4 2.6.9-55.0.6.EL (RHEL 4),
5 2.6.16.46-0.14 (SLES 10),
6 2.6.18-8.1.10.EL5 (RHEL 5).
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
15 Severity : enhancement
17 Description: self-adjustable client's lru lists
18 Details : use adaptive algorithm for managing client cached locks lru
19 lists according to current server load, other client's work
20 pattern, memory activities, etc. Both, server and client
21 side namespaces provide number of proc tunables for controlling
24 Severity : enhancement
26 Description: light-weight GSS support
27 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
28 and reduce performance overhead.
30 Severity : enhancement
32 Description: Linux keyring support
33 Details : Support using service of Linux keyring for Lustre GSS internal
34 context refresh/cache mechanism.
38 Description: Fix errors in lfs documentation
39 Details : Fixes man pages
43 Description: don't use GFP_* in generic Lustre code.
44 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
48 Description: obdclass is limited by single OBD_ALLOC(idarray)
49 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
53 Description: warnings with build patchless client with vanila 2.6.19 and up
54 Details : change old ctl_table style and replace ctl_table/ctl_table_header
55 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
59 Frequency : Only for SLES
60 Description: sanity_quota fail test_1
61 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
62 file, which makes TSTID[2] inunique.
66 Frequency : Only for SLES9
67 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
68 Details : Update target/ChangeLog/which_patch .
72 Frequency : Only for test_10 in sanity-quota.sh
73 Description: a bug in quota test code
74 Details : checking a test flag in a wrong place
78 Frequency : Only for test_13 in sanity-quota.sh
79 Description: a bug in quota test script
80 Details : shouldn't check overfull allocated quota
84 Description: jbd statistics
85 Details : Port older jbd statistics patch for sles10
87 Severity : enhancement
89 Description: Add printing inode info into message about error in writepage.
94 Description: reply_lock_interpret crash due to race with it and lock cancel.
95 Details : Do not replay locks that are being cancelled. Do not reference
96 locks by their address during replay, just by their handle.
100 Frequency : When flocks are used.
101 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
102 Details : Flock locks might destroy just granted lock if it could be merged
103 with another existing flock, this is done in completion handler,
104 so teach ldlm_cli_enquque_fini that this is a valid case for
110 Description: Oops in read and write path when failing to allocate lock.
111 Details : Check if lock allocation failed and return error back.
115 Description: lstripe command fails for valid OST index
116 Details : The stripe offset is compared to lov->desc.ld_tgt_count
117 instead of lov->desc.ld_active_tgt_count.
121 Description: sanity.sh failed test 103
122 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
127 Description: df doesn't work properly if diskfs blocksize != 4K
128 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
132 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
133 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
137 Description: obd_health_check_timeout too short
138 Details : set obd_health_check_timeout as 1.5x of obd_timeout
142 Description: llapi_file_create() does not allow some changes
143 Details : add llapi_file_open() that allows specifying the mode and
144 open flags, and also returns an open file handle.
148 Description: merge and cleanup kernel patches.
149 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
153 Description: Add journal checksum support.(Kernel part)
154 Details : The journal checksum feature adds two new flags i.e
155 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
156 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
157 indicates that the commit block contains the checksum for
158 the blocks described by the descriptor blocks. Now commit
159 record can be sent to disk without waiting for descriptor
160 blocks to be written to disk. This behavior is controlled
161 using JBD2_FEATURE_ASYNC_COMMIT flag.
165 Description: OSS needs mutliple precreate threads
166 Details : Add ability to start more than one create thread per OSS.
170 Description: Sanity.sh test_65k failed: setstripe should have succeeded
171 Details : execute lfs setstripe on client
175 Description: mds_obd_create error creating tmp object
176 Details : When the user sets quota on root, llog will be affected and can't
177 create files and write files.
180 Frequency : Always on ia64 patchless client, and possibly others.
182 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
183 Details : This allows the patchless client to be loaded on architectures
188 Description: RedHat Update kernel for RHEL5
189 Details : Add the kernel config file for RHEL5.
193 Description: RedHat Update kernel for RHEL5
194 Details : Modify the kernel config file more closer RHEL5.
198 Description: Build failure against Centos5 (RHEL5)
199 Details : Define PAGE_SIZE when it isn't present.
203 Description: test_7 on sanity-quota.sh can't be used on separate machines
204 Details : add facet to handle it
208 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
209 Details : with NFS, the anon dentry's parent was set to itself in
210 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
211 corresponding dentry other than use rec->ur_name.
213 Severity : enhancement
215 Description: lfs setstripe enhancement
216 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
220 Description: testing performance impact of enabling checksumming
221 Details : enable checksum by default, allow --disable-checksum
222 configure option and "-o nochecksum" mount option
226 Description: lustre support for RHEL5
227 Details : Add support for RHEL5.
231 Description: Client eviction due to failover config
232 Details : after a connection loss, the lustre client should attempt to
233 reconnect to the last active server first before trying the
234 other potential connections.
238 Description: when mds and osts use different quota unit(32bit and 64bit),
239 quota will be released repeatly.
240 Details : void sending multiple quota reqs to mds, which will keep the
241 status between the reqs.
245 Description: osts not allocated evenly to files
246 Details : change the condition to increase offset_idx
249 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
251 Description: Data corruption for OSTs that are formatted larger than 2TB
253 Details : When generating the bio request for lustre file writes the
254 sector number would overflow a temporary variable before being
255 used for the IO. The data reads correctly from Lustre (which
256 will overflow in a similar manner) but other file data or
257 filesystem metadata may be corrupted in some cases.
261 Description: Tune the kernel for good SCSI performance.
262 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
263 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
268 Description: rewrite ext2-derived code in obdclass/uuid.c
269 Details : rewrite inherited code (uuid parsing code from ext2 utils)
270 from scratch preserving functionality.
274 Description: "lfs find -obd UUID" prints directories
275 Details : "lfs find -obd UUID" will return all directory names instead
276 of just file names. It is incorrect because the directories
277 do not reside on the OSTs.
281 Description: use i_size_read and i_size_write in 2.6 port
282 Details : replace inode->i_size access with i_size_read/write()
285 Frequency : when removing large files
287 Description: scheduling issue during removal of large Lustre files
288 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
289 It causes scheduling issues when removing large files (17TB in the
293 Frequency : only with liblustre clients on XT3
295 Description: evictions taking too long
296 Details : allow llrd to evict clients directly on OSTs
299 Frequency : only on ppc
301 Description: /proc/fs/lustre/devices broken on ppc
302 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
303 We should make sure the type of 'index' is loff_t and then cast
304 explicitly as needed below. Do not assign an explicitly cast
308 Frequency : only for rhel5
310 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
311 Details : Modify the target file & which_kernel.
313 Severity : enhancement
315 Description: omit set fsid for export NFS
316 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
317 as FS_REQUIRES_DEV for avoid problems with generate fsid.
321 Frequency : Always, for kernels after 2.6.16
322 Description: Fix warning idr_remove called for id=.. which is not
324 Details : Last kernels save old s_dev before kill super and not allow
325 to restore from callback - restore it before call kill_anon_super.
329 Description: buffer overruns could theoretically occur
330 Details : llapi_semantic_traverse() modifies the "path" argument by
331 appending values to the end of the origin string, and a
332 overrun may occur. Adding buffer overrun check in liblustreapi.
336 Description: client sometimes tried revalidate root dentry.
337 Details : not revalidate dentry if it root dentry.
341 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
342 Details : Modify vm-tunables-rhel4.patch.
346 Description: Kernel config for 2.6.18-vanilla.
347 Details : Modify targets/2.6-vanilla.target.in.
348 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
349 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
350 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
351 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
353 --------------------------------------------------------------------------------
355 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
357 * Support for kernels:
358 2.6.5-7.283 (SLES 9),
359 2.6.9-55.EL (RHEL 4),
360 2.6.16.46-0.14 (SLES 10),
361 2.6.18.8 vanilla (kernel.org)
362 * Client support for unpatched kernels:
363 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
364 2.6.16 - 2.6.22 vanilla (kernel.org)
365 * Due to recently discovered recovery problems, we do not recommend
366 using patchless RHEL 4 clients with this or any earlier release.
367 * Recommended e2fsprogs version: 1.39.cfs8
368 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
369 * Starting with this release, the ldiskfs backing filesystem required
370 by Lustre is now in its own package, lustre-ldiskfs. This package
371 should be installed. It is versioned separately from Lustre and
372 may be released separately in future.
376 Description: block reactivating mgc import until all deactivates complete
377 Details : Fix race when failing back MDT/MGS to itself (testing)
379 Severity : enhancement
381 Description: add optional extra BUILD_VERSION info
382 Details : add a new environment variable (namely LUSTRE_VERS) which allows
383 to override the lustre version.
386 Frequency : 2.6.18 servers only
388 Description: ll_kern_mount() doesn't release the module reference
389 Details : The ldiskfs module reference count never drops down to 0
390 because ll_kern_mount() doesn't release the module reference.
395 Description: server LBUG when using old ost_num_threads parameter
396 Details : Accept the old ost_num_threads parameter but warn that it
397 is deprecated, and fix an off-by-one error that caused an LBUG.
402 Description: Transient SCSI error results in persistent IO issue
403 Details : iobuf->dr_error is not reinitialized to 0 between two
407 Frequency : sometimes when underlying device returns I/O errors
409 Description: OSTs not going read-only during write failures
410 Details : OSTs are not remounted read-only when the journal commit threads
411 get I/O errors because fsfilt_ext3 calls journal_start/stop()
412 instead of the ext3 wrappers.
416 Description: poor connect scaling with increasing client count
417 Details : Don't run filter_grant_sanity_check for more than 100 exports
418 to improve scaling for large numbers of clients.
421 Frequency : SLES10 only
423 Description: sanity-quota.sh quotacheck failed: rc = -22
424 Details : Quotas cannot be enabled on SLES10.
427 Frequency : liblustre clients only
429 Description: getdirentries does not give error when run on compute nodes
430 Details : getdirentries does not fail when the size specified as an argument
431 is too small to contain at least one entry
433 Severity : enhancement
435 Description: Add LNET router traceability for debug purposes
436 Details : If a checksum failure occurs with a router as part of the
437 IO path, the NID of the last router that forwarded the bulk data
438 is printed so it can be identified.
443 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
444 Details : Due to a race condition, liblustre clients were occasionally
447 Severity : enhancement
449 Description: lfs setstripe use optional parameters instead of postional
452 Severity : enhancement
454 Description: Nanosecond timestamp support for ldiskfs
455 Details : The on-disk ldiskfs filesystem has added support for nanosecond
456 resolution timestamps. There is not yet support for this at
457 the Lustre filesystem level.
460 Frequency : during server recovery
462 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
463 Details : request with rq_no_resend flag not awake l_wait_event if they get a
467 Frequency : nfs export on patchless client
469 Description: connectathon hang when test nfs export over patchless client
470 Details : Disconnected dentry cannot be found with lookup, so we do not need
471 to unhash it or make it invalid
474 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
475 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
476 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
481 Description: open req refcounting wrong on reconnect
482 Details : If reconnect happened between getting open reply from server and
483 call to mdc_set_replay_data in ll_file_open, we will schedule
484 replay for unreferenced request that we are about to free.
485 Subsequent close will crash in variety of ways.
486 Check that request is still eligible for replay in
487 mdc_set_replay_data().
492 Description: disable writes to filesystem when reading health_check file
493 Details : the default for reading the health_check proc file has changed
494 to NOT do a journal transaction and write to disk, because this
495 can cause reads of the /proc file to hang and block HA state
496 checking on a healthy but otherwise heavily loaded system. It
497 is possible to return to the previous behaviour during configure
498 with --enable-health-write.
500 Severity : enhancement
502 Description: 64-bit inode version
503 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
504 to the inode. This will be required for version-based recovery.
509 Description: MDS fails to start if a duplicate client export is detected
510 Details : in some rare cases it was possible for a client to connect to
511 an MDS multiple times. Upon recovery the MDS would detect this
512 and fail during startup. Handle this more gracefully.
514 Severity : enhancement
516 Description: Add -o localflock option to simulate old noflock
518 Details : This will achieve local-only flock/fcntl locks
524 Description: log_commit_thread vs filter_destroy race leads to crash
525 Details : Take import reference before releasing llog record semaphore
530 Description: Wrong request locking in request set processing
531 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
532 imp_delayed_list, in this place should be used imp_lock.
535 Frequency : when reconnection
537 Description: Grant Leak when osc reconnect to OST
538 Details : When osc reconnect ost, OST(filter) should check whether it
539 should grant more space to client by comparing fed_grant and
540 cl_avail_grant, and return the granted space to client instead
541 of "new granted" space, because client will call osc_init_grant
542 to update the client grant space info.
545 Frequency : when client reconnect to OST
547 Description: Grant Leak when osc do resend and replay bulk write
548 Details : When osc reconnect to OST, OST(filter)should clear grant info of
549 bulk write request, because the grant info while be sync between
550 OSC and OST when reconnect, and we should ignore the grant info
551 these of resend/replay write req.
556 Description: Grant space more than avaiable left space sometimes.
557 Details : When then OST is about to be full, if two bulk writing from
558 different clients came to OST. Accord the avaliable space of the
559 OST, the first req should be permitted, and the second one
560 should be denied by ENOSPC. But if the seconde arrived before
561 the first one is commited. The OST might wrongly permit second
562 writing, which will cause grant space > avaiable space.
565 Frequency : when client is evicted
567 Description: Grant might be wrongly erased when osc is evicted by OST
568 Details : when the import is evicted by server, it will fork another
569 thread ptlrpc_invalidate_import_thread to invalidate the
570 import, where the grant will be set to 0. While the original
571 thread will update the grant it got when connecting. So if
572 the former happened latter, the grant will be wrongly errased
573 because of this race.
578 Description: Checking Stale with correct fid
579 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
580 to check whether it is stale, because sometimes, we want the
581 enqueue happened anyway, and op_data.fid2 will not be initialized.
583 Severity : enhancement
585 Description: update patchless client
586 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
589 Frequency : only with 2.4 kernel
591 Description: random memory corruption
592 Details : size of struct ll_inode_info is to big for union inode.u and this
593 can be cause of random memory corruption.
598 Description: Memory leak in recovery
599 Details : Lov_mds_md was not free in an error handler in mds_create_object.
600 It should also check obd_fail before fsfilt_start, otherwise if
601 fsfilt_start return -EROFS,(failover mds during mds recovery).
602 then the req will return with repmsg->transno = 0 and rc = EROFS.
603 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
604 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
605 be freed no matter whether fsfilt_commit success or not.
608 Frequency : only with huge count clients
610 Description: Prevents from taking the superblock lock in llap_from_page for
612 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
613 for llap_from_page call in ll_removepage prevents from taking the
614 superblock lock for a soon died page.
619 Description: Not check open intent error before release open handle
620 Details : in some rare cases, the open intent error is not checked before
621 release open handle, which may cause
622 ASSERTION(open_req->rq_transno != 0), because it tries to release
623 the failed open handle.
628 Description: Set cat log bitmap only after create log success.
629 Details : in some rare cases, the cat log bitmap is set too early. and it
630 should be set only after create log success.
634 Description: Accessing a block bevice can re-enable I/O when Lustre is
635 tearing down a device.
636 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
641 Description: service threads may hog cpus when there are a lot of requests
643 Details : Insert cond_resched to give other threads a chance to use some of
649 Description: the cat log was not initialized in recovery
650 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
651 unlink log on mds will not be initialized until mds post
652 recovery. And also in mds post recovery, the unlink log will
653 initialization will be done asynchronausly, so there will be race
654 between add unlink log and unlink log initialization.
658 Description: brw_stats were being printed incorrectly
659 Details : brw_stats were being printed as log2 but all of them were not
660 recorded as log2. Also remove some code duplication arising from
661 filter_tally_{read,write}.
665 Frequency : rare, only in recovery.
666 Description: ASSERTION(req->rq_type != LI_POISON) failed
667 Details : imp_lock should be held while iterating over imp_sending_list for
668 prevent destroy request after get timeout in ptlrpc_queue_wait.
672 Description: replay-single.sh test 52 fails
673 Details : A lock's skiplist need to be cleanup when it being unlinked
674 from its resource list.
678 Description: Short directio read returns full requested size rather than
680 Details : Direct I/O operations should return actual amount of bytes
681 transferred rather than requested size.
683 Severity : enhancement
685 Description: metadata RPC reduction (e.g. for rm performance)
686 Details : decrease the amount of synchronous RPC between clients and servers
687 by canceling conflicing lock before the operation on the client side
688 and packing thier handles into the main operation RPC to server.
690 Severity : enhancement
692 Description: Async OSC create to avoid the blocking unnecessarily.
693 Details : If a OST has no remain object, system will block on the creating
694 when need to create a new object on this OST. Now, ways use
695 pre-created objects when available, instead of blocking on an
696 empty osc while others are not empty. If we must block, we block
697 for the shortest possible period of time.
699 --------------------------------------------------------------------------------
701 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
706 Frequency : on some architectures
708 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
709 Details : Uninitialized flags sometimes cause configuration commands to
713 Frequency : patchless clients only
715 Description: missing __iget() symbol export
716 Details : The __iget() symbol export is missing. To avoid the need for
717 this on patchless clients the deathrow inode reaper is turned
718 off, and we depend on the VM to clean up old inodes. This
719 dependency was during via the fix for bug 12181.
721 --------------------------------------------------------------------------------
723 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
725 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
726 INTEROPERATE with older versions automatically. In many cases a
727 special upgrade step is needed. Please read the
728 user documentation before upgrading any part of a live system.
729 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
730 version will not interoperate with 1.6 betas before beta5 (1.5.95).
731 * WARNING: Lustre configuration and startup changes are required with
732 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
738 Severity : enhancement
740 Description: MountConf
741 Details : Lustre configuration is now managed via mkfs and mount
742 commands instead of lmc and lconf. New obd types (MGS, MGC)
743 are added for dynamic configuration management. See
744 https://mail.clusterfs.com/wikis/lustre/MountConf for
747 Severity : enhancement
749 Description: dynamic OST addition
750 Details : OSTs can now be added to a live filesystem
752 Severity : enhancement
754 Description: startup order invariance
755 Details : MDTs and OSTs can be started in any order. Clients only
756 require the MDT to complete startup.
758 Severity : enhancement
760 Description: parallel, asynchronous orphan cleanup
761 Details : orphan cleanup is now performed in separate threads for each
762 OST, allowing parallel non-blocking operation.
764 Severity : enhancement
766 Description: optimized stripe assignment
767 Details : stripe assignments are now made based on ost space available,
768 ost previous usage, and OSS previous usage, in order to try
769 to optimize storage space and networking resources.
771 Severity : enhancement
773 Description: Permanently set tunables
774 Details : All writable /proc/fs/lustre tunables can now be permanently
775 set on a per-server basis, at mkfs time or on a live
778 Severity : enhancement
780 Description: Lustre message v2
781 Details : Add lustre message format v2.
783 Severity : enhancement
785 Description: client OST exclusion list
786 Details : Clients can be started with a list of OSTs that should be
787 declared "inactive" for known non-responsive OSTs.
791 Description: SPEC SFS validation failure on NFS v2 over lustre.
792 Details : Changes the blocksize for regular files to be 2x RPC size,
793 and not depend on stripe size.
795 Severity : enhancement
797 Description: Multiple MD RPCs in flight.
798 Details : Further unserialise some read-only MDS RPCs - learn about intents.
799 To avoid overly-overloading MDS, introduce a limit on number of
800 MDS RPCs in flight for a single client and add /proc controls
801 to adjust this limit.
803 Severity : enhancement
805 Description: client read/write statistics
806 Details : Add client read/write call usage stats for performance
807 analysis of user processes.
808 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
809 file access. extents_stats shows chunk size distribution.
810 extents_stats_per_process show chunk size distribution per
813 Severity : enhancement
815 Description: mds statistics
816 Details : Add detailed mds operations statistics in
817 /proc/fs/lustre/mds/*/stats.
821 Description: Failure of copying files with lustre special EAs.
822 Details : Client side always return success for setxattr call for lustre
823 special xattr (currently only "trusted.lov").
828 Description: Refcount LNET uuids
829 Details : The global LNET uuid list grew linearly with every startup;
830 refcount repeated list entries instead of always adding to
833 Severity : enhancement
835 Description: Dynamic service threads
836 Details : Within a small range, start extra service threads
837 automatically when the request queue builds up.
840 Frequency : mixed-endian client/server environments
842 Description: mixed-endian crashes
843 Details : The new msg_v2 system had some failures in mixed-endian
846 Severity : enhancement
848 Description: Easy OST removal
849 Details : OSTs can be permanently deactivated with e.g. 'lctl
850 conf_param lustre-OST0001.osc.active=0'
852 Severity : enhancement
854 Description: MGS proc entries
855 Details : Added basic proc entries for the MGS showing what filesystems
858 Severity : enhancement
860 Description: provide MGS failover
861 Details : Added config lock reacquisition after MGS server failover.
863 Severity : enhancement
865 Description: add Linux 2.4 support
866 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
870 Description: a large application tries to do I/O to the same resource and dies
872 Details : Check the req->rq_arrival time after the call to
873 ost_brw_lock_get(), but before we do anything about
874 processing it & sending the BULK transfer request. This
875 should help move old stale pending locks off the queue as
876 quickly as obd_timeout.
879 Frequency : when an incorrect nid is specified during startup
881 Description: ptlrpc connect to non-existant node causes kernel crash
882 Details : LNET can't be re-entered from an event callback, which
883 happened when we expire a message after the export has been
884 cleaned up. Instead, hand the zombie cleanup off to another
887 Severity : enhancement
889 Description: plain/inodebits lock performance improvement
890 Details : Grouping plain/inodebits in granted list by their request modes
891 and bits policy, thus improving the performance of search through
895 Frequency : only if OST filesystem is corrupted
897 Description: client incorrectly hits assertion in ptlrpc_replay_req()
898 Details : for a short time RPCs with bulk IO are in the replay list,
899 but replay of bulk IOs is unimplemented. If the OST filesystem
900 is corrupted due to disk cache incoherency and then replay is
901 started it is possible to trip an assertion. Avoid putting
902 committed RPCs into the replay list at all to avoid this issue.
905 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
907 Description: System hang on startup
908 Details : This bug allowed the liblustre (e.g. catamount) client to
909 return to the app before handling all startup RPCs. This
910 could leave the node unresponsive to lustre network traffic
911 and manifested as a server ptllnd timeout.
913 Severity : enhancement
915 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
916 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
917 boolean to control whether to print peer debug info when a
918 client's RPC times out.
921 Frequency : only for kernels with patches from Lustre below 1.4.3
923 Description: Remove old rdonly API
924 Details : Remove old rdonly API which unused from at least lustre 1.4.3
927 Frequency : only for devices with external journals
929 Description: Set external device read-only also
930 Details : During a commanded failover stop, we set the disk device
931 read-only while the server shuts down. We now also set any
932 external journal device read-only at the same time.
935 Frequency : when upgrading from 1.4 while trying to change parameters
937 Description: The wrong (new) MDC name was used when setting parameters for
938 upgraded MDT's. Also allows changing of OSC (and MDC)
939 parameters if --writeconf is specified at tunefs upgrade time.
942 Frequency : when setting specific ost indicies
944 Description: QOS code breaks on skipped indicies
945 Details : Add checks for missing OST indicies in the QOS code, so OSTs
946 created with --index need not be sequential.
948 Severity : enhancement
950 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
951 Details : The uninit_groups feature works in conjunction with the kernel
952 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
953 up the pass1 processing of e2fsck. This is a read-only feature
954 in ldiskfs2 only, so older kernels and current ldiskfs cannot
955 mount filesystems that have had this feature enabled.
957 Severity : enhancement
959 Description: Improve multi-block allocation algorithm to avoid fragmentation
960 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
961 allocation locality and avoid filesystem fragmentation.
963 ------------------------------------------------------------------------------
965 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
967 * Support for kernels:
968 2.4.21-47.0.1.EL (RHEL 3)
970 2.6.9-42.0.10.EL (RHEL 4)
971 2.6.12.6 vanilla (kernel.org)
972 2.6.16.27-0.9 (SLES 10)
973 * Recommended e2fsprogs version: 1.39.cfs5
975 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
979 Frequency : occasional, depends on client load and configuration
980 Bugzilla : 12181, 12203
981 Description: data loss for recently-modified files
983 Details : In some cases it is possible that recently written or created
984 files may not be written to disk in a timely manner (this should
985 normally be within 30s unless client IO load is very high).
986 The problem appears as zero-length files or files that are a
987 multiple of 1MB in size after a client crash or client eviction
988 that are missing data at the end of the file.
990 This problem is more likely to be hit on clients where files are
991 repeatedly created and unlinked in the same directory, clients
992 have a large amount of RAM, have many CPUs, the filesystem has
993 many OSTs, the clients are rebooted frequently, and/or the files
994 are not accessed by other nodes after being written.
996 The presence of the problem can be detected by looking at
997 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
998 smaller than the second (nr_unused) then dirty files will not
999 be flushed automatically to disk. "sync; sleep 10" should be
1000 run several times on the node before unmounting it to update
1001 Lustre (this is also safe to run on nodes without this problem).
1003 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1004 that can cause this same problem, so customers using that kernel
1005 also need to update the kernel in addition to Lustre. In order
1006 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1008 It is normal that files written just before a client crash (less
1009 than 30s) may not yet have been flushed to disk, even for local
1013 Frequency : frequent on thin XT3 nodes
1015 Description: UUID collision on thin XT3 Linux nodes
1016 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1017 randomly, since we relied on an insufficiently-seeded PRNG.
1022 Description: OSS hangs after "All ost request buffers busy"
1023 Details : A deadlock between quota and journal operations caused OSS
1024 hangs after printing "All ost request buffers busy."
1027 Frequency : always on liblustre builds
1029 Description: Cleanup compiler warnings on liblustre
1032 Frequency : always on liblustre builds on XT3
1034 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1038 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1039 > 32000 subdirectories
1040 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1041 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1042 subdirectory count crosses 32000. This will aid e2fsck to
1043 correctly handle more than 32000 subdirectories.
1046 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1048 Description: System hang on startup
1049 Details : This bug allowed the liblustre (e.g. catamount) client to
1050 return to the app before handling all startup RPCs. This
1051 could leave the node unresponsive to lustre network traffic
1052 and manifested as a server ptllnd timeout.
1054 Severity : enhancement
1056 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1057 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1058 boolean to control whether to print peer debug info when a
1059 client's RPC times out.
1064 Description: make O_SYNC working on 2.6 kernels
1065 Details : 2.6 kernels use different method for mark pages for write,
1066 so need add a code to lustre for O_SYNC work.
1071 Description: Failure to close file and release space on NFS
1072 Details : Put inode details into lock acquired in ll_intent_file_open.
1073 Use mdc_intent_lock in ll_intent_open to properly
1074 detect all kind of errors unhandled by mdc_enqueue.
1079 Description: proc file read during shutdown sometimes raced obd removal,
1081 Details : Add lock to prevent obd access after proc file removal.
1084 Frequency : Only for files larger than 4GB on 32-bit clients.
1086 Description: improperly doing page alignment of locks
1087 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1088 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1093 Description: RPCs being resent when they shouldn't be
1094 Details : Some RPCs that should not be resent are being resent. This
1095 can cause inconsistencies in the RPC state machine. Do not
1096 resend such requests.
1099 Frequency : rare, only with NFS export
1101 Description: Crash on NFS re-export node
1102 Details : under very unusual load conditions an assertion is hit in
1103 ll_intent_file_open()
1106 Frequency : only if OST filesystem is corrupted
1108 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1109 Details : for a short time RPCs with bulk IO are in the replay list,
1110 but replay of bulk IOs is unimplemented. If the OST filesystem
1111 is corrupted due to disk cache incoherency and then replay is
1112 started it is possible to trip an assertion. Avoid putting
1113 committed RPCs into the replay list at all to avoid this issue.
1118 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1119 Details : Large single O_DIRECT read and write calls can fail to allocate
1120 a sufficiently large buffer to process the request. In case of
1121 allocation failure the allocation is retried with a smaller
1122 buffer and broken into smaller requests.
1124 Severity : enhancement
1126 Description: Add -o localflock option to simulate old noflock behaviour.
1127 Details : This will achieve local-only flock/fcntl locks coherentness.
1132 Description: versioning check is incomplete
1133 Details : Checking the version difference of client vs. server, report
1134 error if the gap is too big.
1139 Description: add support PG_writeback bit
1140 Details : add support for PG_writeback bit for Lustre, for more carefull
1141 work with page cache in 2.6 kernel. This also fix some deadlocks
1142 and remove hack for work O_SYNC with 2.6 kernel.
1144 Severity : enhancement
1146 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1147 Details : The uninit_groups feature works in conjunction with the kernel
1148 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1149 up the pass1 processing of e2fsck. This is a read-only feature
1150 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1151 mount filesystems that have had this feature enabled.
1153 Severity : enhancement
1155 Description: Improve multi-block allocation algorithm to avoid fragmentation
1156 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1157 allocation locality and avoid filesystem fragmentation.
1159 ------------------------------------------------------------------------------
1161 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1163 * Support for kernels:
1164 2.6.9-42.0.3.EL (RHEL 4)
1165 2.6.5-7.276 (SLES 9)
1166 2.4.21-47.0.1.EL (RHEL 3)
1167 2.6.12.6 vanilla (kernel.org)
1168 2.6.16.21-0.8 (SLES10)
1169 * Recommended e2fsprogs version: 1.39.cfs2-0
1171 * The backwards-compatible /proc/sys/portals symlink has been removed
1172 in this release. Before upgrading, please ensure that you change
1173 any configuration scripts or /etc/sysctl.conf files that access
1174 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1175 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1176 in advance of the upgrade on any system running Lustre 1.4.6 or
1177 newer, since /proc/sys/lnet was added in that version.
1178 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1182 Frequency : only when quota is used
1184 Description: avoid scanning export list for quota master
1185 Details : Change the algorithms to avoid scanning export list in order
1186 to improve the efficiency.
1189 Frequency : MDS failover only, very rarely
1191 Description: "went back in time" messages on mds failover
1192 Details : The greatest transno may be lost when the current operation
1193 finishes with an error (transno==0) and the client's last_rcvd
1194 record is over-written. Save the greatest transno in the
1195 mds_last_transno for this case.
1198 Frequency : always for specific kernels and striping counts
1200 Description: client may get "Matching packet too big" without ACL support
1201 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1202 when trying to access files in certain configurations. The
1203 clients should in fact be denied when mounting because they do
1204 not understand ACLs.
1207 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1209 Description: many clients connecting with IO in progress causes connect timeouts
1210 Details : Avoid synchronous journal commits to avoid delays caused by many
1211 clients connecting/disconnecting when bulk IO is in progress.
1212 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1213 OST_IO_PORTAL to avoid delays behind potentially many pending
1217 Frequency : occasionally with multiple writers to a single file
1219 Description: shared writes to file may result in wrong size reported by stat()
1220 Details : Allow growing of kms when extent lock is cancelled
1223 Frequency : always with random mmap IO to multi-striped file
1225 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1226 Details : Only if the hole is at the end of OST object so that kms is too
1227 small. Fix is to increase kms accordingly in ll_nopage.
1230 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1232 Description: writes to a missing object would leak memory on the OST
1233 Details : If there is an inconsistency between the MDS and OST filesystems,
1234 such that the MDS references an object that doesn't exist, writes
1235 to that object will leak memory due to incorrect cleanup in the
1236 error handling path, eventually running out of memory on the OST.
1241 Description: Creating too long symlink causes lustre errors
1242 Details : Check symlink and name lengths before sending requests to MDS.
1245 Frequency : only if flock is enabled (not on by default)
1247 Description: posix locks not released on fd closure on 2.6.9+
1248 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1249 kernels, this caused such locks not to be released on fd close and
1250 then assertions on fs unmount about still used locks.
1253 Frequency : MDS failover only, very rarely
1255 Description: clients may get ASSERTION(granted_lock != NULL)
1256 Details : When request was taking a long time, and a client was resending
1257 a getattr by name lock request. The were multiple lock requests
1258 with the same client lock handle and
1259 mds_getattr_name->fixup_handle_for_resent_request found one of the
1260 lock handles but later failed with ASSERTION(granted_lock != NULL).
1265 Description: handle->h_buffer_credits > 0, assertion failure
1266 Details : h_buffer_credits is zero after truncate, causing assertion
1267 failure. This patch extends the transaction or creates a new
1271 Frequency : NFS re-export or patchless client
1272 Bugzilla : 11179, 10796
1273 Description: Crash on NFS re-export node (__d_move)
1274 Details : We do not want to hash the dentry if we don't have a lock.
1275 But if this dentry is later used in d_move, we'd hit uninitialised
1276 list head d_hash, so we just do this to init d_hash field but
1277 leave dentry unhashed.
1280 Frequency : NFS re-export or patchless client
1282 Description: NFS exports has problem with symbolic link
1283 Details : lustre client didn't properly install dentry when re-exported
1284 to NFS or running patchless client.
1287 Frequency : NFS re-export or patchless client
1289 Description: Various nfs/patchless fixes.
1290 Details : fixes reuse disconected alias for lookup process - this fixes
1291 warning "find_exported_dentry: npd != pd",
1292 fix permission error with open files at nfs.
1293 fix apply umask when do revalidate.
1296 Frequency : occasional
1298 Description: Crash on NFS re-export node
1299 Details : calling clear_page() on the wrong pointer triggered oops in
1300 generic_mapping_read().
1303 Frequency : rarely, using O_DIRECT IO
1305 Description: unaligned directio crashes client with LASSERT
1306 Details : check for unaligned buffers before trying any requests.
1309 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1311 Description: stale data returned from RAID cache
1312 Details : If only a small amount of IO is done to the RAID device before
1313 reading it again it is possible to get stale data from the RAID
1314 cache instead of reading it from disk.
1317 Frequency : always for sles10 kernel
1319 Description: sles10 support
1320 Details : ll_follow_link: compile fixes and using of nd_set_link
1321 under newer kernels.
1324 Frequency : depends on arch, kernel and compiler version, always on sles10
1327 Description: recursive or deep enough symlinks cause stack overflow
1328 Details : getting rid of large stack-allocated variable in
1332 Frequency : depends on hardware
1334 Description: lustre write performance loss in the SLES10 kernel
1335 Details : the performance loss is caused by using of write barriers in the
1336 ext3 code. The SLES10 kernel turns barrier support on by
1337 default. The fix is to undo that change for ldiskfs.
1339 ------------------------------------------------------------------------------
1341 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1343 * Support for kernels:
1344 2.6.9-42.0.3EL (RHEL 4)
1345 2.6.5-7.276 (SLES 9)
1346 2.4.21-47.0.1.EL (RHEL 3)
1347 2.6.12.6 vanilla (kernel.org)
1351 Frequency : quota enabled and large files being deleted
1353 Description: releasing more than 4GB of quota at once hangs OST
1354 Details : If a user deletes more than 4GB of files on a single OST it
1355 will cause the OST to spin in an infinite loop. Release
1356 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1361 Description: statfs data retrieved from /proc may be stale or zero
1362 Details : When reading per-device statfs data from /proc, in the
1363 {kbytes,files}_{total,free,avail} files, it may appear
1364 as zero or be out of date.
1367 Frequency : systems with MD RAID1 external journal devices
1369 Description: lconf's call to blkid is confused by RAID1 journal devices
1370 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1371 of returning all block devices that match the journal UUID.
1374 Frequency : always, for aggregate stripe size over 4GB
1376 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1377 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1378 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1381 Frequency : always if "lfs find" used on a local file/directory
1383 Description: "lfs find" segfaults if used on a local file/directory
1384 Details : The case where a directory component was not specified wasn't
1385 handled correctly. Handle this properly.
1388 Frequency : always on ppc64
1390 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1391 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1392 perform endianness conversion.
1395 Frequency : rarely (truncate to non-zero file size after write under load)
1396 Bugzilla : 10730, 10687
1397 Description: Files padded with zeros to next 4K multiple
1398 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1399 it is possible that files that are truncated to a non-zero size
1400 immediately after being written are filled with zero bytes beyond
1401 the truncated size. No file data is lost.
1403 Severity : enhancement
1405 Description: Allow recovery/failover for liblustre clients.
1406 Details : liblustre clients were unaware of failover configurations until
1409 Severity : enhancement
1411 Description: user file locks should fail when not mounting with flock option
1412 Details : Set up an error-returning stub in ll_file_operations.lock field
1413 to prevent incorrect behaviour when client is mounted without
1414 flock option. Also, set up properly f_op->flock field for
1418 Frequency : always on ia64
1420 Description: "lfs df" loops on printing out MDS statfs information
1421 Details : The obd_ioctl_data was not initialized and in some systems
1422 this caused a failure during the ioctl that did not return
1423 an error. Initialize the struct and return an error on failure.
1426 Frequency : SLES 9 only
1428 Description: Error of copying files with lustre special EAs as root
1429 Details : Client side always return success for setxattr call for lustre
1430 special xattr (currently only "trusted.lov").
1433 Frequency : rarely on clusters with both ia64+i386 clients
1435 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1436 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1437 clients are doing concurrent writes to the same file it is
1438 possible that the ia64 clients may LASSERT because the OST
1439 extent locks are not PAGE_SIZE aligned. Ensure that grown
1440 locks are always aligned on the request boundary.
1443 Frequency : specific use, occasional
1445 Description: Overwriting in use executable truncates on-disk binary image
1446 Details : If one node attempts to overwrite an executable in use by
1447 another node, we now correctly return ETXTBSY instead of
1448 truncating the file.
1450 Severity : enhancement
1452 Description: Async OSC create to avoid the blocking unnecessarily.
1453 Details : If a OST has no remain object, system will block on the creating
1454 when need to create a new object on this OST. Now, ways use
1455 pre-created objects when available, instead of blocking on an
1456 empty osc while others are not empty. If we must block, we block
1457 for the shortest possible period of time.
1462 Description: chmod on Lustre root is propagated to other clients
1463 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1464 invalid by the follow_mount time.
1467 Frequency : liblustre clients only
1469 Description: Race in 'instant cancel' lock handling could lead to such locks
1470 never to be granted in case of SMP MDS
1471 Details : Do not destroy not yet granted but cbpending locks in
1475 Frequency : replay/resend of open
1477 Description: non null lock assetion failure in mds_intent_policy
1478 Details : Trying to replay/resend lockless open requests resulted in
1479 mds_open() returning 0 with no lock. Now it sets a flag if
1480 a lock is going to be returned.
1482 Severity : enhancement
1484 Description: Checksum enhancements
1485 Details : New checksum enhancements allow for resending RPCs that failed
1488 Severity : enhancement
1490 Description: Tunables on number of dirty pages in cacche
1491 Details : Allow to set limit on number of dirty pages cached.
1496 Description: client crash on unmount - lock still has references
1497 Details : In some error handling cases it was possible to leak a lock
1498 reference on a client while accessing a file. This was not
1499 harmful to the client during operation, but would cause the
1500 client to crash when the filesystem is unmounted.
1503 Frequency : specific case, rare
1505 Description: ETXTBSY on mds though file not in use by client
1506 Details : ETXTBSY is no longer incorrectly returned when attempting to
1507 chmod or chown a directory that the user previously tried to
1508 execute or a currently-executing binary.
1511 Frequency : extremely rare except on liblustre-based clients
1513 Description: Lustre space not freed when files are deleted
1514 Details : Clean up open-unlinked files after client eviction. Previously
1515 the unlink was skipped and the files remained as orphans.
1520 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1521 Details : In some cases it was possible to send a blocking callback to
1522 a client doing a glimpse, even though that client didn't get
1523 a lock granted. When the glimpse lock is cancelled on the OST
1524 the freed lock is left on the waiting list and corrupted the list.
1527 Frequency : all core dumps
1529 Description: Broke core dumps to lustre
1530 Details : Negative dentry may be unhashed if parent does not have UPDATE
1531 lock, but some callers, e.g. do_coredump, expect dentry to be
1532 hashed after successful create, hash it in ll_create_it.
1534 ------------------------------------------------------------------------------
1536 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1538 * Support for kernels:
1539 2.6.9-42.0.2.EL (RHEL 4)
1540 2.6.5-7.276 (SLES 9)
1541 2.4.21-40.EL (RHEL 3)
1542 2.6.12.6 vanilla (kernel.org)
1546 Frequency : always on RHEL 3
1548 Description: Number of open files grows over time
1549 Details : The number of open files grows over time, whether or not
1550 Lustre is started. This was due to a filp leak introduced
1551 by one of our kernel patches.
1553 ------------------------------------------------------------------------------
1555 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1557 * Support for kernels:
1558 2.6.9-42.EL (RHEL 4)
1559 2.6.5-7.276 (SLES 9)
1560 2.4.21-40.EL (RHEL 3)
1561 2.6.12.6 vanilla (kernel.org)
1566 Bugzilla : 5719, 9635, 9792, 9684
1567 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1568 Details : If a server is under heavy load and cannot reply to new
1569 connection requests before the client resends the (re)connect,
1570 the connection handling code can behave badly if two service
1571 threads are concurrently handing separate (re)connections from
1572 the same client. Add better locking to the connection handling
1573 code, and ensure that only a single connection will be processed
1574 for a given client UUID, even if the lock is dropped.
1576 Severity : enhancement
1578 Description: add TCP zero-copy support to kernel
1579 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1580 sends if the hardware supports scatter-gather and checksumming.
1581 This allows socklnd to do client-write and server-read more
1582 efficiently and reduce CPU utilization from skbuf copying.
1585 Frequency : only if NFS exporting from client
1587 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1588 Details : If setting attributes on a file created under NFS that had
1589 never been opened it would be possible to oops the client
1590 if the file had no objects.
1593 Frequency : always for liblustre
1595 Description: liblustre client does MDS+OSTs setattr RPC for each write
1596 Details : When doing a write from a liblustre client, the client
1597 incorrectly issued an RPC to the MDS and each OST the file was
1598 striped over in order to update the timestamps. When writing
1599 with small chunks and many clients this could overwhelm the MDS
1600 with RPCs. In all cases it would slow down the write because
1601 these RPCs are unnecessary.
1603 Severity : enhancement
1605 Description: allow number of MDS service threads to be changed at module load
1606 Details : It is now possible to change the number of MDS service threads
1607 running. Adding "options mds mds_num_threads={N}" to the MDS's
1608 /etc/modprobe.conf will set the number of threads for the next
1609 time Lustre is restarted (assuming the "mds" module is also
1610 reloaded at that time). The default number of threads will
1611 stay the same, 32 for most systems.
1616 Description: OST crash if filesystem is unformatted or corrupt
1617 Details : If an OST is started on a device that has never been formatted
1618 or if the filesystem is corrupt and cannot even mount then the
1619 error handling cleanup routines would dereference a NULL pointer.
1624 Description: NULL pointer deref in llap_from_page.
1625 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1626 incorrect) mapping if the page was truncated/reclaimed while it was
1627 searched for. Check for this condition and skip such pages when
1628 doing readahead. Introduce extra check to llap_from_page() to
1629 verify page->mapping->host is non-NULL (so page is not anonymous).
1632 Frequency : Sometimes when using sys_sendfile
1634 Description: "page not covered by a lock" warnings from ll_readpage
1635 Details : sendfile called ll_readpage without right page locks present.
1636 Now we introduced ll_file_sendfile that does necessary locking
1637 around call to generic_file_sendfile() much like we do in
1641 Frequency : with certain MDS communication failures at client mount time
1643 Description: NULL pointer deref after failed client mount
1644 Details : a client connection request may delayed by the network layer
1645 and not be sent until after the PTLRPC layer has timed out the
1646 request. If the client fails the mount immediately it will try
1647 to clean up before the network times out the request. Add a
1648 reference from the request import to the obd device and delay
1649 the cleanup until the network drops the request.
1652 Frequency : occasionally during client (re)connect
1654 Description: assertion failure during client (re)connect
1655 Details : processing a client connection request may be delayed by the
1656 client or server longer than the client connect timeout. This
1657 causes the client to resend the connection request. If the
1658 original connection request is replied in this interval, the
1659 client may trip an assertion failure in ptlrpc_connect_interpret()
1660 which thought it would be the only running connect process.
1663 Frequency : only with obd_echo servers and clients that are rebooted
1665 Description: kernel BUG accessing uninitialized data structure
1666 Details : When running an obd_echo server it did not start the ping_evictor
1667 thread, and when a client was evicted an uninitialized data
1668 structure was accessed. Start the ping_evictor in the RPC
1669 service startup instead of the OBD startup.
1671 Severity : enhancement
1672 Bugzilla : 10193 (patchless)
1673 Description: Remove dependency on various unexported kernel interfaces.
1674 Details : No longer need reparent_to_init, exit_mm, exit_files,
1675 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1678 Frequency : rare (only users of deprecated and unsupported LDAP config)
1680 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1681 Details : LDAP apparently contains 'lustreName' attributes instead of
1682 'name'. A simple remapping of the name is sufficient.
1685 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1687 Description: waiting_locks_callback trips kernel BUG if client is evicted
1688 Details : Running with the dump_on_timeout debug flag turned on makes
1689 it possible that the waiting_locks_callback() can try to dump
1690 the Lustre kernel debug logs from an interrupt handler. Defer
1691 this log dumping to the expired_lock_main() thread.
1693 Severity : enhancement
1695 Description: Support NFS exporting on 2.6 kernels.
1696 Details : Implement non-rawops metadata methods for NFS server to use without
1697 changing NFS server code.
1700 Frequency : very rare (synthetic metadata workload only)
1702 Description: two racing renames might cause an MDS thread to deadlock
1703 Details : Running the "racer" program may cause one MDS thread to rename
1704 a file from being the source of a rename to being the target of
1705 a rename at exactly the same time that another thread is doing
1706 so, and the second thread has already enqueued these locks after
1707 doing a lookup of the target and is trying to relock them in
1708 order. Ensure that we don't try to re-lock the same resource.
1711 Frequency : only very large systems with liblustre clients
1713 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1714 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1715 all OSTs in parallel. This allows the eviction of stale liblustre
1716 clients to proceed much faster than if they were done in series,
1717 and also offers similar improvements for other set_info RPCs.
1722 Description: excessive CPU usage during initial read phase on client
1723 Details : During the initial read phase on a client, it would agressively
1724 retry readahead on the file, consuming too much CPU and impacting
1725 performance (since 1.4.5.8). Improve the readahead algorithm
1726 to avoid this, and also improve some other common cases (read
1727 of small files in particular, where "small" is files smaller than
1728 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1733 Description: MDS crash when receiving packet with unknown intent.
1734 Details : Do not LBUG in unknown intent case, just return -EFAULT
1736 Severity : enhancement
1737 Bugzilla : 9293, 9385
1738 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1739 Details : Do not serialize getattr (non-intent version) and statfs.
1742 Frequency : occasional, when OST network is overloaded/intermittent
1744 Description: client evicted by OST after bulk IO timeout
1745 Details : If a client sends a bulk IO request (read or write) the OST
1746 may evict the client if it is unresposive to its data GET/PUT
1747 request. This is incorrect if the network is overloaded (takes
1748 too long to transfer the RPC data) or dropped the OST GET/PUT
1749 request. There is no need to evict the client at all, since
1750 the pinger and/or lock callbacks will handle this, and the
1751 client can restart the bulk request.
1754 Frequency : Always when mmapping file with no objects
1756 Description: client crashes when mmapping file with no objects
1757 Details : Check that we actually have objects in a file before doing any
1758 operations on objects in ll_vm_open, ll_vm_close and
1764 Description: Request leak when working with deleted CWD
1765 Details : Introduce advanced request refcount tracking for requests
1766 referenced from lustre intent.
1768 Severity : Enhancement
1770 Description: Cache open file handles on client.
1771 Details : MDS now will return special lock along with openhandle, if
1772 requested and client is allowed to hold openhandle, even if unused,
1773 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1774 closing files for every read/write openration.
1776 Severity : Enhancement
1778 Description: Cache open negative dentries on client when possible.
1779 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1780 negative dentries on lock revocation.
1785 Description: Remounting a client read-only wasn't possible with a zconf mount
1786 Details : It wasn't possible to remount a client read-only with llmount.
1788 Severity : enhancement
1789 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1790 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1791 MPICH is included to add Lustre-specific ADIO interfaces.
1792 This is based closely on the UFS ADIO layer and only differs
1793 in file creation, in order to allow the OST striping to be set.
1794 This is user-contributed code and not supported by CFS.
1799 Description: extended inode attributes (immutable, append-only) work improperly
1800 when 2.4 and 2.6 kernels are used on client/server or vice versa
1801 Details : Introduce kernel-independent values for these flags.
1803 Severity : enhancement
1806 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1807 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1808 granularity. Now, Fractional MB granularity is be supported,
1809 this is very useful for low memory system.
1811 Severity : enhancement
1813 Description: Getattr by fid
1814 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1815 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1816 have locks and avoids getattr rpc requests for it on every stat.
1819 Frequency : Always, for filesystems larger than 2TB
1821 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1822 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1823 filesystems larger than 2TB to be mounted with mballoc enabled.
1826 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1828 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1829 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1830 filesystems larger than 2TB is mounted then the kernel will
1831 silently corrupt the start of the filesystem. CONFIG_LBD is
1832 enabled for all CFS-supported kernels, but the possibility of
1833 this happening with a modified kernel config exists.
1835 Severity : enhancement
1837 Description: add client O_DIRECT support for 2.6 kernels
1838 Details : It is now possible to do O_DIRECT reads and writes to files
1839 in the Lustre client mountpoint on 2.6 kernel clients.
1841 Severity : enhancement
1843 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1844 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1845 parallel, not waiting for response from every OST before sending
1846 a rpc to the next OST.
1851 Description: setattr vs write race when updating file timestamps
1852 Details : Client processes that update a file timestamp into the past
1853 right after writing to the file (e.g. tar) it is possible that
1854 the updated file modification time can be reset to the current
1855 time due to a race between processing the setattr and write RPC.
1857 Severity : enhancement
1859 Description: Bring 'lfs find' closer in line with regular Linux find.
1860 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1861 -print0 options and obtains all the needed info through the lustre
1864 Severity : enhancement
1866 Description: support up to 1024 configured devices on one node
1867 Details : change obd_dev array from statically allocated to dynamically
1868 allocated structs as they are first used to reduce memory usage
1873 Description: Flush dirty partially truncated pages during truncate
1874 Details : Immediatelly flush partially truncated pages in filter_setattr,
1875 this way we completely avoid having any pages in page cache on OST
1876 and can retire ugly workarounds during writes to flush such pages.
1881 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1882 Details : protect inode from truncation within vfs_unlink() context
1883 just take a reference before calling vfs_unlink() and release it
1884 when parent's i_sem is free.
1889 Description: last_id value checked outside lock on OST caused LASSERT failure
1890 Details : If there were multiple MDS->OST object precreate requests in
1891 flight, it was possible that the OST's last object id was checked
1892 outside a lock and incorrectly tripped an assertion. Move checks
1893 inside locks, and discard old precreate requests.
1896 Frequency : always, if extents are used on OSTs
1898 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1899 Details : OSTs using the extents format would not zero the high 16 bits of
1900 the index physical block number. This is not a problem for any
1901 OST filesystems smaller than 16TB, and no kernels support ext3
1902 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1903 new/modified files) and can be fixed for existing filesystems
1904 with e2fsprogs-1.39-cfs1.
1909 Description: import connection selection may be incorrect if timer wraps
1910 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1911 import connections to be ignored if the 32-bit jiffies counter
1912 wraps. Use a 64-bit jiffies counter.
1915 Frequency : very large clusters immediately after boot
1917 Description: LNET request buffers exhausted under heavy short-term load
1918 Details : If a large number of client requests are generated on a service
1919 that has previously never seen so many requests it is possible
1920 that the request buffer growth cannot keep up with the spike in
1921 demand. Instead of dropping incoming requests, they are held in
1922 the LND until the RPC service can accept more requests.
1925 Frequency : Sometimes during replay
1927 Description: Assertion failure in ll_local_open after replay.
1928 Details : If replay happened on an open request reply before we were able
1929 to set replay handler, reply will become not swabbed tripping the
1930 assertion in ll_local_open. Now we set the handler right after
1931 recognising of open request
1934 Frequency : very rare
1936 Description: kernel reports "badness in vsnprintf"
1937 Details : Reading from the "recovery_status" /proc file in small chunks
1938 may cause a negative length in lprocfs_obd_rd_recovery_status()
1939 call to vsnprintf() (which is otherwise harmless). Exit early
1940 if there is no more space in the output buffer.
1942 Severity : enhancement
1944 Description: clear OBD RPC statistics by writing to them
1945 Details : It is now possible to clear the OBD RPC statistics by writing
1946 to the "stats" file.
1951 Description: Client mtime is not the same on different clients after utimes
1952 Details : In some cases, the client was using the utimes() syscall on
1953 a file cached on another node. The clients now validate the
1954 ctime from the MDS + OSTs to determine which one is right.
1959 Description: Inability to activate failout mode
1960 Details : lconf script incorrectly assumed that in pythong string's numeric
1961 value is used in comparisons.
1964 Frequency : always with multiple stripes per file
1966 Description: Inefficient object allocation for mutli-stripe files
1967 Details : When selecting which OSTs to stripe files over, for files with
1968 a stripe count that divides evenly into the number of OSTs,
1969 the MDS is always picking the same starting OST for each file.
1970 Return the OST selection heuristic to the original design.
1975 Description: mount failures may take full timeout to return an error
1976 Details : Under some heavy load conditions it is possible that a
1977 failed mount can wait for the full obd_timeout interval,
1978 possibly several minutes, before reporting an error.
1979 Instead return an error as soon as the status is known.
1981 Frequency : quota enabled and large files being deleted
1983 Description: releasing more than 4GB of quota at once hangs OST
1984 Details : If a user deletes more than 4GB of files on a single OST it
1985 will cause the OST to spin in an infinite loop. Release
1986 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1991 Description: statfs data retrieved from /proc may be stale or zero
1992 Details : When reading per-device statfs data from /proc, in the
1993 {kbytes,files}_{total,free,avail} files, it may appear
1994 as zero or be out of date.
1997 Frequency : systems with MD RAID1 external journal devices
1999 Description: lconf's call to blkid is confused by RAID1 journal devices
2000 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2001 of returning all block devices that match the journal UUID.
2004 Frequency : always, for aggregate stripe size over 4GB
2006 Description: assertion fails when trying to use 4GB stripe size
2007 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2008 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2011 Frequency : always on ppc64
2013 Description: the first write on an ext3 filesystem with mballoc got stuck
2014 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2015 perform endianness conversion.
2017 ------------------------------------------------------------------------------
2019 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2021 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2022 INTEROPERATE with older versions automatically. Please read the
2023 user documentation before upgrading any part of a live system.
2024 * WARNING: Lustre networking configuration changes are required with
2025 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2028 * Support for kernels:
2029 2.6.9-22.0.2.EL (RHEL 4)
2030 2.6.5-7.244 (SLES 9)
2031 2.6.12.6 vanilla (kernel.org)
2034 Severity : enhancement
2035 Bugzilla : 7981/8208
2036 Description: Introduced Lustre Networking (LNET)
2037 Details : LNET is new networking infrastructure for Lustre, it includes
2038 a reorganized network configuration mode (see the user
2039 documentation for full details) as well as support for routing
2040 between different network fabrics. Lustre Networking Devices
2041 (LNDs) for the supported network fabrics have also been
2042 created for this new infrastructure.
2044 Severity : enhancement
2045 Description: Introduced Access control lists
2046 Details : clients can set ACLs on files and directories in order to have
2047 more fine-grained permissions than the standard Unix UGO+RWX.
2048 The MDS must be started with the "-o acl" mount option.
2050 Severity : enhancement
2051 Description: Introduced filesystem quotas
2052 Details : Administrators may now establish per-user quotas on the
2055 Severity : enhancement
2057 Description: Configuration change for the XT3
2058 The PTLLND is now used to run Lustre over Portals on the XT3
2059 The configure option(s) --with-cray-portals are no longer used.
2060 Rather --with-portals=<path-to-portals-includes> is used to
2061 enable building on the XT3. In addition to enable XT3 specific
2062 features the option --enable-cray-xt3 must be used.
2067 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2068 Details : A race between allocating a new llcd and re-getting the llcd_lock
2069 allowed another thread to grab newly-allocated llcd.
2071 Severity : enhancement
2073 Description: 2.6 OST async journal commit and locking fix to improve performance
2074 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2075 2.6 kernels are now async as they already were in 2.4 kernels so
2076 that they can commit concurrently with the network bulk transfer.
2077 For block-allocated files the filter allocation semaphore is held
2078 to avoid filesystem fragmentation during allocation. BKL lock
2079 removed for 2.6 xattr operations where it is no longer needed.
2084 Description: lconf incorrectly determined whether two IP networks could talk
2085 Details : In some more complicated routing and multiple-network
2086 configurations, lconf will avoid trying to make a network
2087 connection to a disjoint part of the IP space. It was doing the
2088 math incorrectly for one set of cases.
2093 Description: Fix for potential infinite loop processing records in an llog.
2094 Details : If an llog record is corrupted/zeroed, it is possible to loop
2095 forever in llog_process(). Validate the llog record length
2096 and skip the remainder of the block on error.
2099 Frequency : occasional (liblustre only)
2101 Description: liblustre could not open files whose last component is a symlink
2102 Details : sysio_path_walk() would incorrectly pass the open intent to
2103 intermediate path components.
2106 Frequency : rare (liblustre only with non-standard tuning)
2107 Bugzilla : 7201 (7350)
2108 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2109 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2110 after it has been released, only do this for referenced locks
2112 Severity : enhancement
2114 Description: specifying an (invalid) directory default stripe_size of -1
2115 would reset the directory default striping
2116 Details : stripe_size -1 was used internally to signal directory stripe
2117 removal, now use "all default" to signal dir stripe removal
2118 as a directory striping of "all default" is not useful
2121 Frequency : common for large clusters running liblustre clients
2123 Description: doing an ls when liblustre clients are running is slow
2124 Details : sending a glimpse AST to a liblustre client waits for every AST
2125 to time out, as liblustre clients will not respond. Since they
2126 cannot cache data we refresh the OST lock LVB from disk instead.
2128 Severity : enhancement
2130 Description: doing an ls at the same time as file IO can be slow
2131 Details : enqueue and other "small" requests can be blocked behind many
2132 large IO requests. Create a new OST IO portal for non-IO
2133 requests so they can be processed faster.
2136 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2138 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2139 created file with mode 000
2140 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2141 allow an ftruncate() as a non-root user to a file with mode 000.
2142 Lustre now respects this flag to disable mode checks when
2143 truncating a file owned by the user
2146 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2149 Description: Revoking locks from clients that went dead or catatonic might take
2151 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2152 cancellation of such locks instant on servers without waiting for
2153 any reply from clients. Clients drops these locks when cancel
2154 notification from server is received without replying.
2157 Frequency : liblustre-only, when liblustre client dies or becomes busy
2159 Description: Doing ls on Linux clients can take a long time with active
2161 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2162 granting such locks to it in the first place if possible. Locks
2163 are taken by proxy on the OST during the read or write and
2164 dropped immediately afterward. Add connect flags handling, do
2165 not grant locks to liblustre clients for glimpse ASTs.
2167 Severity : enhancement
2169 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2170 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2171 sequential reads and behaves badly with applications doing random
2172 reads. Improve it by reading ahead at least read region, and
2173 avoiding excessive large RPC for small reads.
2175 Severity : enhancement
2177 Description: Creating more than 1000 files for a single job may cause a load
2178 imbalance on the OSTs if there are also a large number of OSTs.
2179 Details : qos_prep_create() uses an OST index reseed value that is an
2180 even multiple of the number of available OSTs so that if the
2181 reseed happens in the middle of the object allocation it will
2182 still utilize the OSTs as uniformly as possible.
2187 Description: OST or MDS may oops in ping_evictor_main()
2188 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2189 but doesn't restart at beginning of obd_exports_timed list
2192 Severity : enhancement
2194 Description: improve by-nid export eviction on the MDS and OST
2195 Details : allow multiple exports with the same NID to be evicted at one
2196 time without re-searching the exports list.
2199 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2201 Description: MDS may oops in groups_free()
2202 Details : in rare race conditions a newly allocated group_info struct is
2203 freed again, and this can be NULL. The 2.4 compatibility code
2204 for groups_free() checked for a NULL pointer, but 2.6 did not.
2207 Frequency : common for liblustre clients doing little filesystem IO
2208 Bugzilla : 9352, 7313
2209 Description: server may evict liblustre clients accessing contended locks
2210 Details : if a client is granted a lock or receives a completion AST
2211 with a blocking AST already set it would not reply to the AST
2212 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2215 Frequency : lfs setstripe, only systems with more than 160 OSTs
2217 Description: unable to set striping with a starting offset beyond OST 160
2218 Details : llapi_create_file() incorrectly limited the starting stripe
2219 index to the maximum single-file stripe count.
2222 Frequency : LDAP users only
2224 Description: lconf did not handle in-kernel recovery with LDAP properly
2225 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2227 Severity : enhancement
2229 Description: bind OST threads to NUMA nodes to improve performance
2230 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2231 node and do their allocations there to localize memory access
2233 Severity : enhancement
2235 Description: llmount can determine client NID directly from Myrinet (GM)
2236 Details : the client NID code from gmnalnid was moved directly into
2237 llmount, removing the need to use this or specifying the
2238 client NID explicitly when mounting GM clients with zeroconf
2241 Frequency : if client is started with down MDS
2243 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2244 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2245 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2246 ptlrpc_import_delayed_req() because we didn't check if the
2247 request was interrupted, and we also didn't break out of the
2248 event loop if there was no timeout
2253 Description: data loss during non-page-aligned writes to a single file from
2254 both multiple nodes and multiple threads on one node at same time
2255 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2256 inconsistency led to false short-reads, that were cached and later
2257 used by ->prepare_write() to fill in partially written page,
2258 leading to data loss.
2261 Frequency : always, if lconf --abort_recovery used
2263 Description: lconf --abort_recovery fails with 'Operation not supported'
2264 Details : lconf was attempting to abort recovery on the MDT device and not
2267 Severity : enhancement
2269 Description: remove cleanup logs
2270 Details : replace lconf-generated cleanup logs with lustre internal
2271 cleanup routines. Eliminates the need for client-cleanup and
2274 Severity : enhancement
2276 Description: add support for EAs (user and system) on lustre filesystems
2277 Details : it is now possible to store extended attributes in the Lustre
2278 client filesystem, and with the user_xattr mount option it
2279 is possible to allow users to store EAs on their files also
2281 Severity : enhancement
2283 Description: Add possibility (config option) to show minimal available OST free
2285 Details : When compiled with --enable-mindf configure option, statfs(2)
2286 (and so, df) will return least minimal free space available from
2287 all OSTs as amount of free space on FS, instead of summary of
2288 free spaces of all OSTs.
2290 Severity : enhancement
2292 Description: do not expand extent locks acquired on OST-side
2293 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2294 by server: they are not cached anyway.
2297 Frequency : when mmap is used/binaries executed from Lustre
2299 Description: Unmmap pages before throwing them away from read cache.
2300 Details : llap_shrink cache now attempts to unmap pages before discarding
2301 them (if unmapping failed - do not discard). SLES9 kernel has
2302 extra checks that trigger if this unmapping is not done first.
2307 Description: lconf didn't resolve symlinks before checking to see whether a
2308 given mountpoint was already in use
2311 Frequency : when migrating failover services
2312 Bugzilla : 6395, 9514
2313 Description: When migrating a subset of services from a node (e.g. failback
2314 from a failover service node) the remaining services would
2315 time out and evict clients.
2316 Details : lconf --force (implied by --failover) sets the global obd_timeout
2317 to 5 seconds in order to quickly disconnect, but this caused
2318 other RPCs to time out too quickly. Do not change the global
2319 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2321 Severity : enhancement
2322 Frequency : if MDS is started with down OST
2323 Bugzilla : 9439,5706
2324 Description: Allow startup/shutdown of an MDS without depending on the
2325 availability of the OSTs.
2326 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2327 Add appropriate locking and lov-osc refcounts for safe
2328 cleaning. Add osc abort_inflight calls in case the
2329 synchronize never started.
2332 Frequency : occasional (Cray XT3 only)
2334 Description: root not authorized to access files in CRAY_PORTALS environment
2335 Details : The client process capabilities were not honoured on the MDS in
2336 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2337 been accessed by an authorized user then root was able to access
2338 the file on the local client also. The root user capabilities
2339 are now allowed on the MDS, as this environment has secure UID.
2342 Frequency : occasional
2344 Description: ldiskfs "too long searching" message happens too often
2345 Details : A debugging message (otherwise harmless) prints too often on
2346 the OST console. This has been reduced to only happen when
2347 there are fragmentation problems on the filesystem.
2352 Description: Division by zero in statfs when all OSCs are inactive
2353 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2354 lov_statfs divides by value returned by lov_get_stripecnt().
2358 Bugzilla : 9489, 3273
2359 Description: First write from each client to each OST was only 4kB in size,
2360 to initialize client writeback cache, which caused sub-optimal
2361 RPCs and poor layout on disk for the first writen file.
2362 Details : Clients now request an initial cache grant at (re)connect time
2363 and so that they can start streaming writes to the cache right
2364 away and always do full-sized RPCs if there is enough data.
2365 If the OST is rebooted the client also re-establishes its grant
2366 so that client cached writes will be honoured under the grant.
2371 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2372 Details : Now I/O RPCs go to different portal number and (presumably) fast
2373 lock requests (and glimses) and other RPCs get their own service
2374 threads pool that should be able to service those RPCs
2377 Severity : enhancement
2379 Description: Ability to exchange lustre version between client and servers and
2380 issue warnings at client side if client is too old. Also for
2381 liblustre clients there is ability to refuse connection of too old
2383 Details : New 'version' field is added to connect data structure that is
2384 filled with version info. That info is later checked by server and
2388 Frequency : rare, liblustre only.
2389 Bugzilla : 9296, 9581
2390 Description: Two simultaneous writes from liblustre at offset within same page
2391 might proceed at the same time overwriting eachother with stale
2393 Details : I/O lock withing llu_file_prwv was released too early, before data
2394 actually was hitting the wire. Extended lock-holding time until
2395 server acknowledges receiving data.
2398 Frequency : extremely rare. Never observed in practice.
2400 Description: avoid generating lustre_handle cookie of 0.
2401 Details : class_handle_hash() generates handle cookies by incrementing
2402 global counter, and can hit 0 occasionaly (this is unlikely, but
2403 not impossible, because initial value of cookie counter is
2404 selected randonly). Value of 0 is used as a sentinel meaning
2405 "unassigned handle" --- avoid it. Also coalesce two critical
2406 sections in this function into one.
2408 Severity : enhancement
2410 Description: allow liblustre clients to delegate truncate locking to OST
2411 Details : To avoid overhead of locking, liblustre client instructs OST to
2412 take extent lock in ost_punch() on client's behalf. New connection
2413 flag is added to handle backward compatibility.
2415 Severity : enhancement
2416 Bugzilla : 4928, 7341, 9758
2417 Description: allow number of OST service threads to be specified
2418 Details : a module parameter allows the number of OST service threads
2419 to be specified via "options ost ost_num_threads={N}" in the
2420 OSS's /etc/modules.conf or /etc/modprobe.conf.
2424 Bugzilla : 6146, 9635, 9895
2425 Description: servers crash with bad pointer in target_handle_connect()
2426 Details : In rare cases when a client is reconnecting it was possible that
2427 the connection request was the last reference for that export.
2428 We would temporarily drop the export reference and get a new
2429 one, but this may have been the last reference and the export
2430 was just destroyed. Get new reference before dropping old one.
2432 Severity : enhancement
2433 Frequency : if client is started with failover MDS
2435 Description: Allow multiple MDS hostnames in the mount command
2436 Details : Try to read the configuration from all specified MDS
2437 hostnames during a client mount in case the "primary"
2440 Severity : enhancement
2442 Description: Stop sending data to evicted clients as soon as possible.
2443 Details : Check if the client we are about to send or are sending data to
2444 was evicted already. (Check is done every second of waiting,
2445 for which l_wait_event interface was extended to allow checking
2446 of exit condition at specified intervals).
2449 Frequency : rare, normally only when NFS exporting is done from client
2451 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2453 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2456 Frequency : always, if srand() is called before liblustre initialization
2458 Description: Liblustre uses system PRNG disturbing its usage by user application
2459 Details : Introduce internal to lustre fast and high-quality PRNG for
2460 lustre usage and make liblustre and some other places in generic
2461 lustre code to use it.
2463 Severity : enhancement
2464 Bugzilla : 9477, 9557, 9870
2465 Description: Verify that the MDS configuration logs are updated when xml is
2466 Details : Check if the .xml configuration logs are newer than the config
2467 logs stored on the MDS and report an error if this is the case.
2468 Request --write-conf, or allow starting with --old_conf.
2470 Severity : enhancement
2472 Description: Handle symlinks in the path when checking if Lustre is mounted.
2473 Details : Resolve intermediate symlinks when checking if a client has
2474 mounted a filesystem to avoid duplicate client mounts.
2479 Description: lconf can hit an error exception but still return success.
2480 Details : The lconf command catches the Command error exception at the top
2481 level script context and will exit with the associated exit
2482 status, but doesn't ensure that this exit status is non-zero.
2487 Description: failure of ptlrpc thread startup can cause oops
2488 Details : Starting a ptlrpc service thread can fail if there are a large
2489 number of threads or the server memory is very fragmented.
2490 Handle this without oopsing.
2493 Frequency : always, only if liblustre and non-default acceptor port was used
2495 Description: liblustre cannot connect to servers with non-default acceptor port
2496 Details : tcpnal_set_default_params() was not called and was therefore
2497 ignoring the environment varaible TCPNAL_PORT, as well as other
2498 TCPNAL_ environment variables
2503 Description: two objects could be created on the same OST for a single file
2504 Details : If an OST is down, in some cases it was possible to create two
2505 objects on a single OST for a single file. No problems other
2506 than potential performance impact and spurious error messages.
2510 Bugzilla : 5681, 9562
2511 Description: Client may oops in ll_unhash_aliases
2512 Details : Client dcache may become inconsistent in race condition.
2513 In some cases "getcwd" can fail if the current directory is
2519 Description: Inode refcounting problems in NFS export code
2520 Details : link_raw functions used to call d_instantiate without obtaining
2521 extra inode reference first.
2525 Bugzilla : 9942, 9903
2526 Description: Referencing freed requests leading to crash, memleaks with NFS.
2527 Details : We used to require that call to ll_revalidate_it was always
2528 followed by ll_lookup_it. Also with revalidate_special() it is
2529 possible to call ll_revalidate_it() twice for the same dentry
2530 even if first occurence returned success. This fix changes semantic
2531 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2532 reference on a request referred from the intent.
2533 ll_intent_release() then releases such a request.
2536 Frequency : rare, normally benchmark loads only
2538 Description: unlinked inodes were kept in memory on the client
2539 Details : If a client is repeatedly creating and unlinking files it
2540 can accumulate a lot of stale inodes in the inode slab cache.
2541 If there is no other client load running this can cause the
2542 client node to run out of memory. Instead flush old inodes
2543 from client cache that have the same inode number as a new inode.
2546 Frequency : SLES9 2.6.5 kernel and long filenames only
2547 Bugzilla : 9969, 10379
2548 Description: utime reports stale NFS file handle
2549 Details : SLES9 uses out-of-dentry names in some cases, which confused
2550 the lustre dentry revalidation. Change it to always use the
2554 Frequency : rare, unless heavy write-truncate concurrency is continuous
2555 Bugzilla : 4180, 6984, 7171, 9963, 9331
2556 Description: OST becomes very slow and/or deadlocked during object unlink
2557 Details : filter_destroy() was holding onto the parent directory lock
2558 while truncating+unlinking objects. For very large objects this
2559 may block other threads for a long time and slow overall OST
2560 responsiveness. It may also be possible to get a lock ordering
2561 deadlock in this case, or run out of journal credits because of
2562 the combined truncate+unlink. Solution is to do object truncate
2563 first in one transaction without parent lock, and then do the
2564 final unlink in a new transaction with the parent lock. This
2565 reduces the lock hold time dramatically.
2568 Frequency : rare, 2.4 kernels only
2570 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2571 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2572 before unmounting their filesystem in order to avoid deadlock.
2573 The kernel_locked() function in 2.4 kernels only checks whether
2574 the kernel lock is held, not whether it is this process that is
2575 holding it as 2.6 kernels do.
2580 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2581 Details : The client ptlrpc code may be trying to reconnect to a down
2582 server before a previous connection attempt has timed out.
2583 Increase the reconnect interval to be longer than the connection
2584 timeout interval to avoid sending duplicate connections to
2588 Frequency : echo_client brw_test command
2590 Description: fix echo_client to work with OST preallocated code
2591 Details : OST preallocation code (5137) didn't take echo_client IO path
2592 into account: echo_client calls filter methods outside of any
2593 OST thread and, hence, there is no per-thread preallocated
2594 pages and buffers to use. Solution: hijack pga pages for IO. As
2595 a byproduct, this avoids unnecessary data copying.
2599 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2600 Description: Client can oops in mdc_commit_close() after open replay
2601 Details : It was possible for the MDS to return an open request with no
2602 transaction number in mds_finish_transno() if the client was
2603 evicted, but without actually returning an error. Clients
2604 would later try to replay that open and may trip an assertion
2605 Simplify the client close codepath, and always return an error
2606 from the MDS in case the open is not successful.
2609 Frequency : rare, 2.6 OSTs only
2611 Description: OST may deadlock under high load on fragmented files
2612 Details : If there was a heavy load and highly-fragmented OST filesystems
2613 it was possible to have all the OST threads deadlock waiting on
2614 allocation of biovecs, because the biovecs were not released
2615 until the entire RPC IO was completed. Instead, release biovecs
2616 as soon as they are complete to ensure forward IO progress.
2618 Severity : enhancement
2620 Description: Support for specifying external journal device at mount
2621 Details : If an OST or MDS device is formatted with an external journal
2622 device, this device major/minor is stored in the ext3 superblock
2623 and may not be valid for failover. Allow detecting and
2624 specifying the external journal at mount time.
2629 Description: Mounting an MDS with pending unlinked files may cause oops
2630 Details : target_finish_recovery() calls mds_postrecov() which returned
2631 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2632 considers this an error and immediately begins cleaning up the
2633 lov, just after starting the mds_lov process
2635 Severity : enhancement
2637 Description: Implement 'lfs df' to report actual free space on per-OST basis
2638 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2639 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2640 usage of objects; '-h' to report in human readable format.
2642 ------------------------------------------------------------------------------
2644 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2651 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2652 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2653 at startup during orphan unlinking. Instead, skip buddy update
2654 before setup, as it will be regenerated after recovery is complete.
2659 Description: If an OST is inactive, its locks might reference stale inodes.
2660 Details : lov_change_cbdata() must iterate over all namespaces, even if
2661 they are inactive to clear inode references from the lock.
2663 Severity : enhancement
2664 Frequency : occasional, if non-standard max_dirty_mb used
2666 Description: Client will block write RPCs if not enough grant
2667 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2668 then the client will block writes while waiting for another RPC
2669 to complete instead of consuming its dirty limit. With change
2670 we get improved performance when max_dirty_mb is small.
2672 Severity : enhancement
2673 Bugzilla : 3389, 6253
2674 Description: Add support for supplementary groups on the MDS.
2675 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2676 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2677 lookups for user supplementary groups into a cache.
2681 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2682 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2683 for liblustre O_EXCL is enforced by MDS.
2687 Description: readdir never returns NULL in liblustre.
2688 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2689 dentry in struct dirent.
2693 Frequency : liblustre only, depends on application IO pattern
2694 Description: liblustre clients evicted if not contacting servers
2695 Details : Don't put liblustre clients into the ping_evictor list, so
2696 they will not be evicted by the pinger ever.
2698 Severity : enhancement
2700 Description: Add ability to evict clients by NID from MDS.
2701 Details : By echoing "nid:$NID" string into
2702 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2703 $NID would be instantly evicted from this MDS and from all active
2704 OSTs connected to it.
2708 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2709 Details : lookup_it_finish() used to query file size from OSTs that was not
2714 Description: service threads change working directory to that of init
2715 Details : Starting lustre service threads may pin the working directory
2716 of the parent thread, making that filesystem busy. Threads
2717 now change to the working directory of init to avoid this.
2721 Frequency : during shutdown only
2722 Description: shutdown with a failed MDS or OST can cause unmount to hang
2723 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2728 Frequency : 2.6 only
2729 Description: chmod/chown may include an extra supplementary group
2730 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2731 supplementary group and if none is specified this is used.
2734 Bugzilla : 5479 (6816)
2736 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2737 Details : If lookup is in progress on a file that is unlinked we might try
2738 to revalidate the inode and fail in revalidate after lookup is
2739 complete and ll_file_open() enqueues the open again but
2740 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2743 Frequency : always, if lconf --abort_recovery used
2745 Description: lconf --abort_recovery fails with 'Operation not supported'
2746 Details : lconf was attempting to abort recovery on the MDT device and not
2749 ------------------------------------------------------------------------------
2751 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2756 Frequency : rare (only unsupported configurations with a node running as an
2758 Bugzilla : 6514, 5137
2759 Description: Mounting a Lustre file system on a node running as an OST could
2761 Details : OSTs now preallocates memory needed to write out data at
2762 startup, instead of when needed, to avoid having to
2763 allocate memory in possibly low memory situations.
2764 Specifically, if the file system is mounted on on OST,
2765 memory pressure could force it to try to write out data,
2766 which it needed to allocate memory to do. Due to the low
2767 memory, it would be unable to do so and the node would
2768 become unresponsive.
2770 Severity : enhancement
2772 Description: Addition of lconf --service command line option
2773 Details : lconf now accepts a '--service <arg>' option, which is
2774 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2776 Severity : enhancement
2778 Description: Failover mode is now the default for OSTs.
2779 Details : By default, OSTs will now run in failover mode. To return to
2780 the old behaviour, add '--failout' to the lmc line for OSTs.
2782 Severity : enhancement
2784 Description: Health checks are now provided for MDS and OSTs
2785 Details : Additional detailed health check information on MSD and OSTs
2786 is now provided through the procfs health_check value.
2789 Frequency : occasional, depends on IO load
2791 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2792 after numerous create/delete cycles
2793 Details : The ext3 inode allocation policy would not allocate new inodes
2794 very well on the OSTs because there are no new directories
2795 being created. Instead we look for groups with free space if
2796 the parent directories are nearly full.
2801 Description: Network or server problems during mount may cause partially
2802 mounted clients instead of returning an error.
2803 Details : The config llog parsing code may overwrite the error return
2804 code during mount error handling, returning success instead
2810 Description: MDS can fail to allocate large reply buffers
2811 Details : After long uptimes the MDS can fail to allocate large reply
2812 buffers (e.g. zconf client mount config records) due to memory
2813 fragmentation or consumption by the buffer cache. Preallocate
2814 some large reply buffers so that these replies can be sent even
2815 under memory pressure.
2819 Frequency : rare (liblustre)
2820 Description: fsx running with liblustre complained that using truncate() to
2821 extend the file doesn't work. This patch corrects that issue.
2822 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2823 ATTR_SIZE and lsm use in llu_setattr_raw.
2827 Frequency : rare, only 2.6 kernels
2828 Description: Unusual file access patterns on the MDS may result in inode
2829 data being lost in very rare circumstances.
2830 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2831 code in the 2.6 kernel caused the inode and/or EA data not to
2832 be read from disk, causing single-file corruption.
2836 Frequency : rare, only 2.6 filesystems using extents
2837 Description: Heavy concurrent write and delete load may cause data corruption.
2838 Details : It was possible under high-load situations to have an extent
2839 metadata block in the block device cache from a just-unlinked
2840 file overwrite a newly-allocated data block. We now unmap any
2841 metadata buffers that alias just-allocated data blocks.
2845 Frequency : filesystems with default stripe_count larger than 77
2846 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2847 Details : lconf specifies an inode size of 4096 bytes when the default
2848 stripe_count is larger than 77. This conflicts with the default
2849 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2850 this case to avoid pinning too much memory for large EAs.
2852 ------------------------------------------------------------------------------
2854 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2859 Frequency : rare (extremely heavy IO load with hundreds of clients)
2861 Description: Client is evicted, gets IO error writing to file
2862 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2863 caused clients to be evicted for AST timeouts. The fixes in
2864 bug 5192 mean we no longer need to have such short AST timeouts
2865 so ldlm_timeout has been increased.
2868 Frequency : occasional during --force or --failover shutdown under load
2869 Bugzilla : 5949, 4834
2870 Description: Server oops/LBUG if stopped with --force or --failover under load
2871 Details : a collection of import/export refcount and cleanup ordering
2872 issues fixed for safer force cleanup
2875 Frequency : only filesystems larger than 120 OSTs
2876 Bugzilla : 5990, 6223
2877 Description: lfs getstripe would oops on a very large filesystem
2878 Details : lov_getconfig used kfree on vmalloc'd memory
2881 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2882 Bugzilla : 6242, 6243
2883 Description: reading from files that had been truncated to a non-zero size
2884 but never opened returned no data
2885 Details : ll_file_read() reads zeros from no-object files to EOF
2890 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2891 Details : There's little harm in aborting MDS/OSS recovery and letting it
2892 try again, so I removed the LASSERT and return an error instead.
2894 Severity : enhancement
2896 Description: New debugging infrastructure for tracking down data corruption
2897 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2898 (b) cover more of the client-side code path, and (c) try to narrow
2899 down where problems occurred
2903 Bugzilla : 3819, 4364, 4397, 6313
2904 Description: Racing close and eviction MDS could cause assertion in mds_close
2905 Details : It was possible to get multiple mfd references during close and
2906 client eviction, leading to one thread referencing a freed mfd.
2908 Severity: : enhancement
2909 Bugzilla : 3262, 6359
2910 Description: Attempts to reconnect to servers are now more aggressive.
2911 Details : This builds on the enhanced upcall-less recovery that was added
2912 in 1.4.2. When trying to reconnect to servers, clients will
2913 now try each server in the failover group every 10 seconds. By
2914 default, clients would previously try one server every 25 seconds.
2919 Description: After recovery, certain operations trigger a failed
2920 assertion on a client.
2921 Details : Failing over an mds, using lconf -d --failover, while a
2922 client was doing a readdir() call would cause the client to
2923 LBUG after recovery completed and the readdir() was resent.
2925 Severity : enhancement
2927 Description: Default groups are now added by lconf
2928 Details : You can now run lconf --group <servicename> without having to
2929 manually add groups with lmc.
2932 Frequency : occasional
2934 Description: Nodes with an elan id of 0 trigger a failed assertion
2937 Frequency : always when accessing e.g. tty/console device nodes
2939 Description: tty and some other devices nodes cannot be used on lustre
2940 Details : file's private_data field is used by device data and lustre
2941 values in there got lost. New field was added to struct file to
2942 store fs-specific private data.
2945 Frequency : when exporting Lustre via NFS
2947 Description: NFSD failed occasionally when looking up a path component
2948 Details : NFSD is looking up ".." which was broken in ext3 directories
2949 that had grown large enough to become hashed.
2952 Frequency : Clusters with multiple interfaces not on the same subnet
2954 Description: Nodes will repeatedly try to reconnect to an interface which it
2955 cannot reach and report an error to the log.
2956 Details : Extra peer list entries will be created by lconf with some peers
2957 unreachable. lconf now validates the peer before adding it.
2960 Frequency : Only if a default stripe is set on the filesystem root.
2962 Description: Setting a default stripe on the filesystem root prevented the
2963 filesystem from being remounted.
2964 Details : The client was sending extra request flags in the root getattr
2965 request and did not allocate a reply buffer for the dir EA.
2968 Frequency : occasional, higher if lots of files are accessed by one client
2969 Bugzilla : 6159, 6097
2970 Description: Client trips assertion regarding lsm mismatch/magic
2971 Details : While revalidating inodes the VFS looks up inodes with ifind()
2972 and in rare cases can find an inode that is being freed.
2973 The ll_test_inode() code will free the lsm during ifind()
2974 when it finds an existing inode and then the VFS later attaches
2975 this free lsm to a new inode.
2979 Bugzilla : 6422, 7030
2980 Description: MDS deadlock between mkdir and client eviction
2981 Details : Creating a new file via mkdir or mknod (starting a transaction
2982 and getting the ns lock) can deadlock with client eviction
2983 (gets ns lock and trying to finish a synchronous transaction).
2986 Frequency : occasional
2987 Description: While starting a server, the fsfilt_ext3 module could not be
2989 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2990 kernels. Previously, lconf would still use the ext3 name
2991 when trying to load modules. Now, it will correctly use
2992 ext3 on 2.4 and ldiskfs on 2.6.
2994 Severity : enhancement
2995 Description: The default stripe count has been changed to 1
2996 Details : The interpretation of the default stripe count (0, to lfs
2997 or lmc) has been changed to mean striping across a single
2998 OST, rather than all available. For general usage we have
2999 found a stripe count of 1 or 2 works best.
3001 Severity : enhancement
3002 Description: Add support for compiling against Cray portals.
3003 Details : Conditional compiling for some areas that are different
3007 Frequency : occasional
3008 Bugzilla : 6409, 6834
3009 Description: Creating files with an explicit stripe count may lead to
3010 a failed assertion on the MDS
3011 Details : If some OSTs are full or unavailable, creating files may
3012 trigger a failed assertion on the MDS. Now, Lustre will
3013 try to use other servers or return an error to the
3017 Frequency : occasional
3019 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3020 caused lock timeout during readahead (since 1.4.2).
3021 Details : Processes doing readahead might match a lock that hasn't been
3022 granted yet if there are overlapping and conflicting lock
3023 requests. The readahead process waits on ungranted lock
3024 (original lock is CBPENDING), while OST waits for that process
3025 to cancel CBPENDING read lock and eventually evicts client.
3027 Severity : enhancement
3029 Description: Initial enabling of flock support for clients
3030 Details : Implements fcntl advisory locking and file status functions.
3031 This feature is provided as an optional mount flag (default
3032 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3033 locking are implemented yet, and those that are are not guaranteed
3034 to be completely correct in production environments.
3035 mount -t lustre -o [flock|noflock] ...
3038 Frequency : occasional
3040 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3041 assertion in the ext3 JBD (journaling) layer.
3042 Details : The b_committed_data struct is protected by the big kernel lock
3043 in 2.4 kernels, serializing journal_commit_transaction() and
3044 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3045 access to this struct. In 2.6 kernels there is finer grained
3046 locking to improve SMP performance of the JBD layer.
3050 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3055 Description: merge and cleanup kernel patches.
3058 -----------------------------------------------------------------------------
3060 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3062 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3063 versions of Lustre. Please update all servers and clients to
3064 version 1.4.2 or later at the same time. You must also run
3065 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3066 to update the configuration logs.
3068 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3069 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3070 - swab llog records of type '0' so we get proper header size/idx (5861)
3071 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3072 - fix rename of one directory over another leaking an inode (5953)
3073 - avoid SetPageDirty on 2.6 (5981)
3074 - don't re-add just-being-destroyed locks to the waiting list (5653)
3075 - when creating new directories, inherit the parent's custom
3076 striping settings if present parent (3048)
3077 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3078 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3079 - don't hold client locks on temporary worklist from l_lru (5666)
3080 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3081 - automatically evict dead clients (5921)
3082 - Update file size properly in create+truncate+fstat case (6196)
3083 - Do not unhash mountpoint dentries, do not allow removal of
3085 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3086 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3087 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3088 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3089 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3090 - Correctly handle reads of files with no objects (6243)
3091 - lctl recover will also mark a device active if deactivate used (5933)
3093 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3094 - allow --write-conf on an MDS with different nettype than client (5619)
3095 - don't write config llogs to MDS for mounts not from that MDS (5617)
3096 - lconf should create multiple TCP connections from a client (5201)
3097 - init scripts are now turned off by default; run chkconfig --on
3098 lustre and chkconfig --on lustrefs to use them
3099 - upcalls are no longer needed for clients to recover to failover
3101 - add --abort-recovery option to lconf to abort recovery on device
3103 - add support for an arbitrary number of OSTs (3026)
3104 - Quota support protocol changes.
3105 - forward compatibility changes to wire structs (6007)
3106 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3107 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3108 - improved "lustre status" script
3109 - initialize blocksize for non-regular files (6062)
3110 - added --disable-server and --disable-client configure options (5782)
3111 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3112 - Vanilla 2.4.29 support
3113 - increase maximum number of obd devices to 520 (6242)
3114 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3115 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3116 SLES 9 2.6.5 kernels
3117 - limit stripes per file to 160 (the maximum EA size) (6093)
3119 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3122 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3123 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3124 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3125 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3126 - small ext3 extents cleanups and fixes (5733)
3127 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3128 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3129 - avoid needless client->OST connect, fix handle mismatch (5317)
3130 - fix DLM error path that led to out-of-sync client, long delays (5779)
3131 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3132 - fix several locking issues related to i_size (5492,5624,5654,5672)
3133 - don't move pending lock onto export if it is already evicted (5683)
3134 - fix kernel oops when creating .foo in unlinked directory (5548)
3135 - fix deadlock in obdfilter statistics vs. object create (5811)
3136 - use time_{before,after} to avoid timer jiffies wrap (5882)
3137 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3138 - Do not leak request if server was not able to process it (5154)
3139 - If mds_open unable to find parent dir, make that negative lookup(5154)
3140 - don't create new directories with extent-mapping (5909, 5936)
3142 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3143 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3144 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3145 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3146 - support for RHEL4 kernel on i686 (5773)
3147 - provide error messages when incompatible logs are encountered (5898)
3149 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3150 * version 1.4.0.10 (1.4.1 release candidate 1)
3152 - don't keep a lock reference when lock is not granted (4238)
3153 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3154 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3155 - drop import inflight refcount on signal_completed_replay error (5255)
3156 - unlock page after async write error during send (3677)
3157 - handle missing objects in filter_preprw_read properly (5265)
3158 - no transno return for symlink open, don't save no-trasno open (3440)
3159 - don't try to complete elan receive that already failed (4012)
3160 - free RPC server reply state on error (5406)
3161 - clean up thread from ptlrpc_start_thread() on error (5160)
3162 - readahead could read extra page into cache that wasn't ejected (5388)
3163 - prevent races in class_attach/setup/cleanup/detach (5260)
3164 - don't dereference de->d_inode after l_dput of de (5458)
3165 - use "int" for stripe value returned from lock_to_stripe (5544)
3166 - mballoc allocation and error-checking fixes in 2.6 (5504)
3167 - block device patches to fix I/O request sizes in 2.6 (5482)
3168 - look up hostnames for IB nals (5602)
3169 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3170 - don't start multiple acceptors for the same port (5277)
3171 - fix incorrect LASSERT in mds_getattr_name (5635)
3172 - export a proc file for general "ping" checking (5628)
3173 - fix "lfs check" to not block when the MDS is down (5628)
3175 - service request history (4965)
3176 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3177 - create an "evict_client" /proc entry on OSTs, like the MDS has
3178 - fix mount usage message, return errors per mount(8) (5168)
3179 - change grep [] to grep "[]" in tests so they work in more UMLs
3180 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3181 - remove ancient LOV_MAGIC_V0 EA support (5047)
3182 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3183 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3184 - allow readahead allocations to fail when low on memory (5383)
3185 - mmap locking landed again, after considerable improvement (2828)
3186 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3188 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3191 - send OST transaction number in read/write reply to free req (4966)
3192 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3193 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3194 - the watchdog thread now runs as interruptible (5246)
3195 - flock/lockf fixes (but it's still disabled, pending 5135)
3196 - don't use EXT3 constants in llite code (5094)
3197 - memory shortage at startup could cause assertion (5176)
3199 - reorganization of lov code
3200 - single portals codebase
3202 - add extents/mballoc support (5025)
3203 - direct I/O reads in the obdfilter (4048)
3204 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3206 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3209 - send OST transaction number in read/write reply to free req (4966)
3210 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3211 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3212 - the watchdog thread now runs as interruptible (5246)
3213 - handle missing objects in filter_preprw_read properly (5265)
3214 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3215 - drop import inflight refcount on signal_completed_replay error (5255)
3216 - unlock page after async write error during send (3677)
3217 - return original error code on reconstructed replies (3761)
3218 - no transno return for symlink open, don't save no-trasno open (3440)
3220 - add pid to ldlm debugging output (4922)
3221 - bump the watchdog timeouts -- we can't handle 30sec yet
3222 - extra debugging for orphan dentry/inode bug (5259)
3224 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3227 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3228 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3229 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3230 - fix NULL dereference in /proc/sys/portals/routes (4827)
3231 - allow failed mdc_close() operations to be interrupted (4561)
3232 - stop precreate on OST before MDS would time out on it (4778)
3233 - don't send partial-page writes before EOF from client (4410)
3234 - discard client grant for sub-page writes on large-page clients (4520)
3235 - don't free dentries not owned by NFS code, check generation (4806)
3236 - fix lsm leak if mds_create_objects() fails (4801)
3237 - limit debug_daemon file size, always print CERROR messages (4789)
3238 - use transno after validating reply (3892)
3239 - process timed out requests if import state changes (3754)
3240 - update mtime on OST during writes, return in glimpse (4829)
3241 - add mkfsoptions to LDAP (4679)
3242 - use ->max_readahead method instead of zapping global ra (5039)
3243 - don't interrupt __l_wait_event() during strace
3245 - add software watchdogs to catch hung threads quickly (4941)
3246 - make lustrefs init script start after nfs is mounted
3247 - fix CWARN/ERROR duplication (4930)
3248 - return async write errors to application if possible (2248)
3249 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3250 - print NAL number in %x format (4645)
3251 - update barely-supported suse-2.4.21-171 series (4842)
3252 - support for sles 9 %post scripts
3253 - support for building 2.6 kernel-source packages
3254 - support for sles km_* packages
3256 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3259 - ignore -ENOENT errors in osc_destroy (3639)
3260 - notify osc create thread that OSC is being cleaned up (4600)
3261 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3262 - reconstruct ost_handle() like mds_handle() (4657)
3263 - create a new thread to do import eviction to avoid deadlock (3969)
3264 - let lconf resolve symlinked-to devices (4629)
3265 - don't unlink "objects" from directory with default EA (4554)
3266 - hold socknal file ref over connect in case target is down (4394)
3267 - allow more than 32000 subdirectories in a single directory (3244)
3268 - fix blocks count for O_DIRECT writes (3751)
3269 - OST returns ENOSPC from object create when no space left (4539)
3270 - don't send truncate RPC if file size isn't changing (4410)
3271 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3272 - bind to privileged port in socknal and tcpnal (3689)
3274 - rate limit CERROR/CWARN console message to avoid overload (4519)
3275 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3276 - basic mmap support (3918)
3277 - kernel patch series update from b1_4 (4711)
3279 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3282 - avoid crash during MDS cleanup with OST shut down (2775)
3283 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3284 - don't use bad inodes on the MDS (3744)
3285 - dynamic object preallocation to improve recovery speed (4236)
3286 - don't hold spinlock over lock dumping or change debug flags (4401)
3287 - don't zero obd_dev when it is force cleaned (3651)
3288 - print grants to console if they go negative (4431)
3289 - "lctl deactivate" will stop automatic recovery attempts (3406)
3290 - look for existing locks in ldlm_handle_enqueue() (3764)
3291 - don't resolve lock handle twice in recovery avoiding race (4401)
3292 - revalidate should check working dir is a directory (4134)
3294 - don't always mark "slow" obdfilter messages as errors (4418)
3296 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3299 - don't close LustreDB during write_conf until it is done (3860)
3300 - fix typo in lconf for_each_profile (3821)
3301 - allow dumping logs from multiple threads at one time (3820)
3302 - don't allow multiple threads in OSC recovery (3812)
3303 - fix debug_size parameters (3864)
3304 - fix mds_postrecov to initialize import for llog ctxt (3121)
3305 - replace config semaphore with spinlock (3306)
3306 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3307 - don't allow enqueue to complete on a destroyed export (3822)
3308 - down write_lock before checking llog header bitmap (3825)
3309 - recover from lock replay timeout (3764)
3310 - up llog sem before sending rpc (3652)
3311 - reduce ns lock hold times when setting kms (3267)
3312 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3313 - fix NULL deref and obd_dev leak on setup error (3312)
3314 - replace some LBUG about llog ops with error handling (3841)
3315 - don't match INVALID dentries from d_lookup and spin (3784)
3316 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3317 - fix invalid assertion in ptlrpc_set_wait (3880)
3319 - add libwrap support for the TCP acceptor (3996)
3320 - add /proc/sys/portals/routes for non-root route listing (3994)
3321 - allow setting MDS UUID in .xml (2580)
3322 - print the stack of a process that LBUGs (4228)
3324 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3327 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3328 - make sure to unset replay flag from failed open requests (3440)
3329 - if default stripe count is 0, use OST count for inode size (3636)
3330 - update parent mtime/ctime on client for create/unlink (2611)
3331 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3332 - free recovery state on server during a forced cleanup (3571)
3333 - unregister_reply for resent reqs (3063)
3334 - loop back devices mounting and status check on 2.6 (3563)
3335 - fix resource-creation race that can provoke i_size == 0 (3513)
3336 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3337 - more debugging for page-accounting assertion (3746)
3338 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3339 - avoid deadlock after precreation failure (3758)
3340 - fix race and lock order deadlock in orphan handling (3450, 3750)
3341 - add validity checks when grabbing inodes from l_ast_data (3599)
3343 - add /proc/.../recovery_status to obdfilter (3428)
3344 - lightweight CDEBUG infrastructure, debug daemon (3668)
3345 - change default OSC RPC parameters to be better on small clusters
3346 - turn off OST read cache for files smaller than 32MB
3347 - install man pages and include them in rpms (3100)
3348 - add new init script for (un)mounting lustre filesystems (2593)
3349 - run chkconfig in %post for init scripts (3701)
3350 - drop scimac NAL (unmaintained)
3352 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3355 - clean kiobufs before and after use (3485)
3356 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3357 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3358 - change init script to fail more clearly if not run as root (1528)
3359 - allow clients to reconnect during replay (1742)
3360 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3361 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3362 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3363 - fix cond_resched (3554)
3364 - only evict unfinished clients after recovery (3515)
3365 - allow bulk resend, prevent data loss (3570)
3366 - dynamic ptlrpc request buffer allocation (2102)
3367 - don't allow unlinking open directory if it isn't empty (2904)
3368 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3369 - remove extraneous obd dereference causing LASSERT failure (3334)
3370 - don't use get_cycles() when creating temp. files on the mds (3156)
3371 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3372 - handle EEXIST for set-stripe, set proper directory name (3336)
3374 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3375 - fix ksocknal_fmb_callback() error messages (2918)
3377 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3380 - don't copy lvb into (possibly NULL) reply on error (2983)
3381 - don't deref dentry after dput, don't free lvb on error (2922)
3382 - use the kms to determine writeback rpc length (2947)
3383 - increment oti_logcookies when osc is inactive (2948)
3384 - update client's i_blocks count via lvb messages (2543)
3385 - handle intent open/close of special files properly (1557)
3386 - mount MDS with errors=remount-ro, like obdfilter (2009)
3387 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3388 - don't use cancelling-locks' kms values (2947)
3389 - use highest lock extent for kms, not last one (2925)
3390 - don't dereference ERR_PTR() dentry in error handling path (3107)
3391 - fix thread race in portals_debug_dumplog() (3122)
3392 - create lprocfs device entries at setup instead of at attach (1519)
3393 - common AST error handler, don't evict client on completion race (3145)
3394 - zero nameidata in detach_mnt in 2.6 (3118)
3395 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3396 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3397 - initialize RPC timeout timer earlier for 2.6 (3219)
3398 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3399 - print nal/nid for unknown nid (3258)
3400 - additional checks for oscc recovery before doing precreate (3284)
3401 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3402 - don't crash in mdc_close for bad permissions on open (3285)
3403 - zero i_rdev for non-device files (3147)
3404 - clear page->private before handing to FS, better assertion (3119)
3405 - tune the read pipeline (3236)
3406 - fix incorrect decref of invalidated dentry (2350)
3407 - provide read-ahead stats and refine rpc in flight stats (3328)
3408 - don't hold journal transaction open across create RPC (3313)
3409 - update atime on MDS at close time (3265)
3410 - close LDAP connection when recovering to avoid server load (3315)
3411 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3412 - don't leak open file on MDS after open resend (3325)
3413 - serialize filter_precreate and filter_destroy_precreated (3329)
3414 - loop device shouldn't call sync_dev() for nul device (3092)
3415 - clear page cache after eviction (2766)
3416 - resynchronize MDS->OST in background (2824)
3417 - refuse to mount the same filesystem twice on same mountpoint (3394)
3418 - allow llmount to create routes for mounting behind routers (3320)
3419 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3420 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3421 - fix rare problem with rename on htree directories (3417)
3423 - allow default OST striping configuration per directory (1414)
3424 - fix compilation for qswnal for 2.6 kernels (3125)
3425 - increase maximum number of MDS request buffers for large systems
3426 - change liblustreapi to be useful for external progs like lfsck (3098)
3427 - increase local configuration timeout for slow disks (3353)
3428 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3430 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3433 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3434 - don't overwrite extent policy data in reply if lock was blocked (2901)
3435 - drop filter export grants atomically with removal from device (2663)
3436 - del obd_self_export from work_list in class_disconnect_exports (2908)
3437 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3438 - swab reply message in mdc_close, other PPC fixes (2464)
3439 - fix destroying of named logs (2325)
3440 - overwrite old logs when running lconf --write_conf (2264)
3441 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3442 - fix race in target_handle_connect (2898)
3443 - mds_reint_create() should take same inode create lock (2926)
3444 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3445 - don't close files for self_export to avoid uninitialized obd (2936)
3446 - allow MDS with the same name as client node (2939)
3447 - hold dentry reference for closed log files for unlink (2325)
3448 - reserve space for all logs during transactions (2059)
3449 - don't evict page beyond end of stripe extent (2925)
3450 - don't oops on a deleted current working directory (2399)
3451 - handle hard links to targets without a parent properly (2517)
3452 - don't dereference NULL lock when racing during eviction (2867)
3453 - don't grow lock extents when lots of conflicting locks (2919)
3455 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3458 - account for cache space usage on clients to avoid data loss (974)
3459 - lfsck support in lustre kernel code (2349)
3460 - reduce journal credits needed for BRW writes (2370)
3461 - orphan handling to avoid losing space on client/server crashes
3462 - ptlrpcd can be blocked, stopping ALL progress (2477)
3463 - use lock value blocks to assist in proper KMS, faster stat (1021)
3464 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3465 - recovery for initial connections (2355)
3466 - fixes for mds_cleanup_orphans (1934)
3467 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3468 - block all file creations until orphan recovery completes (1901)
3469 - client remove rq_connection from request struct (2423)
3470 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3471 - recovery timer race (2670)
3472 - mdc_close recovey bug (2532)
3473 - ptlrpc cleanup bug (2710)
3474 - mds timeout on local locks (2588)
3475 - namespace lock held during RPCs (2431)
3476 - handle interrupted sync write properly (2503)
3477 - don't try to handle a message that hasn't been replied to (2699)
3478 - client assert failure during cleanup after abort recovery (2701)
3479 - leak mdc device after failed mount (2712)
3480 - ptlrpc_check_set allows timedout requests to complete (2714)
3481 - wait for inflight reqs when ptlrpcd finishes (2710)
3482 - make sure unregistered services are removed from the srv_list
3483 - reset bulk XID's when resending them (caught by 1138 test)
3484 - unregister_bulk after timeout
3485 - fix lconf error (2694)
3486 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3487 - readahead locks pages, leaves pending causing memory pressure (2673)
3488 - increase OST request buffers to 4096 on large machines (2729)
3489 - fix up permission of existing directories in simple_mkdir (2661)
3490 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3491 - don't assign transno to errored transactions (2742)
3492 - don't delete objects on OST if given a bogus objid from MDS (2751)
3493 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3494 - if rq_no_resend, then timeout request after recovery (2432)
3495 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3496 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3497 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3498 - MDS recovery shouldn't delete objects that it has given out (2730)
3499 - if enqueue arrives after completion, don't clobber LVB (2819)
3500 - don't unlock pages twice when trigger_group_io returns error (2814)
3501 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3502 - don't write pages to disk if there was an error (1450)
3503 - don't ping imports that have recovery disabled (2676)
3504 - take buffered bytes into account when balancing socknal conn (2817)
3505 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3506 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3507 - remove little-endian swabbing of llog records (1987)
3508 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3509 - retry reposting request buffers if they fail (1191)
3510 - grow extent at grant time to avoid granting a revoked lock (2809)
3511 - lock revoke doesn't evict page if covered by a second lock (2765)
3512 - disable VM readahead to avoid reading outside lock extents (2805)
3514 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3515 - updated kernel patches for hp-2.4.20 kernel (2681)
3517 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3520 - fix truncated write corruption (2366)
3521 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3523 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3524 - ensure len(uuid) < 37 in lmc (1171)
3525 - fix ia64 OOPS in llog_test (2255)
3526 - zero end of page at obdfilter for partial page writes (2648)
3527 - don't leave stale dentries around after renames (bug 2428)
3528 - fix timeouts when evicting a client with a single lock held (2642)
3529 - set deadline for the initial HELLO message to drain (2634)
3530 - print out dotted-quad IP addresses in the socknal (2302)
3532 - additional debugging for MDS client eviction problem (2443)
3533 - fix mkfsoptions support for osts (2603, 2604)
3535 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3538 - add series for the vanilla 2.6.0 kernel
3539 - add series for the vanilla 2.4.24 kernel
3540 - add series for a cray x86/64 UL kernel drop
3541 - fix xattr patches for the vanilla 2.4.19 series
3543 - generate true UUIDs in lmc (1171)
3544 - have portals stack dumping break in UML (2466)
3545 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3546 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3547 - wake callback waiting threads on client eviction (2460)
3548 - Add --ptldebug and --subsystem to lmc (1719)
3549 - update assertion to allow safe interrupt allocation
3550 - set rq_no_resend for cancel requests (2432)
3551 - recalculate ptlrpcd timeout after resend (2494)
3552 - call vfs_rmdir when removing pending directories (2368)
3553 - fix renaming a file to itself (2429)
3554 - lmc creates a default one-stripe lov (2454)
3555 - expand procfs space to handle large clusters (2326)
3556 - increase UML stack to avoid overflow
3557 - update lconf's list of debug and subsystem masks
3558 - fix lfs find --obd (2510)
3559 - /proc tunable for disabling filter read caching (2591)
3560 - stop rpm packages from altering slapd.conf (2301)
3561 - disable nagle in the socknal under 0conf (2578)
3562 - choose mds inode size based on stripe count (2572)
3563 - fix kernel-source rpm problems (2516)
3565 - add --disable-doc to avoid pdf generation (2421)
3566 - update documentation, tests, type-os, comments
3567 - avoid format warnings on ia64
3568 - remove the TOE NAL
3569 - tiny code cleanups by removing unused fields
3571 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3574 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3575 - remove the most blatant lies from BUILDING file (2371)
3576 - change default debug level to reasonable production setting
3577 - reduce client side cache size to reduce cache flush time
3578 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3579 - make TCP zerocopy and pinger support enabled by default (2476)
3580 - sync writes completed after process exits caused crashes (2319)
3581 - maintain correct mount count on the MDS (2356)
3582 - backout 1557, because 2316 wasn't really fixed
3583 - better file I/O statistics gathering in /proc
3584 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3585 - another kernel patch to fix zero-copy TCP function export
3586 - don't take duplicate lock when processing re-sent getattr (2420)
3587 - lctl uses obd_self_export instead of creating new conn (2353)
3588 - MDS/OST recovery case which requires object creation asserted (2425)
3589 - move lfs from /usr/sbin to /usr/bin in packages
3590 - fix race between mds_client_add and mds_client_free (2417)
3591 - use kmalloc instead of slabs in portals (2430)
3592 - don't create duplicate records when a failover MDS is present (2442)
3593 - remove unnecessary mount age check (2332)
3594 - don't remove directory inodes from locks prematurely (2451)
3595 - don't break if MDS service name is the same as hostname (2103)
3596 - fix races in client write RPC generation when cache full (2482)
3598 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3601 - remove now-unused request->rq_obd (278)
3602 - if an allocation fails, print out how much memory we've used (1933)
3603 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3604 - add the "configurable stack size" patch to most series files (1256)
3605 - ability to write large log records, for 100+ OST configs (2306)
3606 - fix NULL deref when filter_prep fails (2314)
3607 - fix operator precedence error in filter_sync
3608 - dynamic allocation of socknal TX descriptors (2315)
3609 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3610 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3611 - ensure that CWARN messages in Portals always get to the syslog
3612 - __init/__exit are not for prototype decls (ldlm_init/exit)
3613 - x86-64 compile warning fixes
3614 - fix gateway LMC keyword conflict (2318)
3615 - fix MDS lock inversions in getattr/reint paths (1844)
3616 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3617 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3618 - DLM race condition prevented some lock evictions (2328)
3619 - ENOMEM detection and retry on socknal sends (2230)
3620 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3621 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3622 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3623 - better error messages when a client is rejected during recovery (1505)
3624 - avoid cancelling locks which were never granted, after failure (2330)
3625 - fix i_sem/journal inversion in mds_client_add (2333)
3626 - fix truncate/getattr lock cycle deadlock (2334)
3627 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3628 - fix two rare exit paths which could leak an l_lock() ref (2321)
3629 - fencepost error in MDS/OST orphan recovery (2226)
3630 - make log record alignment 8 bytes (1988)
3631 - lstripe now fails when requested offset > ost_count (2237)
3632 - ensure that all kernel series have a complete list.h (1607)
3633 - fix crashes in special-file operations (2316)
3634 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3636 - allow configurable automake binary, for testing new versions
3637 - small update to the lfs documentation
3639 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3641 * fix negative export reference count in fsfilt_sync (2312)
3643 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3644 * release candidate 0.9.1
3646 - orphans are moved into the PENDING directory for possible recovery
3647 - replayed opens now open by fid for orphan/rename safety (1042)
3648 - last close of an orphan inode generates a transno (683)
3649 - chdir() and mount() now pin the directory entry (1020)
3650 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3651 - discard very old requests without processing them (1502)
3652 - remove some common, well-understood CERRORs (1505)
3653 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3654 - clear "grant" flags in OST replies until OST grant code lands (1644)
3655 - fix read performance by not clobbering i_blksize on client (1598)
3656 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3657 - make LRU size a /proc tunable, clears locks when reduced (707)
3658 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3659 - server locks take references on exports now (1558)
3660 - build fixes for 2.4.20-rh trees (1663)
3661 - return an error from lov_create if all OSCs are inactive (1751)
3662 - fix import levels when a reconnect happens without a timeout (1597)
3663 - exit early from mds_open if we get a lookup error (1749)
3664 - partial page read at EOF wouldn't wait for disk before sending (1642)
3665 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3666 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3667 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3668 - write/truncate lock inversion (1639)
3669 - Don't auto-load obdclass, portals modules during cleanup (1495)
3670 - fix timestamps from jumping to "now" (1763)
3671 - extra journal assertions (1648)
3672 - add an extra multiunlink test (1771)
3673 - fix read_record/write_record API (1776)
3674 - fix leak of offset_extent, possible incorrect i_size later (1772)
3675 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3676 - Debugging for the kqswnal_get_idle_tx problems (1820)
3677 - Allow recovery to be attempted multiple times (1536)
3678 - Write out MDS last_rcvd file after it is first created (1600)
3679 - Fix tx_descriptor leak in failed transmit situations (1827)
3680 - ext3 journaling fixes for assertion failure after IO error (1871)
3681 - class_export_put() on freed export after completion AST error (1896)
3682 - Fix revalidate looping in VFS (1322)
3683 - Don't access a freed export during MDS_REINT timeout (1521)
3684 - Add open-unlink recovery support on the MDS (1673,1764)
3685 - Return an error if no MDS data was read from last_rcvd (1946)
3686 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3687 - Don't setup a disk device that doesn't match exported UUID (317)
3688 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3689 - avoid committing NULL handle in force close
3690 - local.sh is now a one-stripe LOV configuration
3691 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3692 - fix ext3 htree duplicate directory entry corruption (1516)
3693 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3694 - update to new LOV EA format (2097)
3695 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3697 2003-06-15 Phil Schwan <phil@clusterfs.com>
3700 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3701 - per-import/export recovery handling (958, 931, 959)
3702 - multiple last-rcvd slots, for serving multiple FSes (949)
3703 - connections are again shared between multiple imp/exports (963, 964)
3704 - "umount -f" would hang if any requests needed to be sent (393, 978)
3705 - avoid pinning large req buffer by copying for queued messages (989)
3706 - add "uuid" to "lctl device" command to help upcalls (991)
3707 - "open" RPCs with transnos would confuse recovery counters (1037)
3708 - do proper endian conversion of all wire messages (288, 340, 891)
3709 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3710 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3711 - fix ptlrpc_connection leak in target_handle_connect (1174)
3712 - fix import refcounting bug in OST and MDS cleanup (1134)
3713 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3714 - fix very unlikely obd_types race condition (501)
3715 - remove osc_open hack for echo_client (1187)
3716 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3717 - a failure in read_inode2 leads to deadlock (1139)
3718 - cancel ack-locks as soon as transaction is committed (1072)
3719 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3720 - make sure to commitrw after any preprw to avoid deadlock (1162)
3721 - failing to execute a file in a lustre FS would lock inode (1203)
3722 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3723 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3724 - obd_uuid2tgt would walk off the end of the list (1255)
3725 - on IA64 the osc would give portals incorrect bulk size (1258)
3726 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3727 - fix lock inversion caused by new llite matching code (1282)
3728 - limit the number of dirty pages on a client to 10MB (1286)
3729 - timed out locks were not being corrected cancelled (1289)
3730 - fix O_DIRECT above 4GB on IA-32 (1292)
3731 * major user-visible changes
3732 - fail out/fail over policy now controlled by the upcall (993)
3734 - add OBD_PING to check server availability and failure (954)
3735 - lustre messages are now sent in sending host order (288, 340, 891)
3736 - add eadatalen to MDS getattr reply (340)
3737 - OST read replies may contain second buffer, with per-page status (593)
3739 2003-03-11 Phil Schwan <phil@clusterfs.com>
3742 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3743 - failed open()s could cause deadlock; fixed (867, 869)
3744 - stop cancelling OST locks when files are closed (481)
3745 - overlapping XID spaces caused network corruption (851, 853)
3746 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3747 - setattr_raw added, to do single-RPC, server-side setattrs
3748 - lmc/lconf syntax change for OST UUIDs
3749 - fix crashy race condition between ptlrpc_free_req and osc_close
3750 - don't use request in mdc_enqueue if we hit a timeout (889)
3751 - don't set the inode i_size for regular files from the MDS (896)
3752 - handle out of order completion AST (842)
3753 - don't LBUG if a lock request times out after receiving AST (913)
3754 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3755 - if a bad lock AST arrives, send an error instead of dropping entirely
3756 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3757 - fix leak in bulk IO when only partially completed (899, 900, 926)
3758 - fix O_DIRECT for ia64 (55)
3759 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3760 - C-z after timeout could hang a process forever; fixed (977)
3762 - client-side I/O cache (678, 924, 929, 941, 970)
3764 - READPAGE and SETATTRs which don't take server-side locks get
3767 2003-02-11 Phil Schwan <phil@clusterfs.com>
3770 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3771 - Fix fsfilt_extN_readpage() to read a full page of directory
3772 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3773 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3774 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3775 - Most intent operations were removed, in favour of a new RPC mode
3776 that does a single RPC to the server and bypasses most of the VFS
3777 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3778 - Aggressively cancel local locks on DLM servers
3779 - mds_reint_unlink sends EA to the client if it's the last nlink.
3780 client uses that EA to unlink OST objects.
3781 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3782 - recursive symlinks were fixed (439)
3783 - fixed NULL deref in DEBUG_REQ
3784 - filter_update_lastobjid no longer calls sync, which annoyed extN
3785 - fixed multi-client small-writes to a single file problem (445)
3786 - fixed mtime updates during file writes (607)
3787 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3788 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3789 - fixed Linux OST size reporting problem (444, 656)
3790 - OST now updates object mtime with writes or setattr (607, 619)
3791 - client verifies file size before zeroing page past EOF (445)
3792 - OST now writes last allocated objid to disk with allocation (108)
3793 - LOV on echo now works (409)
3795 - mds_reint_unlink sends a new buffer, with the EA included. this
3796 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3797 set if a regular file was being unlinked, and it was the last link
3798 - use PtlGet from the target for bulk writes (315)
3799 - OST now updates object mtime with writes or setattr (607, 619)
3800 - LDLM now has a grant-time callback to revalidate locked items, if
3802 - Many MDS operations were reorganized to combat race conditions
3804 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3805 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3807 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3810 - Fully reactivate OST imports after reconnection (512, others)
3811 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3812 - More graceful error handling for truncating on dead OST (515)
3813 - Don't error out unless we're actually accessing dead stripes (474)
3814 - Fix garbage sizes when stripes are missing (410)
3815 - LRU counters were broken, causing constant lock purge (433, 432)
3816 - garbage on read from stripes with failed OSTs (441)
3817 - mark OSCs as active before reconnecting during recovery (438)
3818 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3819 - lfind did not preserve OST order in output (443)
3820 - symlinks cause hung clients, incorrect data (439)
3821 - stop dereferencing request after dropping refcount (457)
3822 - don't LASSERT(spin_is_locked) on non-SMP (455)
3823 - fixes for many rename() bugs
3824 - fstat didn't correctly synchronize attributes (399)
3825 - server must handle lock cancellation during blocking AST prep (487)
3826 - bulk descriptors were free()d too soon (511)
3827 - fix paths in lconf, which would load incorrect modules (451, 507)
3828 - fix confusing lconf 'host not found' error message (386)
3829 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3830 - fix race condition in mdc_blocking_ast() for inode access (526)
3831 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3832 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3833 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3834 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3835 - fix lstripe in several ways (kernel side) (527)
3836 - fix request leak in ldlm_cli_enqueue (262)
3837 - incorrect OSC was marked inactive after OST failure
3838 - call mds_fs_cleanup before unmounting filesystem (524)
3839 - fix races between taking ns_lock and ldlm_lock_change_resource
3840 - fix races updating LOV export open file list
3841 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3842 - fix recovery NULL deref in ldlm_cli_cancel_unused
3843 - fix some DLM races by using new hash table for lock handles (419)
3844 - permit the client to specify desired inodes, at replay
3845 - duplicate requests when we queue them for replay reintegration
3846 - fix last_rcvd offset calculation
3847 - sync after each recovered transaction, so we always make progress
3848 - never, not always, ERESTART requests without transnos
3849 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3851 - skip replay if the MDS says that the client is already connected
3852 - don't check for a recovery-enabled export to match lctl's UUID
3853 - don't INC_USE_COUNT for phantom exports
3854 - don't crash when cleaning up phantom exports (567)
3855 - don't double-finish or set replay data for errored mdc_open requests
3856 - abort requests when they time out, so we don't get old replies
3857 - send/receive replies for AST messages again
3858 - if the client says that it doesn't have the lock, cancel it on the
3860 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3861 mark it as destroyed, it will all work out when decref is called
3862 - fix module use counts (22, 581)
3864 - ASTs now expect a reply (server cancels lock on error reply)
3866 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3869 - fix many simultaneous client startup (392)
3870 - fix dentry->d_it clobbering
3871 - credentials weren't being shipped for readdir/getattr operations
3872 - remove invalid assertions triggered during some concurrent MD
3874 - proper Lustre versions added (336, 389)
3875 - fix memory leak for create error case (398)
3876 - fix LOV locking bug that would get cli/srv out of sync
3877 - fix echo client over LOV (409)
3878 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3879 - fix double-O_EXCL intent crash (424)
3880 - avoid sending multiple lock CANCELs (352)
3882 - MDS can do multi-client recovery (modulo bugs in new code)
3884 - many updates, edits, cleanups
3886 2002-11-18 Phil Schwan <phil@clusterfs.com>
3889 - fix null d_it dereference (346)
3890 - fix full OST/dbench hang (333)
3891 - fix permission problem with file removal (286)
3892 - fix removal of OSCs from LOV when they fail
3893 - fix NULL deref during bulk timeout (214)
3894 - fix problems related to multiple filesystems on one MDS (241)
3895 - fixed serious subtle metadata locking bugs
3896 - free locks on clients when inodes are removed due to memory
3898 - fix inode pointer in lock data (285)
3899 - partial support for multiple MDS on a single host (241)
3900 - data locks weren't cancelled at clear_inode time (290, 311)
3901 - intent locks could lead to unbounded lock growth (205)
3902 - added a maximum lock count, an LRU list, and a flusher
3903 - fix multiple rename (365)
3904 - properly abstracted the echo client
3905 - OSC locked 1 byte too many; fixed
3906 - rewrote brw callback code:
3907 - fixed recovery bugs related to LOVs (306)
3908 - fixed too-many-pages-in-one-write crash (191)
3909 - fixed (again) crash in sync_io_timeout (214)
3910 - probably fixed callback-related race (385)
3912 - Add capability to MDS protocol
3913 - LDLM cancellations and callbacks on different portals
3915 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3918 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3919 - defer open object destruction to close time (601981, 138)
3920 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3921 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3922 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3923 - don't call obd_setattr in ll_file_release for destroyed objects
3924 * protocol change to lustre_msg: move |version| and add |flags|
3925 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3926 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3927 * added replay of create, unlink, link and rename operations during
3928 MDS failover; recovery should be much more robust now
3929 * remove failed OSCs from LOVs (only lov_create uses this so far)
3930 * the lustre-HOWTO was brought (more) up to date (582544)
3932 2002-10-23 Phil Schwan <phil@clusterfs.com>
3935 - in-use dentries weren't being reused properly (617851)
3936 - prevent multiple LDLM setup (599178)
3937 - fix LOV size calculations for truncate (617853)
3938 - fix client handling of MDS intent errors (POSIX)
3939 - fix permission bug in lovstripe.c test (624321)
3940 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3941 - truncate past end of file could corrupt data
3942 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3943 - a race in recovery could return ETIMEDOUT to apps (623947)
3944 - building outside the source directory was fixed
3945 * the lustre-HOWTO was brought (more) up to date (582544)
3946 * major progress was made on recovery functionality
3948 2002-10-10 Phil Schwan <phil@clusterfs.com>
3951 - recovery deadlock fix
3952 - rm -rf causes LBUG fix (617817)
3953 - file open by multiple tasks fix (618962)
3954 - directory permissions bugs (602707 and 620007)
3955 - journal_stop fixed with locking (611313)
3956 - O_APPEND failures resolved (618273, perhaps 614459)
3957 - lconf PATH fix (619770)
3958 - IA64 build fix (621450)
3959 - RPC buffer sizes scale with amount of memory
3961 2002-10-01 Phil Schwan <phil@clusterfs.com>
3964 - locks would be cancelled without throwing away data pages,
3965 resulting in inconsistent data (605627)
3966 - inode attributes were not always being refreshed (605627, 612449)
3967 - lconf now continues to cleanup after lctl reports an error
3968 - MDS now enforces user permissions (602707)
3969 - lprocfs cleanup fixed, but not yet enabled (614157)
3970 - fixed infinite server hang, should a client not respond to an AST
3971 - avoid going into recovery if user calls readlink() with a buffer
3972 that's too small (613941)
3973 - AST RPCs no longer require replies (614867) -- this may be changed
3974 - don't crash server if client sends an IOV that's too big (611336)
3975 - fixed lock conversion deadlock (611892)
3976 - fixed the following of symlinks (614622)
3977 * recovery: the server can remove locks from a client that dies, other
3978 clients can make progress
3979 * more extN patch fixes
3980 * compile-time configurable ptlrpc buffer allocations
3982 - collaborative read cache document
3983 - Lustre Lite Performance CDR document-in-progress
3985 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3988 - fix typo in patch-2.4.18
3990 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3993 - clear ptlrpc request each time in handle_incoming_request()
3994 - unlink of files now destroys the object on the OST
3996 2002-09-19 Peter Braam <braam@clusterfs.com>
3998 * add hard link support
3999 * change obdfile creation method
4000 * kernel patch changed
4002 2002-09-19 Peter Braam <braam@clusterfs.com>
4005 - stack overflow bug in extN fixed
4007 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4009 * documentation updates
4010 - add man pages for config tools
4011 - update tests/README to describe testing with new config tools
4012 - finish metadata API descriptions
4013 * bug fixes and cleanups
4014 - statfs workaround for 16TB limit
4015 - LOV stripe allocation improved, can stripe on subset of OSTs
4016 - LOV file size/IO offset was wrong for files > 4GB in size
4017 - object EA data was being dropped, caused files to be unreadable
4018 - memory overflow with non-LOV OST caused memory corruption
4019 - fixed regression tests to work with new config tools, obdfilter
4020 - fixed bug when directory size became larger than 1 block
4021 - fixed bug (for single client case) when PWD was deleted
4022 - invalidate local directory pages when doing intent-based ops
4023 - avoid LDLM oops when lock callback contained bad data
4025 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4027 * documentation updates
4028 * bug fixes and cleanups
4029 - configuration tools
4032 - 64-bit compile warnings
4033 - 64-bit internal statfs data
4035 * test_brw on persistent OST devices
4037 * lprocfs (disabled)
4039 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4041 * documentation updates
4042 * bug fixes and cleanups
4043 * configuration tools
4045 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4048 * many small fixes to 0_5_4
4049 * io/network handling
4050 * thinkos in MDS operations
4052 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4055 * crucial basic fixes to 0.5.3
4056 * IOR, Iozone work over Elan
4059 2002-08-07 Phil Schwan <phil@clusterfs.com>
4060 * version 0_5_3, our first alpha
4061 * we use the new Portals iovs
4062 * documentation updates
4063 * bug fixes and cleanups
4064 * small changes in the DLM wire protocol
4066 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4067 * version 0_5_1 with some initial stability,
4068 * locking on MD and file I/O.
4069 * documentation updates
4070 * several bug fixes since 0.5.0
4071 * small changes in wire protocol
4073 2002-07-18 Phil Schwan <phil@clusterfs.com>
4075 * delivered as Lustre Light Alpha
4076 * fixed a crash after handling invalid MDS requests
4077 * fixed directory pages for architectures with non-4k pages sizes
4079 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4080 * release version v0_4_4
4081 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4083 * Fixes a number of interruption problems with OST operations.
4084 * Update documentation for portals header changes
4085 * Move all wire protocol structs/defines to lustre_idl.h
4086 * Fixes symlink length bug.
4087 * Add tcpdump to repository.
4089 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4090 * release version v0_4_3
4091 * Fixes statfs for inodes on extN.
4092 * Fixes bug in runtests which would delete /etc/hosts.
4093 * Use 64-bit object IDs wherever possible (not into VFS though)
4094 Remove ost_get_info, which is unused by lustre, and out of date.
4096 2002-07-03 Peter Braam <braam@clusterfs.com>
4097 * release version v0_4_2 Fixes a lookup error (type not passed)
4098 * move forward to head of Portals
4099 * move forward to latest Lustre kernel
4101 2002-06-25 Peter Braam <braam@clusterfs.com>
4102 * release version v0_4_1. Hopefully stable on single node use.