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.
339 --------------------------------------------------------------------------------
341 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
343 * Support for kernels:
344 2.6.5-7.283 (SLES 9),
345 2.6.9-55.EL (RHEL 4),
346 2.6.16.46-0.14 (SLES 10),
347 2.6.18.8 vanilla (kernel.org)
348 * Client support for unpatched kernels:
349 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
350 2.6.16 - 2.6.22 vanilla (kernel.org)
351 * Due to recently discovered recovery problems, we do not recommend
352 using patchless RHEL 4 clients with this or any earlier release.
353 * Recommended e2fsprogs version: 1.39.cfs8
354 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
355 * Starting with this release, the ldiskfs backing filesystem required
356 by Lustre is now in its own package, lustre-ldiskfs. This package
357 should be installed. It is versioned separately from Lustre and
358 may be released separately in future.
362 Description: block reactivating mgc import until all deactivates complete
363 Details : Fix race when failing back MDT/MGS to itself (testing)
365 Severity : enhancement
367 Description: add optional extra BUILD_VERSION info
368 Details : add a new environment variable (namely LUSTRE_VERS) which allows
369 to override the lustre version.
372 Frequency : 2.6.18 servers only
374 Description: ll_kern_mount() doesn't release the module reference
375 Details : The ldiskfs module reference count never drops down to 0
376 because ll_kern_mount() doesn't release the module reference.
381 Description: server LBUG when using old ost_num_threads parameter
382 Details : Accept the old ost_num_threads parameter but warn that it
383 is deprecated, and fix an off-by-one error that caused an LBUG.
388 Description: Transient SCSI error results in persistent IO issue
389 Details : iobuf->dr_error is not reinitialized to 0 between two
393 Frequency : sometimes when underlying device returns I/O errors
395 Description: OSTs not going read-only during write failures
396 Details : OSTs are not remounted read-only when the journal commit threads
397 get I/O errors because fsfilt_ext3 calls journal_start/stop()
398 instead of the ext3 wrappers.
402 Description: poor connect scaling with increasing client count
403 Details : Don't run filter_grant_sanity_check for more than 100 exports
404 to improve scaling for large numbers of clients.
407 Frequency : SLES10 only
409 Description: sanity-quota.sh quotacheck failed: rc = -22
410 Details : Quotas cannot be enabled on SLES10.
413 Frequency : liblustre clients only
415 Description: getdirentries does not give error when run on compute nodes
416 Details : getdirentries does not fail when the size specified as an argument
417 is too small to contain at least one entry
419 Severity : enhancement
421 Description: Add LNET router traceability for debug purposes
422 Details : If a checksum failure occurs with a router as part of the
423 IO path, the NID of the last router that forwarded the bulk data
424 is printed so it can be identified.
429 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
430 Details : Due to a race condition, liblustre clients were occasionally
433 Severity : enhancement
435 Description: lfs setstripe use optional parameters instead of postional
438 Severity : enhancement
440 Description: Nanosecond timestamp support for ldiskfs
441 Details : The on-disk ldiskfs filesystem has added support for nanosecond
442 resolution timestamps. There is not yet support for this at
443 the Lustre filesystem level.
446 Frequency : during server recovery
448 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
449 Details : request with rq_no_resend flag not awake l_wait_event if they get a
453 Frequency : nfs export on patchless client
455 Description: connectathon hang when test nfs export over patchless client
456 Details : Disconnected dentry cannot be found with lookup, so we do not need
457 to unhash it or make it invalid
460 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
461 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
462 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
467 Description: open req refcounting wrong on reconnect
468 Details : If reconnect happened between getting open reply from server and
469 call to mdc_set_replay_data in ll_file_open, we will schedule
470 replay for unreferenced request that we are about to free.
471 Subsequent close will crash in variety of ways.
472 Check that request is still eligible for replay in
473 mdc_set_replay_data().
478 Description: disable writes to filesystem when reading health_check file
479 Details : the default for reading the health_check proc file has changed
480 to NOT do a journal transaction and write to disk, because this
481 can cause reads of the /proc file to hang and block HA state
482 checking on a healthy but otherwise heavily loaded system. It
483 is possible to return to the previous behaviour during configure
484 with --enable-health-write.
486 Severity : enhancement
488 Description: 64-bit inode version
489 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
490 to the inode. This will be required for version-based recovery.
495 Description: MDS fails to start if a duplicate client export is detected
496 Details : in some rare cases it was possible for a client to connect to
497 an MDS multiple times. Upon recovery the MDS would detect this
498 and fail during startup. Handle this more gracefully.
500 Severity : enhancement
502 Description: Add -o localflock option to simulate old noflock
504 Details : This will achieve local-only flock/fcntl locks
510 Description: log_commit_thread vs filter_destroy race leads to crash
511 Details : Take import reference before releasing llog record semaphore
516 Description: Wrong request locking in request set processing
517 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
518 imp_delayed_list, in this place should be used imp_lock.
521 Frequency : when reconnection
523 Description: Grant Leak when osc reconnect to OST
524 Details : When osc reconnect ost, OST(filter) should check whether it
525 should grant more space to client by comparing fed_grant and
526 cl_avail_grant, and return the granted space to client instead
527 of "new granted" space, because client will call osc_init_grant
528 to update the client grant space info.
531 Frequency : when client reconnect to OST
533 Description: Grant Leak when osc do resend and replay bulk write
534 Details : When osc reconnect to OST, OST(filter)should clear grant info of
535 bulk write request, because the grant info while be sync between
536 OSC and OST when reconnect, and we should ignore the grant info
537 these of resend/replay write req.
542 Description: Grant space more than avaiable left space sometimes.
543 Details : When then OST is about to be full, if two bulk writing from
544 different clients came to OST. Accord the avaliable space of the
545 OST, the first req should be permitted, and the second one
546 should be denied by ENOSPC. But if the seconde arrived before
547 the first one is commited. The OST might wrongly permit second
548 writing, which will cause grant space > avaiable space.
551 Frequency : when client is evicted
553 Description: Grant might be wrongly erased when osc is evicted by OST
554 Details : when the import is evicted by server, it will fork another
555 thread ptlrpc_invalidate_import_thread to invalidate the
556 import, where the grant will be set to 0. While the original
557 thread will update the grant it got when connecting. So if
558 the former happened latter, the grant will be wrongly errased
559 because of this race.
564 Description: Checking Stale with correct fid
565 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
566 to check whether it is stale, because sometimes, we want the
567 enqueue happened anyway, and op_data.fid2 will not be initialized.
569 Severity : enhancement
571 Description: update patchless client
572 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
575 Frequency : only with 2.4 kernel
577 Description: random memory corruption
578 Details : size of struct ll_inode_info is to big for union inode.u and this
579 can be cause of random memory corruption.
584 Description: Memory leak in recovery
585 Details : Lov_mds_md was not free in an error handler in mds_create_object.
586 It should also check obd_fail before fsfilt_start, otherwise if
587 fsfilt_start return -EROFS,(failover mds during mds recovery).
588 then the req will return with repmsg->transno = 0 and rc = EROFS.
589 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
590 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
591 be freed no matter whether fsfilt_commit success or not.
594 Frequency : only with huge count clients
596 Description: Prevents from taking the superblock lock in llap_from_page for
598 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
599 for llap_from_page call in ll_removepage prevents from taking the
600 superblock lock for a soon died page.
605 Description: Not check open intent error before release open handle
606 Details : in some rare cases, the open intent error is not checked before
607 release open handle, which may cause
608 ASSERTION(open_req->rq_transno != 0), because it tries to release
609 the failed open handle.
614 Description: Set cat log bitmap only after create log success.
615 Details : in some rare cases, the cat log bitmap is set too early. and it
616 should be set only after create log success.
620 Description: Accessing a block bevice can re-enable I/O when Lustre is
621 tearing down a device.
622 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
627 Description: service threads may hog cpus when there are a lot of requests
629 Details : Insert cond_resched to give other threads a chance to use some of
635 Description: the cat log was not initialized in recovery
636 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
637 unlink log on mds will not be initialized until mds post
638 recovery. And also in mds post recovery, the unlink log will
639 initialization will be done asynchronausly, so there will be race
640 between add unlink log and unlink log initialization.
644 Description: brw_stats were being printed incorrectly
645 Details : brw_stats were being printed as log2 but all of them were not
646 recorded as log2. Also remove some code duplication arising from
647 filter_tally_{read,write}.
651 Frequency : rare, only in recovery.
652 Description: ASSERTION(req->rq_type != LI_POISON) failed
653 Details : imp_lock should be held while iterating over imp_sending_list for
654 prevent destroy request after get timeout in ptlrpc_queue_wait.
658 Description: replay-single.sh test 52 fails
659 Details : A lock's skiplist need to be cleanup when it being unlinked
660 from its resource list.
664 Description: Short directio read returns full requested size rather than
666 Details : Direct I/O operations should return actual amount of bytes
667 transferred rather than requested size.
669 Severity : enhancement
671 Description: metadata RPC reduction (e.g. for rm performance)
672 Details : decrease the amount of synchronous RPC between clients and servers
673 by canceling conflicing lock before the operation on the client side
674 and packing thier handles into the main operation RPC to server.
676 Severity : enhancement
678 Description: Async OSC create to avoid the blocking unnecessarily.
679 Details : If a OST has no remain object, system will block on the creating
680 when need to create a new object on this OST. Now, ways use
681 pre-created objects when available, instead of blocking on an
682 empty osc while others are not empty. If we must block, we block
683 for the shortest possible period of time.
685 --------------------------------------------------------------------------------
687 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
692 Frequency : on some architectures
694 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
695 Details : Uninitialized flags sometimes cause configuration commands to
699 Frequency : patchless clients only
701 Description: missing __iget() symbol export
702 Details : The __iget() symbol export is missing. To avoid the need for
703 this on patchless clients the deathrow inode reaper is turned
704 off, and we depend on the VM to clean up old inodes. This
705 dependency was during via the fix for bug 12181.
707 --------------------------------------------------------------------------------
709 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
711 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
712 INTEROPERATE with older versions automatically. In many cases a
713 special upgrade step is needed. Please read the
714 user documentation before upgrading any part of a live system.
715 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
716 version will not interoperate with 1.6 betas before beta5 (1.5.95).
717 * WARNING: Lustre configuration and startup changes are required with
718 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
724 Severity : enhancement
726 Description: MountConf
727 Details : Lustre configuration is now managed via mkfs and mount
728 commands instead of lmc and lconf. New obd types (MGS, MGC)
729 are added for dynamic configuration management. See
730 https://mail.clusterfs.com/wikis/lustre/MountConf for
733 Severity : enhancement
735 Description: dynamic OST addition
736 Details : OSTs can now be added to a live filesystem
738 Severity : enhancement
740 Description: startup order invariance
741 Details : MDTs and OSTs can be started in any order. Clients only
742 require the MDT to complete startup.
744 Severity : enhancement
746 Description: parallel, asynchronous orphan cleanup
747 Details : orphan cleanup is now performed in separate threads for each
748 OST, allowing parallel non-blocking operation.
750 Severity : enhancement
752 Description: optimized stripe assignment
753 Details : stripe assignments are now made based on ost space available,
754 ost previous usage, and OSS previous usage, in order to try
755 to optimize storage space and networking resources.
757 Severity : enhancement
759 Description: Permanently set tunables
760 Details : All writable /proc/fs/lustre tunables can now be permanently
761 set on a per-server basis, at mkfs time or on a live
764 Severity : enhancement
766 Description: Lustre message v2
767 Details : Add lustre message format v2.
769 Severity : enhancement
771 Description: client OST exclusion list
772 Details : Clients can be started with a list of OSTs that should be
773 declared "inactive" for known non-responsive OSTs.
777 Description: SPEC SFS validation failure on NFS v2 over lustre.
778 Details : Changes the blocksize for regular files to be 2x RPC size,
779 and not depend on stripe size.
781 Severity : enhancement
783 Description: Multiple MD RPCs in flight.
784 Details : Further unserialise some read-only MDS RPCs - learn about intents.
785 To avoid overly-overloading MDS, introduce a limit on number of
786 MDS RPCs in flight for a single client and add /proc controls
787 to adjust this limit.
789 Severity : enhancement
791 Description: client read/write statistics
792 Details : Add client read/write call usage stats for performance
793 analysis of user processes.
794 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
795 file access. extents_stats shows chunk size distribution.
796 extents_stats_per_process show chunk size distribution per
799 Severity : enhancement
801 Description: mds statistics
802 Details : Add detailed mds operations statistics in
803 /proc/fs/lustre/mds/*/stats.
807 Description: Failure of copying files with lustre special EAs.
808 Details : Client side always return success for setxattr call for lustre
809 special xattr (currently only "trusted.lov").
814 Description: Refcount LNET uuids
815 Details : The global LNET uuid list grew linearly with every startup;
816 refcount repeated list entries instead of always adding to
819 Severity : enhancement
821 Description: Dynamic service threads
822 Details : Within a small range, start extra service threads
823 automatically when the request queue builds up.
826 Frequency : mixed-endian client/server environments
828 Description: mixed-endian crashes
829 Details : The new msg_v2 system had some failures in mixed-endian
832 Severity : enhancement
834 Description: Easy OST removal
835 Details : OSTs can be permanently deactivated with e.g. 'lctl
836 conf_param lustre-OST0001.osc.active=0'
838 Severity : enhancement
840 Description: MGS proc entries
841 Details : Added basic proc entries for the MGS showing what filesystems
844 Severity : enhancement
846 Description: provide MGS failover
847 Details : Added config lock reacquisition after MGS server failover.
849 Severity : enhancement
851 Description: add Linux 2.4 support
852 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
856 Description: a large application tries to do I/O to the same resource and dies
858 Details : Check the req->rq_arrival time after the call to
859 ost_brw_lock_get(), but before we do anything about
860 processing it & sending the BULK transfer request. This
861 should help move old stale pending locks off the queue as
862 quickly as obd_timeout.
865 Frequency : when an incorrect nid is specified during startup
867 Description: ptlrpc connect to non-existant node causes kernel crash
868 Details : LNET can't be re-entered from an event callback, which
869 happened when we expire a message after the export has been
870 cleaned up. Instead, hand the zombie cleanup off to another
873 Severity : enhancement
875 Description: plain/inodebits lock performance improvement
876 Details : Grouping plain/inodebits in granted list by their request modes
877 and bits policy, thus improving the performance of search through
881 Frequency : only if OST filesystem is corrupted
883 Description: client incorrectly hits assertion in ptlrpc_replay_req()
884 Details : for a short time RPCs with bulk IO are in the replay list,
885 but replay of bulk IOs is unimplemented. If the OST filesystem
886 is corrupted due to disk cache incoherency and then replay is
887 started it is possible to trip an assertion. Avoid putting
888 committed RPCs into the replay list at all to avoid this issue.
891 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
893 Description: System hang on startup
894 Details : This bug allowed the liblustre (e.g. catamount) client to
895 return to the app before handling all startup RPCs. This
896 could leave the node unresponsive to lustre network traffic
897 and manifested as a server ptllnd timeout.
899 Severity : enhancement
901 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
902 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
903 boolean to control whether to print peer debug info when a
904 client's RPC times out.
907 Frequency : only for kernels with patches from Lustre below 1.4.3
909 Description: Remove old rdonly API
910 Details : Remove old rdonly API which unused from at least lustre 1.4.3
913 Frequency : only for devices with external journals
915 Description: Set external device read-only also
916 Details : During a commanded failover stop, we set the disk device
917 read-only while the server shuts down. We now also set any
918 external journal device read-only at the same time.
921 Frequency : when upgrading from 1.4 while trying to change parameters
923 Description: The wrong (new) MDC name was used when setting parameters for
924 upgraded MDT's. Also allows changing of OSC (and MDC)
925 parameters if --writeconf is specified at tunefs upgrade time.
928 Frequency : when setting specific ost indicies
930 Description: QOS code breaks on skipped indicies
931 Details : Add checks for missing OST indicies in the QOS code, so OSTs
932 created with --index need not be sequential.
934 Severity : enhancement
936 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
937 Details : The uninit_groups feature works in conjunction with the kernel
938 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
939 up the pass1 processing of e2fsck. This is a read-only feature
940 in ldiskfs2 only, so older kernels and current ldiskfs cannot
941 mount filesystems that have had this feature enabled.
943 Severity : enhancement
945 Description: Improve multi-block allocation algorithm to avoid fragmentation
946 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
947 allocation locality and avoid filesystem fragmentation.
949 ------------------------------------------------------------------------------
951 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
953 * Support for kernels:
954 2.4.21-47.0.1.EL (RHEL 3)
956 2.6.9-42.0.10.EL (RHEL 4)
957 2.6.12.6 vanilla (kernel.org)
958 2.6.16.27-0.9 (SLES 10)
959 * Recommended e2fsprogs version: 1.39.cfs5
961 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
965 Frequency : occasional, depends on client load and configuration
966 Bugzilla : 12181, 12203
967 Description: data loss for recently-modified files
969 Details : In some cases it is possible that recently written or created
970 files may not be written to disk in a timely manner (this should
971 normally be within 30s unless client IO load is very high).
972 The problem appears as zero-length files or files that are a
973 multiple of 1MB in size after a client crash or client eviction
974 that are missing data at the end of the file.
976 This problem is more likely to be hit on clients where files are
977 repeatedly created and unlinked in the same directory, clients
978 have a large amount of RAM, have many CPUs, the filesystem has
979 many OSTs, the clients are rebooted frequently, and/or the files
980 are not accessed by other nodes after being written.
982 The presence of the problem can be detected by looking at
983 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
984 smaller than the second (nr_unused) then dirty files will not
985 be flushed automatically to disk. "sync; sleep 10" should be
986 run several times on the node before unmounting it to update
987 Lustre (this is also safe to run on nodes without this problem).
989 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
990 that can cause this same problem, so customers using that kernel
991 also need to update the kernel in addition to Lustre. In order
992 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
994 It is normal that files written just before a client crash (less
995 than 30s) may not yet have been flushed to disk, even for local
999 Frequency : frequent on thin XT3 nodes
1001 Description: UUID collision on thin XT3 Linux nodes
1002 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1003 randomly, since we relied on an insufficiently-seeded PRNG.
1008 Description: OSS hangs after "All ost request buffers busy"
1009 Details : A deadlock between quota and journal operations caused OSS
1010 hangs after printing "All ost request buffers busy."
1013 Frequency : always on liblustre builds
1015 Description: Cleanup compiler warnings on liblustre
1018 Frequency : always on liblustre builds on XT3
1020 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1024 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1025 > 32000 subdirectories
1026 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1027 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1028 subdirectory count crosses 32000. This will aid e2fsck to
1029 correctly handle more than 32000 subdirectories.
1032 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1034 Description: System hang on startup
1035 Details : This bug allowed the liblustre (e.g. catamount) client to
1036 return to the app before handling all startup RPCs. This
1037 could leave the node unresponsive to lustre network traffic
1038 and manifested as a server ptllnd timeout.
1040 Severity : enhancement
1042 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1043 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1044 boolean to control whether to print peer debug info when a
1045 client's RPC times out.
1050 Description: make O_SYNC working on 2.6 kernels
1051 Details : 2.6 kernels use different method for mark pages for write,
1052 so need add a code to lustre for O_SYNC work.
1057 Description: Failure to close file and release space on NFS
1058 Details : Put inode details into lock acquired in ll_intent_file_open.
1059 Use mdc_intent_lock in ll_intent_open to properly
1060 detect all kind of errors unhandled by mdc_enqueue.
1065 Description: proc file read during shutdown sometimes raced obd removal,
1067 Details : Add lock to prevent obd access after proc file removal.
1070 Frequency : Only for files larger than 4GB on 32-bit clients.
1072 Description: improperly doing page alignment of locks
1073 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1074 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1079 Description: RPCs being resent when they shouldn't be
1080 Details : Some RPCs that should not be resent are being resent. This
1081 can cause inconsistencies in the RPC state machine. Do not
1082 resend such requests.
1085 Frequency : rare, only with NFS export
1087 Description: Crash on NFS re-export node
1088 Details : under very unusual load conditions an assertion is hit in
1089 ll_intent_file_open()
1092 Frequency : only if OST filesystem is corrupted
1094 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1095 Details : for a short time RPCs with bulk IO are in the replay list,
1096 but replay of bulk IOs is unimplemented. If the OST filesystem
1097 is corrupted due to disk cache incoherency and then replay is
1098 started it is possible to trip an assertion. Avoid putting
1099 committed RPCs into the replay list at all to avoid this issue.
1104 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1105 Details : Large single O_DIRECT read and write calls can fail to allocate
1106 a sufficiently large buffer to process the request. In case of
1107 allocation failure the allocation is retried with a smaller
1108 buffer and broken into smaller requests.
1110 Severity : enhancement
1112 Description: Add -o localflock option to simulate old noflock behaviour.
1113 Details : This will achieve local-only flock/fcntl locks coherentness.
1118 Description: versioning check is incomplete
1119 Details : Checking the version difference of client vs. server, report
1120 error if the gap is too big.
1125 Description: add support PG_writeback bit
1126 Details : add support for PG_writeback bit for Lustre, for more carefull
1127 work with page cache in 2.6 kernel. This also fix some deadlocks
1128 and remove hack for work O_SYNC with 2.6 kernel.
1130 Severity : enhancement
1132 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1133 Details : The uninit_groups feature works in conjunction with the kernel
1134 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1135 up the pass1 processing of e2fsck. This is a read-only feature
1136 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1137 mount filesystems that have had this feature enabled.
1139 Severity : enhancement
1141 Description: Improve multi-block allocation algorithm to avoid fragmentation
1142 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1143 allocation locality and avoid filesystem fragmentation.
1145 ------------------------------------------------------------------------------
1147 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1149 * Support for kernels:
1150 2.6.9-42.0.3.EL (RHEL 4)
1151 2.6.5-7.276 (SLES 9)
1152 2.4.21-47.0.1.EL (RHEL 3)
1153 2.6.12.6 vanilla (kernel.org)
1154 2.6.16.21-0.8 (SLES10)
1155 * Recommended e2fsprogs version: 1.39.cfs2-0
1157 * The backwards-compatible /proc/sys/portals symlink has been removed
1158 in this release. Before upgrading, please ensure that you change
1159 any configuration scripts or /etc/sysctl.conf files that access
1160 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1161 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1162 in advance of the upgrade on any system running Lustre 1.4.6 or
1163 newer, since /proc/sys/lnet was added in that version.
1164 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1168 Frequency : only when quota is used
1170 Description: avoid scanning export list for quota master
1171 Details : Change the algorithms to avoid scanning export list in order
1172 to improve the efficiency.
1175 Frequency : MDS failover only, very rarely
1177 Description: "went back in time" messages on mds failover
1178 Details : The greatest transno may be lost when the current operation
1179 finishes with an error (transno==0) and the client's last_rcvd
1180 record is over-written. Save the greatest transno in the
1181 mds_last_transno for this case.
1184 Frequency : always for specific kernels and striping counts
1186 Description: client may get "Matching packet too big" without ACL support
1187 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1188 when trying to access files in certain configurations. The
1189 clients should in fact be denied when mounting because they do
1190 not understand ACLs.
1193 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1195 Description: many clients connecting with IO in progress causes connect timeouts
1196 Details : Avoid synchronous journal commits to avoid delays caused by many
1197 clients connecting/disconnecting when bulk IO is in progress.
1198 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1199 OST_IO_PORTAL to avoid delays behind potentially many pending
1203 Frequency : occasionally with multiple writers to a single file
1205 Description: shared writes to file may result in wrong size reported by stat()
1206 Details : Allow growing of kms when extent lock is cancelled
1209 Frequency : always with random mmap IO to multi-striped file
1211 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1212 Details : Only if the hole is at the end of OST object so that kms is too
1213 small. Fix is to increase kms accordingly in ll_nopage.
1216 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1218 Description: writes to a missing object would leak memory on the OST
1219 Details : If there is an inconsistency between the MDS and OST filesystems,
1220 such that the MDS references an object that doesn't exist, writes
1221 to that object will leak memory due to incorrect cleanup in the
1222 error handling path, eventually running out of memory on the OST.
1227 Description: Creating too long symlink causes lustre errors
1228 Details : Check symlink and name lengths before sending requests to MDS.
1231 Frequency : only if flock is enabled (not on by default)
1233 Description: posix locks not released on fd closure on 2.6.9+
1234 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1235 kernels, this caused such locks not to be released on fd close and
1236 then assertions on fs unmount about still used locks.
1239 Frequency : MDS failover only, very rarely
1241 Description: clients may get ASSERTION(granted_lock != NULL)
1242 Details : When request was taking a long time, and a client was resending
1243 a getattr by name lock request. The were multiple lock requests
1244 with the same client lock handle and
1245 mds_getattr_name->fixup_handle_for_resent_request found one of the
1246 lock handles but later failed with ASSERTION(granted_lock != NULL).
1251 Description: handle->h_buffer_credits > 0, assertion failure
1252 Details : h_buffer_credits is zero after truncate, causing assertion
1253 failure. This patch extends the transaction or creates a new
1257 Frequency : NFS re-export or patchless client
1258 Bugzilla : 11179, 10796
1259 Description: Crash on NFS re-export node (__d_move)
1260 Details : We do not want to hash the dentry if we don't have a lock.
1261 But if this dentry is later used in d_move, we'd hit uninitialised
1262 list head d_hash, so we just do this to init d_hash field but
1263 leave dentry unhashed.
1266 Frequency : NFS re-export or patchless client
1268 Description: NFS exports has problem with symbolic link
1269 Details : lustre client didn't properly install dentry when re-exported
1270 to NFS or running patchless client.
1273 Frequency : NFS re-export or patchless client
1275 Description: Various nfs/patchless fixes.
1276 Details : fixes reuse disconected alias for lookup process - this fixes
1277 warning "find_exported_dentry: npd != pd",
1278 fix permission error with open files at nfs.
1279 fix apply umask when do revalidate.
1282 Frequency : occasional
1284 Description: Crash on NFS re-export node
1285 Details : calling clear_page() on the wrong pointer triggered oops in
1286 generic_mapping_read().
1289 Frequency : rarely, using O_DIRECT IO
1291 Description: unaligned directio crashes client with LASSERT
1292 Details : check for unaligned buffers before trying any requests.
1295 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1297 Description: stale data returned from RAID cache
1298 Details : If only a small amount of IO is done to the RAID device before
1299 reading it again it is possible to get stale data from the RAID
1300 cache instead of reading it from disk.
1303 Frequency : always for sles10 kernel
1305 Description: sles10 support
1306 Details : ll_follow_link: compile fixes and using of nd_set_link
1307 under newer kernels.
1310 Frequency : depends on arch, kernel and compiler version, always on sles10
1313 Description: recursive or deep enough symlinks cause stack overflow
1314 Details : getting rid of large stack-allocated variable in
1318 Frequency : depends on hardware
1320 Description: lustre write performance loss in the SLES10 kernel
1321 Details : the performance loss is caused by using of write barriers in the
1322 ext3 code. The SLES10 kernel turns barrier support on by
1323 default. The fix is to undo that change for ldiskfs.
1325 ------------------------------------------------------------------------------
1327 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1329 * Support for kernels:
1330 2.6.9-42.0.3EL (RHEL 4)
1331 2.6.5-7.276 (SLES 9)
1332 2.4.21-47.0.1.EL (RHEL 3)
1333 2.6.12.6 vanilla (kernel.org)
1337 Frequency : quota enabled and large files being deleted
1339 Description: releasing more than 4GB of quota at once hangs OST
1340 Details : If a user deletes more than 4GB of files on a single OST it
1341 will cause the OST to spin in an infinite loop. Release
1342 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1347 Description: statfs data retrieved from /proc may be stale or zero
1348 Details : When reading per-device statfs data from /proc, in the
1349 {kbytes,files}_{total,free,avail} files, it may appear
1350 as zero or be out of date.
1353 Frequency : systems with MD RAID1 external journal devices
1355 Description: lconf's call to blkid is confused by RAID1 journal devices
1356 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1357 of returning all block devices that match the journal UUID.
1360 Frequency : always, for aggregate stripe size over 4GB
1362 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1363 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1364 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1367 Frequency : always if "lfs find" used on a local file/directory
1369 Description: "lfs find" segfaults if used on a local file/directory
1370 Details : The case where a directory component was not specified wasn't
1371 handled correctly. Handle this properly.
1374 Frequency : always on ppc64
1376 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1377 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1378 perform endianness conversion.
1381 Frequency : rarely (truncate to non-zero file size after write under load)
1382 Bugzilla : 10730, 10687
1383 Description: Files padded with zeros to next 4K multiple
1384 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1385 it is possible that files that are truncated to a non-zero size
1386 immediately after being written are filled with zero bytes beyond
1387 the truncated size. No file data is lost.
1389 Severity : enhancement
1391 Description: Allow recovery/failover for liblustre clients.
1392 Details : liblustre clients were unaware of failover configurations until
1395 Severity : enhancement
1397 Description: user file locks should fail when not mounting with flock option
1398 Details : Set up an error-returning stub in ll_file_operations.lock field
1399 to prevent incorrect behaviour when client is mounted without
1400 flock option. Also, set up properly f_op->flock field for
1404 Frequency : always on ia64
1406 Description: "lfs df" loops on printing out MDS statfs information
1407 Details : The obd_ioctl_data was not initialized and in some systems
1408 this caused a failure during the ioctl that did not return
1409 an error. Initialize the struct and return an error on failure.
1412 Frequency : SLES 9 only
1414 Description: Error of copying files with lustre special EAs as root
1415 Details : Client side always return success for setxattr call for lustre
1416 special xattr (currently only "trusted.lov").
1419 Frequency : rarely on clusters with both ia64+i386 clients
1421 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1422 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1423 clients are doing concurrent writes to the same file it is
1424 possible that the ia64 clients may LASSERT because the OST
1425 extent locks are not PAGE_SIZE aligned. Ensure that grown
1426 locks are always aligned on the request boundary.
1429 Frequency : specific use, occasional
1431 Description: Overwriting in use executable truncates on-disk binary image
1432 Details : If one node attempts to overwrite an executable in use by
1433 another node, we now correctly return ETXTBSY instead of
1434 truncating the file.
1436 Severity : enhancement
1438 Description: Async OSC create to avoid the blocking unnecessarily.
1439 Details : If a OST has no remain object, system will block on the creating
1440 when need to create a new object on this OST. Now, ways use
1441 pre-created objects when available, instead of blocking on an
1442 empty osc while others are not empty. If we must block, we block
1443 for the shortest possible period of time.
1448 Description: chmod on Lustre root is propagated to other clients
1449 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1450 invalid by the follow_mount time.
1453 Frequency : liblustre clients only
1455 Description: Race in 'instant cancel' lock handling could lead to such locks
1456 never to be granted in case of SMP MDS
1457 Details : Do not destroy not yet granted but cbpending locks in
1461 Frequency : replay/resend of open
1463 Description: non null lock assetion failure in mds_intent_policy
1464 Details : Trying to replay/resend lockless open requests resulted in
1465 mds_open() returning 0 with no lock. Now it sets a flag if
1466 a lock is going to be returned.
1468 Severity : enhancement
1470 Description: Checksum enhancements
1471 Details : New checksum enhancements allow for resending RPCs that failed
1474 Severity : enhancement
1476 Description: Tunables on number of dirty pages in cacche
1477 Details : Allow to set limit on number of dirty pages cached.
1482 Description: client crash on unmount - lock still has references
1483 Details : In some error handling cases it was possible to leak a lock
1484 reference on a client while accessing a file. This was not
1485 harmful to the client during operation, but would cause the
1486 client to crash when the filesystem is unmounted.
1489 Frequency : specific case, rare
1491 Description: ETXTBSY on mds though file not in use by client
1492 Details : ETXTBSY is no longer incorrectly returned when attempting to
1493 chmod or chown a directory that the user previously tried to
1494 execute or a currently-executing binary.
1497 Frequency : extremely rare except on liblustre-based clients
1499 Description: Lustre space not freed when files are deleted
1500 Details : Clean up open-unlinked files after client eviction. Previously
1501 the unlink was skipped and the files remained as orphans.
1506 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1507 Details : In some cases it was possible to send a blocking callback to
1508 a client doing a glimpse, even though that client didn't get
1509 a lock granted. When the glimpse lock is cancelled on the OST
1510 the freed lock is left on the waiting list and corrupted the list.
1513 Frequency : all core dumps
1515 Description: Broke core dumps to lustre
1516 Details : Negative dentry may be unhashed if parent does not have UPDATE
1517 lock, but some callers, e.g. do_coredump, expect dentry to be
1518 hashed after successful create, hash it in ll_create_it.
1520 ------------------------------------------------------------------------------
1522 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1524 * Support for kernels:
1525 2.6.9-42.0.2.EL (RHEL 4)
1526 2.6.5-7.276 (SLES 9)
1527 2.4.21-40.EL (RHEL 3)
1528 2.6.12.6 vanilla (kernel.org)
1532 Frequency : always on RHEL 3
1534 Description: Number of open files grows over time
1535 Details : The number of open files grows over time, whether or not
1536 Lustre is started. This was due to a filp leak introduced
1537 by one of our kernel patches.
1539 ------------------------------------------------------------------------------
1541 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1543 * Support for kernels:
1544 2.6.9-42.EL (RHEL 4)
1545 2.6.5-7.276 (SLES 9)
1546 2.4.21-40.EL (RHEL 3)
1547 2.6.12.6 vanilla (kernel.org)
1552 Bugzilla : 5719, 9635, 9792, 9684
1553 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1554 Details : If a server is under heavy load and cannot reply to new
1555 connection requests before the client resends the (re)connect,
1556 the connection handling code can behave badly if two service
1557 threads are concurrently handing separate (re)connections from
1558 the same client. Add better locking to the connection handling
1559 code, and ensure that only a single connection will be processed
1560 for a given client UUID, even if the lock is dropped.
1562 Severity : enhancement
1564 Description: add TCP zero-copy support to kernel
1565 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1566 sends if the hardware supports scatter-gather and checksumming.
1567 This allows socklnd to do client-write and server-read more
1568 efficiently and reduce CPU utilization from skbuf copying.
1571 Frequency : only if NFS exporting from client
1573 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1574 Details : If setting attributes on a file created under NFS that had
1575 never been opened it would be possible to oops the client
1576 if the file had no objects.
1579 Frequency : always for liblustre
1581 Description: liblustre client does MDS+OSTs setattr RPC for each write
1582 Details : When doing a write from a liblustre client, the client
1583 incorrectly issued an RPC to the MDS and each OST the file was
1584 striped over in order to update the timestamps. When writing
1585 with small chunks and many clients this could overwhelm the MDS
1586 with RPCs. In all cases it would slow down the write because
1587 these RPCs are unnecessary.
1589 Severity : enhancement
1591 Description: allow number of MDS service threads to be changed at module load
1592 Details : It is now possible to change the number of MDS service threads
1593 running. Adding "options mds mds_num_threads={N}" to the MDS's
1594 /etc/modprobe.conf will set the number of threads for the next
1595 time Lustre is restarted (assuming the "mds" module is also
1596 reloaded at that time). The default number of threads will
1597 stay the same, 32 for most systems.
1602 Description: OST crash if filesystem is unformatted or corrupt
1603 Details : If an OST is started on a device that has never been formatted
1604 or if the filesystem is corrupt and cannot even mount then the
1605 error handling cleanup routines would dereference a NULL pointer.
1610 Description: NULL pointer deref in llap_from_page.
1611 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1612 incorrect) mapping if the page was truncated/reclaimed while it was
1613 searched for. Check for this condition and skip such pages when
1614 doing readahead. Introduce extra check to llap_from_page() to
1615 verify page->mapping->host is non-NULL (so page is not anonymous).
1618 Frequency : Sometimes when using sys_sendfile
1620 Description: "page not covered by a lock" warnings from ll_readpage
1621 Details : sendfile called ll_readpage without right page locks present.
1622 Now we introduced ll_file_sendfile that does necessary locking
1623 around call to generic_file_sendfile() much like we do in
1627 Frequency : with certain MDS communication failures at client mount time
1629 Description: NULL pointer deref after failed client mount
1630 Details : a client connection request may delayed by the network layer
1631 and not be sent until after the PTLRPC layer has timed out the
1632 request. If the client fails the mount immediately it will try
1633 to clean up before the network times out the request. Add a
1634 reference from the request import to the obd device and delay
1635 the cleanup until the network drops the request.
1638 Frequency : occasionally during client (re)connect
1640 Description: assertion failure during client (re)connect
1641 Details : processing a client connection request may be delayed by the
1642 client or server longer than the client connect timeout. This
1643 causes the client to resend the connection request. If the
1644 original connection request is replied in this interval, the
1645 client may trip an assertion failure in ptlrpc_connect_interpret()
1646 which thought it would be the only running connect process.
1649 Frequency : only with obd_echo servers and clients that are rebooted
1651 Description: kernel BUG accessing uninitialized data structure
1652 Details : When running an obd_echo server it did not start the ping_evictor
1653 thread, and when a client was evicted an uninitialized data
1654 structure was accessed. Start the ping_evictor in the RPC
1655 service startup instead of the OBD startup.
1657 Severity : enhancement
1658 Bugzilla : 10193 (patchless)
1659 Description: Remove dependency on various unexported kernel interfaces.
1660 Details : No longer need reparent_to_init, exit_mm, exit_files,
1661 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1664 Frequency : rare (only users of deprecated and unsupported LDAP config)
1666 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1667 Details : LDAP apparently contains 'lustreName' attributes instead of
1668 'name'. A simple remapping of the name is sufficient.
1671 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1673 Description: waiting_locks_callback trips kernel BUG if client is evicted
1674 Details : Running with the dump_on_timeout debug flag turned on makes
1675 it possible that the waiting_locks_callback() can try to dump
1676 the Lustre kernel debug logs from an interrupt handler. Defer
1677 this log dumping to the expired_lock_main() thread.
1679 Severity : enhancement
1681 Description: Support NFS exporting on 2.6 kernels.
1682 Details : Implement non-rawops metadata methods for NFS server to use without
1683 changing NFS server code.
1686 Frequency : very rare (synthetic metadata workload only)
1688 Description: two racing renames might cause an MDS thread to deadlock
1689 Details : Running the "racer" program may cause one MDS thread to rename
1690 a file from being the source of a rename to being the target of
1691 a rename at exactly the same time that another thread is doing
1692 so, and the second thread has already enqueued these locks after
1693 doing a lookup of the target and is trying to relock them in
1694 order. Ensure that we don't try to re-lock the same resource.
1697 Frequency : only very large systems with liblustre clients
1699 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1700 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1701 all OSTs in parallel. This allows the eviction of stale liblustre
1702 clients to proceed much faster than if they were done in series,
1703 and also offers similar improvements for other set_info RPCs.
1708 Description: excessive CPU usage during initial read phase on client
1709 Details : During the initial read phase on a client, it would agressively
1710 retry readahead on the file, consuming too much CPU and impacting
1711 performance (since 1.4.5.8). Improve the readahead algorithm
1712 to avoid this, and also improve some other common cases (read
1713 of small files in particular, where "small" is files smaller than
1714 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1719 Description: MDS crash when receiving packet with unknown intent.
1720 Details : Do not LBUG in unknown intent case, just return -EFAULT
1722 Severity : enhancement
1723 Bugzilla : 9293, 9385
1724 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1725 Details : Do not serialize getattr (non-intent version) and statfs.
1728 Frequency : occasional, when OST network is overloaded/intermittent
1730 Description: client evicted by OST after bulk IO timeout
1731 Details : If a client sends a bulk IO request (read or write) the OST
1732 may evict the client if it is unresposive to its data GET/PUT
1733 request. This is incorrect if the network is overloaded (takes
1734 too long to transfer the RPC data) or dropped the OST GET/PUT
1735 request. There is no need to evict the client at all, since
1736 the pinger and/or lock callbacks will handle this, and the
1737 client can restart the bulk request.
1740 Frequency : Always when mmapping file with no objects
1742 Description: client crashes when mmapping file with no objects
1743 Details : Check that we actually have objects in a file before doing any
1744 operations on objects in ll_vm_open, ll_vm_close and
1750 Description: Request leak when working with deleted CWD
1751 Details : Introduce advanced request refcount tracking for requests
1752 referenced from lustre intent.
1754 Severity : Enhancement
1756 Description: Cache open file handles on client.
1757 Details : MDS now will return special lock along with openhandle, if
1758 requested and client is allowed to hold openhandle, even if unused,
1759 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1760 closing files for every read/write openration.
1762 Severity : Enhancement
1764 Description: Cache open negative dentries on client when possible.
1765 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1766 negative dentries on lock revocation.
1771 Description: Remounting a client read-only wasn't possible with a zconf mount
1772 Details : It wasn't possible to remount a client read-only with llmount.
1774 Severity : enhancement
1775 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1776 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1777 MPICH is included to add Lustre-specific ADIO interfaces.
1778 This is based closely on the UFS ADIO layer and only differs
1779 in file creation, in order to allow the OST striping to be set.
1780 This is user-contributed code and not supported by CFS.
1785 Description: extended inode attributes (immutable, append-only) work improperly
1786 when 2.4 and 2.6 kernels are used on client/server or vice versa
1787 Details : Introduce kernel-independent values for these flags.
1789 Severity : enhancement
1792 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1793 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1794 granularity. Now, Fractional MB granularity is be supported,
1795 this is very useful for low memory system.
1797 Severity : enhancement
1799 Description: Getattr by fid
1800 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1801 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1802 have locks and avoids getattr rpc requests for it on every stat.
1805 Frequency : Always, for filesystems larger than 2TB
1807 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1808 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1809 filesystems larger than 2TB to be mounted with mballoc enabled.
1812 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1814 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1815 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1816 filesystems larger than 2TB is mounted then the kernel will
1817 silently corrupt the start of the filesystem. CONFIG_LBD is
1818 enabled for all CFS-supported kernels, but the possibility of
1819 this happening with a modified kernel config exists.
1821 Severity : enhancement
1823 Description: add client O_DIRECT support for 2.6 kernels
1824 Details : It is now possible to do O_DIRECT reads and writes to files
1825 in the Lustre client mountpoint on 2.6 kernel clients.
1827 Severity : enhancement
1829 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1830 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1831 parallel, not waiting for response from every OST before sending
1832 a rpc to the next OST.
1837 Description: setattr vs write race when updating file timestamps
1838 Details : Client processes that update a file timestamp into the past
1839 right after writing to the file (e.g. tar) it is possible that
1840 the updated file modification time can be reset to the current
1841 time due to a race between processing the setattr and write RPC.
1843 Severity : enhancement
1845 Description: Bring 'lfs find' closer in line with regular Linux find.
1846 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1847 -print0 options and obtains all the needed info through the lustre
1850 Severity : enhancement
1852 Description: support up to 1024 configured devices on one node
1853 Details : change obd_dev array from statically allocated to dynamically
1854 allocated structs as they are first used to reduce memory usage
1859 Description: Flush dirty partially truncated pages during truncate
1860 Details : Immediatelly flush partially truncated pages in filter_setattr,
1861 this way we completely avoid having any pages in page cache on OST
1862 and can retire ugly workarounds during writes to flush such pages.
1867 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1868 Details : protect inode from truncation within vfs_unlink() context
1869 just take a reference before calling vfs_unlink() and release it
1870 when parent's i_sem is free.
1875 Description: last_id value checked outside lock on OST caused LASSERT failure
1876 Details : If there were multiple MDS->OST object precreate requests in
1877 flight, it was possible that the OST's last object id was checked
1878 outside a lock and incorrectly tripped an assertion. Move checks
1879 inside locks, and discard old precreate requests.
1882 Frequency : always, if extents are used on OSTs
1884 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1885 Details : OSTs using the extents format would not zero the high 16 bits of
1886 the index physical block number. This is not a problem for any
1887 OST filesystems smaller than 16TB, and no kernels support ext3
1888 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1889 new/modified files) and can be fixed for existing filesystems
1890 with e2fsprogs-1.39-cfs1.
1895 Description: import connection selection may be incorrect if timer wraps
1896 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1897 import connections to be ignored if the 32-bit jiffies counter
1898 wraps. Use a 64-bit jiffies counter.
1901 Frequency : very large clusters immediately after boot
1903 Description: LNET request buffers exhausted under heavy short-term load
1904 Details : If a large number of client requests are generated on a service
1905 that has previously never seen so many requests it is possible
1906 that the request buffer growth cannot keep up with the spike in
1907 demand. Instead of dropping incoming requests, they are held in
1908 the LND until the RPC service can accept more requests.
1911 Frequency : Sometimes during replay
1913 Description: Assertion failure in ll_local_open after replay.
1914 Details : If replay happened on an open request reply before we were able
1915 to set replay handler, reply will become not swabbed tripping the
1916 assertion in ll_local_open. Now we set the handler right after
1917 recognising of open request
1920 Frequency : very rare
1922 Description: kernel reports "badness in vsnprintf"
1923 Details : Reading from the "recovery_status" /proc file in small chunks
1924 may cause a negative length in lprocfs_obd_rd_recovery_status()
1925 call to vsnprintf() (which is otherwise harmless). Exit early
1926 if there is no more space in the output buffer.
1928 Severity : enhancement
1930 Description: clear OBD RPC statistics by writing to them
1931 Details : It is now possible to clear the OBD RPC statistics by writing
1932 to the "stats" file.
1937 Description: Client mtime is not the same on different clients after utimes
1938 Details : In some cases, the client was using the utimes() syscall on
1939 a file cached on another node. The clients now validate the
1940 ctime from the MDS + OSTs to determine which one is right.
1945 Description: Inability to activate failout mode
1946 Details : lconf script incorrectly assumed that in pythong string's numeric
1947 value is used in comparisons.
1950 Frequency : always with multiple stripes per file
1952 Description: Inefficient object allocation for mutli-stripe files
1953 Details : When selecting which OSTs to stripe files over, for files with
1954 a stripe count that divides evenly into the number of OSTs,
1955 the MDS is always picking the same starting OST for each file.
1956 Return the OST selection heuristic to the original design.
1961 Description: mount failures may take full timeout to return an error
1962 Details : Under some heavy load conditions it is possible that a
1963 failed mount can wait for the full obd_timeout interval,
1964 possibly several minutes, before reporting an error.
1965 Instead return an error as soon as the status is known.
1967 Frequency : quota enabled and large files being deleted
1969 Description: releasing more than 4GB of quota at once hangs OST
1970 Details : If a user deletes more than 4GB of files on a single OST it
1971 will cause the OST to spin in an infinite loop. Release
1972 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1977 Description: statfs data retrieved from /proc may be stale or zero
1978 Details : When reading per-device statfs data from /proc, in the
1979 {kbytes,files}_{total,free,avail} files, it may appear
1980 as zero or be out of date.
1983 Frequency : systems with MD RAID1 external journal devices
1985 Description: lconf's call to blkid is confused by RAID1 journal devices
1986 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1987 of returning all block devices that match the journal UUID.
1990 Frequency : always, for aggregate stripe size over 4GB
1992 Description: assertion fails when trying to use 4GB stripe size
1993 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1994 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1997 Frequency : always on ppc64
1999 Description: the first write on an ext3 filesystem with mballoc got stuck
2000 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2001 perform endianness conversion.
2003 ------------------------------------------------------------------------------
2005 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2007 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2008 INTEROPERATE with older versions automatically. Please read the
2009 user documentation before upgrading any part of a live system.
2010 * WARNING: Lustre networking configuration changes are required with
2011 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2014 * Support for kernels:
2015 2.6.9-22.0.2.EL (RHEL 4)
2016 2.6.5-7.244 (SLES 9)
2017 2.6.12.6 vanilla (kernel.org)
2020 Severity : enhancement
2021 Bugzilla : 7981/8208
2022 Description: Introduced Lustre Networking (LNET)
2023 Details : LNET is new networking infrastructure for Lustre, it includes
2024 a reorganized network configuration mode (see the user
2025 documentation for full details) as well as support for routing
2026 between different network fabrics. Lustre Networking Devices
2027 (LNDs) for the supported network fabrics have also been
2028 created for this new infrastructure.
2030 Severity : enhancement
2031 Description: Introduced Access control lists
2032 Details : clients can set ACLs on files and directories in order to have
2033 more fine-grained permissions than the standard Unix UGO+RWX.
2034 The MDS must be started with the "-o acl" mount option.
2036 Severity : enhancement
2037 Description: Introduced filesystem quotas
2038 Details : Administrators may now establish per-user quotas on the
2041 Severity : enhancement
2043 Description: Configuration change for the XT3
2044 The PTLLND is now used to run Lustre over Portals on the XT3
2045 The configure option(s) --with-cray-portals are no longer used.
2046 Rather --with-portals=<path-to-portals-includes> is used to
2047 enable building on the XT3. In addition to enable XT3 specific
2048 features the option --enable-cray-xt3 must be used.
2053 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2054 Details : A race between allocating a new llcd and re-getting the llcd_lock
2055 allowed another thread to grab newly-allocated llcd.
2057 Severity : enhancement
2059 Description: 2.6 OST async journal commit and locking fix to improve performance
2060 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2061 2.6 kernels are now async as they already were in 2.4 kernels so
2062 that they can commit concurrently with the network bulk transfer.
2063 For block-allocated files the filter allocation semaphore is held
2064 to avoid filesystem fragmentation during allocation. BKL lock
2065 removed for 2.6 xattr operations where it is no longer needed.
2070 Description: lconf incorrectly determined whether two IP networks could talk
2071 Details : In some more complicated routing and multiple-network
2072 configurations, lconf will avoid trying to make a network
2073 connection to a disjoint part of the IP space. It was doing the
2074 math incorrectly for one set of cases.
2079 Description: Fix for potential infinite loop processing records in an llog.
2080 Details : If an llog record is corrupted/zeroed, it is possible to loop
2081 forever in llog_process(). Validate the llog record length
2082 and skip the remainder of the block on error.
2085 Frequency : occasional (liblustre only)
2087 Description: liblustre could not open files whose last component is a symlink
2088 Details : sysio_path_walk() would incorrectly pass the open intent to
2089 intermediate path components.
2092 Frequency : rare (liblustre only with non-standard tuning)
2093 Bugzilla : 7201 (7350)
2094 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2095 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2096 after it has been released, only do this for referenced locks
2098 Severity : enhancement
2100 Description: specifying an (invalid) directory default stripe_size of -1
2101 would reset the directory default striping
2102 Details : stripe_size -1 was used internally to signal directory stripe
2103 removal, now use "all default" to signal dir stripe removal
2104 as a directory striping of "all default" is not useful
2107 Frequency : common for large clusters running liblustre clients
2109 Description: doing an ls when liblustre clients are running is slow
2110 Details : sending a glimpse AST to a liblustre client waits for every AST
2111 to time out, as liblustre clients will not respond. Since they
2112 cannot cache data we refresh the OST lock LVB from disk instead.
2114 Severity : enhancement
2116 Description: doing an ls at the same time as file IO can be slow
2117 Details : enqueue and other "small" requests can be blocked behind many
2118 large IO requests. Create a new OST IO portal for non-IO
2119 requests so they can be processed faster.
2122 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2124 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2125 created file with mode 000
2126 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2127 allow an ftruncate() as a non-root user to a file with mode 000.
2128 Lustre now respects this flag to disable mode checks when
2129 truncating a file owned by the user
2132 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2135 Description: Revoking locks from clients that went dead or catatonic might take
2137 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2138 cancellation of such locks instant on servers without waiting for
2139 any reply from clients. Clients drops these locks when cancel
2140 notification from server is received without replying.
2143 Frequency : liblustre-only, when liblustre client dies or becomes busy
2145 Description: Doing ls on Linux clients can take a long time with active
2147 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2148 granting such locks to it in the first place if possible. Locks
2149 are taken by proxy on the OST during the read or write and
2150 dropped immediately afterward. Add connect flags handling, do
2151 not grant locks to liblustre clients for glimpse ASTs.
2153 Severity : enhancement
2155 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2156 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2157 sequential reads and behaves badly with applications doing random
2158 reads. Improve it by reading ahead at least read region, and
2159 avoiding excessive large RPC for small reads.
2161 Severity : enhancement
2163 Description: Creating more than 1000 files for a single job may cause a load
2164 imbalance on the OSTs if there are also a large number of OSTs.
2165 Details : qos_prep_create() uses an OST index reseed value that is an
2166 even multiple of the number of available OSTs so that if the
2167 reseed happens in the middle of the object allocation it will
2168 still utilize the OSTs as uniformly as possible.
2173 Description: OST or MDS may oops in ping_evictor_main()
2174 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2175 but doesn't restart at beginning of obd_exports_timed list
2178 Severity : enhancement
2180 Description: improve by-nid export eviction on the MDS and OST
2181 Details : allow multiple exports with the same NID to be evicted at one
2182 time without re-searching the exports list.
2185 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2187 Description: MDS may oops in groups_free()
2188 Details : in rare race conditions a newly allocated group_info struct is
2189 freed again, and this can be NULL. The 2.4 compatibility code
2190 for groups_free() checked for a NULL pointer, but 2.6 did not.
2193 Frequency : common for liblustre clients doing little filesystem IO
2194 Bugzilla : 9352, 7313
2195 Description: server may evict liblustre clients accessing contended locks
2196 Details : if a client is granted a lock or receives a completion AST
2197 with a blocking AST already set it would not reply to the AST
2198 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2201 Frequency : lfs setstripe, only systems with more than 160 OSTs
2203 Description: unable to set striping with a starting offset beyond OST 160
2204 Details : llapi_create_file() incorrectly limited the starting stripe
2205 index to the maximum single-file stripe count.
2208 Frequency : LDAP users only
2210 Description: lconf did not handle in-kernel recovery with LDAP properly
2211 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2213 Severity : enhancement
2215 Description: bind OST threads to NUMA nodes to improve performance
2216 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2217 node and do their allocations there to localize memory access
2219 Severity : enhancement
2221 Description: llmount can determine client NID directly from Myrinet (GM)
2222 Details : the client NID code from gmnalnid was moved directly into
2223 llmount, removing the need to use this or specifying the
2224 client NID explicitly when mounting GM clients with zeroconf
2227 Frequency : if client is started with down MDS
2229 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2230 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2231 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2232 ptlrpc_import_delayed_req() because we didn't check if the
2233 request was interrupted, and we also didn't break out of the
2234 event loop if there was no timeout
2239 Description: data loss during non-page-aligned writes to a single file from
2240 both multiple nodes and multiple threads on one node at same time
2241 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2242 inconsistency led to false short-reads, that were cached and later
2243 used by ->prepare_write() to fill in partially written page,
2244 leading to data loss.
2247 Frequency : always, if lconf --abort_recovery used
2249 Description: lconf --abort_recovery fails with 'Operation not supported'
2250 Details : lconf was attempting to abort recovery on the MDT device and not
2253 Severity : enhancement
2255 Description: remove cleanup logs
2256 Details : replace lconf-generated cleanup logs with lustre internal
2257 cleanup routines. Eliminates the need for client-cleanup and
2260 Severity : enhancement
2262 Description: add support for EAs (user and system) on lustre filesystems
2263 Details : it is now possible to store extended attributes in the Lustre
2264 client filesystem, and with the user_xattr mount option it
2265 is possible to allow users to store EAs on their files also
2267 Severity : enhancement
2269 Description: Add possibility (config option) to show minimal available OST free
2271 Details : When compiled with --enable-mindf configure option, statfs(2)
2272 (and so, df) will return least minimal free space available from
2273 all OSTs as amount of free space on FS, instead of summary of
2274 free spaces of all OSTs.
2276 Severity : enhancement
2278 Description: do not expand extent locks acquired on OST-side
2279 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2280 by server: they are not cached anyway.
2283 Frequency : when mmap is used/binaries executed from Lustre
2285 Description: Unmmap pages before throwing them away from read cache.
2286 Details : llap_shrink cache now attempts to unmap pages before discarding
2287 them (if unmapping failed - do not discard). SLES9 kernel has
2288 extra checks that trigger if this unmapping is not done first.
2293 Description: lconf didn't resolve symlinks before checking to see whether a
2294 given mountpoint was already in use
2297 Frequency : when migrating failover services
2298 Bugzilla : 6395, 9514
2299 Description: When migrating a subset of services from a node (e.g. failback
2300 from a failover service node) the remaining services would
2301 time out and evict clients.
2302 Details : lconf --force (implied by --failover) sets the global obd_timeout
2303 to 5 seconds in order to quickly disconnect, but this caused
2304 other RPCs to time out too quickly. Do not change the global
2305 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2307 Severity : enhancement
2308 Frequency : if MDS is started with down OST
2309 Bugzilla : 9439,5706
2310 Description: Allow startup/shutdown of an MDS without depending on the
2311 availability of the OSTs.
2312 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2313 Add appropriate locking and lov-osc refcounts for safe
2314 cleaning. Add osc abort_inflight calls in case the
2315 synchronize never started.
2318 Frequency : occasional (Cray XT3 only)
2320 Description: root not authorized to access files in CRAY_PORTALS environment
2321 Details : The client process capabilities were not honoured on the MDS in
2322 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2323 been accessed by an authorized user then root was able to access
2324 the file on the local client also. The root user capabilities
2325 are now allowed on the MDS, as this environment has secure UID.
2328 Frequency : occasional
2330 Description: ldiskfs "too long searching" message happens too often
2331 Details : A debugging message (otherwise harmless) prints too often on
2332 the OST console. This has been reduced to only happen when
2333 there are fragmentation problems on the filesystem.
2338 Description: Division by zero in statfs when all OSCs are inactive
2339 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2340 lov_statfs divides by value returned by lov_get_stripecnt().
2344 Bugzilla : 9489, 3273
2345 Description: First write from each client to each OST was only 4kB in size,
2346 to initialize client writeback cache, which caused sub-optimal
2347 RPCs and poor layout on disk for the first writen file.
2348 Details : Clients now request an initial cache grant at (re)connect time
2349 and so that they can start streaming writes to the cache right
2350 away and always do full-sized RPCs if there is enough data.
2351 If the OST is rebooted the client also re-establishes its grant
2352 so that client cached writes will be honoured under the grant.
2357 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2358 Details : Now I/O RPCs go to different portal number and (presumably) fast
2359 lock requests (and glimses) and other RPCs get their own service
2360 threads pool that should be able to service those RPCs
2363 Severity : enhancement
2365 Description: Ability to exchange lustre version between client and servers and
2366 issue warnings at client side if client is too old. Also for
2367 liblustre clients there is ability to refuse connection of too old
2369 Details : New 'version' field is added to connect data structure that is
2370 filled with version info. That info is later checked by server and
2374 Frequency : rare, liblustre only.
2375 Bugzilla : 9296, 9581
2376 Description: Two simultaneous writes from liblustre at offset within same page
2377 might proceed at the same time overwriting eachother with stale
2379 Details : I/O lock withing llu_file_prwv was released too early, before data
2380 actually was hitting the wire. Extended lock-holding time until
2381 server acknowledges receiving data.
2384 Frequency : extremely rare. Never observed in practice.
2386 Description: avoid generating lustre_handle cookie of 0.
2387 Details : class_handle_hash() generates handle cookies by incrementing
2388 global counter, and can hit 0 occasionaly (this is unlikely, but
2389 not impossible, because initial value of cookie counter is
2390 selected randonly). Value of 0 is used as a sentinel meaning
2391 "unassigned handle" --- avoid it. Also coalesce two critical
2392 sections in this function into one.
2394 Severity : enhancement
2396 Description: allow liblustre clients to delegate truncate locking to OST
2397 Details : To avoid overhead of locking, liblustre client instructs OST to
2398 take extent lock in ost_punch() on client's behalf. New connection
2399 flag is added to handle backward compatibility.
2401 Severity : enhancement
2402 Bugzilla : 4928, 7341, 9758
2403 Description: allow number of OST service threads to be specified
2404 Details : a module parameter allows the number of OST service threads
2405 to be specified via "options ost ost_num_threads={N}" in the
2406 OSS's /etc/modules.conf or /etc/modprobe.conf.
2410 Bugzilla : 6146, 9635, 9895
2411 Description: servers crash with bad pointer in target_handle_connect()
2412 Details : In rare cases when a client is reconnecting it was possible that
2413 the connection request was the last reference for that export.
2414 We would temporarily drop the export reference and get a new
2415 one, but this may have been the last reference and the export
2416 was just destroyed. Get new reference before dropping old one.
2418 Severity : enhancement
2419 Frequency : if client is started with failover MDS
2421 Description: Allow multiple MDS hostnames in the mount command
2422 Details : Try to read the configuration from all specified MDS
2423 hostnames during a client mount in case the "primary"
2426 Severity : enhancement
2428 Description: Stop sending data to evicted clients as soon as possible.
2429 Details : Check if the client we are about to send or are sending data to
2430 was evicted already. (Check is done every second of waiting,
2431 for which l_wait_event interface was extended to allow checking
2432 of exit condition at specified intervals).
2435 Frequency : rare, normally only when NFS exporting is done from client
2437 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2439 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2442 Frequency : always, if srand() is called before liblustre initialization
2444 Description: Liblustre uses system PRNG disturbing its usage by user application
2445 Details : Introduce internal to lustre fast and high-quality PRNG for
2446 lustre usage and make liblustre and some other places in generic
2447 lustre code to use it.
2449 Severity : enhancement
2450 Bugzilla : 9477, 9557, 9870
2451 Description: Verify that the MDS configuration logs are updated when xml is
2452 Details : Check if the .xml configuration logs are newer than the config
2453 logs stored on the MDS and report an error if this is the case.
2454 Request --write-conf, or allow starting with --old_conf.
2456 Severity : enhancement
2458 Description: Handle symlinks in the path when checking if Lustre is mounted.
2459 Details : Resolve intermediate symlinks when checking if a client has
2460 mounted a filesystem to avoid duplicate client mounts.
2465 Description: lconf can hit an error exception but still return success.
2466 Details : The lconf command catches the Command error exception at the top
2467 level script context and will exit with the associated exit
2468 status, but doesn't ensure that this exit status is non-zero.
2473 Description: failure of ptlrpc thread startup can cause oops
2474 Details : Starting a ptlrpc service thread can fail if there are a large
2475 number of threads or the server memory is very fragmented.
2476 Handle this without oopsing.
2479 Frequency : always, only if liblustre and non-default acceptor port was used
2481 Description: liblustre cannot connect to servers with non-default acceptor port
2482 Details : tcpnal_set_default_params() was not called and was therefore
2483 ignoring the environment varaible TCPNAL_PORT, as well as other
2484 TCPNAL_ environment variables
2489 Description: two objects could be created on the same OST for a single file
2490 Details : If an OST is down, in some cases it was possible to create two
2491 objects on a single OST for a single file. No problems other
2492 than potential performance impact and spurious error messages.
2496 Bugzilla : 5681, 9562
2497 Description: Client may oops in ll_unhash_aliases
2498 Details : Client dcache may become inconsistent in race condition.
2499 In some cases "getcwd" can fail if the current directory is
2505 Description: Inode refcounting problems in NFS export code
2506 Details : link_raw functions used to call d_instantiate without obtaining
2507 extra inode reference first.
2511 Bugzilla : 9942, 9903
2512 Description: Referencing freed requests leading to crash, memleaks with NFS.
2513 Details : We used to require that call to ll_revalidate_it was always
2514 followed by ll_lookup_it. Also with revalidate_special() it is
2515 possible to call ll_revalidate_it() twice for the same dentry
2516 even if first occurence returned success. This fix changes semantic
2517 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2518 reference on a request referred from the intent.
2519 ll_intent_release() then releases such a request.
2522 Frequency : rare, normally benchmark loads only
2524 Description: unlinked inodes were kept in memory on the client
2525 Details : If a client is repeatedly creating and unlinking files it
2526 can accumulate a lot of stale inodes in the inode slab cache.
2527 If there is no other client load running this can cause the
2528 client node to run out of memory. Instead flush old inodes
2529 from client cache that have the same inode number as a new inode.
2532 Frequency : SLES9 2.6.5 kernel and long filenames only
2533 Bugzilla : 9969, 10379
2534 Description: utime reports stale NFS file handle
2535 Details : SLES9 uses out-of-dentry names in some cases, which confused
2536 the lustre dentry revalidation. Change it to always use the
2540 Frequency : rare, unless heavy write-truncate concurrency is continuous
2541 Bugzilla : 4180, 6984, 7171, 9963, 9331
2542 Description: OST becomes very slow and/or deadlocked during object unlink
2543 Details : filter_destroy() was holding onto the parent directory lock
2544 while truncating+unlinking objects. For very large objects this
2545 may block other threads for a long time and slow overall OST
2546 responsiveness. It may also be possible to get a lock ordering
2547 deadlock in this case, or run out of journal credits because of
2548 the combined truncate+unlink. Solution is to do object truncate
2549 first in one transaction without parent lock, and then do the
2550 final unlink in a new transaction with the parent lock. This
2551 reduces the lock hold time dramatically.
2554 Frequency : rare, 2.4 kernels only
2556 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2557 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2558 before unmounting their filesystem in order to avoid deadlock.
2559 The kernel_locked() function in 2.4 kernels only checks whether
2560 the kernel lock is held, not whether it is this process that is
2561 holding it as 2.6 kernels do.
2566 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2567 Details : The client ptlrpc code may be trying to reconnect to a down
2568 server before a previous connection attempt has timed out.
2569 Increase the reconnect interval to be longer than the connection
2570 timeout interval to avoid sending duplicate connections to
2574 Frequency : echo_client brw_test command
2576 Description: fix echo_client to work with OST preallocated code
2577 Details : OST preallocation code (5137) didn't take echo_client IO path
2578 into account: echo_client calls filter methods outside of any
2579 OST thread and, hence, there is no per-thread preallocated
2580 pages and buffers to use. Solution: hijack pga pages for IO. As
2581 a byproduct, this avoids unnecessary data copying.
2585 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2586 Description: Client can oops in mdc_commit_close() after open replay
2587 Details : It was possible for the MDS to return an open request with no
2588 transaction number in mds_finish_transno() if the client was
2589 evicted, but without actually returning an error. Clients
2590 would later try to replay that open and may trip an assertion
2591 Simplify the client close codepath, and always return an error
2592 from the MDS in case the open is not successful.
2595 Frequency : rare, 2.6 OSTs only
2597 Description: OST may deadlock under high load on fragmented files
2598 Details : If there was a heavy load and highly-fragmented OST filesystems
2599 it was possible to have all the OST threads deadlock waiting on
2600 allocation of biovecs, because the biovecs were not released
2601 until the entire RPC IO was completed. Instead, release biovecs
2602 as soon as they are complete to ensure forward IO progress.
2604 Severity : enhancement
2606 Description: Support for specifying external journal device at mount
2607 Details : If an OST or MDS device is formatted with an external journal
2608 device, this device major/minor is stored in the ext3 superblock
2609 and may not be valid for failover. Allow detecting and
2610 specifying the external journal at mount time.
2615 Description: Mounting an MDS with pending unlinked files may cause oops
2616 Details : target_finish_recovery() calls mds_postrecov() which returned
2617 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2618 considers this an error and immediately begins cleaning up the
2619 lov, just after starting the mds_lov process
2621 Severity : enhancement
2623 Description: Implement 'lfs df' to report actual free space on per-OST basis
2624 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2625 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2626 usage of objects; '-h' to report in human readable format.
2628 ------------------------------------------------------------------------------
2630 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2637 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2638 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2639 at startup during orphan unlinking. Instead, skip buddy update
2640 before setup, as it will be regenerated after recovery is complete.
2645 Description: If an OST is inactive, its locks might reference stale inodes.
2646 Details : lov_change_cbdata() must iterate over all namespaces, even if
2647 they are inactive to clear inode references from the lock.
2649 Severity : enhancement
2650 Frequency : occasional, if non-standard max_dirty_mb used
2652 Description: Client will block write RPCs if not enough grant
2653 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2654 then the client will block writes while waiting for another RPC
2655 to complete instead of consuming its dirty limit. With change
2656 we get improved performance when max_dirty_mb is small.
2658 Severity : enhancement
2659 Bugzilla : 3389, 6253
2660 Description: Add support for supplementary groups on the MDS.
2661 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2662 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2663 lookups for user supplementary groups into a cache.
2667 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2668 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2669 for liblustre O_EXCL is enforced by MDS.
2673 Description: readdir never returns NULL in liblustre.
2674 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2675 dentry in struct dirent.
2679 Frequency : liblustre only, depends on application IO pattern
2680 Description: liblustre clients evicted if not contacting servers
2681 Details : Don't put liblustre clients into the ping_evictor list, so
2682 they will not be evicted by the pinger ever.
2684 Severity : enhancement
2686 Description: Add ability to evict clients by NID from MDS.
2687 Details : By echoing "nid:$NID" string into
2688 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2689 $NID would be instantly evicted from this MDS and from all active
2690 OSTs connected to it.
2694 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2695 Details : lookup_it_finish() used to query file size from OSTs that was not
2700 Description: service threads change working directory to that of init
2701 Details : Starting lustre service threads may pin the working directory
2702 of the parent thread, making that filesystem busy. Threads
2703 now change to the working directory of init to avoid this.
2707 Frequency : during shutdown only
2708 Description: shutdown with a failed MDS or OST can cause unmount to hang
2709 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2714 Frequency : 2.6 only
2715 Description: chmod/chown may include an extra supplementary group
2716 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2717 supplementary group and if none is specified this is used.
2720 Bugzilla : 5479 (6816)
2722 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2723 Details : If lookup is in progress on a file that is unlinked we might try
2724 to revalidate the inode and fail in revalidate after lookup is
2725 complete and ll_file_open() enqueues the open again but
2726 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2729 Frequency : always, if lconf --abort_recovery used
2731 Description: lconf --abort_recovery fails with 'Operation not supported'
2732 Details : lconf was attempting to abort recovery on the MDT device and not
2735 ------------------------------------------------------------------------------
2737 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2742 Frequency : rare (only unsupported configurations with a node running as an
2744 Bugzilla : 6514, 5137
2745 Description: Mounting a Lustre file system on a node running as an OST could
2747 Details : OSTs now preallocates memory needed to write out data at
2748 startup, instead of when needed, to avoid having to
2749 allocate memory in possibly low memory situations.
2750 Specifically, if the file system is mounted on on OST,
2751 memory pressure could force it to try to write out data,
2752 which it needed to allocate memory to do. Due to the low
2753 memory, it would be unable to do so and the node would
2754 become unresponsive.
2756 Severity : enhancement
2758 Description: Addition of lconf --service command line option
2759 Details : lconf now accepts a '--service <arg>' option, which is
2760 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2762 Severity : enhancement
2764 Description: Failover mode is now the default for OSTs.
2765 Details : By default, OSTs will now run in failover mode. To return to
2766 the old behaviour, add '--failout' to the lmc line for OSTs.
2768 Severity : enhancement
2770 Description: Health checks are now provided for MDS and OSTs
2771 Details : Additional detailed health check information on MSD and OSTs
2772 is now provided through the procfs health_check value.
2775 Frequency : occasional, depends on IO load
2777 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2778 after numerous create/delete cycles
2779 Details : The ext3 inode allocation policy would not allocate new inodes
2780 very well on the OSTs because there are no new directories
2781 being created. Instead we look for groups with free space if
2782 the parent directories are nearly full.
2787 Description: Network or server problems during mount may cause partially
2788 mounted clients instead of returning an error.
2789 Details : The config llog parsing code may overwrite the error return
2790 code during mount error handling, returning success instead
2796 Description: MDS can fail to allocate large reply buffers
2797 Details : After long uptimes the MDS can fail to allocate large reply
2798 buffers (e.g. zconf client mount config records) due to memory
2799 fragmentation or consumption by the buffer cache. Preallocate
2800 some large reply buffers so that these replies can be sent even
2801 under memory pressure.
2805 Frequency : rare (liblustre)
2806 Description: fsx running with liblustre complained that using truncate() to
2807 extend the file doesn't work. This patch corrects that issue.
2808 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2809 ATTR_SIZE and lsm use in llu_setattr_raw.
2813 Frequency : rare, only 2.6 kernels
2814 Description: Unusual file access patterns on the MDS may result in inode
2815 data being lost in very rare circumstances.
2816 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2817 code in the 2.6 kernel caused the inode and/or EA data not to
2818 be read from disk, causing single-file corruption.
2822 Frequency : rare, only 2.6 filesystems using extents
2823 Description: Heavy concurrent write and delete load may cause data corruption.
2824 Details : It was possible under high-load situations to have an extent
2825 metadata block in the block device cache from a just-unlinked
2826 file overwrite a newly-allocated data block. We now unmap any
2827 metadata buffers that alias just-allocated data blocks.
2831 Frequency : filesystems with default stripe_count larger than 77
2832 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2833 Details : lconf specifies an inode size of 4096 bytes when the default
2834 stripe_count is larger than 77. This conflicts with the default
2835 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2836 this case to avoid pinning too much memory for large EAs.
2838 ------------------------------------------------------------------------------
2840 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2845 Frequency : rare (extremely heavy IO load with hundreds of clients)
2847 Description: Client is evicted, gets IO error writing to file
2848 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2849 caused clients to be evicted for AST timeouts. The fixes in
2850 bug 5192 mean we no longer need to have such short AST timeouts
2851 so ldlm_timeout has been increased.
2854 Frequency : occasional during --force or --failover shutdown under load
2855 Bugzilla : 5949, 4834
2856 Description: Server oops/LBUG if stopped with --force or --failover under load
2857 Details : a collection of import/export refcount and cleanup ordering
2858 issues fixed for safer force cleanup
2861 Frequency : only filesystems larger than 120 OSTs
2862 Bugzilla : 5990, 6223
2863 Description: lfs getstripe would oops on a very large filesystem
2864 Details : lov_getconfig used kfree on vmalloc'd memory
2867 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2868 Bugzilla : 6242, 6243
2869 Description: reading from files that had been truncated to a non-zero size
2870 but never opened returned no data
2871 Details : ll_file_read() reads zeros from no-object files to EOF
2876 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2877 Details : There's little harm in aborting MDS/OSS recovery and letting it
2878 try again, so I removed the LASSERT and return an error instead.
2880 Severity : enhancement
2882 Description: New debugging infrastructure for tracking down data corruption
2883 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2884 (b) cover more of the client-side code path, and (c) try to narrow
2885 down where problems occurred
2889 Bugzilla : 3819, 4364, 4397, 6313
2890 Description: Racing close and eviction MDS could cause assertion in mds_close
2891 Details : It was possible to get multiple mfd references during close and
2892 client eviction, leading to one thread referencing a freed mfd.
2894 Severity: : enhancement
2895 Bugzilla : 3262, 6359
2896 Description: Attempts to reconnect to servers are now more aggressive.
2897 Details : This builds on the enhanced upcall-less recovery that was added
2898 in 1.4.2. When trying to reconnect to servers, clients will
2899 now try each server in the failover group every 10 seconds. By
2900 default, clients would previously try one server every 25 seconds.
2905 Description: After recovery, certain operations trigger a failed
2906 assertion on a client.
2907 Details : Failing over an mds, using lconf -d --failover, while a
2908 client was doing a readdir() call would cause the client to
2909 LBUG after recovery completed and the readdir() was resent.
2911 Severity : enhancement
2913 Description: Default groups are now added by lconf
2914 Details : You can now run lconf --group <servicename> without having to
2915 manually add groups with lmc.
2918 Frequency : occasional
2920 Description: Nodes with an elan id of 0 trigger a failed assertion
2923 Frequency : always when accessing e.g. tty/console device nodes
2925 Description: tty and some other devices nodes cannot be used on lustre
2926 Details : file's private_data field is used by device data and lustre
2927 values in there got lost. New field was added to struct file to
2928 store fs-specific private data.
2931 Frequency : when exporting Lustre via NFS
2933 Description: NFSD failed occasionally when looking up a path component
2934 Details : NFSD is looking up ".." which was broken in ext3 directories
2935 that had grown large enough to become hashed.
2938 Frequency : Clusters with multiple interfaces not on the same subnet
2940 Description: Nodes will repeatedly try to reconnect to an interface which it
2941 cannot reach and report an error to the log.
2942 Details : Extra peer list entries will be created by lconf with some peers
2943 unreachable. lconf now validates the peer before adding it.
2946 Frequency : Only if a default stripe is set on the filesystem root.
2948 Description: Setting a default stripe on the filesystem root prevented the
2949 filesystem from being remounted.
2950 Details : The client was sending extra request flags in the root getattr
2951 request and did not allocate a reply buffer for the dir EA.
2954 Frequency : occasional, higher if lots of files are accessed by one client
2955 Bugzilla : 6159, 6097
2956 Description: Client trips assertion regarding lsm mismatch/magic
2957 Details : While revalidating inodes the VFS looks up inodes with ifind()
2958 and in rare cases can find an inode that is being freed.
2959 The ll_test_inode() code will free the lsm during ifind()
2960 when it finds an existing inode and then the VFS later attaches
2961 this free lsm to a new inode.
2965 Bugzilla : 6422, 7030
2966 Description: MDS deadlock between mkdir and client eviction
2967 Details : Creating a new file via mkdir or mknod (starting a transaction
2968 and getting the ns lock) can deadlock with client eviction
2969 (gets ns lock and trying to finish a synchronous transaction).
2972 Frequency : occasional
2973 Description: While starting a server, the fsfilt_ext3 module could not be
2975 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2976 kernels. Previously, lconf would still use the ext3 name
2977 when trying to load modules. Now, it will correctly use
2978 ext3 on 2.4 and ldiskfs on 2.6.
2980 Severity : enhancement
2981 Description: The default stripe count has been changed to 1
2982 Details : The interpretation of the default stripe count (0, to lfs
2983 or lmc) has been changed to mean striping across a single
2984 OST, rather than all available. For general usage we have
2985 found a stripe count of 1 or 2 works best.
2987 Severity : enhancement
2988 Description: Add support for compiling against Cray portals.
2989 Details : Conditional compiling for some areas that are different
2993 Frequency : occasional
2994 Bugzilla : 6409, 6834
2995 Description: Creating files with an explicit stripe count may lead to
2996 a failed assertion on the MDS
2997 Details : If some OSTs are full or unavailable, creating files may
2998 trigger a failed assertion on the MDS. Now, Lustre will
2999 try to use other servers or return an error to the
3003 Frequency : occasional
3005 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3006 caused lock timeout during readahead (since 1.4.2).
3007 Details : Processes doing readahead might match a lock that hasn't been
3008 granted yet if there are overlapping and conflicting lock
3009 requests. The readahead process waits on ungranted lock
3010 (original lock is CBPENDING), while OST waits for that process
3011 to cancel CBPENDING read lock and eventually evicts client.
3013 Severity : enhancement
3015 Description: Initial enabling of flock support for clients
3016 Details : Implements fcntl advisory locking and file status functions.
3017 This feature is provided as an optional mount flag (default
3018 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3019 locking are implemented yet, and those that are are not guaranteed
3020 to be completely correct in production environments.
3021 mount -t lustre -o [flock|noflock] ...
3024 Frequency : occasional
3026 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3027 assertion in the ext3 JBD (journaling) layer.
3028 Details : The b_committed_data struct is protected by the big kernel lock
3029 in 2.4 kernels, serializing journal_commit_transaction() and
3030 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3031 access to this struct. In 2.6 kernels there is finer grained
3032 locking to improve SMP performance of the JBD layer.
3036 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3041 Description: merge and cleanup kernel patches.
3044 -----------------------------------------------------------------------------
3046 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3048 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3049 versions of Lustre. Please update all servers and clients to
3050 version 1.4.2 or later at the same time. You must also run
3051 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3052 to update the configuration logs.
3054 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3055 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3056 - swab llog records of type '0' so we get proper header size/idx (5861)
3057 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3058 - fix rename of one directory over another leaking an inode (5953)
3059 - avoid SetPageDirty on 2.6 (5981)
3060 - don't re-add just-being-destroyed locks to the waiting list (5653)
3061 - when creating new directories, inherit the parent's custom
3062 striping settings if present parent (3048)
3063 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3064 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3065 - don't hold client locks on temporary worklist from l_lru (5666)
3066 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3067 - automatically evict dead clients (5921)
3068 - Update file size properly in create+truncate+fstat case (6196)
3069 - Do not unhash mountpoint dentries, do not allow removal of
3071 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3072 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3073 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3074 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3075 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3076 - Correctly handle reads of files with no objects (6243)
3077 - lctl recover will also mark a device active if deactivate used (5933)
3079 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3080 - allow --write-conf on an MDS with different nettype than client (5619)
3081 - don't write config llogs to MDS for mounts not from that MDS (5617)
3082 - lconf should create multiple TCP connections from a client (5201)
3083 - init scripts are now turned off by default; run chkconfig --on
3084 lustre and chkconfig --on lustrefs to use them
3085 - upcalls are no longer needed for clients to recover to failover
3087 - add --abort-recovery option to lconf to abort recovery on device
3089 - add support for an arbitrary number of OSTs (3026)
3090 - Quota support protocol changes.
3091 - forward compatibility changes to wire structs (6007)
3092 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3093 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3094 - improved "lustre status" script
3095 - initialize blocksize for non-regular files (6062)
3096 - added --disable-server and --disable-client configure options (5782)
3097 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3098 - Vanilla 2.4.29 support
3099 - increase maximum number of obd devices to 520 (6242)
3100 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3101 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3102 SLES 9 2.6.5 kernels
3103 - limit stripes per file to 160 (the maximum EA size) (6093)
3105 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3108 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3109 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3110 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3111 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3112 - small ext3 extents cleanups and fixes (5733)
3113 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3114 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3115 - avoid needless client->OST connect, fix handle mismatch (5317)
3116 - fix DLM error path that led to out-of-sync client, long delays (5779)
3117 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3118 - fix several locking issues related to i_size (5492,5624,5654,5672)
3119 - don't move pending lock onto export if it is already evicted (5683)
3120 - fix kernel oops when creating .foo in unlinked directory (5548)
3121 - fix deadlock in obdfilter statistics vs. object create (5811)
3122 - use time_{before,after} to avoid timer jiffies wrap (5882)
3123 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3124 - Do not leak request if server was not able to process it (5154)
3125 - If mds_open unable to find parent dir, make that negative lookup(5154)
3126 - don't create new directories with extent-mapping (5909, 5936)
3128 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3129 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3130 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3131 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3132 - support for RHEL4 kernel on i686 (5773)
3133 - provide error messages when incompatible logs are encountered (5898)
3135 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3136 * version 1.4.0.10 (1.4.1 release candidate 1)
3138 - don't keep a lock reference when lock is not granted (4238)
3139 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3140 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3141 - drop import inflight refcount on signal_completed_replay error (5255)
3142 - unlock page after async write error during send (3677)
3143 - handle missing objects in filter_preprw_read properly (5265)
3144 - no transno return for symlink open, don't save no-trasno open (3440)
3145 - don't try to complete elan receive that already failed (4012)
3146 - free RPC server reply state on error (5406)
3147 - clean up thread from ptlrpc_start_thread() on error (5160)
3148 - readahead could read extra page into cache that wasn't ejected (5388)
3149 - prevent races in class_attach/setup/cleanup/detach (5260)
3150 - don't dereference de->d_inode after l_dput of de (5458)
3151 - use "int" for stripe value returned from lock_to_stripe (5544)
3152 - mballoc allocation and error-checking fixes in 2.6 (5504)
3153 - block device patches to fix I/O request sizes in 2.6 (5482)
3154 - look up hostnames for IB nals (5602)
3155 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3156 - don't start multiple acceptors for the same port (5277)
3157 - fix incorrect LASSERT in mds_getattr_name (5635)
3158 - export a proc file for general "ping" checking (5628)
3159 - fix "lfs check" to not block when the MDS is down (5628)
3161 - service request history (4965)
3162 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3163 - create an "evict_client" /proc entry on OSTs, like the MDS has
3164 - fix mount usage message, return errors per mount(8) (5168)
3165 - change grep [] to grep "[]" in tests so they work in more UMLs
3166 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3167 - remove ancient LOV_MAGIC_V0 EA support (5047)
3168 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3169 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3170 - allow readahead allocations to fail when low on memory (5383)
3171 - mmap locking landed again, after considerable improvement (2828)
3172 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3174 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3177 - send OST transaction number in read/write reply to free req (4966)
3178 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3179 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3180 - the watchdog thread now runs as interruptible (5246)
3181 - flock/lockf fixes (but it's still disabled, pending 5135)
3182 - don't use EXT3 constants in llite code (5094)
3183 - memory shortage at startup could cause assertion (5176)
3185 - reorganization of lov code
3186 - single portals codebase
3188 - add extents/mballoc support (5025)
3189 - direct I/O reads in the obdfilter (4048)
3190 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3192 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3195 - send OST transaction number in read/write reply to free req (4966)
3196 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3197 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3198 - the watchdog thread now runs as interruptible (5246)
3199 - handle missing objects in filter_preprw_read properly (5265)
3200 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3201 - drop import inflight refcount on signal_completed_replay error (5255)
3202 - unlock page after async write error during send (3677)
3203 - return original error code on reconstructed replies (3761)
3204 - no transno return for symlink open, don't save no-trasno open (3440)
3206 - add pid to ldlm debugging output (4922)
3207 - bump the watchdog timeouts -- we can't handle 30sec yet
3208 - extra debugging for orphan dentry/inode bug (5259)
3210 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3213 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3214 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3215 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3216 - fix NULL dereference in /proc/sys/portals/routes (4827)
3217 - allow failed mdc_close() operations to be interrupted (4561)
3218 - stop precreate on OST before MDS would time out on it (4778)
3219 - don't send partial-page writes before EOF from client (4410)
3220 - discard client grant for sub-page writes on large-page clients (4520)
3221 - don't free dentries not owned by NFS code, check generation (4806)
3222 - fix lsm leak if mds_create_objects() fails (4801)
3223 - limit debug_daemon file size, always print CERROR messages (4789)
3224 - use transno after validating reply (3892)
3225 - process timed out requests if import state changes (3754)
3226 - update mtime on OST during writes, return in glimpse (4829)
3227 - add mkfsoptions to LDAP (4679)
3228 - use ->max_readahead method instead of zapping global ra (5039)
3229 - don't interrupt __l_wait_event() during strace
3231 - add software watchdogs to catch hung threads quickly (4941)
3232 - make lustrefs init script start after nfs is mounted
3233 - fix CWARN/ERROR duplication (4930)
3234 - return async write errors to application if possible (2248)
3235 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3236 - print NAL number in %x format (4645)
3237 - update barely-supported suse-2.4.21-171 series (4842)
3238 - support for sles 9 %post scripts
3239 - support for building 2.6 kernel-source packages
3240 - support for sles km_* packages
3242 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3245 - ignore -ENOENT errors in osc_destroy (3639)
3246 - notify osc create thread that OSC is being cleaned up (4600)
3247 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3248 - reconstruct ost_handle() like mds_handle() (4657)
3249 - create a new thread to do import eviction to avoid deadlock (3969)
3250 - let lconf resolve symlinked-to devices (4629)
3251 - don't unlink "objects" from directory with default EA (4554)
3252 - hold socknal file ref over connect in case target is down (4394)
3253 - allow more than 32000 subdirectories in a single directory (3244)
3254 - fix blocks count for O_DIRECT writes (3751)
3255 - OST returns ENOSPC from object create when no space left (4539)
3256 - don't send truncate RPC if file size isn't changing (4410)
3257 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3258 - bind to privileged port in socknal and tcpnal (3689)
3260 - rate limit CERROR/CWARN console message to avoid overload (4519)
3261 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3262 - basic mmap support (3918)
3263 - kernel patch series update from b1_4 (4711)
3265 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3268 - avoid crash during MDS cleanup with OST shut down (2775)
3269 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3270 - don't use bad inodes on the MDS (3744)
3271 - dynamic object preallocation to improve recovery speed (4236)
3272 - don't hold spinlock over lock dumping or change debug flags (4401)
3273 - don't zero obd_dev when it is force cleaned (3651)
3274 - print grants to console if they go negative (4431)
3275 - "lctl deactivate" will stop automatic recovery attempts (3406)
3276 - look for existing locks in ldlm_handle_enqueue() (3764)
3277 - don't resolve lock handle twice in recovery avoiding race (4401)
3278 - revalidate should check working dir is a directory (4134)
3280 - don't always mark "slow" obdfilter messages as errors (4418)
3282 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3285 - don't close LustreDB during write_conf until it is done (3860)
3286 - fix typo in lconf for_each_profile (3821)
3287 - allow dumping logs from multiple threads at one time (3820)
3288 - don't allow multiple threads in OSC recovery (3812)
3289 - fix debug_size parameters (3864)
3290 - fix mds_postrecov to initialize import for llog ctxt (3121)
3291 - replace config semaphore with spinlock (3306)
3292 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3293 - don't allow enqueue to complete on a destroyed export (3822)
3294 - down write_lock before checking llog header bitmap (3825)
3295 - recover from lock replay timeout (3764)
3296 - up llog sem before sending rpc (3652)
3297 - reduce ns lock hold times when setting kms (3267)
3298 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3299 - fix NULL deref and obd_dev leak on setup error (3312)
3300 - replace some LBUG about llog ops with error handling (3841)
3301 - don't match INVALID dentries from d_lookup and spin (3784)
3302 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3303 - fix invalid assertion in ptlrpc_set_wait (3880)
3305 - add libwrap support for the TCP acceptor (3996)
3306 - add /proc/sys/portals/routes for non-root route listing (3994)
3307 - allow setting MDS UUID in .xml (2580)
3308 - print the stack of a process that LBUGs (4228)
3310 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3313 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3314 - make sure to unset replay flag from failed open requests (3440)
3315 - if default stripe count is 0, use OST count for inode size (3636)
3316 - update parent mtime/ctime on client for create/unlink (2611)
3317 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3318 - free recovery state on server during a forced cleanup (3571)
3319 - unregister_reply for resent reqs (3063)
3320 - loop back devices mounting and status check on 2.6 (3563)
3321 - fix resource-creation race that can provoke i_size == 0 (3513)
3322 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3323 - more debugging for page-accounting assertion (3746)
3324 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3325 - avoid deadlock after precreation failure (3758)
3326 - fix race and lock order deadlock in orphan handling (3450, 3750)
3327 - add validity checks when grabbing inodes from l_ast_data (3599)
3329 - add /proc/.../recovery_status to obdfilter (3428)
3330 - lightweight CDEBUG infrastructure, debug daemon (3668)
3331 - change default OSC RPC parameters to be better on small clusters
3332 - turn off OST read cache for files smaller than 32MB
3333 - install man pages and include them in rpms (3100)
3334 - add new init script for (un)mounting lustre filesystems (2593)
3335 - run chkconfig in %post for init scripts (3701)
3336 - drop scimac NAL (unmaintained)
3338 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3341 - clean kiobufs before and after use (3485)
3342 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3343 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3344 - change init script to fail more clearly if not run as root (1528)
3345 - allow clients to reconnect during replay (1742)
3346 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3347 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3348 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3349 - fix cond_resched (3554)
3350 - only evict unfinished clients after recovery (3515)
3351 - allow bulk resend, prevent data loss (3570)
3352 - dynamic ptlrpc request buffer allocation (2102)
3353 - don't allow unlinking open directory if it isn't empty (2904)
3354 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3355 - remove extraneous obd dereference causing LASSERT failure (3334)
3356 - don't use get_cycles() when creating temp. files on the mds (3156)
3357 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3358 - handle EEXIST for set-stripe, set proper directory name (3336)
3360 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3361 - fix ksocknal_fmb_callback() error messages (2918)
3363 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3366 - don't copy lvb into (possibly NULL) reply on error (2983)
3367 - don't deref dentry after dput, don't free lvb on error (2922)
3368 - use the kms to determine writeback rpc length (2947)
3369 - increment oti_logcookies when osc is inactive (2948)
3370 - update client's i_blocks count via lvb messages (2543)
3371 - handle intent open/close of special files properly (1557)
3372 - mount MDS with errors=remount-ro, like obdfilter (2009)
3373 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3374 - don't use cancelling-locks' kms values (2947)
3375 - use highest lock extent for kms, not last one (2925)
3376 - don't dereference ERR_PTR() dentry in error handling path (3107)
3377 - fix thread race in portals_debug_dumplog() (3122)
3378 - create lprocfs device entries at setup instead of at attach (1519)
3379 - common AST error handler, don't evict client on completion race (3145)
3380 - zero nameidata in detach_mnt in 2.6 (3118)
3381 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3382 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3383 - initialize RPC timeout timer earlier for 2.6 (3219)
3384 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3385 - print nal/nid for unknown nid (3258)
3386 - additional checks for oscc recovery before doing precreate (3284)
3387 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3388 - don't crash in mdc_close for bad permissions on open (3285)
3389 - zero i_rdev for non-device files (3147)
3390 - clear page->private before handing to FS, better assertion (3119)
3391 - tune the read pipeline (3236)
3392 - fix incorrect decref of invalidated dentry (2350)
3393 - provide read-ahead stats and refine rpc in flight stats (3328)
3394 - don't hold journal transaction open across create RPC (3313)
3395 - update atime on MDS at close time (3265)
3396 - close LDAP connection when recovering to avoid server load (3315)
3397 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3398 - don't leak open file on MDS after open resend (3325)
3399 - serialize filter_precreate and filter_destroy_precreated (3329)
3400 - loop device shouldn't call sync_dev() for nul device (3092)
3401 - clear page cache after eviction (2766)
3402 - resynchronize MDS->OST in background (2824)
3403 - refuse to mount the same filesystem twice on same mountpoint (3394)
3404 - allow llmount to create routes for mounting behind routers (3320)
3405 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3406 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3407 - fix rare problem with rename on htree directories (3417)
3409 - allow default OST striping configuration per directory (1414)
3410 - fix compilation for qswnal for 2.6 kernels (3125)
3411 - increase maximum number of MDS request buffers for large systems
3412 - change liblustreapi to be useful for external progs like lfsck (3098)
3413 - increase local configuration timeout for slow disks (3353)
3414 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3416 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3419 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3420 - don't overwrite extent policy data in reply if lock was blocked (2901)
3421 - drop filter export grants atomically with removal from device (2663)
3422 - del obd_self_export from work_list in class_disconnect_exports (2908)
3423 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3424 - swab reply message in mdc_close, other PPC fixes (2464)
3425 - fix destroying of named logs (2325)
3426 - overwrite old logs when running lconf --write_conf (2264)
3427 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3428 - fix race in target_handle_connect (2898)
3429 - mds_reint_create() should take same inode create lock (2926)
3430 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3431 - don't close files for self_export to avoid uninitialized obd (2936)
3432 - allow MDS with the same name as client node (2939)
3433 - hold dentry reference for closed log files for unlink (2325)
3434 - reserve space for all logs during transactions (2059)
3435 - don't evict page beyond end of stripe extent (2925)
3436 - don't oops on a deleted current working directory (2399)
3437 - handle hard links to targets without a parent properly (2517)
3438 - don't dereference NULL lock when racing during eviction (2867)
3439 - don't grow lock extents when lots of conflicting locks (2919)
3441 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3444 - account for cache space usage on clients to avoid data loss (974)
3445 - lfsck support in lustre kernel code (2349)
3446 - reduce journal credits needed for BRW writes (2370)
3447 - orphan handling to avoid losing space on client/server crashes
3448 - ptlrpcd can be blocked, stopping ALL progress (2477)
3449 - use lock value blocks to assist in proper KMS, faster stat (1021)
3450 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3451 - recovery for initial connections (2355)
3452 - fixes for mds_cleanup_orphans (1934)
3453 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3454 - block all file creations until orphan recovery completes (1901)
3455 - client remove rq_connection from request struct (2423)
3456 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3457 - recovery timer race (2670)
3458 - mdc_close recovey bug (2532)
3459 - ptlrpc cleanup bug (2710)
3460 - mds timeout on local locks (2588)
3461 - namespace lock held during RPCs (2431)
3462 - handle interrupted sync write properly (2503)
3463 - don't try to handle a message that hasn't been replied to (2699)
3464 - client assert failure during cleanup after abort recovery (2701)
3465 - leak mdc device after failed mount (2712)
3466 - ptlrpc_check_set allows timedout requests to complete (2714)
3467 - wait for inflight reqs when ptlrpcd finishes (2710)
3468 - make sure unregistered services are removed from the srv_list
3469 - reset bulk XID's when resending them (caught by 1138 test)
3470 - unregister_bulk after timeout
3471 - fix lconf error (2694)
3472 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3473 - readahead locks pages, leaves pending causing memory pressure (2673)
3474 - increase OST request buffers to 4096 on large machines (2729)
3475 - fix up permission of existing directories in simple_mkdir (2661)
3476 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3477 - don't assign transno to errored transactions (2742)
3478 - don't delete objects on OST if given a bogus objid from MDS (2751)
3479 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3480 - if rq_no_resend, then timeout request after recovery (2432)
3481 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3482 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3483 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3484 - MDS recovery shouldn't delete objects that it has given out (2730)
3485 - if enqueue arrives after completion, don't clobber LVB (2819)
3486 - don't unlock pages twice when trigger_group_io returns error (2814)
3487 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3488 - don't write pages to disk if there was an error (1450)
3489 - don't ping imports that have recovery disabled (2676)
3490 - take buffered bytes into account when balancing socknal conn (2817)
3491 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3492 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3493 - remove little-endian swabbing of llog records (1987)
3494 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3495 - retry reposting request buffers if they fail (1191)
3496 - grow extent at grant time to avoid granting a revoked lock (2809)
3497 - lock revoke doesn't evict page if covered by a second lock (2765)
3498 - disable VM readahead to avoid reading outside lock extents (2805)
3500 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3501 - updated kernel patches for hp-2.4.20 kernel (2681)
3503 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3506 - fix truncated write corruption (2366)
3507 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3509 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3510 - ensure len(uuid) < 37 in lmc (1171)
3511 - fix ia64 OOPS in llog_test (2255)
3512 - zero end of page at obdfilter for partial page writes (2648)
3513 - don't leave stale dentries around after renames (bug 2428)
3514 - fix timeouts when evicting a client with a single lock held (2642)
3515 - set deadline for the initial HELLO message to drain (2634)
3516 - print out dotted-quad IP addresses in the socknal (2302)
3518 - additional debugging for MDS client eviction problem (2443)
3519 - fix mkfsoptions support for osts (2603, 2604)
3521 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3524 - add series for the vanilla 2.6.0 kernel
3525 - add series for the vanilla 2.4.24 kernel
3526 - add series for a cray x86/64 UL kernel drop
3527 - fix xattr patches for the vanilla 2.4.19 series
3529 - generate true UUIDs in lmc (1171)
3530 - have portals stack dumping break in UML (2466)
3531 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3532 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3533 - wake callback waiting threads on client eviction (2460)
3534 - Add --ptldebug and --subsystem to lmc (1719)
3535 - update assertion to allow safe interrupt allocation
3536 - set rq_no_resend for cancel requests (2432)
3537 - recalculate ptlrpcd timeout after resend (2494)
3538 - call vfs_rmdir when removing pending directories (2368)
3539 - fix renaming a file to itself (2429)
3540 - lmc creates a default one-stripe lov (2454)
3541 - expand procfs space to handle large clusters (2326)
3542 - increase UML stack to avoid overflow
3543 - update lconf's list of debug and subsystem masks
3544 - fix lfs find --obd (2510)
3545 - /proc tunable for disabling filter read caching (2591)
3546 - stop rpm packages from altering slapd.conf (2301)
3547 - disable nagle in the socknal under 0conf (2578)
3548 - choose mds inode size based on stripe count (2572)
3549 - fix kernel-source rpm problems (2516)
3551 - add --disable-doc to avoid pdf generation (2421)
3552 - update documentation, tests, type-os, comments
3553 - avoid format warnings on ia64
3554 - remove the TOE NAL
3555 - tiny code cleanups by removing unused fields
3557 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3560 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3561 - remove the most blatant lies from BUILDING file (2371)
3562 - change default debug level to reasonable production setting
3563 - reduce client side cache size to reduce cache flush time
3564 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3565 - make TCP zerocopy and pinger support enabled by default (2476)
3566 - sync writes completed after process exits caused crashes (2319)
3567 - maintain correct mount count on the MDS (2356)
3568 - backout 1557, because 2316 wasn't really fixed
3569 - better file I/O statistics gathering in /proc
3570 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3571 - another kernel patch to fix zero-copy TCP function export
3572 - don't take duplicate lock when processing re-sent getattr (2420)
3573 - lctl uses obd_self_export instead of creating new conn (2353)
3574 - MDS/OST recovery case which requires object creation asserted (2425)
3575 - move lfs from /usr/sbin to /usr/bin in packages
3576 - fix race between mds_client_add and mds_client_free (2417)
3577 - use kmalloc instead of slabs in portals (2430)
3578 - don't create duplicate records when a failover MDS is present (2442)
3579 - remove unnecessary mount age check (2332)
3580 - don't remove directory inodes from locks prematurely (2451)
3581 - don't break if MDS service name is the same as hostname (2103)
3582 - fix races in client write RPC generation when cache full (2482)
3584 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3587 - remove now-unused request->rq_obd (278)
3588 - if an allocation fails, print out how much memory we've used (1933)
3589 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3590 - add the "configurable stack size" patch to most series files (1256)
3591 - ability to write large log records, for 100+ OST configs (2306)
3592 - fix NULL deref when filter_prep fails (2314)
3593 - fix operator precedence error in filter_sync
3594 - dynamic allocation of socknal TX descriptors (2315)
3595 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3596 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3597 - ensure that CWARN messages in Portals always get to the syslog
3598 - __init/__exit are not for prototype decls (ldlm_init/exit)
3599 - x86-64 compile warning fixes
3600 - fix gateway LMC keyword conflict (2318)
3601 - fix MDS lock inversions in getattr/reint paths (1844)
3602 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3603 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3604 - DLM race condition prevented some lock evictions (2328)
3605 - ENOMEM detection and retry on socknal sends (2230)
3606 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3607 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3608 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3609 - better error messages when a client is rejected during recovery (1505)
3610 - avoid cancelling locks which were never granted, after failure (2330)
3611 - fix i_sem/journal inversion in mds_client_add (2333)
3612 - fix truncate/getattr lock cycle deadlock (2334)
3613 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3614 - fix two rare exit paths which could leak an l_lock() ref (2321)
3615 - fencepost error in MDS/OST orphan recovery (2226)
3616 - make log record alignment 8 bytes (1988)
3617 - lstripe now fails when requested offset > ost_count (2237)
3618 - ensure that all kernel series have a complete list.h (1607)
3619 - fix crashes in special-file operations (2316)
3620 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3622 - allow configurable automake binary, for testing new versions
3623 - small update to the lfs documentation
3625 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3627 * fix negative export reference count in fsfilt_sync (2312)
3629 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3630 * release candidate 0.9.1
3632 - orphans are moved into the PENDING directory for possible recovery
3633 - replayed opens now open by fid for orphan/rename safety (1042)
3634 - last close of an orphan inode generates a transno (683)
3635 - chdir() and mount() now pin the directory entry (1020)
3636 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3637 - discard very old requests without processing them (1502)
3638 - remove some common, well-understood CERRORs (1505)
3639 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3640 - clear "grant" flags in OST replies until OST grant code lands (1644)
3641 - fix read performance by not clobbering i_blksize on client (1598)
3642 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3643 - make LRU size a /proc tunable, clears locks when reduced (707)
3644 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3645 - server locks take references on exports now (1558)
3646 - build fixes for 2.4.20-rh trees (1663)
3647 - return an error from lov_create if all OSCs are inactive (1751)
3648 - fix import levels when a reconnect happens without a timeout (1597)
3649 - exit early from mds_open if we get a lookup error (1749)
3650 - partial page read at EOF wouldn't wait for disk before sending (1642)
3651 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3652 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3653 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3654 - write/truncate lock inversion (1639)
3655 - Don't auto-load obdclass, portals modules during cleanup (1495)
3656 - fix timestamps from jumping to "now" (1763)
3657 - extra journal assertions (1648)
3658 - add an extra multiunlink test (1771)
3659 - fix read_record/write_record API (1776)
3660 - fix leak of offset_extent, possible incorrect i_size later (1772)
3661 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3662 - Debugging for the kqswnal_get_idle_tx problems (1820)
3663 - Allow recovery to be attempted multiple times (1536)
3664 - Write out MDS last_rcvd file after it is first created (1600)
3665 - Fix tx_descriptor leak in failed transmit situations (1827)
3666 - ext3 journaling fixes for assertion failure after IO error (1871)
3667 - class_export_put() on freed export after completion AST error (1896)
3668 - Fix revalidate looping in VFS (1322)
3669 - Don't access a freed export during MDS_REINT timeout (1521)
3670 - Add open-unlink recovery support on the MDS (1673,1764)
3671 - Return an error if no MDS data was read from last_rcvd (1946)
3672 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3673 - Don't setup a disk device that doesn't match exported UUID (317)
3674 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3675 - avoid committing NULL handle in force close
3676 - local.sh is now a one-stripe LOV configuration
3677 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3678 - fix ext3 htree duplicate directory entry corruption (1516)
3679 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3680 - update to new LOV EA format (2097)
3681 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3683 2003-06-15 Phil Schwan <phil@clusterfs.com>
3686 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3687 - per-import/export recovery handling (958, 931, 959)
3688 - multiple last-rcvd slots, for serving multiple FSes (949)
3689 - connections are again shared between multiple imp/exports (963, 964)
3690 - "umount -f" would hang if any requests needed to be sent (393, 978)
3691 - avoid pinning large req buffer by copying for queued messages (989)
3692 - add "uuid" to "lctl device" command to help upcalls (991)
3693 - "open" RPCs with transnos would confuse recovery counters (1037)
3694 - do proper endian conversion of all wire messages (288, 340, 891)
3695 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3696 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3697 - fix ptlrpc_connection leak in target_handle_connect (1174)
3698 - fix import refcounting bug in OST and MDS cleanup (1134)
3699 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3700 - fix very unlikely obd_types race condition (501)
3701 - remove osc_open hack for echo_client (1187)
3702 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3703 - a failure in read_inode2 leads to deadlock (1139)
3704 - cancel ack-locks as soon as transaction is committed (1072)
3705 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3706 - make sure to commitrw after any preprw to avoid deadlock (1162)
3707 - failing to execute a file in a lustre FS would lock inode (1203)
3708 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3709 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3710 - obd_uuid2tgt would walk off the end of the list (1255)
3711 - on IA64 the osc would give portals incorrect bulk size (1258)
3712 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3713 - fix lock inversion caused by new llite matching code (1282)
3714 - limit the number of dirty pages on a client to 10MB (1286)
3715 - timed out locks were not being corrected cancelled (1289)
3716 - fix O_DIRECT above 4GB on IA-32 (1292)
3717 * major user-visible changes
3718 - fail out/fail over policy now controlled by the upcall (993)
3720 - add OBD_PING to check server availability and failure (954)
3721 - lustre messages are now sent in sending host order (288, 340, 891)
3722 - add eadatalen to MDS getattr reply (340)
3723 - OST read replies may contain second buffer, with per-page status (593)
3725 2003-03-11 Phil Schwan <phil@clusterfs.com>
3728 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3729 - failed open()s could cause deadlock; fixed (867, 869)
3730 - stop cancelling OST locks when files are closed (481)
3731 - overlapping XID spaces caused network corruption (851, 853)
3732 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3733 - setattr_raw added, to do single-RPC, server-side setattrs
3734 - lmc/lconf syntax change for OST UUIDs
3735 - fix crashy race condition between ptlrpc_free_req and osc_close
3736 - don't use request in mdc_enqueue if we hit a timeout (889)
3737 - don't set the inode i_size for regular files from the MDS (896)
3738 - handle out of order completion AST (842)
3739 - don't LBUG if a lock request times out after receiving AST (913)
3740 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3741 - if a bad lock AST arrives, send an error instead of dropping entirely
3742 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3743 - fix leak in bulk IO when only partially completed (899, 900, 926)
3744 - fix O_DIRECT for ia64 (55)
3745 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3746 - C-z after timeout could hang a process forever; fixed (977)
3748 - client-side I/O cache (678, 924, 929, 941, 970)
3750 - READPAGE and SETATTRs which don't take server-side locks get
3753 2003-02-11 Phil Schwan <phil@clusterfs.com>
3756 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3757 - Fix fsfilt_extN_readpage() to read a full page of directory
3758 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3759 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3760 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3761 - Most intent operations were removed, in favour of a new RPC mode
3762 that does a single RPC to the server and bypasses most of the VFS
3763 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3764 - Aggressively cancel local locks on DLM servers
3765 - mds_reint_unlink sends EA to the client if it's the last nlink.
3766 client uses that EA to unlink OST objects.
3767 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3768 - recursive symlinks were fixed (439)
3769 - fixed NULL deref in DEBUG_REQ
3770 - filter_update_lastobjid no longer calls sync, which annoyed extN
3771 - fixed multi-client small-writes to a single file problem (445)
3772 - fixed mtime updates during file writes (607)
3773 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3774 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3775 - fixed Linux OST size reporting problem (444, 656)
3776 - OST now updates object mtime with writes or setattr (607, 619)
3777 - client verifies file size before zeroing page past EOF (445)
3778 - OST now writes last allocated objid to disk with allocation (108)
3779 - LOV on echo now works (409)
3781 - mds_reint_unlink sends a new buffer, with the EA included. this
3782 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3783 set if a regular file was being unlinked, and it was the last link
3784 - use PtlGet from the target for bulk writes (315)
3785 - OST now updates object mtime with writes or setattr (607, 619)
3786 - LDLM now has a grant-time callback to revalidate locked items, if
3788 - Many MDS operations were reorganized to combat race conditions
3790 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3791 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3793 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3796 - Fully reactivate OST imports after reconnection (512, others)
3797 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3798 - More graceful error handling for truncating on dead OST (515)
3799 - Don't error out unless we're actually accessing dead stripes (474)
3800 - Fix garbage sizes when stripes are missing (410)
3801 - LRU counters were broken, causing constant lock purge (433, 432)
3802 - garbage on read from stripes with failed OSTs (441)
3803 - mark OSCs as active before reconnecting during recovery (438)
3804 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3805 - lfind did not preserve OST order in output (443)
3806 - symlinks cause hung clients, incorrect data (439)
3807 - stop dereferencing request after dropping refcount (457)
3808 - don't LASSERT(spin_is_locked) on non-SMP (455)
3809 - fixes for many rename() bugs
3810 - fstat didn't correctly synchronize attributes (399)
3811 - server must handle lock cancellation during blocking AST prep (487)
3812 - bulk descriptors were free()d too soon (511)
3813 - fix paths in lconf, which would load incorrect modules (451, 507)
3814 - fix confusing lconf 'host not found' error message (386)
3815 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3816 - fix race condition in mdc_blocking_ast() for inode access (526)
3817 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3818 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3819 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3820 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3821 - fix lstripe in several ways (kernel side) (527)
3822 - fix request leak in ldlm_cli_enqueue (262)
3823 - incorrect OSC was marked inactive after OST failure
3824 - call mds_fs_cleanup before unmounting filesystem (524)
3825 - fix races between taking ns_lock and ldlm_lock_change_resource
3826 - fix races updating LOV export open file list
3827 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3828 - fix recovery NULL deref in ldlm_cli_cancel_unused
3829 - fix some DLM races by using new hash table for lock handles (419)
3830 - permit the client to specify desired inodes, at replay
3831 - duplicate requests when we queue them for replay reintegration
3832 - fix last_rcvd offset calculation
3833 - sync after each recovered transaction, so we always make progress
3834 - never, not always, ERESTART requests without transnos
3835 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3837 - skip replay if the MDS says that the client is already connected
3838 - don't check for a recovery-enabled export to match lctl's UUID
3839 - don't INC_USE_COUNT for phantom exports
3840 - don't crash when cleaning up phantom exports (567)
3841 - don't double-finish or set replay data for errored mdc_open requests
3842 - abort requests when they time out, so we don't get old replies
3843 - send/receive replies for AST messages again
3844 - if the client says that it doesn't have the lock, cancel it on the
3846 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3847 mark it as destroyed, it will all work out when decref is called
3848 - fix module use counts (22, 581)
3850 - ASTs now expect a reply (server cancels lock on error reply)
3852 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3855 - fix many simultaneous client startup (392)
3856 - fix dentry->d_it clobbering
3857 - credentials weren't being shipped for readdir/getattr operations
3858 - remove invalid assertions triggered during some concurrent MD
3860 - proper Lustre versions added (336, 389)
3861 - fix memory leak for create error case (398)
3862 - fix LOV locking bug that would get cli/srv out of sync
3863 - fix echo client over LOV (409)
3864 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3865 - fix double-O_EXCL intent crash (424)
3866 - avoid sending multiple lock CANCELs (352)
3868 - MDS can do multi-client recovery (modulo bugs in new code)
3870 - many updates, edits, cleanups
3872 2002-11-18 Phil Schwan <phil@clusterfs.com>
3875 - fix null d_it dereference (346)
3876 - fix full OST/dbench hang (333)
3877 - fix permission problem with file removal (286)
3878 - fix removal of OSCs from LOV when they fail
3879 - fix NULL deref during bulk timeout (214)
3880 - fix problems related to multiple filesystems on one MDS (241)
3881 - fixed serious subtle metadata locking bugs
3882 - free locks on clients when inodes are removed due to memory
3884 - fix inode pointer in lock data (285)
3885 - partial support for multiple MDS on a single host (241)
3886 - data locks weren't cancelled at clear_inode time (290, 311)
3887 - intent locks could lead to unbounded lock growth (205)
3888 - added a maximum lock count, an LRU list, and a flusher
3889 - fix multiple rename (365)
3890 - properly abstracted the echo client
3891 - OSC locked 1 byte too many; fixed
3892 - rewrote brw callback code:
3893 - fixed recovery bugs related to LOVs (306)
3894 - fixed too-many-pages-in-one-write crash (191)
3895 - fixed (again) crash in sync_io_timeout (214)
3896 - probably fixed callback-related race (385)
3898 - Add capability to MDS protocol
3899 - LDLM cancellations and callbacks on different portals
3901 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3904 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3905 - defer open object destruction to close time (601981, 138)
3906 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3907 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3908 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3909 - don't call obd_setattr in ll_file_release for destroyed objects
3910 * protocol change to lustre_msg: move |version| and add |flags|
3911 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3912 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3913 * added replay of create, unlink, link and rename operations during
3914 MDS failover; recovery should be much more robust now
3915 * remove failed OSCs from LOVs (only lov_create uses this so far)
3916 * the lustre-HOWTO was brought (more) up to date (582544)
3918 2002-10-23 Phil Schwan <phil@clusterfs.com>
3921 - in-use dentries weren't being reused properly (617851)
3922 - prevent multiple LDLM setup (599178)
3923 - fix LOV size calculations for truncate (617853)
3924 - fix client handling of MDS intent errors (POSIX)
3925 - fix permission bug in lovstripe.c test (624321)
3926 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3927 - truncate past end of file could corrupt data
3928 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3929 - a race in recovery could return ETIMEDOUT to apps (623947)
3930 - building outside the source directory was fixed
3931 * the lustre-HOWTO was brought (more) up to date (582544)
3932 * major progress was made on recovery functionality
3934 2002-10-10 Phil Schwan <phil@clusterfs.com>
3937 - recovery deadlock fix
3938 - rm -rf causes LBUG fix (617817)
3939 - file open by multiple tasks fix (618962)
3940 - directory permissions bugs (602707 and 620007)
3941 - journal_stop fixed with locking (611313)
3942 - O_APPEND failures resolved (618273, perhaps 614459)
3943 - lconf PATH fix (619770)
3944 - IA64 build fix (621450)
3945 - RPC buffer sizes scale with amount of memory
3947 2002-10-01 Phil Schwan <phil@clusterfs.com>
3950 - locks would be cancelled without throwing away data pages,
3951 resulting in inconsistent data (605627)
3952 - inode attributes were not always being refreshed (605627, 612449)
3953 - lconf now continues to cleanup after lctl reports an error
3954 - MDS now enforces user permissions (602707)
3955 - lprocfs cleanup fixed, but not yet enabled (614157)
3956 - fixed infinite server hang, should a client not respond to an AST
3957 - avoid going into recovery if user calls readlink() with a buffer
3958 that's too small (613941)
3959 - AST RPCs no longer require replies (614867) -- this may be changed
3960 - don't crash server if client sends an IOV that's too big (611336)
3961 - fixed lock conversion deadlock (611892)
3962 - fixed the following of symlinks (614622)
3963 * recovery: the server can remove locks from a client that dies, other
3964 clients can make progress
3965 * more extN patch fixes
3966 * compile-time configurable ptlrpc buffer allocations
3968 - collaborative read cache document
3969 - Lustre Lite Performance CDR document-in-progress
3971 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3974 - fix typo in patch-2.4.18
3976 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3979 - clear ptlrpc request each time in handle_incoming_request()
3980 - unlink of files now destroys the object on the OST
3982 2002-09-19 Peter Braam <braam@clusterfs.com>
3984 * add hard link support
3985 * change obdfile creation method
3986 * kernel patch changed
3988 2002-09-19 Peter Braam <braam@clusterfs.com>
3991 - stack overflow bug in extN fixed
3993 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3995 * documentation updates
3996 - add man pages for config tools
3997 - update tests/README to describe testing with new config tools
3998 - finish metadata API descriptions
3999 * bug fixes and cleanups
4000 - statfs workaround for 16TB limit
4001 - LOV stripe allocation improved, can stripe on subset of OSTs
4002 - LOV file size/IO offset was wrong for files > 4GB in size
4003 - object EA data was being dropped, caused files to be unreadable
4004 - memory overflow with non-LOV OST caused memory corruption
4005 - fixed regression tests to work with new config tools, obdfilter
4006 - fixed bug when directory size became larger than 1 block
4007 - fixed bug (for single client case) when PWD was deleted
4008 - invalidate local directory pages when doing intent-based ops
4009 - avoid LDLM oops when lock callback contained bad data
4011 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4013 * documentation updates
4014 * bug fixes and cleanups
4015 - configuration tools
4018 - 64-bit compile warnings
4019 - 64-bit internal statfs data
4021 * test_brw on persistent OST devices
4023 * lprocfs (disabled)
4025 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4027 * documentation updates
4028 * bug fixes and cleanups
4029 * configuration tools
4031 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4034 * many small fixes to 0_5_4
4035 * io/network handling
4036 * thinkos in MDS operations
4038 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4041 * crucial basic fixes to 0.5.3
4042 * IOR, Iozone work over Elan
4045 2002-08-07 Phil Schwan <phil@clusterfs.com>
4046 * version 0_5_3, our first alpha
4047 * we use the new Portals iovs
4048 * documentation updates
4049 * bug fixes and cleanups
4050 * small changes in the DLM wire protocol
4052 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4053 * version 0_5_1 with some initial stability,
4054 * locking on MD and file I/O.
4055 * documentation updates
4056 * several bug fixes since 0.5.0
4057 * small changes in wire protocol
4059 2002-07-18 Phil Schwan <phil@clusterfs.com>
4061 * delivered as Lustre Light Alpha
4062 * fixed a crash after handling invalid MDS requests
4063 * fixed directory pages for architectures with non-4k pages sizes
4065 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4066 * release version v0_4_4
4067 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4069 * Fixes a number of interruption problems with OST operations.
4070 * Update documentation for portals header changes
4071 * Move all wire protocol structs/defines to lustre_idl.h
4072 * Fixes symlink length bug.
4073 * Add tcpdump to repository.
4075 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4076 * release version v0_4_3
4077 * Fixes statfs for inodes on extN.
4078 * Fixes bug in runtests which would delete /etc/hosts.
4079 * Use 64-bit object IDs wherever possible (not into VFS though)
4080 Remove ost_get_info, which is unused by lustre, and out of date.
4082 2002-07-03 Peter Braam <braam@clusterfs.com>
4083 * release version v0_4_2 Fixes a lookup error (type not passed)
4084 * move forward to head of Portals
4085 * move forward to latest Lustre kernel
4087 2002-06-25 Peter Braam <braam@clusterfs.com>
4088 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob