd
1185 record is over-written. Save the greatest transno in the
1186 mds_last_transno for this case.
1189 Frequency : always for specific kernels and striping counts
1191 Description: client may get "Matching packet too big" without ACL support
1192 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1193 when trying to access files in certain configurations. The
1194 clients should in fact be denied when mounting because they do
1195 not understand ACLs.
1198 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1200 Description: many clients connecting with IO in progress causes connect timeouts
1201 Details : Avoid synchronous journal commits to avoid delays caused by many
1202 clients connecting/disconnecting when bulk IO is in progress.
1203 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1204 OST_IO_PORTAL to avoid delays behind potentially many pending
1208 Frequency : occasionally with multiple writers to a single file
1210 Description: shared writes to file may result in wrong size reported by stat()
1211 Details : Allow growing of kms when extent lock is cancelled
1214 Frequency : always with random mmap IO to multi-striped file
1216 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1217 Details : Only if the hole is at the end of OST object so that kms is too
1218 small. Fix is to increase kms accordingly in ll_nopage.
1221 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1223 Description: writes to a missing object would leak memory on the OST
1224 Details : If there is an inconsistency between the MDS and OST filesystems,
1225 such that the MDS references an object that doesn't exist, writes
1226 to that object will leak memory due to incorrect cleanup in the
1227 error handling path, eventually running out of memory on the OST.
1232 Description: Creating too long symlink causes lustre errors
1233 Details : Check symlink and name lengths before sending requests to MDS.
1236 Frequency : only if flock is enabled (not on by default)
1238 Description: posix locks not released on fd closure on 2.6.9+
1239 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1240 kernels, this caused such locks not to be released on fd close and
1241 then assertions on fs unmount about still used locks.
1244 Frequency : MDS failover only, very rarely
1246 Description: clients may get ASSERTION(granted_lock != NULL)
1247 Details : When request was taking a long time, and a client was resending
1248 a getattr by name lock request. The were multiple lock requests
1249 with the same client lock handle and
1250 mds_getattr_name->fixup_handle_for_resent_request found one of the
1251 lock handles but later failed with ASSERTION(granted_lock != NULL).
1256 Description: handle->h_buffer_credits > 0, assertion failure
1257 Details : h_buffer_credits is zero after truncate, causing assertion
1258 failure. This patch extends the transaction or creates a new
1262 Frequency : NFS re-export or patchless client
1263 Bugzilla : 11179, 10796
1264 Description: Crash on NFS re-export node (__d_move)
1265 Details : We do not want to hash the dentry if we don't have a lock.
1266 But if this dentry is later used in d_move, we'd hit uninitialised
1267 list head d_hash, so we just do this to init d_hash field but
1268 leave dentry unhashed.
1271 Frequency : NFS re-export or patchless client
1273 Description: NFS exports has problem with symbolic link
1274 Details : lustre client didn't properly install dentry when re-exported
1275 to NFS or running patchless client.
1278 Frequency : NFS re-export or patchless client
1280 Description: Various nfs/patchless fixes.
1281 Details : fixes reuse disconected alias for lookup process - this fixes
1282 warning "find_exported_dentry: npd != pd",
1283 fix permission error with open files at nfs.
1284 fix apply umask when do revalidate.
1287 Frequency : occasional
1289 Description: Crash on NFS re-export node
1290 Details : calling clear_page() on the wrong pointer triggered oops in
1291 generic_mapping_read().
1294 Frequency : rarely, using O_DIRECT IO
1296 Description: unaligned directio crashes client with LASSERT
1297 Details : check for unaligned buffers before trying any requests.
1300 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1302 Description: stale data returned from RAID cache
1303 Details : If only a small amount of IO is done to the RAID device before
1304 reading it again it is possible to get stale data from the RAID
1305 cache instead of reading it from disk.
1308 Frequency : always for sles10 kernel
1310 Description: sles10 support
1311 Details : ll_follow_link: compile fixes and using of nd_set_link
1312 under newer kernels.
1315 Frequency : depends on arch, kernel and compiler version, always on sles10
1318 Description: recursive or deep enough symlinks cause stack overflow
1319 Details : getting rid of large stack-allocated variable in
1323 Frequency : depends on hardware
1325 Description: lustre write performance loss in the SLES10 kernel
1326 Details : the performance loss is caused by using of write barriers in the
1327 ext3 code. The SLES10 kernel turns barrier support on by
1328 default. The fix is to undo that change for ldiskfs.
1330 ------------------------------------------------------------------------------
1332 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1334 * Support for kernels:
1335 2.6.9-42.0.3EL (RHEL 4)
1336 2.6.5-7.276 (SLES 9)
1337 2.4.21-47.0.1.EL (RHEL 3)
1338 2.6.12.6 vanilla (kernel.org)
1342 Frequency : quota enabled and large files being deleted
1344 Description: releasing more than 4GB of quota at once hangs OST
1345 Details : If a user deletes more than 4GB of files on a single OST it
1346 will cause the OST to spin in an infinite loop. Release
1347 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1352 Description: statfs data retrieved from /proc may be stale or zero
1353 Details : When reading per-device statfs data from /proc, in the
1354 {kbytes,files}_{total,free,avail} files, it may appear
1355 as zero or be out of date.
1358 Frequency : systems with MD RAID1 external journal devices
1360 Description: lconf's call to blkid is confused by RAID1 journal devices
1361 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1362 of returning all block devices that match the journal UUID.
1365 Frequency : always, for aggregate stripe size over 4GB
1367 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1368 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1369 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1372 Frequency : always if "lfs find" used on a local file/directory
1374 Description: "lfs find" segfaults if used on a local file/directory
1375 Details : The case where a directory component was not specified wasn't
1376 handled correctly. Handle this properly.
1379 Frequency : always on ppc64
1381 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1382 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1383 perform endianness conversion.
1386 Frequency : rarely (truncate to non-zero file size after write under load)
1387 Bugzilla : 10730, 10687
1388 Description: Files padded with zeros to next 4K multiple
1389 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1390 it is possible that files that are truncated to a non-zero size
1391 immediately after being written are filled with zero bytes beyond
1392 the truncated size. No file data is lost.
1394 Severity : enhancement
1396 Description: Allow recovery/failover for liblustre clients.
1397 Details : liblustre clients were unaware of failover configurations until
1400 Severity : enhancement
1402 Description: user file locks should fail when not mounting with flock option
1403 Details : Set up an error-returning stub in ll_file_operations.lock field
1404 to prevent incorrect behaviour when client is mounted without
1405 flock option. Also, set up properly f_op->flock field for
1409 Frequency : always on ia64
1411 Description: "lfs df" loops on printing out MDS statfs information
1412 Details : The obd_ioctl_data was not initialized and in some systems
1413 this caused a failure during the ioctl that did not return
1414 an error. Initialize the struct and return an error on failure.
1417 Frequency : SLES 9 only
1419 Description: Error of copying files with lustre special EAs as root
1420 Details : Client side always return success for setxattr call for lustre
1421 special xattr (currently only "trusted.lov").
1424 Frequency : rarely on clusters with both ia64+i386 clients
1426 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1427 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1428 clients are doing concurrent writes to the same file it is
1429 possible that the ia64 clients may LASSERT because the OST
1430 extent locks are not PAGE_SIZE aligned. Ensure that grown
1431 locks are always aligned on the request boundary.
1434 Frequency : specific use, occasional
1436 Description: Overwriting in use executable truncates on-disk binary image
1437 Details : If one node attempts to overwrite an executable in use by
1438 another node, we now correctly return ETXTBSY instead of
1439 truncating the file.
1441 Severity : enhancement
1443 Description: Async OSC create to avoid the blocking unnecessarily.
1444 Details : If a OST has no remain object, system will block on the creating
1445 when need to create a new object on this OST. Now, ways use
1446 pre-created objects when available, instead of blocking on an
1447 empty osc while others are not empty. If we must block, we block
1448 for the shortest possible period of time.
1453 Description: chmod on Lustre root is propagated to other clients
1454 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1455 invalid by the follow_mount time.
1458 Frequency : liblustre clients only
1460 Description: Race in 'instant cancel' lock handling could lead to such locks
1461 never to be granted in case of SMP MDS
1462 Details : Do not destroy not yet granted but cbpending locks in
1466 Frequency : replay/resend of open
1468 Description: non null lock assetion failure in mds_intent_policy
1469 Details : Trying to replay/resend lockless open requests resulted in
1470 mds_open() returning 0 with no lock. Now it sets a flag if
1471 a lock is going to be returned.
1473 Severity : enhancement
1475 Description: Checksum enhancements
1476 Details : New checksum enhancements allow for resending RPCs that failed
1479 Severity : enhancement
1481 Description: Tunables on number of dirty pages in cacche
1482 Details : Allow to set limit on number of dirty pages cached.
1487 Description: client crash on unmount - lock still has references
1488 Details : In some error handling cases it was possible to leak a lock
1489 reference on a client while accessing a file. This was not
1490 harmful to the client during operation, but would cause the
1491 client to crash when the filesystem is unmounted.
1494 Frequency : specific case, rare
1496 Description: ETXTBSY on mds though file not in use by client
1497 Details : ETXTBSY is no longer incorrectly returned when attempting to
1498 chmod or chown a directory that the user previously tried to
1499 execute or a currently-executing binary.
1502 Frequency : extremely rare except on liblustre-based clients
1504 Description: Lustre space not freed when files are deleted
1505 Details : Clean up open-unlinked files after client eviction. Previously
1506 the unlink was skipped and the files remained as orphans.
1511 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1512 Details : In some cases it was possible to send a blocking callback to
1513 a client doing a glimpse, even though that client didn't get
1514 a lock granted. When the glimpse lock is cancelled on the OST
1515 the freed lock is left on the waiting list and corrupted the list.
1518 Frequency : all core dumps
1520 Description: Broke core dumps to lustre
1521 Details : Negative dentry may be unhashed if parent does not have UPDATE
1522 lock, but some callers, e.g. do_coredump, expect dentry to be
1523 hashed after successful create, hash it in ll_create_it.
1525 ------------------------------------------------------------------------------
1527 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1529 * Support for kernels:
1530 2.6.9-42.0.2.EL (RHEL 4)
1531 2.6.5-7.276 (SLES 9)
1532 2.4.21-40.EL (RHEL 3)
1533 2.6.12.6 vanilla (kernel.org)
1537 Frequency : always on RHEL 3
1539 Description: Number of open files grows over time
1540 Details : The number of open files grows over time, whether or not
1541 Lustre is started. This was due to a filp leak introduced
1542 by one of our kernel patches.
1544 ------------------------------------------------------------------------------
1546 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1548 * Support for kernels:
1549 2.6.9-42.EL (RHEL 4)
1550 2.6.5-7.276 (SLES 9)
1551 2.4.21-40.EL (RHEL 3)
1552 2.6.12.6 vanilla (kernel.org)
1557 Bugzilla : 5719, 9635, 9792, 9684
1558 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1559 Details : If a server is under heavy load and cannot reply to new
1560 connection requests before the client resends the (re)connect,
1561 the connection handling code can behave badly if two service
1562 threads are concurrently handing separate (re)connections from
1563 the same client. Add better locking to the connection handling
1564 code, and ensure that only a single connection will be processed
1565 for a given client UUID, even if the lock is dropped.
1567 Severity : enhancement
1569 Description: add TCP zero-copy support to kernel
1570 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1571 sends if the hardware supports scatter-gather and checksumming.
1572 This allows socklnd to do client-write and server-read more
1573 efficiently and reduce CPU utilization from skbuf copying.
1576 Frequency : only if NFS exporting from client
1578 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1579 Details : If setting attributes on a file created under NFS that had
1580 never been opened it would be possible to oops the client
1581 if the file had no objects.
1584 Frequency : always for liblustre
1586 Description: liblustre client does MDS+OSTs setattr RPC for each write
1587 Details : When doing a write from a liblustre client, the client
1588 incorrectly issued an RPC to the MDS and each OST the file was
1589 striped over in order to update the timestamps. When writing
1590 with small chunks and many clients this could overwhelm the MDS
1591 with RPCs. In all cases it would slow down the write because
1592 these RPCs are unnecessary.
1594 Severity : enhancement
1596 Description: allow number of MDS service threads to be changed at module load
1597 Details : It is now possible to change the number of MDS service threads
1598 running. Adding "options mds mds_num_threads={N}" to the MDS's
1599 /etc/modprobe.conf will set the number of threads for the next
1600 time Lustre is restarted (assuming the "mds" module is also
1601 reloaded at that time). The default number of threads will
1602 stay the same, 32 for most systems.
1607 Description: OST crash if filesystem is unformatted or corrupt
1608 Details : If an OST is started on a device that has never been formatted
1609 or if the filesystem is corrupt and cannot even mount then the
1610 error handling cleanup routines would dereference a NULL pointer.
1615 Description: NULL pointer deref in llap_from_page.
1616 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1617 incorrect) mapping if the page was truncated/reclaimed while it was
1618 searched for. Check for this condition and skip such pages when
1619 doing readahead. Introduce extra check to llap_from_page() to
1620 verify page->mapping->host is non-NULL (so page is not anonymous).
1623 Frequency : Sometimes when using sys_sendfile
1625 Description: "page not covered by a lock" warnings from ll_readpage
1626 Details : sendfile called ll_readpage without right page locks present.
1627 Now we introduced ll_file_sendfile that does necessary locking
1628 around call to generic_file_sendfile() much like we do in
1632 Frequency : with certain MDS communication failures at client mount time
1634 Description: NULL pointer deref after failed client mount
1635 Details : a client connection request may delayed by the network layer
1636 and not be sent until after the PTLRPC layer has timed out the
1637 request. If the client fails the mount immediately it will try
1638 to clean up before the network times out the request. Add a
1639 reference from the request import to the obd device and delay
1640 the cleanup until the network drops the request.
1643 Frequency : occasionally during client (re)connect
1645 Description: assertion failure during client (re)connect
1646 Details : processing a client connection request may be delayed by the
1647 client or server longer than the client connect timeout. This
1648 causes the client to resend the connection request. If the
1649 original connection request is replied in this interval, the
1650 client may trip an assertion failure in ptlrpc_connect_interpret()
1651 which thought it would be the only running connect process.
1654 Frequency : only with obd_echo servers and clients that are rebooted
1656 Description: kernel BUG accessing uninitialized data structure
1657 Details : When running an obd_echo server it did not start the ping_evictor
1658 thread, and when a client was evicted an uninitialized data
1659 structure was accessed. Start the ping_evictor in the RPC
1660 service startup instead of the OBD startup.
1662 Severity : enhancement
1663 Bugzilla : 10193 (patchless)
1664 Description: Remove dependency on various unexported kernel interfaces.
1665 Details : No longer need reparent_to_init, exit_mm, exit_files,
1666 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1669 Frequency : rare (only users of deprecated and unsupported LDAP config)
1671 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1672 Details : LDAP apparently contains 'lustreName' attributes instead of
1673 'name'. A simple remapping of the name is sufficient.
1676 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1678 Description: waiting_locks_callback trips kernel BUG if client is evicted
1679 Details : Running with the dump_on_timeout debug flag turned on makes
1680 it possible that the waiting_locks_callback() can try to dump
1681 the Lustre kernel debug logs from an interrupt handler. Defer
1682 this log dumping to the expired_lock_main() thread.
1684 Severity : enhancement
1686 Description: Support NFS exporting on 2.6 kernels.
1687 Details : Implement non-rawops metadata methods for NFS server to use without
1688 changing NFS server code.
1691 Frequency : very rare (synthetic metadata workload only)
1693 Description: two racing renames might cause an MDS thread to deadlock
1694 Details : Running the "racer" program may cause one MDS thread to rename
1695 a file from being the source of a rename to being the target of
1696 a rename at exactly the same time that another thread is doing
1697 so, and the second thread has already enqueued these locks after
1698 doing a lookup of the target and is trying to relock them in
1699 order. Ensure that we don't try to re-lock the same resource.
1702 Frequency : only very large systems with liblustre clients
1704 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1705 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1706 all OSTs in parallel. This allows the eviction of stale liblustre
1707 clients to proceed much faster than if they were done in series,
1708 and also offers similar improvements for other set_info RPCs.
1713 Description: excessive CPU usage during initial read phase on client
1714 Details : During the initial read phase on a client, it would agressively
1715 retry readahead on the file, consuming too much CPU and impacting
1716 performance (since 1.4.5.8). Improve the readahead algorithm
1717 to avoid this, and also improve some other common cases (read
1718 of small files in particular, where "small" is files smaller than
1719 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1724 Description: MDS crash when receiving packet with unknown intent.
1725 Details : Do not LBUG in unknown intent case, just return -EFAULT
1727 Severity : enhancement
1728 Bugzilla : 9293, 9385
1729 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1730 Details : Do not serialize getattr (non-intent version) and statfs.
1733 Frequency : occasional, when OST network is overloaded/intermittent
1735 Description: client evicted by OST after bulk IO timeout
1736 Details : If a client sends a bulk IO request (read or write) the OST
1737 may evict the client if it is unresposive to its data GET/PUT
1738 request. This is incorrect if the network is overloaded (takes
1739 too long to transfer the RPC data) or dropped the OST GET/PUT
1740 request. There is no need to evict the client at all, since
1741 the pinger and/or lock callbacks will handle this, and the
1742 client can restart the bulk request.
1745 Frequency : Always when mmapping file with no objects
1747 Description: client crashes when mmapping file with no objects
1748 Details : Check that we actually have objects in a file before doing any
1749 operations on objects in ll_vm_open, ll_vm_close and
1755 Description: Request leak when working with deleted CWD
1756 Details : Introduce advanced request refcount tracking for requests
1757 referenced from lustre intent.
1759 Severity : Enhancement
1761 Description: Cache open file handles on client.
1762 Details : MDS now will return special lock along with openhandle, if
1763 requested and client is allowed to hold openhandle, even if unused,
1764 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1765 closing files for every read/write openration.
1767 Severity : Enhancement
1769 Description: Cache open negative dentries on client when possible.
1770 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1771 negative dentries on lock revocation.
1776 Description: Remounting a client read-only wasn't possible with a zconf mount
1777 Details : It wasn't possible to remount a client read-only with llmount.
1779 Severity : enhancement
1780 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1781 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1782 MPICH is included to add Lustre-specific ADIO interfaces.
1783 This is based closely on the UFS ADIO layer and only differs
1784 in file creation, in order to allow the OST striping to be set.
1785 This is user-contributed code and not supported by CFS.
1790 Description: extended inode attributes (immutable, append-only) work improperly
1791 when 2.4 and 2.6 kernels are used on client/server or vice versa
1792 Details : Introduce kernel-independent values for these flags.
1794 Severity : enhancement
1797 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1798 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1799 granularity. Now, Fractional MB granularity is be supported,
1800 this is very useful for low memory system.
1802 Severity : enhancement
1804 Description: Getattr by fid
1805 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1806 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1807 have locks and avoids getattr rpc requests for it on every stat.
1810 Frequency : Always, for filesystems larger than 2TB
1812 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1813 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1814 filesystems larger than 2TB to be mounted with mballoc enabled.
1817 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1819 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1820 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1821 filesystems larger than 2TB is mounted then the kernel will
1822 silently corrupt the start of the filesystem. CONFIG_LBD is
1823 enabled for all CFS-supported kernels, but the possibility of
1824 this happening with a modified kernel config exists.
1826 Severity : enhancement
1828 Description: add client O_DIRECT support for 2.6 kernels
1829 Details : It is now possible to do O_DIRECT reads and writes to files
1830 in the Lustre client mountpoint on 2.6 kernel clients.
1832 Severity : enhancement
1834 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1835 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1836 parallel, not waiting for response from every OST before sending
1837 a rpc to the next OST.
1842 Description: setattr vs write race when updating file timestamps
1843 Details : Client processes that update a file timestamp into the past
1844 right after writing to the file (e.g. tar) it is possible that
1845 the updated file modification time can be reset to the current
1846 time due to a race between processing the setattr and write RPC.
1848 Severity : enhancement
1850 Description: Bring 'lfs find' closer in line with regular Linux find.
1851 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1852 -print0 options and obtains all the needed info through the lustre
1855 Severity : enhancement
1857 Description: support up to 1024 configured devices on one node
1858 Details : change obd_dev array from statically allocated to dynamically
1859 allocated structs as they are first used to reduce memory usage
1864 Description: Flush dirty partially truncated pages during truncate
1865 Details : Immediatelly flush partially truncated pages in filter_setattr,
1866 this way we completely avoid having any pages in page cache on OST
1867 and can retire ugly workarounds during writes to flush such pages.
1872 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1873 Details : protect inode from truncation within vfs_unlink() context
1874 just take a reference before calling vfs_unlink() and release it
1875 when parent's i_sem is free.
1880 Description: last_id value checked outside lock on OST caused LASSERT failure
1881 Details : If there were multiple MDS->OST object precreate requests in
1882 flight, it was possible that the OST's last object id was checked
1883 outside a lock and incorrectly tripped an assertion. Move checks
1884 inside locks, and discard old precreate requests.
1887 Frequency : always, if extents are used on OSTs
1889 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1890 Details : OSTs using the extents format would not zero the high 16 bits of
1891 the index physical block number. This is not a problem for any
1892 OST filesystems smaller than 16TB, and no kernels support ext3
1893 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1894 new/modified files) and can be fixed for existing filesystems
1895 with e2fsprogs-1.39-cfs1.
1900 Description: import connection selection may be incorrect if timer wraps
1901 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1902 import connections to be ignored if the 32-bit jiffies counter
1903 wraps. Use a 64-bit jiffies counter.
1906 Frequency : very large clusters immediately after boot
1908 Description: LNET request buffers exhausted under heavy short-term load
1909 Details : If a large number of client requests are generated on a service
1910 that has previously never seen so many requests it is possible
1911 that the request buffer growth cannot keep up with the spike in
1912 demand. Instead of dropping incoming requests, they are held in
1913 the LND until the RPC service can accept more requests.
1916 Frequency : Sometimes during replay
1918 Description: Assertion failure in ll_local_open after replay.
1919 Details : If replay happened on an open request reply before we were able
1920 to set replay handler, reply will become not swabbed tripping the
1921 assertion in ll_local_open. Now we set the handler right after
1922 recognising of open request
1925 Frequency : very rare
1927 Description: kernel reports "badness in vsnprintf"
1928 Details : Reading from the "recovery_status" /proc file in small chunks
1929 may cause a negative length in lprocfs_obd_rd_recovery_status()
1930 call to vsnprintf() (which is otherwise harmless). Exit early
1931 if there is no more space in the output buffer.
1933 Severity : enhancement
1935 Description: clear OBD RPC statistics by writing to them
1936 Details : It is now possible to clear the OBD RPC statistics by writing
1937 to the "stats" file.
1942 Description: Client mtime is not the same on different clients after utimes
1943 Details : In some cases, the client was using the utimes() syscall on
1944 a file cached on another node. The clients now validate the
1945 ctime from the MDS + OSTs to determine which one is right.
1950 Description: Inability to activate failout mode
1951 Details : lconf script incorrectly assumed that in pythong string's numeric
1952 value is used in comparisons.
1955 Frequency : always with multiple stripes per file
1957 Description: Inefficient object allocation for mutli-stripe files
1958 Details : When selecting which OSTs to stripe files over, for files with
1959 a stripe count that divides evenly into the number of OSTs,
1960 the MDS is always picking the same starting OST for each file.
1961 Return the OST selection heuristic to the original design.
1966 Description: mount failures may take full timeout to return an error
1967 Details : Under some heavy load conditions it is possible that a
1968 failed mount can wait for the full obd_timeout interval,
1969 possibly several minutes, before reporting an error.
1970 Instead return an error as soon as the status is known.
1972 Frequency : quota enabled and large files being deleted
1974 Description: releasing more than 4GB of quota at once hangs OST
1975 Details : If a user deletes more than 4GB of files on a single OST it
1976 will cause the OST to spin in an infinite loop. Release
1977 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1982 Description: statfs data retrieved from /proc may be stale or zero
1983 Details : When reading per-device statfs data from /proc, in the
1984 {kbytes,files}_{total,free,avail} files, it may appear
1985 as zero or be out of date.
1988 Frequency : systems with MD RAID1 external journal devices
1990 Description: lconf's call to blkid is confused by RAID1 journal devices
1991 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1992 of returning all block devices that match the journal UUID.
1995 Frequency : always, for aggregate stripe size over 4GB
1997 Description: assertion fails when trying to use 4GB stripe size
1998 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1999 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2002 Frequency : always on ppc64
2004 Description: the first write on an ext3 filesystem with mballoc got stuck
2005 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2006 perform endianness conversion.
2008 ------------------------------------------------------------------------------
2010 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2012 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2013 INTEROPERATE with older versions automatically. Please read the
2014 user documentation before upgrading any part of a live system.
2015 * WARNING: Lustre networking configuration changes are required with
2016 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2019 * Support for kernels:
2020 2.6.9-22.0.2.EL (RHEL 4)
2021 2.6.5-7.244 (SLES 9)
2022 2.6.12.6 vanilla (kernel.org)
2025 Severity : enhancement
2026 Bugzilla : 7981/8208
2027 Description: Introduced Lustre Networking (LNET)
2028 Details : LNET is new networking infrastructure for Lustre, it includes
2029 a reorganized network configuration mode (see the user
2030 documentation for full details) as well as support for routing
2031 between different network fabrics. Lustre Networking Devices
2032 (LNDs) for the supported network fabrics have also been
2033 created for this new infrastructure.
2035 Severity : enhancement
2036 Description: Introduced Access control lists
2037 Details : clients can set ACLs on files and directories in order to have
2038 more fine-grained permissions than the standard Unix UGO+RWX.
2039 The MDS must be started with the "-o acl" mount option.
2041 Severity : enhancement
2042 Description: Introduced filesystem quotas
2043 Details : Administrators may now establish per-user quotas on the
2046 Severity : enhancement
2048 Description: Configuration change for the XT3
2049 The PTLLND is now used to run Lustre over Portals on the XT3
2050 The configure option(s) --with-cray-portals are no longer used.
2051 Rather --with-portals=<path-to-portals-includes> is used to
2052 enable building on the XT3. In addition to enable XT3 specific
2053 features the option --enable-cray-xt3 must be used.
2058 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2059 Details : A race between allocating a new llcd and re-getting the llcd_lock
2060 allowed another thread to grab newly-allocated llcd.
2062 Severity : enhancement
2064 Description: 2.6 OST async journal commit and locking fix to improve performance
2065 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2066 2.6 kernels are now async as they already were in 2.4 kernels so
2067 that they can commit concurrently with the network bulk transfer.
2068 For block-allocated files the filter allocation semaphore is held
2069 to avoid filesystem fragmentation during allocation. BKL lock
2070 removed for 2.6 xattr operations where it is no longer needed.
2075 Description: lconf incorrectly determined whether two IP networks could talk
2076 Details : In some more complicated routing and multiple-network
2077 configurations, lconf will avoid trying to make a network
2078 connection to a disjoint part of the IP space. It was doing the
2079 math incorrectly for one set of cases.
2084 Description: Fix for potential infinite loop processing records in an llog.
2085 Details : If an llog record is corrupted/zeroed, it is possible to loop
2086 forever in llog_process(). Validate the llog record length
2087 and skip the remainder of the block on error.
2090 Frequency : occasional (liblustre only)
2092 Description: liblustre could not open files whose last component is a symlink
2093 Details : sysio_path_walk() would incorrectly pass the open intent to
2094 intermediate path components.
2097 Frequency : rare (liblustre only with non-standard tuning)
2098 Bugzilla : 7201 (7350)
2099 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2100 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2101 after it has been released, only do this for referenced locks
2103 Severity : enhancement
2105 Description: specifying an (invalid) directory default stripe_size of -1
2106 would reset the directory default striping
2107 Details : stripe_size -1 was used internally to signal directory stripe
2108 removal, now use "all default" to signal dir stripe removal
2109 as a directory striping of "all default" is not useful
2112 Frequency : common for large clusters running liblustre clients
2114 Description: doing an ls when liblustre clients are running is slow
2115 Details : sending a glimpse AST to a liblustre client waits for every AST
2116 to time out, as liblustre clients will not respond. Since they
2117 cannot cache data we refresh the OST lock LVB from disk instead.
2119 Severity : enhancement
2121 Description: doing an ls at the same time as file IO can be slow
2122 Details : enqueue and other "small" requests can be blocked behind many
2123 large IO requests. Create a new OST IO portal for non-IO
2124 requests so they can be processed faster.
2127 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2129 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2130 created file with mode 000
2131 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2132 allow an ftruncate() as a non-root user to a file with mode 000.
2133 Lustre now respects this flag to disable mode checks when
2134 truncating a file owned by the user
2137 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2140 Description: Revoking locks from clients that went dead or catatonic might take
2142 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2143 cancellation of such locks instant on servers without waiting for
2144 any reply from clients. Clients drops these locks when cancel
2145 notification from server is received without replying.
2148 Frequency : liblustre-only, when liblustre client dies or becomes busy
2150 Description: Doing ls on Linux clients can take a long time with active
2152 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2153 granting such locks to it in the first place if possible. Locks
2154 are taken by proxy on the OST during the read or write and
2155 dropped immediately afterward. Add connect flags handling, do
2156 not grant locks to liblustre clients for glimpse ASTs.
2158 Severity : enhancement
2160 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2161 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2162 sequential reads and behaves badly with applications doing random
2163 reads. Improve it by reading ahead at least read region, and
2164 avoiding excessive large RPC for small reads.
2166 Severity : enhancement
2168 Description: Creating more than 1000 files for a single job may cause a load
2169 imbalance on the OSTs if there are also a large number of OSTs.
2170 Details : qos_prep_create() uses an OST index reseed value that is an
2171 even multiple of the number of available OSTs so that if the
2172 reseed happens in the middle of the object allocation it will
2173 still utilize the OSTs as uniformly as possible.
2178 Description: OST or MDS may oops in ping_evictor_main()
2179 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2180 but doesn't restart at beginning of obd_exports_timed list
2183 Severity : enhancement
2185 Description: improve by-nid export eviction on the MDS and OST
2186 Details : allow multiple exports with the same NID to be evicted at one
2187 time without re-searching the exports list.
2190 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2192 Description: MDS may oops in groups_free()
2193 Details : in rare race conditions a newly allocated group_info struct is
2194 freed again, and this can be NULL. The 2.4 compatibility code
2195 for groups_free() checked for a NULL pointer, but 2.6 did not.
2198 Frequency : common for liblustre clients doing little filesystem IO
2199 Bugzilla : 9352, 7313
2200 Description: server may evict liblustre clients accessing contended locks
2201 Details : if a client is granted a lock or receives a completion AST
2202 with a blocking AST already set it would not reply to the AST
2203 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2206 Frequency : lfs setstripe, only systems with more than 160 OSTs
2208 Description: unable to set striping with a starting offset beyond OST 160
2209 Details : llapi_create_file() incorrectly limited the starting stripe
2210 index to the maximum single-file stripe count.
2213 Frequency : LDAP users only
2215 Description: lconf did not handle in-kernel recovery with LDAP properly
2216 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2218 Severity : enhancement
2220 Description: bind OST threads to NUMA nodes to improve performance
2221 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2222 node and do their allocations there to localize memory access
2224 Severity : enhancement
2226 Description: llmount can determine client NID directly from Myrinet (GM)
2227 Details : the client NID code from gmnalnid was moved directly into
2228 llmount, removing the need to use this or specifying the
2229 client NID explicitly when mounting GM clients with zeroconf
2232 Frequency : if client is started with down MDS
2234 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2235 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2236 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2237 ptlrpc_import_delayed_req() because we didn't check if the
2238 request was interrupted, and we also didn't break out of the
2239 event loop if there was no timeout
2244 Description: data loss during non-page-aligned writes to a single file from
2245 both multiple nodes and multiple threads on one node at same time
2246 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2247 inconsistency led to false short-reads, that were cached and later
2248 used by ->prepare_write() to fill in partially written page,
2249 leading to data loss.
2252 Frequency : always, if lconf --abort_recovery used
2254 Description: lconf --abort_recovery fails with 'Operation not supported'
2255 Details : lconf was attempting to abort recovery on the MDT device and not
2258 Severity : enhancement
2260 Description: remove cleanup logs
2261 Details : replace lconf-generated cleanup logs with lustre internal
2262 cleanup routines. Eliminates the need for client-cleanup and
2265 Severity : enhancement
2267 Description: add support for EAs (user and system) on lustre filesystems
2268 Details : it is now possible to store extended attributes in the Lustre
2269 client filesystem, and with the user_xattr mount option it
2270 is possible to allow users to store EAs on their files also
2272 Severity : enhancement
2274 Description: Add possibility (config option) to show minimal available OST free
2276 Details : When compiled with --enable-mindf configure option, statfs(2)
2277 (and so, df) will return least minimal free space available from
2278 all OSTs as amount of free space on FS, instead of summary of
2279 free spaces of all OSTs.
2281 Severity : enhancement
2283 Description: do not expand extent locks acquired on OST-side
2284 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2285 by server: they are not cached anyway.
2288 Frequency : when mmap is used/binaries executed from Lustre
2290 Description: Unmmap pages before throwing them away from read cache.
2291 Details : llap_shrink cache now attempts to unmap pages before discarding
2292 them (if unmapping failed - do not discard). SLES9 kernel has
2293 extra checks that trigger if this unmapping is not done first.
2298 Description: lconf didn't resolve symlinks before checking to see whether a
2299 given mountpoint was already in use
2302 Frequency : when migrating failover services
2303 Bugzilla : 6395, 9514
2304 Description: When migrating a subset of services from a node (e.g. failback
2305 from a failover service node) the remaining services would
2306 time out and evict clients.
2307 Details : lconf --force (implied by --failover) sets the global obd_timeout
2308 to 5 seconds in order to quickly disconnect, but this caused
2309 other RPCs to time out too quickly. Do not change the global
2310 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2312 Severity : enhancement
2313 Frequency : if MDS is started with down OST
2314 Bugzilla : 9439,5706
2315 Description: Allow startup/shutdown of an MDS without depending on the
2316 availability of the OSTs.
2317 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2318 Add appropriate locking and lov-osc refcounts for safe
2319 cleaning. Add osc abort_inflight calls in case the
2320 synchronize never started.
2323 Frequency : occasional (Cray XT3 only)
2325 Description: root not authorized to access files in CRAY_PORTALS environment
2326 Details : The client process capabilities were not honoured on the MDS in
2327 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2328 been accessed by an authorized user then root was able to access
2329 the file on the local client also. The root user capabilities
2330 are now allowed on the MDS, as this environment has secure UID.
2333 Frequency : occasional
2335 Description: ldiskfs "too long searching" message happens too often
2336 Details : A debugging message (otherwise harmless) prints too often on
2337 the OST console. This has been reduced to only happen when
2338 there are fragmentation problems on the filesystem.
2343 Description: Division by zero in statfs when all OSCs are inactive
2344 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2345 lov_statfs divides by value returned by lov_get_stripecnt().
2349 Bugzilla : 9489, 3273
2350 Description: First write from each client to each OST was only 4kB in size,
2351 to initialize client writeback cache, which caused sub-optimal
2352 RPCs and poor layout on disk for the first writen file.
2353 Details : Clients now request an initial cache grant at (re)connect time
2354 and so that they can start streaming writes to the cache right
2355 away and always do full-sized RPCs if there is enough data.
2356 If the OST is rebooted the client also re-establishes its grant
2357 so that client cached writes will be honoured under the grant.
2362 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2363 Details : Now I/O RPCs go to different portal number and (presumably) fast
2364 lock requests (and glimses) and other RPCs get their own service
2365 threads pool that should be able to service those RPCs
2368 Severity : enhancement
2370 Description: Ability to exchange lustre version between client and servers and
2371 issue warnings at client side if client is too old. Also for
2372 liblustre clients there is ability to refuse connection of too old
2374 Details : New 'version' field is added to connect data structure that is
2375 filled with version info. That info is later checked by server and
2379 Frequency : rare, liblustre only.
2380 Bugzilla : 9296, 9581
2381 Description: Two simultaneous writes from liblustre at offset within same page
2382 might proceed at the same time overwriting eachother with stale
2384 Details : I/O lock withing llu_file_prwv was released too early, before data
2385 actually was hitting the wire. Extended lock-holding time until
2386 server acknowledges receiving data.
2389 Frequency : extremely rare. Never observed in practice.
2391 Description: avoid generating lustre_handle cookie of 0.
2392 Details : class_handle_hash() generates handle cookies by incrementing
2393 global counter, and can hit 0 occasionaly (this is unlikely, but
2394 not impossible, because initial value of cookie counter is
2395 selected randonly). Value of 0 is used as a sentinel meaning
2396 "unassigned handle" --- avoid it. Also coalesce two critical
2397 sections in this function into one.
2399 Severity : enhancement
2401 Description: allow liblustre clients to delegate truncate locking to OST
2402 Details : To avoid overhead of locking, liblustre client instructs OST to
2403 take extent lock in ost_punch() on client's behalf. New connection
2404 flag is added to handle backward compatibility.
2406 Severity : enhancement
2407 Bugzilla : 4928, 7341, 9758
2408 Description: allow number of OST service threads to be specified
2409 Details : a module parameter allows the number of OST service threads
2410 to be specified via "options ost ost_num_threads={N}" in the
2411 OSS's /etc/modules.conf or /etc/modprobe.conf.
2415 Bugzilla : 6146, 9635, 9895
2416 Description: servers crash with bad pointer in target_handle_connect()
2417 Details : In rare cases when a client is reconnecting it was possible that
2418 the connection request was the last reference for that export.
2419 We would temporarily drop the export reference and get a new
2420 one, but this may have been the last reference and the export
2421 was just destroyed. Get new reference before dropping old one.
2423 Severity : enhancement
2424 Frequency : if client is started with failover MDS
2426 Description: Allow multiple MDS hostnames in the mount command
2427 Details : Try to read the configuration from all specified MDS
2428 hostnames during a client mount in case the "primary"
2431 Severity : enhancement
2433 Description: Stop sending data to evicted clients as soon as possible.
2434 Details : Check if the client we are about to send or are sending data to
2435 was evicted already. (Check is done every second of waiting,
2436 for which l_wait_event interface was extended to allow checking
2437 of exit condition at specified intervals).
2440 Frequency : rare, normally only when NFS exporting is done from client
2442 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2444 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2447 Frequency : always, if srand() is called before liblustre initialization
2449 Description: Liblustre uses system PRNG disturbing its usage by user application
2450 Details : Introduce internal to lustre fast and high-quality PRNG for
2451 lustre usage and make liblustre and some other places in generic
2452 lustre code to use it.
2454 Severity : enhancement
2455 Bugzilla : 9477, 9557, 9870
2456 Description: Verify that the MDS configuration logs are updated when xml is
2457 Details : Check if the .xml configuration logs are newer than the config
2458 logs stored on the MDS and report an error if this is the case.
2459 Request --write-conf, or allow starting with --old_conf.
2461 Severity : enhancement
2463 Description: Handle symlinks in the path when checking if Lustre is mounted.
2464 Details : Resolve intermediate symlinks when checking if a client has
2465 mounted a filesystem to avoid duplicate client mounts.
2470 Description: lconf can hit an error exception but still return success.
2471 Details : The lconf command catches the Command error exception at the top
2472 level script context and will exit with the associated exit
2473 status, but doesn't ensure that this exit status is non-zero.
2478 Description: failure of ptlrpc thread startup can cause oops
2479 Details : Starting a ptlrpc service thread can fail if there are a large
2480 number of threads or the server memory is very fragmented.
2481 Handle this without oopsing.
2484 Frequency : always, only if liblustre and non-default acceptor port was used
2486 Description: liblustre cannot connect to servers with non-default acceptor port
2487 Details : tcpnal_set_default_params() was not called and was therefore
2488 ignoring the environment varaible TCPNAL_PORT, as well as other
2489 TCPNAL_ environment variables
2494 Description: two objects could be created on the same OST for a single file
2495 Details : If an OST is down, in some cases it was possible to create two
2496 objects on a single OST for a single file. No problems other
2497 than potential performance impact and spurious error messages.
2501 Bugzilla : 5681, 9562
2502 Description: Client may oops in ll_unhash_aliases
2503 Details : Client dcache may become inconsistent in race condition.
2504 In some cases "getcwd" can fail if the current directory is
2510 Description: Inode refcounting problems in NFS export code
2511 Details : link_raw functions used to call d_instantiate without obtaining
2512 extra inode reference first.
2516 Bugzilla : 9942, 9903
2517 Description: Referencing freed requests leading to crash, memleaks with NFS.
2518 Details : We used to require that call to ll_revalidate_it was always
2519 followed by ll_lookup_it. Also with revalidate_special() it is
2520 possible to call ll_revalidate_it() twice for the same dentry
2521 even if first occurence returned success. This fix changes semantic
2522 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2523 reference on a request referred from the intent.
2524 ll_intent_release() then releases such a request.
2527 Frequency : rare, normally benchmark loads only
2529 Description: unlinked inodes were kept in memory on the client
2530 Details : If a client is repeatedly creating and unlinking files it
2531 can accumulate a lot of stale inodes in the inode slab cache.
2532 If there is no other client load running this can cause the
2533 client node to run out of memory. Instead flush old inodes
2534 from client cache that have the same inode number as a new inode.
2537 Frequency : SLES9 2.6.5 kernel and long filenames only
2538 Bugzilla : 9969, 10379
2539 Description: utime reports stale NFS file handle
2540 Details : SLES9 uses out-of-dentry names in some cases, which confused
2541 the lustre dentry revalidation. Change it to always use the
2545 Frequency : rare, unless heavy write-truncate concurrency is continuous
2546 Bugzilla : 4180, 6984, 7171, 9963, 9331
2547 Description: OST becomes very slow and/or deadlocked during object unlink
2548 Details : filter_destroy() was holding onto the parent directory lock
2549 while truncating+unlinking objects. For very large objects this
2550 may block other threads for a long time and slow overall OST
2551 responsiveness. It may also be possible to get a lock ordering
2552 deadlock in this case, or run out of journal credits because of
2553 the combined truncate+unlink. Solution is to do object truncate
2554 first in one transaction without parent lock, and then do the
2555 final unlink in a new transaction with the parent lock. This
2556 reduces the lock hold time dramatically.
2559 Frequency : rare, 2.4 kernels only
2561 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2562 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2563 before unmounting their filesystem in order to avoid deadlock.
2564 The kernel_locked() function in 2.4 kernels only checks whether
2565 the kernel lock is held, not whether it is this process that is
2566 holding it as 2.6 kernels do.
2571 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2572 Details : The client ptlrpc code may be trying to reconnect to a down
2573 server before a previous connection attempt has timed out.
2574 Increase the reconnect interval to be longer than the connection
2575 timeout interval to avoid sending duplicate connections to
2579 Frequency : echo_client brw_test command
2581 Description: fix echo_client to work with OST preallocated code
2582 Details : OST preallocation code (5137) didn't take echo_client IO path
2583 into account: echo_client calls filter methods outside of any
2584 OST thread and, hence, there is no per-thread preallocated
2585 pages and buffers to use. Solution: hijack pga pages for IO. As
2586 a byproduct, this avoids unnecessary data copying.
2590 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2591 Description: Client can oops in mdc_commit_close() after open replay
2592 Details : It was possible for the MDS to return an open request with no
2593 transaction number in mds_finish_transno() if the client was
2594 evicted, but without actually returning an error. Clients
2595 would later try to replay that open and may trip an assertion
2596 Simplify the client close codepath, and always return an error
2597 from the MDS in case the open is not successful.
2600 Frequency : rare, 2.6 OSTs only
2602 Description: OST may deadlock under high load on fragmented files
2603 Details : If there was a heavy load and highly-fragmented OST filesystems
2604 it was possible to have all the OST threads deadlock waiting on
2605 allocation of biovecs, because the biovecs were not released
2606 until the entire RPC IO was completed. Instead, release biovecs
2607 as soon as they are complete to ensure forward IO progress.
2609 Severity : enhancement
2611 Description: Support for specifying external journal device at mount
2612 Details : If an OST or MDS device is formatted with an external journal
2613 device, this device major/minor is stored in the ext3 superblock
2614 and may not be valid for failover. Allow detecting and
2615 specifying the external journal at mount time.
2620 Description: Mounting an MDS with pending unlinked files may cause oops
2621 Details : target_finish_recovery() calls mds_postrecov() which returned
2622 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2623 considers this an error and immediately begins cleaning up the
2624 lov, just after starting the mds_lov process
2626 Severity : enhancement
2628 Description: Implement 'lfs df' to report actual free space on per-OST basis
2629 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2630 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2631 usage of objects; '-h' to report in human readable format.
2633 ------------------------------------------------------------------------------
2635 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2642 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2643 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2644 at startup during orphan unlinking. Instead, skip buddy update
2645 before setup, as it will be regenerated after recovery is complete.
2650 Description: If an OST is inactive, its locks might reference stale inodes.
2651 Details : lov_change_cbdata() must iterate over all namespaces, even if
2652 they are inactive to clear inode references from the lock.
2654 Severity : enhancement
2655 Frequency : occasional, if non-standard max_dirty_mb used
2657 Description: Client will block write RPCs if not enough grant
2658 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2659 then the client will block writes while waiting for another RPC
2660 to complete instead of consuming its dirty limit. With change
2661 we get improved performance when max_dirty_mb is small.
2663 Severity : enhancement
2664 Bugzilla : 3389, 6253
2665 Description: Add support for supplementary groups on the MDS.
2666 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2667 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2668 lookups for user supplementary groups into a cache.
2672 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2673 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2674 for liblustre O_EXCL is enforced by MDS.
2678 Description: readdir never returns NULL in liblustre.
2679 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2680 dentry in struct dirent.
2684 Frequency : liblustre only, depends on application IO pattern
2685 Description: liblustre clients evicted if not contacting servers
2686 Details : Don't put liblustre clients into the ping_evictor list, so
2687 they will not be evicted by the pinger ever.
2689 Severity : enhancement
2691 Description: Add ability to evict clients by NID from MDS.
2692 Details : By echoing "nid:$NID" string into
2693 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2694 $NID would be instantly evicted from this MDS and from all active
2695 OSTs connected to it.
2699 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2700 Details : lookup_it_finish() used to query file size from OSTs that was not
2705 Description: service threads change working directory to that of init
2706 Details : Starting lustre service threads may pin the working directory
2707 of the parent thread, making that filesystem busy. Threads
2708 now change to the working directory of init to avoid this.
2712 Frequency : during shutdown only
2713 Description: shutdown with a failed MDS or OST can cause unmount to hang
2714 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2719 Frequency : 2.6 only
2720 Description: chmod/chown may include an extra supplementary group
2721 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2722 supplementary group and if none is specified this is used.
2725 Bugzilla : 5479 (6816)
2727 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2728 Details : If lookup is in progress on a file that is unlinked we might try
2729 to revalidate the inode and fail in revalidate after lookup is
2730 complete and ll_file_open() enqueues the open again but
2731 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2734 Frequency : always, if lconf --abort_recovery used
2736 Description: lconf --abort_recovery fails with 'Operation not supported'
2737 Details : lconf was attempting to abort recovery on the MDT device and not
2740 ------------------------------------------------------------------------------
2742 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2747 Frequency : rare (only unsupported configurations with a node running as an
2749 Bugzilla : 6514, 5137
2750 Description: Mounting a Lustre file system on a node running as an OST could
2752 Details : OSTs now preallocates memory needed to write out data at
2753 startup, instead of when needed, to avoid having to
2754 allocate memory in possibly low memory situations.
2755 Specifically, if the file system is mounted on on OST,
2756 memory pressure could force it to try to write out data,
2757 which it needed to allocate memory to do. Due to the low
2758 memory, it would be unable to do so and the node would
2759 become unresponsive.
2761 Severity : enhancement
2763 Description: Addition of lconf --service command line option
2764 Details : lconf now accepts a '--service <arg>' option, which is
2765 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2767 Severity : enhancement
2769 Description: Failover mode is now the default for OSTs.
2770 Details : By default, OSTs will now run in failover mode. To return to
2771 the old behaviour, add '--failout' to the lmc line for OSTs.
2773 Severity : enhancement
2775 Description: Health checks are now provided for MDS and OSTs
2776 Details : Additional detailed health check information on MSD and OSTs
2777 is now provided through the procfs health_check value.
2780 Frequency : occasional, depends on IO load
2782 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2783 after numerous create/delete cycles
2784 Details : The ext3 inode allocation policy would not allocate new inodes
2785 very well on the OSTs because there are no new directories
2786 being created. Instead we look for groups with free space if
2787 the parent directories are nearly full.
2792 Description: Network or server problems during mount may cause partially
2793 mounted clients instead of returning an error.
2794 Details : The config llog parsing code may overwrite the error return
2795 code during mount error handling, returning success instead
2801 Description: MDS can fail to allocate large reply buffers
2802 Details : After long uptimes the MDS can fail to allocate large reply
2803 buffers (e.g. zconf client mount config records) due to memory
2804 fragmentation or consumption by the buffer cache. Preallocate
2805 some large reply buffers so that these replies can be sent even
2806 under memory pressure.
2810 Frequency : rare (liblustre)
2811 Description: fsx running with liblustre complained that using truncate() to
2812 extend the file doesn't work. This patch corrects that issue.
2813 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2814 ATTR_SIZE and lsm use in llu_setattr_raw.
2818 Frequency : rare, only 2.6 kernels
2819 Description: Unusual file access patterns on the MDS may result in inode
2820 data being lost in very rare circumstances.
2821 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2822 code in the 2.6 kernel caused the inode and/or EA data not to
2823 be read from disk, causing single-file corruption.
2827 Frequency : rare, only 2.6 filesystems using extents
2828 Description: Heavy concurrent write and delete load may cause data corruption.
2829 Details : It was possible under high-load situations to have an extent
2830 metadata block in the block device cache from a just-unlinked
2831 file overwrite a newly-allocated data block. We now unmap any
2832 metadata buffers that alias just-allocated data blocks.
2836 Frequency : filesystems with default stripe_count larger than 77
2837 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2838 Details : lconf specifies an inode size of 4096 bytes when the default
2839 stripe_count is larger than 77. This conflicts with the default
2840 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2841 this case to avoid pinning too much memory for large EAs.
2843 ------------------------------------------------------------------------------
2845 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2850 Frequency : rare (extremely heavy IO load with hundreds of clients)
2852 Description: Client is evicted, gets IO error writing to file
2853 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2854 caused clients to be evicted for AST timeouts. The fixes in
2855 bug 5192 mean we no longer need to have such short AST timeouts
2856 so ldlm_timeout has been increased.
2859 Frequency : occasional during --force or --failover shutdown under load
2860 Bugzilla : 5949, 4834
2861 Description: Server oops/LBUG if stopped with --force or --failover under load
2862 Details : a collection of import/export refcount and cleanup ordering
2863 issues fixed for safer force cleanup
2866 Frequency : only filesystems larger than 120 OSTs
2867 Bugzilla : 5990, 6223
2868 Description: lfs getstripe would oops on a very large filesystem
2869 Details : lov_getconfig used kfree on vmalloc'd memory
2872 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2873 Bugzilla : 6242, 6243
2874 Description: reading from files that had been truncated to a non-zero size
2875 but never opened returned no data
2876 Details : ll_file_read() reads zeros from no-object files to EOF
2881 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2882 Details : There's little harm in aborting MDS/OSS recovery and letting it
2883 try again, so I removed the LASSERT and return an error instead.
2885 Severity : enhancement
2887 Description: New debugging infrastructure for tracking down data corruption
2888 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2889 (b) cover more of the client-side code path, and (c) try to narrow
2890 down where problems occurred
2894 Bugzilla : 3819, 4364, 4397, 6313
2895 Description: Racing close and eviction MDS could cause assertion in mds_close
2896 Details : It was possible to get multiple mfd references during close and
2897 client eviction, leading to one thread referencing a freed mfd.
2899 Severity: : enhancement
2900 Bugzilla : 3262, 6359
2901 Description: Attempts to reconnect to servers are now more aggressive.
2902 Details : This builds on the enhanced upcall-less recovery that was added
2903 in 1.4.2. When trying to reconnect to servers, clients will
2904 now try each server in the failover group every 10 seconds. By
2905 default, clients would previously try one server every 25 seconds.
2910 Description: After recovery, certain operations trigger a failed
2911 assertion on a client.
2912 Details : Failing over an mds, using lconf -d --failover, while a
2913 client was doing a readdir() call would cause the client to
2914 LBUG after recovery completed and the readdir() was resent.
2916 Severity : enhancement
2918 Description: Default groups are now added by lconf
2919 Details : You can now run lconf --group <servicename> without having to
2920 manually add groups with lmc.
2923 Frequency : occasional
2925 Description: Nodes with an elan id of 0 trigger a failed assertion
2928 Frequency : always when accessing e.g. tty/console device nodes
2930 Description: tty and some other devices nodes cannot be used on lustre
2931 Details : file's private_data field is used by device data and lustre
2932 values in there got lost. New field was added to struct file to
2933 store fs-specific private data.
2936 Frequency : when exporting Lustre via NFS
2938 Description: NFSD failed occasionally when looking up a path component
2939 Details : NFSD is looking up ".." which was broken in ext3 directories
2940 that had grown large enough to become hashed.
2943 Frequency : Clusters with multiple interfaces not on the same subnet
2945 Description: Nodes will repeatedly try to reconnect to an interface which it
2946 cannot reach and report an error to the log.
2947 Details : Extra peer list entries will be created by lconf with some peers
2948 unreachable. lconf now validates the peer before adding it.
2951 Frequency : Only if a default stripe is set on the filesystem root.
2953 Description: Setting a default stripe on the filesystem root prevented the
2954 filesystem from being remounted.
2955 Details : The client was sending extra request flags in the root getattr
2956 request and did not allocate a reply buffer for the dir EA.
2959 Frequency : occasional, higher if lots of files are accessed by one client
2960 Bugzilla : 6159, 6097
2961 Description: Client trips assertion regarding lsm mismatch/magic
2962 Details : While revalidating inodes the VFS looks up inodes with ifind()
2963 and in rare cases can find an inode that is being freed.
2964 The ll_test_inode() code will free the lsm during ifind()
2965 when it finds an existing inode and then the VFS later attaches
2966 this free lsm to a new inode.
2970 Bugzilla : 6422, 7030
2971 Description: MDS deadlock between mkdir and client eviction
2972 Details : Creating a new file via mkdir or mknod (starting a transaction
2973 and getting the ns lock) can deadlock with client eviction
2974 (gets ns lock and trying to finish a synchronous transaction).
2977 Frequency : occasional
2978 Description: While starting a server, the fsfilt_ext3 module could not be
2980 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2981 kernels. Previously, lconf would still use the ext3 name
2982 when trying to load modules. Now, it will correctly use
2983 ext3 on 2.4 and ldiskfs on 2.6.
2985 Severity : enhancement
2986 Description: The default stripe count has been changed to 1
2987 Details : The interpretation of the default stripe count (0, to lfs
2988 or lmc) has been changed to mean striping across a single
2989 OST, rather than all available. For general usage we have
2990 found a stripe count of 1 or 2 works best.
2992 Severity : enhancement
2993 Description: Add support for compiling against Cray portals.
2994 Details : Conditional compiling for some areas that are different
2998 Frequency : occasional
2999 Bugzilla : 6409, 6834
3000 Description: Creating files with an explicit stripe count may lead to
3001 a failed assertion on the MDS
3002 Details : If some OSTs are full or unavailable, creating files may
3003 trigger a failed assertion on the MDS. Now, Lustre will
3004 try to use other servers or return an error to the
3008 Frequency : occasional
3010 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3011 caused lock timeout during readahead (since 1.4.2).
3012 Details : Processes doing readahead might match a lock that hasn't been
3013 granted yet if there are overlapping and conflicting lock
3014 requests. The readahead process waits on ungranted lock
3015 (original lock is CBPENDING), while OST waits for that process
3016 to cancel CBPENDING read lock and eventually evicts client.
3018 Severity : enhancement
3020 Description: Initial enabling of flock support for clients
3021 Details : Implements fcntl advisory locking and file status functions.
3022 This feature is provided as an optional mount flag (default
3023 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3024 locking are implemented yet, and those that are are not guaranteed
3025 to be completely correct in production environments.
3026 mount -t lustre -o [flock|noflock] ...
3029 Frequency : occasional
3031 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3032 assertion in the ext3 JBD (journaling) layer.
3033 Details : The b_committed_data struct is protected by the big kernel lock
3034 in 2.4 kernels, serializing journal_commit_transaction() and
3035 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3036 access to this struct. In 2.6 kernels there is finer grained
3037 locking to improve SMP performance of the JBD layer.
3041 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3046 Description: merge and cleanup kernel patches.
3049 -----------------------------------------------------------------------------
3051 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3053 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3054 versions of Lustre. Please update all servers and clients to
3055 version 1.4.2 or later at the same time. You must also run
3056 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3057 to update the configuration logs.
3059 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3060 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3061 - swab llog records of type '0' so we get proper header size/idx (5861)
3062 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3063 - fix rename of one directory over another leaking an inode (5953)
3064 - avoid SetPageDirty on 2.6 (5981)
3065 - don't re-add just-being-destroyed locks to the waiting list (5653)
3066 - when creating new directories, inherit the parent's custom
3067 striping settings if present parent (3048)
3068 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3069 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3070 - don't hold client locks on temporary worklist from l_lru (5666)
3071 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3072 - automatically evict dead clients (5921)
3073 - Update file size properly in create+truncate+fstat case (6196)
3074 - Do not unhash mountpoint dentries, do not allow removal of
3076 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3077 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3078 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3079 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3080 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3081 - Correctly handle reads of files with no objects (6243)
3082 - lctl recover will also mark a device active if deactivate used (5933)
3084 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3085 - allow --write-conf on an MDS with different nettype than client (5619)
3086 - don't write config llogs to MDS for mounts not from that MDS (5617)
3087 - lconf should create multiple TCP connections from a client (5201)
3088 - init scripts are now turned off by default; run chkconfig --on
3089 lustre and chkconfig --on lustrefs to use them
3090 - upcalls are no longer needed for clients to recover to failover
3092 - add --abort-recovery option to lconf to abort recovery on device
3094 - add support for an arbitrary number of OSTs (3026)
3095 - Quota support protocol changes.
3096 - forward compatibility changes to wire structs (6007)
3097 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3098 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3099 - improved "lustre status" script
3100 - initialize blocksize for non-regular files (6062)
3101 - added --disable-server and --disable-client configure options (5782)
3102 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3103 - Vanilla 2.4.29 support
3104 - increase maximum number of obd devices to 520 (6242)
3105 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3106 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3107 SLES 9 2.6.5 kernels
3108 - limit stripes per file to 160 (the maximum EA size) (6093)
3110 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3113 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3114 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3115 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3116 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3117 - small ext3 extents cleanups and fixes (5733)
3118 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3119 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3120 - avoid needless client->OST connect, fix handle mismatch (5317)
3121 - fix DLM error path that led to out-of-sync client, long delays (5779)
3122 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3123 - fix several locking issues related to i_size (5492,5624,5654,5672)
3124 - don't move pending lock onto export if it is already evicted (5683)
3125 - fix kernel oops when creating .foo in unlinked directory (5548)
3126 - fix deadlock in obdfilter statistics vs. object create (5811)
3127 - use time_{before,after} to avoid timer jiffies wrap (5882)
3128 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3129 - Do not leak request if server was not able to process it (5154)
3130 - If mds_open unable to find parent dir, make that negative lookup(5154)
3131 - don't create new directories with extent-mapping (5909, 5936)
3133 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3134 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3135 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3136 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3137 - support for RHEL4 kernel on i686 (5773)
3138 - provide error messages when incompatible logs are encountered (5898)
3140 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3141 * version 1.4.0.10 (1.4.1 release candidate 1)
3143 - don't keep a lock reference when lock is not granted (4238)
3144 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3145 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3146 - drop import inflight refcount on signal_completed_replay error (5255)
3147 - unlock page after async write error during send (3677)
3148 - handle missing objects in filter_preprw_read properly (5265)
3149 - no transno return for symlink open, don't save no-trasno open (3440)
3150 - don't try to complete elan receive that already failed (4012)
3151 - free RPC server reply state on error (5406)
3152 - clean up thread from ptlrpc_start_thread() on error (5160)
3153 - readahead could read extra page into cache that wasn't ejected (5388)
3154 - prevent races in class_attach/setup/cleanup/detach (5260)
3155 - don't dereference de->d_inode after l_dput of de (5458)
3156 - use "int" for stripe value returned from lock_to_stripe (5544)
3157 - mballoc allocation and error-checking fixes in 2.6 (5504)
3158 - block device patches to fix I/O request sizes in 2.6 (5482)
3159 - look up hostnames for IB nals (5602)
3160 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3161 - don't start multiple acceptors for the same port (5277)
3162 - fix incorrect LASSERT in mds_getattr_name (5635)
3163 - export a proc file for general "ping" checking (5628)
3164 - fix "lfs check" to not block when the MDS is down (5628)
3166 - service request history (4965)
3167 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3168 - create an "evict_client" /proc entry on OSTs, like the MDS has
3169 - fix mount usage message, return errors per mount(8) (5168)
3170 - change grep [] to grep "[]" in tests so they work in more UMLs
3171 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3172 - remove ancient LOV_MAGIC_V0 EA support (5047)
3173 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3174 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3175 - allow readahead allocations to fail when low on memory (5383)
3176 - mmap locking landed again, after considerable improvement (2828)
3177 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3179 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3182 - send OST transaction number in read/write reply to free req (4966)
3183 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3184 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3185 - the watchdog thread now runs as interruptible (5246)
3186 - flock/lockf fixes (but it's still disabled, pending 5135)
3187 - don't use EXT3 constants in llite code (5094)
3188 - memory shortage at startup could cause assertion (5176)
3190 - reorganization of lov code
3191 - single portals codebase
3193 - add extents/mballoc support (5025)
3194 - direct I/O reads in the obdfilter (4048)
3195 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3197 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3200 - send OST transaction number in read/write reply to free req (4966)
3201 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3202 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3203 - the watchdog thread now runs as interruptible (5246)
3204 - handle missing objects in filter_preprw_read properly (5265)
3205 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3206 - drop import inflight refcount on signal_completed_replay error (5255)
3207 - unlock page after async write error during send (3677)
3208 - return original error code on reconstructed replies (3761)
3209 - no transno return for symlink open, don't save no-trasno open (3440)
3211 - add pid to ldlm debugging output (4922)
3212 - bump the watchdog timeouts -- we can't handle 30sec yet
3213 - extra debugging for orphan dentry/inode bug (5259)
3215 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3218 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3219 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3220 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3221 - fix NULL dereference in /proc/sys/portals/routes (4827)
3222 - allow failed mdc_close() operations to be interrupted (4561)
3223 - stop precreate on OST before MDS would time out on it (4778)
3224 - don't send partial-page writes before EOF from client (4410)
3225 - discard client grant for sub-page writes on large-page clients (4520)
3226 - don't free dentries not owned by NFS code, check generation (4806)
3227 - fix lsm leak if mds_create_objects() fails (4801)
3228 - limit debug_daemon file size, always print CERROR messages (4789)
3229 - use transno after validating reply (3892)
3230 - process timed out requests if import state changes (3754)
3231 - update mtime on OST during writes, return in glimpse (4829)
3232 - add mkfsoptions to LDAP (4679)
3233 - use ->max_readahead method instead of zapping global ra (5039)
3234 - don't interrupt __l_wait_event() during strace
3236 - add software watchdogs to catch hung threads quickly (4941)
3237 - make lustrefs init script start after nfs is mounted
3238 - fix CWARN/ERROR duplication (4930)
3239 - return async write errors to application if possible (2248)
3240 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3241 - print NAL number in %x format (4645)
3242 - update barely-supported suse-2.4.21-171 series (4842)
3243 - support for sles 9 %post scripts
3244 - support for building 2.6 kernel-source packages
3245 - support for sles km_* packages
3247 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3250 - ignore -ENOENT errors in osc_destroy (3639)
3251 - notify osc create thread that OSC is being cleaned up (4600)
3252 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3253 - reconstruct ost_handle() like mds_handle() (4657)
3254 - create a new thread to do import eviction to avoid deadlock (3969)
3255 - let lconf resolve symlinked-to devices (4629)
3256 - don't unlink "objects" from directory with default EA (4554)
3257 - hold socknal file ref over connect in case target is down (4394)
3258 - allow more than 32000 subdirectories in a single directory (3244)
3259 - fix blocks count for O_DIRECT writes (3751)
3260 - OST returns ENOSPC from object create when no space left (4539)
3261 - don't send truncate RPC if file size isn't changing (4410)
3262 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3263 - bind to privileged port in socknal and tcpnal (3689)
3265 - rate limit CERROR/CWARN console message to avoid overload (4519)
3266 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3267 - basic mmap support (3918)
3268 - kernel patch series update from b1_4 (4711)
3270 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3273 - avoid crash during MDS cleanup with OST shut down (2775)
3274 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3275 - don't use bad inodes on the MDS (3744)
3276 - dynamic object preallocation to improve recovery speed (4236)
3277 - don't hold spinlock over lock dumping or change debug flags (4401)
3278 - don't zero obd_dev when it is force cleaned (3651)
3279 - print grants to console if they go negative (4431)
3280 - "lctl deactivate" will stop automatic recovery attempts (3406)
3281 - look for existing locks in ldlm_handle_enqueue() (3764)
3282 - don't resolve lock handle twice in recovery avoiding race (4401)
3283 - revalidate should check working dir is a directory (4134)
3285 - don't always mark "slow" obdfilter messages as errors (4418)
3287 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3290 - don't close LustreDB during write_conf until it is done (3860)
3291 - fix typo in lconf for_each_profile (3821)
3292 - allow dumping logs from multiple threads at one time (3820)
3293 - don't allow multiple threads in OSC recovery (3812)
3294 - fix debug_size parameters (3864)
3295 - fix mds_postrecov to initialize import for llog ctxt (3121)
3296 - replace config semaphore with spinlock (3306)
3297 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3298 - don't allow enqueue to complete on a destroyed export (3822)
3299 - down write_lock before checking llog header bitmap (3825)
3300 - recover from lock replay timeout (3764)
3301 - up llog sem before sending rpc (3652)
3302 - reduce ns lock hold times when setting kms (3267)
3303 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3304 - fix NULL deref and obd_dev leak on setup error (3312)
3305 - replace some LBUG about llog ops with error handling (3841)
3306 - don't match INVALID dentries from d_lookup and spin (3784)
3307 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3308 - fix invalid assertion in ptlrpc_set_wait (3880)
3310 - add libwrap support for the TCP acceptor (3996)
3311 - add /proc/sys/portals/routes for non-root route listing (3994)
3312 - allow setting MDS UUID in .xml (2580)
3313 - print the stack of a process that LBUGs (4228)
3315 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3318 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3319 - make sure to unset replay flag from failed open requests (3440)
3320 - if default stripe count is 0, use OST count for inode size (3636)
3321 - update parent mtime/ctime on client for create/unlink (2611)
3322 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3323 - free recovery state on server during a forced cleanup (3571)
3324 - unregister_reply for resent reqs (3063)
3325 - loop back devices mounting and status check on 2.6 (3563)
3326 - fix resource-creation race that can provoke i_size == 0 (3513)
3327 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3328 - more debugging for page-accounting assertion (3746)
3329 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3330 - avoid deadlock after precreation failure (3758)
3331 - fix race and lock order deadlock in orphan handling (3450, 3750)
3332 - add validity checks when grabbing inodes from l_ast_data (3599)
3334 - add /proc/.../recovery_status to obdfilter (3428)
3335 - lightweight CDEBUG infrastructure, debug daemon (3668)
3336 - change default OSC RPC parameters to be better on small clusters
3337 - turn off OST read cache for files smaller than 32MB
3338 - install man pages and include them in rpms (3100)
3339 - add new init script for (un)mounting lustre filesystems (2593)
3340 - run chkconfig in %post for init scripts (3701)
3341 - drop scimac NAL (unmaintained)
3343 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3346 - clean kiobufs before and after use (3485)
3347 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3348 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3349 - change init script to fail more clearly if not run as root (1528)
3350 - allow clients to reconnect during replay (1742)
3351 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3352 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3353 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3354 - fix cond_resched (3554)
3355 - only evict unfinished clients after recovery (3515)
3356 - allow bulk resend, prevent data loss (3570)
3357 - dynamic ptlrpc request buffer allocation (2102)
3358 - don't allow unlinking open directory if it isn't empty (2904)
3359 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3360 - remove extraneous obd dereference causing LASSERT failure (3334)
3361 - don't use get_cycles() when creating temp. files on the mds (3156)
3362 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3363 - handle EEXIST for set-stripe, set proper directory name (3336)
3365 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3366 - fix ksocknal_fmb_callback() error messages (2918)
3368 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3371 - don't copy lvb into (possibly NULL) reply on error (2983)
3372 - don't deref dentry after dput, don't free lvb on error (2922)
3373 - use the kms to determine writeback rpc length (2947)
3374 - increment oti_logcookies when osc is inactive (2948)
3375 - update client's i_blocks count via lvb messages (2543)
3376 - handle intent open/close of special files properly (1557)
3377 - mount MDS with errors=remount-ro, like obdfilter (2009)
3378 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3379 - don't use cancelling-locks' kms values (2947)
3380 - use highest lock extent for kms, not last one (2925)
3381 - don't dereference ERR_PTR() dentry in error handling path (3107)
3382 - fix thread race in portals_debug_dumplog() (3122)
3383 - create lprocfs device entries at setup instead of at attach (1519)
3384 - common AST error handler, don't evict client on completion race (3145)
3385 - zero nameidata in detach_mnt in 2.6 (3118)
3386 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3387 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3388 - initialize RPC timeout timer earlier for 2.6 (3219)
3389 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3390 - print nal/nid for unknown nid (3258)
3391 - additional checks for oscc recovery before doing precreate (3284)
3392 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3393 - don't crash in mdc_close for bad permissions on open (3285)
3394 - zero i_rdev for non-device files (3147)
3395 - clear page->private before handing to FS, better assertion (3119)
3396 - tune the read pipeline (3236)
3397 - fix incorrect decref of invalidated dentry (2350)
3398 - provide read-ahead stats and refine rpc in flight stats (3328)
3399 - don't hold journal transaction open across create RPC (3313)
3400 - update atime on MDS at close time (3265)
3401 - close LDAP connection when recovering to avoid server load (3315)
3402 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3403 - don't leak open file on MDS after open resend (3325)
3404 - serialize filter_precreate and filter_destroy_precreated (3329)
3405 - loop device shouldn't call sync_dev() for nul device (3092)
3406 - clear page cache after eviction (2766)
3407 - resynchronize MDS->OST in background (2824)
3408 - refuse to mount the same filesystem twice on same mountpoint (3394)
3409 - allow llmount to create routes for mounting behind routers (3320)
3410 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3411 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3412 - fix rare problem with rename on htree directories (3417)
3414 - allow default OST striping configuration per directory (1414)
3415 - fix compilation for qswnal for 2.6 kernels (3125)
3416 - increase maximum number of MDS request buffers for large systems
3417 - change liblustreapi to be useful for external progs like lfsck (3098)
3418 - increase local configuration timeout for slow disks (3353)
3419 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3421 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3424 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3425 - don't overwrite extent policy data in reply if lock was blocked (2901)
3426 - drop filter export grants atomically with removal from device (2663)
3427 - del obd_self_export from work_list in class_disconnect_exports (2908)
3428 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3429 - swab reply message in mdc_close, other PPC fixes (2464)
3430 - fix destroying of named logs (2325)
3431 - overwrite old logs when running lconf --write_conf (2264)
3432 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3433 - fix race in target_handle_connect (2898)
3434 - mds_reint_create() should take same inode create lock (2926)
3435 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3436 - don't close files for self_export to avoid uninitialized obd (2936)
3437 - allow MDS with the same name as client node (2939)
3438 - hold dentry reference for closed log files for unlink (2325)
3439 - reserve space for all logs during transactions (2059)
3440 - don't evict page beyond end of stripe extent (2925)
3441 - don't oops on a deleted current working directory (2399)
3442 - handle hard links to targets without a parent properly (2517)
3443 - don't dereference NULL lock when racing during eviction (2867)
3444 - don't grow lock extents when lots of conflicting locks (2919)
3446 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3449 - account for cache space usage on clients to avoid data loss (974)
3450 - lfsck support in lustre kernel code (2349)
3451 - reduce journal credits needed for BRW writes (2370)
3452 - orphan handling to avoid losing space on client/server crashes
3453 - ptlrpcd can be blocked, stopping ALL progress (2477)
3454 - use lock value blocks to assist in proper KMS, faster stat (1021)
3455 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3456 - recovery for initial connections (2355)
3457 - fixes for mds_cleanup_orphans (1934)
3458 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3459 - block all file creations until orphan recovery completes (1901)
3460 - client remove rq_connection from request struct (2423)
3461 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3462 - recovery timer race (2670)
3463 - mdc_close recovey bug (2532)
3464 - ptlrpc cleanup bug (2710)
3465 - mds timeout on local locks (2588)
3466 - namespace lock held during RPCs (2431)
3467 - handle interrupted sync write properly (2503)
3468 - don't try to handle a message that hasn't been replied to (2699)
3469 - client assert failure during cleanup after abort recovery (2701)
3470 - leak mdc device after failed mount (2712)
3471 - ptlrpc_check_set allows timedout requests to complete (2714)
3472 - wait for inflight reqs when ptlrpcd finishes (2710)
3473 - make sure unregistered services are removed from the srv_list
3474 - reset bulk XID's when resending them (caught by 1138 test)
3475 - unregister_bulk after timeout
3476 - fix lconf error (2694)
3477 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3478 - readahead locks pages, leaves pending causing memory pressure (2673)
3479 - increase OST request buffers to 4096 on large machines (2729)
3480 - fix up permission of existing directories in simple_mkdir (2661)
3481 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3482 - don't assign transno to errored transactions (2742)
3483 - don't delete objects on OST if given a bogus objid from MDS (2751)
3484 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3485 - if rq_no_resend, then timeout request after recovery (2432)
3486 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3487 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3488 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3489 - MDS recovery shouldn't delete objects that it has given out (2730)
3490 - if enqueue arrives after completion, don't clobber LVB (2819)
3491 - don't unlock pages twice when trigger_group_io returns error (2814)
3492 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3493 - don't write pages to disk if there was an error (1450)
3494 - don't ping imports that have recovery disabled (2676)
3495 - take buffered bytes into account when balancing socknal conn (2817)
3496 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3497 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3498 - remove little-endian swabbing of llog records (1987)
3499 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3500 - retry reposting request buffers if they fail (1191)
3501 - grow extent at grant time to avoid granting a revoked lock (2809)
3502 - lock revoke doesn't evict page if covered by a second lock (2765)
3503 - disable VM readahead to avoid reading outside lock extents (2805)
3505 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3506 - updated kernel patches for hp-2.4.20 kernel (2681)
3508 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3511 - fix truncated write corruption (2366)
3512 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3514 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3515 - ensure len(uuid) < 37 in lmc (1171)
3516 - fix ia64 OOPS in llog_test (2255)
3517 - zero end of page at obdfilter for partial page writes (2648)
3518 - don't leave stale dentries around after renames (bug 2428)
3519 - fix timeouts when evicting a client with a single lock held (2642)
3520 - set deadline for the initial HELLO message to drain (2634)
3521 - print out dotted-quad IP addresses in the socknal (2302)
3523 - additional debugging for MDS client eviction problem (2443)
3524 - fix mkfsoptions support for osts (2603, 2604)
3526 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3529 - add series for the vanilla 2.6.0 kernel
3530 - add series for the vanilla 2.4.24 kernel
3531 - add series for a cray x86/64 UL kernel drop
3532 - fix xattr patches for the vanilla 2.4.19 series
3534 - generate true UUIDs in lmc (1171)
3535 - have portals stack dumping break in UML (2466)
3536 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3537 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3538 - wake callback waiting threads on client eviction (2460)
3539 - Add --ptldebug and --subsystem to lmc (1719)
3540 - update assertion to allow safe interrupt allocation
3541 - set rq_no_resend for cancel requests (2432)
3542 - recalculate ptlrpcd timeout after resend (2494)
3543 - call vfs_rmdir when removing pending directories (2368)
3544 - fix renaming a file to itself (2429)
3545 - lmc creates a default one-stripe lov (2454)
3546 - expand procfs space to handle large clusters (2326)
3547 - increase UML stack to avoid overflow
3548 - update lconf's list of debug and subsystem masks
3549 - fix lfs find --obd (2510)
3550 - /proc tunable for disabling filter read caching (2591)
3551 - stop rpm packages from altering slapd.conf (2301)
3552 - disable nagle in the socknal under 0conf (2578)
3553 - choose mds inode size based on stripe count (2572)
3554 - fix kernel-source rpm problems (2516)
3556 - add --disable-doc to avoid pdf generation (2421)
3557 - update documentation, tests, type-os, comments
3558 - avoid format warnings on ia64
3559 - remove the TOE NAL
3560 - tiny code cleanups by removing unused fields
3562 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3565 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3566 - remove the most blatant lies from BUILDING file (2371)
3567 - change default debug level to reasonable production setting
3568 - reduce client side cache size to reduce cache flush time
3569 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3570 - make TCP zerocopy and pinger support enabled by default (2476)
3571 - sync writes completed after process exits caused crashes (2319)
3572 - maintain correct mount count on the MDS (2356)
3573 - backout 1557, because 2316 wasn't really fixed
3574 - better file I/O statistics gathering in /proc
3575 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3576 - another kernel patch to fix zero-copy TCP function export
3577 - don't take duplicate lock when processing re-sent getattr (2420)
3578 - lctl uses obd_self_export instead of creating new conn (2353)
3579 - MDS/OST recovery case which requires object creation asserted (2425)
3580 - move lfs from /usr/sbin to /usr/bin in packages
3581 - fix race between mds_client_add and mds_client_free (2417)
3582 - use kmalloc instead of slabs in portals (2430)
3583 - don't create duplicate records when a failover MDS is present (2442)
3584 - remove unnecessary mount age check (2332)
3585 - don't remove directory inodes from locks prematurely (2451)
3586 - don't break if MDS service name is the same as hostname (2103)
3587 - fix races in client write RPC generation when cache full (2482)
3589 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3592 - remove now-unused request->rq_obd (278)
3593 - if an allocation fails, print out how much memory we've used (1933)
3594 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3595 - add the "configurable stack size" patch to most series files (1256)
3596 - ability to write large log records, for 100+ OST configs (2306)
3597 - fix NULL deref when filter_prep fails (2314)
3598 - fix operator precedence error in filter_sync
3599 - dynamic allocation of socknal TX descriptors (2315)
3600 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3601 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3602 - ensure that CWARN messages in Portals always get to the syslog
3603 - __init/__exit are not for prototype decls (ldlm_init/exit)
3604 - x86-64 compile warning fixes
3605 - fix gateway LMC keyword conflict (2318)
3606 - fix MDS lock inversions in getattr/reint paths (1844)
3607 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3608 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3609 - DLM race condition prevented some lock evictions (2328)
3610 - ENOMEM detection and retry on socknal sends (2230)
3611 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3612 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3613 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3614 - better error messages when a client is rejected during recovery (1505)
3615 - avoid cancelling locks which were never granted, after failure (2330)
3616 - fix i_sem/journal inversion in mds_client_add (2333)
3617 - fix truncate/getattr lock cycle deadlock (2334)
3618 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3619 - fix two rare exit paths which could leak an l_lock() ref (2321)
3620 - fencepost error in MDS/OST orphan recovery (2226)
3621 - make log record alignment 8 bytes (1988)
3622 - lstripe now fails when requested offset > ost_count (2237)
3623 - ensure that all kernel series have a complete list.h (1607)
3624 - fix crashes in special-file operations (2316)
3625 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3627 - allow configurable automake binary, for testing new versions
3628 - small update to the lfs documentation
3630 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3632 * fix negative export reference count in fsfilt_sync (2312)
3634 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3635 * release candidate 0.9.1
3637 - orphans are moved into the PENDING directory for possible recovery
3638 - replayed opens now open by fid for orphan/rename safety (1042)
3639 - last close of an orphan inode generates a transno (683)
3640 - chdir() and mount() now pin the directory entry (1020)
3641 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3642 - discard very old requests without processing them (1502)
3643 - remove some common, well-understood CERRORs (1505)
3644 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3645 - clear "grant" flags in OST replies until OST grant code lands (1644)
3646 - fix read performance by not clobbering i_blksize on client (1598)
3647 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3648 - make LRU size a /proc tunable, clears locks when reduced (707)
3649 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3650 - server locks take references on exports now (1558)
3651 - build fixes for 2.4.20-rh trees (1663)
3652 - return an error from lov_create if all OSCs are inactive (1751)
3653 - fix import levels when a reconnect happens without a timeout (1597)
3654 - exit early from mds_open if we get a lookup error (1749)
3655 - partial page read at EOF wouldn't wait for disk before sending (1642)
3656 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3657 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3658 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3659 - write/truncate lock inversion (1639)
3660 - Don't auto-load obdclass, portals modules during cleanup (1495)
3661 - fix timestamps from jumping to "now" (1763)
3662 - extra journal assertions (1648)
3663 - add an extra multiunlink test (1771)
3664 - fix read_record/write_record API (1776)
3665 - fix leak of offset_extent, possible incorrect i_size later (1772)
3666 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3667 - Debugging for the kqswnal_get_idle_tx problems (1820)
3668 - Allow recovery to be attempted multiple times (1536)
3669 - Write out MDS last_rcvd file after it is first created (1600)
3670 - Fix tx_descriptor leak in failed transmit situations (1827)
3671 - ext3 journaling fixes for assertion failure after IO error (1871)
3672 - class_export_put() on freed export after completion AST error (1896)
3673 - Fix revalidate looping in VFS (1322)
3674 - Don't access a freed export during MDS_REINT timeout (1521)
3675 - Add open-unlink recovery support on the MDS (1673,1764)
3676 - Return an error if no MDS data was read from last_rcvd (1946)
3677 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3678 - Don't setup a disk device that doesn't match exported UUID (317)
3679 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3680 - avoid committing NULL handle in force close
3681 - local.sh is now a one-stripe LOV configuration
3682 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3683 - fix ext3 htree duplicate directory entry corruption (1516)
3684 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3685 - update to new LOV EA format (2097)
3686 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3688 2003-06-15 Phil Schwan <phil@clusterfs.com>
3691 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3692 - per-import/export recovery handling (958, 931, 959)
3693 - multiple last-rcvd slots, for serving multiple FSes (949)
3694 - connections are again shared between multiple imp/exports (963, 964)
3695 - "umount -f" would hang if any requests needed to be sent (393, 978)
3696 - avoid pinning large req buffer by copying for queued messages (989)
3697 - add "uuid" to "lctl device" command to help upcalls (991)
3698 - "open" RPCs with transnos would confuse recovery counters (1037)
3699 - do proper endian conversion of all wire messages (288, 340, 891)
3700 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3701 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3702 - fix ptlrpc_connection leak in target_handle_connect (1174)
3703 - fix import refcounting bug in OST and MDS cleanup (1134)
3704 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3705 - fix very unlikely obd_types race condition (501)
3706 - remove osc_open hack for echo_client (1187)
3707 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3708 - a failure in read_inode2 leads to deadlock (1139)
3709 - cancel ack-locks as soon as transaction is committed (1072)
3710 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3711 - make sure to commitrw after any preprw to avoid deadlock (1162)
3712 - failing to execute a file in a lustre FS would lock inode (1203)
3713 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3714 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3715 - obd_uuid2tgt would walk off the end of the list (1255)
3716 - on IA64 the osc would give portals incorrect bulk size (1258)
3717 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3718 - fix lock inversion caused by new llite matching code (1282)
3719 - limit the number of dirty pages on a client to 10MB (1286)
3720 - timed out locks were not being corrected cancelled (1289)
3721 - fix O_DIRECT above 4GB on IA-32 (1292)
3722 * major user-visible changes
3723 - fail out/fail over policy now controlled by the upcall (993)
3725 - add OBD_PING to check server availability and failure (954)
3726 - lustre messages are now sent in sending host order (288, 340, 891)
3727 - add eadatalen to MDS getattr reply (340)
3728 - OST read replies may contain second buffer, with per-page status (593)
3730 2003-03-11 Phil Schwan <phil@clusterfs.com>
3733 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3734 - failed open()s could cause deadlock; fixed (867, 869)
3735 - stop cancelling OST locks when files are closed (481)
3736 - overlapping XID spaces caused network corruption (851, 853)
3737 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3738 - setattr_raw added, to do single-RPC, server-side setattrs
3739 - lmc/lconf syntax change for OST UUIDs
3740 - fix crashy race condition between ptlrpc_free_req and osc_close
3741 - don't use request in mdc_enqueue if we hit a timeout (889)
3742 - don't set the inode i_size for regular files from the MDS (896)
3743 - handle out of order completion AST (842)
3744 - don't LBUG if a lock request times out after receiving AST (913)
3745 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3746 - if a bad lock AST arrives, send an error instead of dropping entirely
3747 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3748 - fix leak in bulk IO when only partially completed (899, 900, 926)
3749 - fix O_DIRECT for ia64 (55)
3750 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3751 - C-z after timeout could hang a process forever; fixed (977)
3753 - client-side I/O cache (678, 924, 929, 941, 970)
3755 - READPAGE and SETATTRs which don't take server-side locks get
3758 2003-02-11 Phil Schwan <phil@clusterfs.com>
3761 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3762 - Fix fsfilt_extN_readpage() to read a full page of directory
3763 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3764 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3765 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3766 - Most intent operations were removed, in favour of a new RPC mode
3767 that does a single RPC to the server and bypasses most of the VFS
3768 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3769 - Aggressively cancel local locks on DLM servers
3770 - mds_reint_unlink sends EA to the client if it's the last nlink.
3771 client uses that EA to unlink OST objects.
3772 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3773 - recursive symlinks were fixed (439)
3774 - fixed NULL deref in DEBUG_REQ
3775 - filter_update_lastobjid no longer calls sync, which annoyed extN
3776 - fixed multi-client small-writes to a single file problem (445)
3777 - fixed mtime updates during file writes (607)
3778 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3779 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3780 - fixed Linux OST size reporting problem (444, 656)
3781 - OST now updates object mtime with writes or setattr (607, 619)
3782 - client verifies file size before zeroing page past EOF (445)
3783 - OST now writes last allocated objid to disk with allocation (108)
3784 - LOV on echo now works (409)
3786 - mds_reint_unlink sends a new buffer, with the EA included. this
3787 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3788 set if a regular file was being unlinked, and it was the last link
3789 - use PtlGet from the target for bulk writes (315)
3790 - OST now updates object mtime with writes or setattr (607, 619)
3791 - LDLM now has a grant-time callback to revalidate locked items, if
3793 - Many MDS operations were reorganized to combat race conditions
3795 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3796 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3798 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3801 - Fully reactivate OST imports after reconnection (512, others)
3802 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3803 - More graceful error handling for truncating on dead OST (515)
3804 - Don't error out unless we're actually accessing dead stripes (474)
3805 - Fix garbage sizes when stripes are missing (410)
3806 - LRU counters were broken, causing constant lock purge (433, 432)
3807 - garbage on read from stripes with failed OSTs (441)
3808 - mark OSCs as active before reconnecting during recovery (438)
3809 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3810 - lfind did not preserve OST order in output (443)
3811 - symlinks cause hung clients, incorrect data (439)
3812 - stop dereferencing request after dropping refcount (457)
3813 - don't LASSERT(spin_is_locked) on non-SMP (455)
3814 - fixes for many rename() bugs
3815 - fstat didn't correctly synchronize attributes (399)
3816 - server must handle lock cancellation during blocking AST prep (487)
3817 - bulk descriptors were free()d too soon (511)
3818 - fix paths in lconf, which would load incorrect modules (451, 507)
3819 - fix confusing lconf 'host not found' error message (386)
3820 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3821 - fix race condition in mdc_blocking_ast() for inode access (526)
3822 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3823 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3824 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3825 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3826 - fix lstripe in several ways (kernel side) (527)
3827 - fix request leak in ldlm_cli_enqueue (262)
3828 - incorrect OSC was marked inactive after OST failure
3829 - call mds_fs_cleanup before unmounting filesystem (524)
3830 - fix races between taking ns_lock and ldlm_lock_change_resource
3831 - fix races updating LOV export open file list
3832 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3833 - fix recovery NULL deref in ldlm_cli_cancel_unused
3834 - fix some DLM races by using new hash table for lock handles (419)
3835 - permit the client to specify desired inodes, at replay
3836 - duplicate requests when we queue them for replay reintegration
3837 - fix last_rcvd offset calculation
3838 - sync after each recovered transaction, so we always make progress
3839 - never, not always, ERESTART requests without transnos
3840 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3842 - skip replay if the MDS says that the client is already connected
3843 - don't check for a recovery-enabled export to match lctl's UUID
3844 - don't INC_USE_COUNT for phantom exports
3845 - don't crash when cleaning up phantom exports (567)
3846 - don't double-finish or set replay data for errored mdc_open requests
3847 - abort requests when they time out, so we don't get old replies
3848 - send/receive replies for AST messages again
3849 - if the client says that it doesn't have the lock, cancel it on the
3851 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3852 mark it as destroyed, it will all work out when decref is called
3853 - fix module use counts (22, 581)
3855 - ASTs now expect a reply (server cancels lock on error reply)
3857 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3860 - fix many simultaneous client startup (392)
3861 - fix dentry->d_it clobbering
3862 - credentials weren't being shipped for readdir/getattr operations
3863 - remove invalid assertions triggered during some concurrent MD
3865 - proper Lustre versions added (336, 389)
3866 - fix memory leak for create error case (398)
3867 - fix LOV locking bug that would get cli/srv out of sync
3868 - fix echo client over LOV (409)
3869 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3870 - fix double-O_EXCL intent crash (424)
3871 - avoid sending multiple lock CANCELs (352)
3873 - MDS can do multi-client recovery (modulo bugs in new code)
3875 - many updates, edits, cleanups
3877 2002-11-18 Phil Schwan <phil@clusterfs.com>
3880 - fix null d_it dereference (346)
3881 - fix full OST/dbench hang (333)
3882 - fix permission problem with file removal (286)
3883 - fix removal of OSCs from LOV when they fail
3884 - fix NULL deref during bulk timeout (214)
3885 - fix problems related to multiple filesystems on one MDS (241)
3886 - fixed serious subtle metadata locking bugs
3887 - free locks on clients when inodes are removed due to memory
3889 - fix inode pointer in lock data (285)
3890 - partial support for multiple MDS on a single host (241)
3891 - data locks weren't cancelled at clear_inode time (290, 311)
3892 - intent locks could lead to unbounded lock growth (205)
3893 - added a maximum lock count, an LRU list, and a flusher
3894 - fix multiple rename (365)
3895 - properly abstracted the echo client
3896 - OSC locked 1 byte too many; fixed
3897 - rewrote brw callback code:
3898 - fixed recovery bugs related to LOVs (306)
3899 - fixed too-many-pages-in-one-write crash (191)
3900 - fixed (again) crash in sync_io_timeout (214)
3901 - probably fixed callback-related race (385)
3903 - Add capability to MDS protocol
3904 - LDLM cancellations and callbacks on different portals
3906 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3909 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3910 - defer open object destruction to close time (601981, 138)
3911 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3912 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3913 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3914 - don't call obd_setattr in ll_file_release for destroyed objects
3915 * protocol change to lustre_msg: move |version| and add |flags|
3916 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3917 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3918 * added replay of create, unlink, link and rename operations during
3919 MDS failover; recovery should be much more robust now
3920 * remove failed OSCs from LOVs (only lov_create uses this so far)
3921 * the lustre-HOWTO was brought (more) up to date (582544)
3923 2002-10-23 Phil Schwan <phil@clusterfs.com>
3926 - in-use dentries weren't being reused properly (617851)
3927 - prevent multiple LDLM setup (599178)
3928 - fix LOV size calculations for truncate (617853)
3929 - fix client handling of MDS intent errors (POSIX)
3930 - fix permission bug in lovstripe.c test (624321)
3931 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3932 - truncate past end of file could corrupt data
3933 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3934 - a race in recovery could return ETIMEDOUT to apps (623947)
3935 - building outside the source directory was fixed
3936 * the lustre-HOWTO was brought (more) up to date (582544)
3937 * major progress was made on recovery functionality
3939 2002-10-10 Phil Schwan <phil@clusterfs.com>
3942 - recovery deadlock fix
3943 - rm -rf causes LBUG fix (617817)
3944 - file open by multiple tasks fix (618962)
3945 - directory permissions bugs (602707 and 620007)
3946 - journal_stop fixed with locking (611313)
3947 - O_APPEND failures resolved (618273, perhaps 614459)
3948 - lconf PATH fix (619770)
3949 - IA64 build fix (621450)
3950 - RPC buffer sizes scale with amount of memory
3952 2002-10-01 Phil Schwan <phil@clusterfs.com>
3955 - locks would be cancelled without throwing away data pages,
3956 resulting in inconsistent data (605627)
3957 - inode attributes were not always being refreshed (605627, 612449)
3958 - lconf now continues to cleanup after lctl reports an error
3959 - MDS now enforces user permissions (602707)
3960 - lprocfs cleanup fixed, but not yet enabled (614157)
3961 - fixed infinite server hang, should a client not respond to an AST
3962 - avoid going into recovery if user calls readlink() with a buffer
3963 that's too small (613941)
3964 - AST RPCs no longer require replies (614867) -- this may be changed
3965 - don't crash server if client sends an IOV that's too big (611336)
3966 - fixed lock conversion deadlock (611892)
3967 - fixed the following of symlinks (614622)
3968 * recovery: the server can remove locks from a client that dies, other
3969 clients can make progress
3970 * more extN patch fixes
3971 * compile-time configurable ptlrpc buffer allocations
3973 - collaborative read cache document
3974 - Lustre Lite Performance CDR document-in-progress
3976 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3979 - fix typo in patch-2.4.18
3981 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3984 - clear ptlrpc request each time in handle_incoming_request()
3985 - unlink of files now destroys the object on the OST
3987 2002-09-19 Peter Braam <braam@clusterfs.com>
3989 * add hard link support
3990 * change obdfile creation method
3991 * kernel patch changed
3993 2002-09-19 Peter Braam <braam@clusterfs.com>
3996 - stack overflow bug in extN fixed
3998 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4000 * documentation updates
4001 - add man pages for config tools
4002 - update tests/README to describe testing with new config tools
4003 - finish metadata API descriptions
4004 * bug fixes and cleanups
4005 - statfs workaround for 16TB limit
4006 - LOV stripe allocation improved, can stripe on subset of OSTs
4007 - LOV file size/IO offset was wrong for files > 4GB in size
4008 - object EA data was being dropped, caused files to be unreadable
4009 - memory overflow with non-LOV OST caused memory corruption
4010 - fixed regression tests to work with new config tools, obdfilter
4011 - fixed bug when directory size became larger than 1 block
4012 - fixed bug (for single client case) when PWD was deleted
4013 - invalidate local directory pages when doing intent-based ops
4014 - avoid LDLM oops when lock callback contained bad data
4016 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4018 * documentation updates
4019 * bug fixes and cleanups
4020 - configuration tools
4023 - 64-bit compile warnings
4024 - 64-bit internal statfs data
4026 * test_brw on persistent OST devices
4028 * lprocfs (disabled)
4030 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4032 * documentation updates
4033 * bug fixes and cleanups
4034 * configuration tools
4036 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4039 * many small fixes to 0_5_4
4040 * io/network handling
4041 * thinkos in MDS operations
4043 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4046 * crucial basic fixes to 0.5.3
4047 * IOR, Iozone work over Elan
4050 2002-08-07 Phil Schwan <phil@clusterfs.com>
4051 * version 0_5_3, our first alpha
4052 * we use the new Portals iovs
4053 * documentation updates
4054 * bug fixes and cleanups
4055 * small changes in the DLM wire protocol
4057 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4058 * version 0_5_1 with some initial stability,
4059 * locking on MD and file I/O.
4060 * documentation updates
4061 * several bug fixes since 0.5.0
4062 * small changes in wire protocol
4064 2002-07-18 Phil Schwan <phil@clusterfs.com>
4066 * delivered as Lustre Light Alpha
4067 * fixed a crash after handling invalid MDS requests
4068 * fixed directory pages for architectures with non-4k pages sizes
4070 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4071 * release version v0_4_4
4072 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4074 * Fixes a number of interruption problems with OST operations.
4075 * Update documentation for portals header changes
4076 * Move all wire protocol structs/defines to lustre_idl.h
4077 * Fixes symlink length bug.
4078 * Add tcpdump to repository.
4080 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4081 * release version v0_4_3
4082 * Fixes statfs for inodes on extN.
4083 * Fixes bug in runtests which would delete /etc/hosts.
4084 * Use 64-bit object IDs wherever possible (not into VFS though)
4085 Remove ost_get_info, which is unused by lustre, and out of date.
4087 2002-07-03 Peter Braam <braam@clusterfs.com>
4088 * release version v0_4_2 Fixes a lookup error (type not passed)
4089 * move forward to head of Portals
4090 * move forward to latest Lustre kernel
4092 2002-06-25 Peter Braam <braam@clusterfs.com>
4093 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob