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: light-weight GSS support
18 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
19 and reduce performance overhead.
21 Severity : enhancement
23 Description: Linux keyring support
24 Details : Support using service of Linux keyring for Lustre GSS internal
25 context refresh/cache mechanism.
29 Description: Fix errors in lfs documentation
30 Details : Fixes man pages
34 Description: don't use GFP_* in generic Lustre code.
35 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
39 Description: obdclass is limited by single OBD_ALLOC(idarray)
40 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
44 Description: warnings with build patchless client with vanila 2.6.19 and up
45 Details : change old ctl_table style and replace ctl_table/ctl_table_header
46 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
50 Frequency : Only for SLES
51 Description: sanity_quota fail test_1
52 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
53 file, which makes TSTID[2] inunique.
57 Frequency : Only for SLES9
58 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
59 Details : Update target/ChangeLog/which_patch .
63 Frequency : Only for test_10 in sanity-quota.sh
64 Description: a bug in quota test code
65 Details : checking a test flag in a wrong place
69 Frequency : Only for test_13 in sanity-quota.sh
70 Description: a bug in quota test script
71 Details : shouldn't check overfull allocated quota
75 Description: jbd statistics
76 Details : Port older jbd statistics patch for sles10
78 Severity : enhancement
80 Description: Add printing inode info into message about error in writepage.
85 Description: reply_lock_interpret crash due to race with it and lock cancel.
86 Details : Do not replay locks that are being cancelled. Do not reference
87 locks by their address during replay, just by their handle.
91 Frequency : When flocks are used.
92 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
93 Details : Flock locks might destroy just granted lock if it could be merged
94 with another existing flock, this is done in completion handler,
95 so teach ldlm_cli_enquque_fini that this is a valid case for
101 Description: Oops in read and write path when failing to allocate lock.
102 Details : Check if lock allocation failed and return error back.
106 Description: lstripe command fails for valid OST index
107 Details : The stripe offset is compared to lov->desc.ld_tgt_count
108 instead of lov->desc.ld_active_tgt_count.
112 Description: sanity.sh failed test 103
113 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
118 Description: df doesn't work properly if diskfs blocksize != 4K
119 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
123 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
124 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
128 Description: obd_health_check_timeout too short
129 Details : set obd_health_check_timeout as 1.5x of obd_timeout
133 Description: llapi_file_create() does not allow some changes
134 Details : add llapi_file_open() that allows specifying the mode and
135 open flags, and also returns an open file handle.
139 Description: merge and cleanup kernel patches.
140 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
144 Description: Add journal checksum support.(Kernel part)
145 Details : The journal checksum feature adds two new flags i.e
146 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
147 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
148 indicates that the commit block contains the checksum for
149 the blocks described by the descriptor blocks. Now commit
150 record can be sent to disk without waiting for descriptor
151 blocks to be written to disk. This behavior is controlled
152 using JBD2_FEATURE_ASYNC_COMMIT flag.
156 Description: OSS needs mutliple precreate threads
157 Details : Add ability to start more than one create thread per OSS.
161 Description: Sanity.sh test_65k failed: setstripe should have succeeded
162 Details : execute lfs setstripe on client
166 Description: mds_obd_create error creating tmp object
167 Details : When the user sets quota on root, llog will be affected and can't
168 create files and write files.
171 Frequency : Always on ia64 patchless client, and possibly others.
173 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
174 Details : This allows the patchless client to be loaded on architectures
179 Description: RedHat Update kernel for RHEL5
180 Details : Add the kernel config file for RHEL5.
184 Description: RedHat Update kernel for RHEL5
185 Details : Modify the kernel config file more closer RHEL5.
189 Description: Build failure against Centos5 (RHEL5)
190 Details : Define PAGE_SIZE when it isn't present.
194 Description: test_7 on sanity-quota.sh can't be used on separate machines
195 Details : add facet to handle it
199 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
200 Details : with NFS, the anon dentry's parent was set to itself in
201 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
202 corresponding dentry other than use rec->ur_name.
204 Severity : enhancement
206 Description: lfs setstripe enhancement
207 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
211 Description: testing performance impact of enabling checksumming
212 Details : enable checksum by default, allow --disable-checksum
213 configure option and "-o nochecksum" mount option
217 Description: lustre support for RHEL5
218 Details : Add support for RHEL5.
222 Description: Client eviction due to failover config
223 Details : after a connection loss, the lustre client should attempt to
224 reconnect to the last active server first before trying the
225 other potential connections.
229 Description: when mds and osts use different quota unit(32bit and 64bit),
230 quota will be released repeatly.
231 Details : void sending multiple quota reqs to mds, which will keep the
232 status between the reqs.
236 Description: osts not allocated evenly to files
237 Details : change the condition to increase offset_idx
240 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
242 Description: Data corruption for OSTs that are formatted larger than 2TB
244 Details : When generating the bio request for lustre file writes the
245 sector number would overflow a temporary variable before being
246 used for the IO. The data reads correctly from Lustre (which
247 will overflow in a similar manner) but other file data or
248 filesystem metadata may be corrupted in some cases.
252 Description: Tune the kernel for good SCSI performance.
253 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
254 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
259 Description: rewrite ext2-derived code in obdclass/uuid.c
260 Details : rewrite inherited code (uuid parsing code from ext2 utils)
261 from scratch preserving functionality.
265 Description: "lfs find -obd UUID" prints directories
266 Details : "lfs find -obd UUID" will return all directory names instead
267 of just file names. It is incorrect because the directories
268 do not reside on the OSTs.
272 Description: use i_size_read and i_size_write in 2.6 port
273 Details : replace inode->i_size access with i_size_read/write()
276 Frequency : when removing large files
278 Description: scheduling issue during removal of large Lustre files
279 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
280 It causes scheduling issues when removing large files (17TB in the
284 Frequency : only with liblustre clients on XT3
286 Description: evictions taking too long
287 Details : allow llrd to evict clients directly on OSTs
290 Frequency : only on ppc
292 Description: /proc/fs/lustre/devices broken on ppc
293 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
294 We should make sure the type of 'index' is loff_t and then cast
295 explicitly as needed below. Do not assign an explicitly cast
299 Frequency : only for rhel5
301 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
302 Details : Modify the target file & which_kernel.
304 Severity : enhancement
306 Description: omit set fsid for export NFS
307 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
308 as FS_REQUIRES_DEV for avoid problems with generate fsid.
312 Frequency : Always, for kernels after 2.6.16
313 Description: Fix warning idr_remove called for id=.. which is not
315 Details : Last kernels save old s_dev before kill super and not allow
316 to restore from callback - restore it before call kill_anon_super.
318 --------------------------------------------------------------------------------
320 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
322 * Support for kernels:
323 2.6.5-7.283 (SLES 9),
324 2.6.9-55.EL (RHEL 4),
325 2.6.16.46-0.14 (SLES 10),
326 2.6.18.8 vanilla (kernel.org)
327 * Client support for unpatched kernels:
328 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
329 2.6.16 - 2.6.22 vanilla (kernel.org)
330 * Due to recently discovered recovery problems, we do not recommend
331 using patchless RHEL 4 clients with this or any earlier release.
332 * Recommended e2fsprogs version: 1.39.cfs8
333 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
334 * Starting with this release, the ldiskfs backing filesystem required
335 by Lustre is now in its own package, lustre-ldiskfs. This package
336 should be installed. It is versioned separately from Lustre and
337 may be released separately in future.
341 Description: block reactivating mgc import until all deactivates complete
342 Details : Fix race when failing back MDT/MGS to itself (testing)
344 Severity : enhancement
346 Description: add optional extra BUILD_VERSION info
347 Details : add a new environment variable (namely LUSTRE_VERS) which allows
348 to override the lustre version.
351 Frequency : 2.6.18 servers only
353 Description: ll_kern_mount() doesn't release the module reference
354 Details : The ldiskfs module reference count never drops down to 0
355 because ll_kern_mount() doesn't release the module reference.
360 Description: server LBUG when using old ost_num_threads parameter
361 Details : Accept the old ost_num_threads parameter but warn that it
362 is deprecated, and fix an off-by-one error that caused an LBUG.
367 Description: Transient SCSI error results in persistent IO issue
368 Details : iobuf->dr_error is not reinitialized to 0 between two
372 Frequency : sometimes when underlying device returns I/O errors
374 Description: OSTs not going read-only during write failures
375 Details : OSTs are not remounted read-only when the journal commit threads
376 get I/O errors because fsfilt_ext3 calls journal_start/stop()
377 instead of the ext3 wrappers.
381 Description: poor connect scaling with increasing client count
382 Details : Don't run filter_grant_sanity_check for more than 100 exports
383 to improve scaling for large numbers of clients.
386 Frequency : SLES10 only
388 Description: sanity-quota.sh quotacheck failed: rc = -22
389 Details : Quotas cannot be enabled on SLES10.
392 Frequency : liblustre clients only
394 Description: getdirentries does not give error when run on compute nodes
395 Details : getdirentries does not fail when the size specified as an argument
396 is too small to contain at least one entry
398 Severity : enhancement
400 Description: Add LNET router traceability for debug purposes
401 Details : If a checksum failure occurs with a router as part of the
402 IO path, the NID of the last router that forwarded the bulk data
403 is printed so it can be identified.
408 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
409 Details : Due to a race condition, liblustre clients were occasionally
412 Severity : enhancement
414 Description: lfs setstripe use optional parameters instead of postional
417 Severity : enhancement
419 Description: Nanosecond timestamp support for ldiskfs
420 Details : The on-disk ldiskfs filesystem has added support for nanosecond
421 resolution timestamps. There is not yet support for this at
422 the Lustre filesystem level.
425 Frequency : during server recovery
427 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
428 Details : request with rq_no_resend flag not awake l_wait_event if they get a
432 Frequency : nfs export on patchless client
434 Description: connectathon hang when test nfs export over patchless client
435 Details : Disconnected dentry cannot be found with lookup, so we do not need
436 to unhash it or make it invalid
439 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
440 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
441 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
446 Description: open req refcounting wrong on reconnect
447 Details : If reconnect happened between getting open reply from server and
448 call to mdc_set_replay_data in ll_file_open, we will schedule
449 replay for unreferenced request that we are about to free.
450 Subsequent close will crash in variety of ways.
451 Check that request is still eligible for replay in
452 mdc_set_replay_data().
457 Description: disable writes to filesystem when reading health_check file
458 Details : the default for reading the health_check proc file has changed
459 to NOT do a journal transaction and write to disk, because this
460 can cause reads of the /proc file to hang and block HA state
461 checking on a healthy but otherwise heavily loaded system. It
462 is possible to return to the previous behaviour during configure
463 with --enable-health-write.
465 Severity : enhancement
467 Description: 64-bit inode version
468 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
469 to the inode. This will be required for version-based recovery.
474 Description: MDS fails to start if a duplicate client export is detected
475 Details : in some rare cases it was possible for a client to connect to
476 an MDS multiple times. Upon recovery the MDS would detect this
477 and fail during startup. Handle this more gracefully.
479 Severity : enhancement
481 Description: Add -o localflock option to simulate old noflock
483 Details : This will achieve local-only flock/fcntl locks
489 Description: log_commit_thread vs filter_destroy race leads to crash
490 Details : Take import reference before releasing llog record semaphore
495 Description: Wrong request locking in request set processing
496 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
497 imp_delayed_list, in this place should be used imp_lock.
500 Frequency : when reconnection
502 Description: Grant Leak when osc reconnect to OST
503 Details : When osc reconnect ost, OST(filter) should check whether it
504 should grant more space to client by comparing fed_grant and
505 cl_avail_grant, and return the granted space to client instead
506 of "new granted" space, because client will call osc_init_grant
507 to update the client grant space info.
510 Frequency : when client reconnect to OST
512 Description: Grant Leak when osc do resend and replay bulk write
513 Details : When osc reconnect to OST, OST(filter)should clear grant info of
514 bulk write request, because the grant info while be sync between
515 OSC and OST when reconnect, and we should ignore the grant info
516 these of resend/replay write req.
521 Description: Grant space more than avaiable left space sometimes.
522 Details : When then OST is about to be full, if two bulk writing from
523 different clients came to OST. Accord the avaliable space of the
524 OST, the first req should be permitted, and the second one
525 should be denied by ENOSPC. But if the seconde arrived before
526 the first one is commited. The OST might wrongly permit second
527 writing, which will cause grant space > avaiable space.
530 Frequency : when client is evicted
532 Description: Grant might be wrongly erased when osc is evicted by OST
533 Details : when the import is evicted by server, it will fork another
534 thread ptlrpc_invalidate_import_thread to invalidate the
535 import, where the grant will be set to 0. While the original
536 thread will update the grant it got when connecting. So if
537 the former happened latter, the grant will be wrongly errased
538 because of this race.
543 Description: Checking Stale with correct fid
544 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
545 to check whether it is stale, because sometimes, we want the
546 enqueue happened anyway, and op_data.fid2 will not be initialized.
548 Severity : enhancement
550 Description: update patchless client
551 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
554 Frequency : only with 2.4 kernel
556 Description: random memory corruption
557 Details : size of struct ll_inode_info is to big for union inode.u and this
558 can be cause of random memory corruption.
563 Description: Memory leak in recovery
564 Details : Lov_mds_md was not free in an error handler in mds_create_object.
565 It should also check obd_fail before fsfilt_start, otherwise if
566 fsfilt_start return -EROFS,(failover mds during mds recovery).
567 then the req will return with repmsg->transno = 0 and rc = EROFS.
568 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
569 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
570 be freed no matter whether fsfilt_commit success or not.
573 Frequency : only with huge count clients
575 Description: Prevents from taking the superblock lock in llap_from_page for
577 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
578 for llap_from_page call in ll_removepage prevents from taking the
579 superblock lock for a soon died page.
584 Description: Not check open intent error before release open handle
585 Details : in some rare cases, the open intent error is not checked before
586 release open handle, which may cause
587 ASSERTION(open_req->rq_transno != 0), because it tries to release
588 the failed open handle.
593 Description: Set cat log bitmap only after create log success.
594 Details : in some rare cases, the cat log bitmap is set too early. and it
595 should be set only after create log success.
599 Description: Accessing a block bevice can re-enable I/O when Lustre is
600 tearing down a device.
601 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
606 Description: service threads may hog cpus when there are a lot of requests
608 Details : Insert cond_resched to give other threads a chance to use some of
614 Description: the cat log was not initialized in recovery
615 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
616 unlink log on mds will not be initialized until mds post
617 recovery. And also in mds post recovery, the unlink log will
618 initialization will be done asynchronausly, so there will be race
619 between add unlink log and unlink log initialization.
623 Description: brw_stats were being printed incorrectly
624 Details : brw_stats were being printed as log2 but all of them were not
625 recorded as log2. Also remove some code duplication arising from
626 filter_tally_{read,write}.
630 Frequency : rare, only in recovery.
631 Description: ASSERTION(req->rq_type != LI_POISON) failed
632 Details : imp_lock should be held while iterating over imp_sending_list for
633 prevent destroy request after get timeout in ptlrpc_queue_wait.
637 Description: replay-single.sh test 52 fails
638 Details : A lock's skiplist need to be cleanup when it being unlinked
639 from its resource list.
643 Description: Short directio read returns full requested size rather than
645 Details : Direct I/O operations should return actual amount of bytes
646 transferred rather than requested size.
648 Severity : enhancement
650 Description: metadata RPC reduction (e.g. for rm performance)
651 Details : decrease the amount of synchronous RPC between clients and servers
652 by canceling conflicing lock before the operation on the client side
653 and packing thier handles into the main operation RPC to server.
655 Severity : enhancement
657 Description: Async OSC create to avoid the blocking unnecessarily.
658 Details : If a OST has no remain object, system will block on the creating
659 when need to create a new object on this OST. Now, ways use
660 pre-created objects when available, instead of blocking on an
661 empty osc while others are not empty. If we must block, we block
662 for the shortest possible period of time.
664 --------------------------------------------------------------------------------
666 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
671 Frequency : on some architectures
673 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
674 Details : Uninitialized flags sometimes cause configuration commands to
678 Frequency : patchless clients only
680 Description: missing __iget() symbol export
681 Details : The __iget() symbol export is missing. To avoid the need for
682 this on patchless clients the deathrow inode reaper is turned
683 off, and we depend on the VM to clean up old inodes. This
684 dependency was during via the fix for bug 12181.
686 --------------------------------------------------------------------------------
688 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
690 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
691 INTEROPERATE with older versions automatically. In many cases a
692 special upgrade step is needed. Please read the
693 user documentation before upgrading any part of a live system.
694 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
695 version will not interoperate with 1.6 betas before beta5 (1.5.95).
696 * WARNING: Lustre configuration and startup changes are required with
697 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
703 Severity : enhancement
705 Description: MountConf
706 Details : Lustre configuration is now managed via mkfs and mount
707 commands instead of lmc and lconf. New obd types (MGS, MGC)
708 are added for dynamic configuration management. See
709 https://mail.clusterfs.com/wikis/lustre/MountConf for
712 Severity : enhancement
714 Description: dynamic OST addition
715 Details : OSTs can now be added to a live filesystem
717 Severity : enhancement
719 Description: startup order invariance
720 Details : MDTs and OSTs can be started in any order. Clients only
721 require the MDT to complete startup.
723 Severity : enhancement
725 Description: parallel, asynchronous orphan cleanup
726 Details : orphan cleanup is now performed in separate threads for each
727 OST, allowing parallel non-blocking operation.
729 Severity : enhancement
731 Description: optimized stripe assignment
732 Details : stripe assignments are now made based on ost space available,
733 ost previous usage, and OSS previous usage, in order to try
734 to optimize storage space and networking resources.
736 Severity : enhancement
738 Description: Permanently set tunables
739 Details : All writable /proc/fs/lustre tunables can now be permanently
740 set on a per-server basis, at mkfs time or on a live
743 Severity : enhancement
745 Description: Lustre message v2
746 Details : Add lustre message format v2.
748 Severity : enhancement
750 Description: client OST exclusion list
751 Details : Clients can be started with a list of OSTs that should be
752 declared "inactive" for known non-responsive OSTs.
756 Description: SPEC SFS validation failure on NFS v2 over lustre.
757 Details : Changes the blocksize for regular files to be 2x RPC size,
758 and not depend on stripe size.
760 Severity : enhancement
762 Description: Multiple MD RPCs in flight.
763 Details : Further unserialise some read-only MDS RPCs - learn about intents.
764 To avoid overly-overloading MDS, introduce a limit on number of
765 MDS RPCs in flight for a single client and add /proc controls
766 to adjust this limit.
768 Severity : enhancement
770 Description: client read/write statistics
771 Details : Add client read/write call usage stats for performance
772 analysis of user processes.
773 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
774 file access. extents_stats shows chunk size distribution.
775 extents_stats_per_process show chunk size distribution per
778 Severity : enhancement
780 Description: mds statistics
781 Details : Add detailed mds operations statistics in
782 /proc/fs/lustre/mds/*/stats.
786 Description: Failure of copying files with lustre special EAs.
787 Details : Client side always return success for setxattr call for lustre
788 special xattr (currently only "trusted.lov").
793 Description: Refcount LNET uuids
794 Details : The global LNET uuid list grew linearly with every startup;
795 refcount repeated list entries instead of always adding to
798 Severity : enhancement
800 Description: Dynamic service threads
801 Details : Within a small range, start extra service threads
802 automatically when the request queue builds up.
805 Frequency : mixed-endian client/server environments
807 Description: mixed-endian crashes
808 Details : The new msg_v2 system had some failures in mixed-endian
811 Severity : enhancement
813 Description: Easy OST removal
814 Details : OSTs can be permanently deactivated with e.g. 'lctl
815 conf_param lustre-OST0001.osc.active=0'
817 Severity : enhancement
819 Description: MGS proc entries
820 Details : Added basic proc entries for the MGS showing what filesystems
823 Severity : enhancement
825 Description: provide MGS failover
826 Details : Added config lock reacquisition after MGS server failover.
828 Severity : enhancement
830 Description: add Linux 2.4 support
831 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
835 Description: a large application tries to do I/O to the same resource and dies
837 Details : Check the req->rq_arrival time after the call to
838 ost_brw_lock_get(), but before we do anything about
839 processing it & sending the BULK transfer request. This
840 should help move old stale pending locks off the queue as
841 quickly as obd_timeout.
844 Frequency : when an incorrect nid is specified during startup
846 Description: ptlrpc connect to non-existant node causes kernel crash
847 Details : LNET can't be re-entered from an event callback, which
848 happened when we expire a message after the export has been
849 cleaned up. Instead, hand the zombie cleanup off to another
852 Severity : enhancement
854 Description: plain/inodebits lock performance improvement
855 Details : Grouping plain/inodebits in granted list by their request modes
856 and bits policy, thus improving the performance of search through
860 Frequency : only if OST filesystem is corrupted
862 Description: client incorrectly hits assertion in ptlrpc_replay_req()
863 Details : for a short time RPCs with bulk IO are in the replay list,
864 but replay of bulk IOs is unimplemented. If the OST filesystem
865 is corrupted due to disk cache incoherency and then replay is
866 started it is possible to trip an assertion. Avoid putting
867 committed RPCs into the replay list at all to avoid this issue.
870 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
872 Description: System hang on startup
873 Details : This bug allowed the liblustre (e.g. catamount) client to
874 return to the app before handling all startup RPCs. This
875 could leave the node unresponsive to lustre network traffic
876 and manifested as a server ptllnd timeout.
878 Severity : enhancement
880 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
881 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
882 boolean to control whether to print peer debug info when a
883 client's RPC times out.
886 Frequency : only for kernels with patches from Lustre below 1.4.3
888 Description: Remove old rdonly API
889 Details : Remove old rdonly API which unused from at least lustre 1.4.3
892 Frequency : only for devices with external journals
894 Description: Set external device read-only also
895 Details : During a commanded failover stop, we set the disk device
896 read-only while the server shuts down. We now also set any
897 external journal device read-only at the same time.
900 Frequency : when upgrading from 1.4 while trying to change parameters
902 Description: The wrong (new) MDC name was used when setting parameters for
903 upgraded MDT's. Also allows changing of OSC (and MDC)
904 parameters if --writeconf is specified at tunefs upgrade time.
907 Frequency : when setting specific ost indicies
909 Description: QOS code breaks on skipped indicies
910 Details : Add checks for missing OST indicies in the QOS code, so OSTs
911 created with --index need not be sequential.
913 Severity : enhancement
915 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
916 Details : The uninit_groups feature works in conjunction with the kernel
917 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
918 up the pass1 processing of e2fsck. This is a read-only feature
919 in ldiskfs2 only, so older kernels and current ldiskfs cannot
920 mount filesystems that have had this feature enabled.
922 Severity : enhancement
924 Description: Improve multi-block allocation algorithm to avoid fragmentation
925 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
926 allocation locality and avoid filesystem fragmentation.
928 ------------------------------------------------------------------------------
930 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
932 * Support for kernels:
933 2.4.21-47.0.1.EL (RHEL 3)
935 2.6.9-42.0.10.EL (RHEL 4)
936 2.6.12.6 vanilla (kernel.org)
937 2.6.16.27-0.9 (SLES 10)
938 * Recommended e2fsprogs version: 1.39.cfs5
940 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
944 Frequency : occasional, depends on client load and configuration
945 Bugzilla : 12181, 12203
946 Description: data loss for recently-modified files
948 Details : In some cases it is possible that recently written or created
949 files may not be written to disk in a timely manner (this should
950 normally be within 30s unless client IO load is very high).
951 The problem appears as zero-length files or files that are a
952 multiple of 1MB in size after a client crash or client eviction
953 that are missing data at the end of the file.
955 This problem is more likely to be hit on clients where files are
956 repeatedly created and unlinked in the same directory, clients
957 have a large amount of RAM, have many CPUs, the filesystem has
958 many OSTs, the clients are rebooted frequently, and/or the files
959 are not accessed by other nodes after being written.
961 The presence of the problem can be detected by looking at
962 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
963 smaller than the second (nr_unused) then dirty files will not
964 be flushed automatically to disk. "sync; sleep 10" should be
965 run several times on the node before unmounting it to update
966 Lustre (this is also safe to run on nodes without this problem).
968 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
969 that can cause this same problem, so customers using that kernel
970 also need to update the kernel in addition to Lustre. In order
971 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
973 It is normal that files written just before a client crash (less
974 than 30s) may not yet have been flushed to disk, even for local
978 Frequency : frequent on thin XT3 nodes
980 Description: UUID collision on thin XT3 Linux nodes
981 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
982 randomly, since we relied on an insufficiently-seeded PRNG.
987 Description: OSS hangs after "All ost request buffers busy"
988 Details : A deadlock between quota and journal operations caused OSS
989 hangs after printing "All ost request buffers busy."
992 Frequency : always on liblustre builds
994 Description: Cleanup compiler warnings on liblustre
997 Frequency : always on liblustre builds on XT3
999 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1003 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1004 > 32000 subdirectories
1005 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1006 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1007 subdirectory count crosses 32000. This will aid e2fsck to
1008 correctly handle more than 32000 subdirectories.
1011 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1013 Description: System hang on startup
1014 Details : This bug allowed the liblustre (e.g. catamount) client to
1015 return to the app before handling all startup RPCs. This
1016 could leave the node unresponsive to lustre network traffic
1017 and manifested as a server ptllnd timeout.
1019 Severity : enhancement
1021 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1022 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1023 boolean to control whether to print peer debug info when a
1024 client's RPC times out.
1029 Description: make O_SYNC working on 2.6 kernels
1030 Details : 2.6 kernels use different method for mark pages for write,
1031 so need add a code to lustre for O_SYNC work.
1036 Description: Failure to close file and release space on NFS
1037 Details : Put inode details into lock acquired in ll_intent_file_open.
1038 Use mdc_intent_lock in ll_intent_open to properly
1039 detect all kind of errors unhandled by mdc_enqueue.
1044 Description: proc file read during shutdown sometimes raced obd removal,
1046 Details : Add lock to prevent obd access after proc file removal.
1049 Frequency : Only for files larger than 4GB on 32-bit clients.
1051 Description: improperly doing page alignment of locks
1052 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1053 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1058 Description: RPCs being resent when they shouldn't be
1059 Details : Some RPCs that should not be resent are being resent. This
1060 can cause inconsistencies in the RPC state machine. Do not
1061 resend such requests.
1064 Frequency : rare, only with NFS export
1066 Description: Crash on NFS re-export node
1067 Details : under very unusual load conditions an assertion is hit in
1068 ll_intent_file_open()
1071 Frequency : only if OST filesystem is corrupted
1073 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1074 Details : for a short time RPCs with bulk IO are in the replay list,
1075 but replay of bulk IOs is unimplemented. If the OST filesystem
1076 is corrupted due to disk cache incoherency and then replay is
1077 started it is possible to trip an assertion. Avoid putting
1078 committed RPCs into the replay list at all to avoid this issue.
1083 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1084 Details : Large single O_DIRECT read and write calls can fail to allocate
1085 a sufficiently large buffer to process the request. In case of
1086 allocation failure the allocation is retried with a smaller
1087 buffer and broken into smaller requests.
1089 Severity : enhancement
1091 Description: Add -o localflock option to simulate old noflock behaviour.
1092 Details : This will achieve local-only flock/fcntl locks coherentness.
1097 Description: versioning check is incomplete
1098 Details : Checking the version difference of client vs. server, report
1099 error if the gap is too big.
1104 Description: add support PG_writeback bit
1105 Details : add support for PG_writeback bit for Lustre, for more carefull
1106 work with page cache in 2.6 kernel. This also fix some deadlocks
1107 and remove hack for work O_SYNC with 2.6 kernel.
1109 Severity : enhancement
1111 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1112 Details : The uninit_groups feature works in conjunction with the kernel
1113 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1114 up the pass1 processing of e2fsck. This is a read-only feature
1115 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1116 mount filesystems that have had this feature enabled.
1118 Severity : enhancement
1120 Description: Improve multi-block allocation algorithm to avoid fragmentation
1121 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1122 allocation locality and avoid filesystem fragmentation.
1124 ------------------------------------------------------------------------------
1126 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1128 * Support for kernels:
1129 2.6.9-42.0.3.EL (RHEL 4)
1130 2.6.5-7.276 (SLES 9)
1131 2.4.21-47.0.1.EL (RHEL 3)
1132 2.6.12.6 vanilla (kernel.org)
1133 2.6.16.21-0.8 (SLES10)
1134 * Recommended e2fsprogs version: 1.39.cfs2-0
1136 * The backwards-compatible /proc/sys/portals symlink has been removed
1137 in this release. Before upgrading, please ensure that you change
1138 any configuration scripts or /etc/sysctl.conf files that access
1139 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1140 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1141 in advance of the upgrade on any system running Lustre 1.4.6 or
1142 newer, since /proc/sys/lnet was added in that version.
1143 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1147 Frequency : only when quota is used
1149 Description: avoid scanning export list for quota master
1150 Details : Change the algorithms to avoid scanning export list in order
1151 to improve the efficiency.
1154 Frequency : MDS failover only, very rarely
1156 Description: "went back in time" messages on mds failover
1157 Details : The greatest transno may be lost when the current operation
1158 finishes with an error (transno==0) and the client's last_rcvd
1159 record is over-written. Save the greatest transno in the
1160 mds_last_transno for this case.
1163 Frequency : always for specific kernels and striping counts
1165 Description: client may get "Matching packet too big" without ACL support
1166 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1167 when trying to access files in certain configurations. The
1168 clients should in fact be denied when mounting because they do
1169 not understand ACLs.
1172 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1174 Description: many clients connecting with IO in progress causes connect timeouts
1175 Details : Avoid synchronous journal commits to avoid delays caused by many
1176 clients connecting/disconnecting when bulk IO is in progress.
1177 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1178 OST_IO_PORTAL to avoid delays behind potentially many pending
1182 Frequency : occasionally with multiple writers to a single file
1184 Description: shared writes to file may result in wrong size reported by stat()
1185 Details : Allow growing of kms when extent lock is cancelled
1188 Frequency : always with random mmap IO to multi-striped file
1190 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1191 Details : Only if the hole is at the end of OST object so that kms is too
1192 small. Fix is to increase kms accordingly in ll_nopage.
1195 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1197 Description: writes to a missing object would leak memory on the OST
1198 Details : If there is an inconsistency between the MDS and OST filesystems,
1199 such that the MDS references an object that doesn't exist, writes
1200 to that object will leak memory due to incorrect cleanup in the
1201 error handling path, eventually running out of memory on the OST.
1206 Description: Creating too long symlink causes lustre errors
1207 Details : Check symlink and name lengths before sending requests to MDS.
1210 Frequency : only if flock is enabled (not on by default)
1212 Description: posix locks not released on fd closure on 2.6.9+
1213 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1214 kernels, this caused such locks not to be released on fd close and
1215 then assertions on fs unmount about still used locks.
1218 Frequency : MDS failover only, very rarely
1220 Description: clients may get ASSERTION(granted_lock != NULL)
1221 Details : When request was taking a long time, and a client was resending
1222 a getattr by name lock request. The were multiple lock requests
1223 with the same client lock handle and
1224 mds_getattr_name->fixup_handle_for_resent_request found one of the
1225 lock handles but later failed with ASSERTION(granted_lock != NULL).
1230 Description: handle->h_buffer_credits > 0, assertion failure
1231 Details : h_buffer_credits is zero after truncate, causing assertion
1232 failure. This patch extends the transaction or creates a new
1236 Frequency : NFS re-export or patchless client
1237 Bugzilla : 11179, 10796
1238 Description: Crash on NFS re-export node (__d_move)
1239 Details : We do not want to hash the dentry if we don't have a lock.
1240 But if this dentry is later used in d_move, we'd hit uninitialised
1241 list head d_hash, so we just do this to init d_hash field but
1242 leave dentry unhashed.
1245 Frequency : NFS re-export or patchless client
1247 Description: NFS exports has problem with symbolic link
1248 Details : lustre client didn't properly install dentry when re-exported
1249 to NFS or running patchless client.
1252 Frequency : NFS re-export or patchless client
1254 Description: Various nfs/patchless fixes.
1255 Details : fixes reuse disconected alias for lookup process - this fixes
1256 warning "find_exported_dentry: npd != pd",
1257 fix permission error with open files at nfs.
1258 fix apply umask when do revalidate.
1261 Frequency : occasional
1263 Description: Crash on NFS re-export node
1264 Details : calling clear_page() on the wrong pointer triggered oops in
1265 generic_mapping_read().
1268 Frequency : rarely, using O_DIRECT IO
1270 Description: unaligned directio crashes client with LASSERT
1271 Details : check for unaligned buffers before trying any requests.
1274 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1276 Description: stale data returned from RAID cache
1277 Details : If only a small amount of IO is done to the RAID device before
1278 reading it again it is possible to get stale data from the RAID
1279 cache instead of reading it from disk.
1282 Frequency : always for sles10 kernel
1284 Description: sles10 support
1285 Details : ll_follow_link: compile fixes and using of nd_set_link
1286 under newer kernels.
1289 Frequency : depends on arch, kernel and compiler version, always on sles10
1292 Description: recursive or deep enough symlinks cause stack overflow
1293 Details : getting rid of large stack-allocated variable in
1297 Frequency : depends on hardware
1299 Description: lustre write performance loss in the SLES10 kernel
1300 Details : the performance loss is caused by using of write barriers in the
1301 ext3 code. The SLES10 kernel turns barrier support on by
1302 default. The fix is to undo that change for ldiskfs.
1304 ------------------------------------------------------------------------------
1306 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1308 * Support for kernels:
1309 2.6.9-42.0.3EL (RHEL 4)
1310 2.6.5-7.276 (SLES 9)
1311 2.4.21-47.0.1.EL (RHEL 3)
1312 2.6.12.6 vanilla (kernel.org)
1316 Frequency : quota enabled and large files being deleted
1318 Description: releasing more than 4GB of quota at once hangs OST
1319 Details : If a user deletes more than 4GB of files on a single OST it
1320 will cause the OST to spin in an infinite loop. Release
1321 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1326 Description: statfs data retrieved from /proc may be stale or zero
1327 Details : When reading per-device statfs data from /proc, in the
1328 {kbytes,files}_{total,free,avail} files, it may appear
1329 as zero or be out of date.
1332 Frequency : systems with MD RAID1 external journal devices
1334 Description: lconf's call to blkid is confused by RAID1 journal devices
1335 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1336 of returning all block devices that match the journal UUID.
1339 Frequency : always, for aggregate stripe size over 4GB
1341 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1342 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1343 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1346 Frequency : always if "lfs find" used on a local file/directory
1348 Description: "lfs find" segfaults if used on a local file/directory
1349 Details : The case where a directory component was not specified wasn't
1350 handled correctly. Handle this properly.
1353 Frequency : always on ppc64
1355 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1356 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1357 perform endianness conversion.
1360 Frequency : rarely (truncate to non-zero file size after write under load)
1361 Bugzilla : 10730, 10687
1362 Description: Files padded with zeros to next 4K multiple
1363 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1364 it is possible that files that are truncated to a non-zero size
1365 immediately after being written are filled with zero bytes beyond
1366 the truncated size. No file data is lost.
1368 Severity : enhancement
1370 Description: Allow recovery/failover for liblustre clients.
1371 Details : liblustre clients were unaware of failover configurations until
1374 Severity : enhancement
1376 Description: user file locks should fail when not mounting with flock option
1377 Details : Set up an error-returning stub in ll_file_operations.lock field
1378 to prevent incorrect behaviour when client is mounted without
1379 flock option. Also, set up properly f_op->flock field for
1383 Frequency : always on ia64
1385 Description: "lfs df" loops on printing out MDS statfs information
1386 Details : The obd_ioctl_data was not initialized and in some systems
1387 this caused a failure during the ioctl that did not return
1388 an error. Initialize the struct and return an error on failure.
1391 Frequency : SLES 9 only
1393 Description: Error of copying files with lustre special EAs as root
1394 Details : Client side always return success for setxattr call for lustre
1395 special xattr (currently only "trusted.lov").
1398 Frequency : rarely on clusters with both ia64+i386 clients
1400 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1401 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1402 clients are doing concurrent writes to the same file it is
1403 possible that the ia64 clients may LASSERT because the OST
1404 extent locks are not PAGE_SIZE aligned. Ensure that grown
1405 locks are always aligned on the request boundary.
1408 Frequency : specific use, occasional
1410 Description: Overwriting in use executable truncates on-disk binary image
1411 Details : If one node attempts to overwrite an executable in use by
1412 another node, we now correctly return ETXTBSY instead of
1413 truncating the file.
1415 Severity : enhancement
1417 Description: Async OSC create to avoid the blocking unnecessarily.
1418 Details : If a OST has no remain object, system will block on the creating
1419 when need to create a new object on this OST. Now, ways use
1420 pre-created objects when available, instead of blocking on an
1421 empty osc while others are not empty. If we must block, we block
1422 for the shortest possible period of time.
1427 Description: chmod on Lustre root is propagated to other clients
1428 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1429 invalid by the follow_mount time.
1432 Frequency : liblustre clients only
1434 Description: Race in 'instant cancel' lock handling could lead to such locks
1435 never to be granted in case of SMP MDS
1436 Details : Do not destroy not yet granted but cbpending locks in
1440 Frequency : replay/resend of open
1442 Description: non null lock assetion failure in mds_intent_policy
1443 Details : Trying to replay/resend lockless open requests resulted in
1444 mds_open() returning 0 with no lock. Now it sets a flag if
1445 a lock is going to be returned.
1447 Severity : enhancement
1449 Description: Checksum enhancements
1450 Details : New checksum enhancements allow for resending RPCs that failed
1453 Severity : enhancement
1455 Description: Tunables on number of dirty pages in cacche
1456 Details : Allow to set limit on number of dirty pages cached.
1461 Description: client crash on unmount - lock still has references
1462 Details : In some error handling cases it was possible to leak a lock
1463 reference on a client while accessing a file. This was not
1464 harmful to the client during operation, but would cause the
1465 client to crash when the filesystem is unmounted.
1468 Frequency : specific case, rare
1470 Description: ETXTBSY on mds though file not in use by client
1471 Details : ETXTBSY is no longer incorrectly returned when attempting to
1472 chmod or chown a directory that the user previously tried to
1473 execute or a currently-executing binary.
1476 Frequency : extremely rare except on liblustre-based clients
1478 Description: Lustre space not freed when files are deleted
1479 Details : Clean up open-unlinked files after client eviction. Previously
1480 the unlink was skipped and the files remained as orphans.
1485 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1486 Details : In some cases it was possible to send a blocking callback to
1487 a client doing a glimpse, even though that client didn't get
1488 a lock granted. When the glimpse lock is cancelled on the OST
1489 the freed lock is left on the waiting list and corrupted the list.
1492 Frequency : all core dumps
1494 Description: Broke core dumps to lustre
1495 Details : Negative dentry may be unhashed if parent does not have UPDATE
1496 lock, but some callers, e.g. do_coredump, expect dentry to be
1497 hashed after successful create, hash it in ll_create_it.
1499 ------------------------------------------------------------------------------
1501 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1503 * Support for kernels:
1504 2.6.9-42.0.2.EL (RHEL 4)
1505 2.6.5-7.276 (SLES 9)
1506 2.4.21-40.EL (RHEL 3)
1507 2.6.12.6 vanilla (kernel.org)
1511 Frequency : always on RHEL 3
1513 Description: Number of open files grows over time
1514 Details : The number of open files grows over time, whether or not
1515 Lustre is started. This was due to a filp leak introduced
1516 by one of our kernel patches.
1518 ------------------------------------------------------------------------------
1520 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1522 * Support for kernels:
1523 2.6.9-42.EL (RHEL 4)
1524 2.6.5-7.276 (SLES 9)
1525 2.4.21-40.EL (RHEL 3)
1526 2.6.12.6 vanilla (kernel.org)
1531 Bugzilla : 5719, 9635, 9792, 9684
1532 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1533 Details : If a server is under heavy load and cannot reply to new
1534 connection requests before the client resends the (re)connect,
1535 the connection handling code can behave badly if two service
1536 threads are concurrently handing separate (re)connections from
1537 the same client. Add better locking to the connection handling
1538 code, and ensure that only a single connection will be processed
1539 for a given client UUID, even if the lock is dropped.
1541 Severity : enhancement
1543 Description: add TCP zero-copy support to kernel
1544 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1545 sends if the hardware supports scatter-gather and checksumming.
1546 This allows socklnd to do client-write and server-read more
1547 efficiently and reduce CPU utilization from skbuf copying.
1550 Frequency : only if NFS exporting from client
1552 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1553 Details : If setting attributes on a file created under NFS that had
1554 never been opened it would be possible to oops the client
1555 if the file had no objects.
1558 Frequency : always for liblustre
1560 Description: liblustre client does MDS+OSTs setattr RPC for each write
1561 Details : When doing a write from a liblustre client, the client
1562 incorrectly issued an RPC to the MDS and each OST the file was
1563 striped over in order to update the timestamps. When writing
1564 with small chunks and many clients this could overwhelm the MDS
1565 with RPCs. In all cases it would slow down the write because
1566 these RPCs are unnecessary.
1568 Severity : enhancement
1570 Description: allow number of MDS service threads to be changed at module load
1571 Details : It is now possible to change the number of MDS service threads
1572 running. Adding "options mds mds_num_threads={N}" to the MDS's
1573 /etc/modprobe.conf will set the number of threads for the next
1574 time Lustre is restarted (assuming the "mds" module is also
1575 reloaded at that time). The default number of threads will
1576 stay the same, 32 for most systems.
1581 Description: OST crash if filesystem is unformatted or corrupt
1582 Details : If an OST is started on a device that has never been formatted
1583 or if the filesystem is corrupt and cannot even mount then the
1584 error handling cleanup routines would dereference a NULL pointer.
1589 Description: NULL pointer deref in llap_from_page.
1590 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1591 incorrect) mapping if the page was truncated/reclaimed while it was
1592 searched for. Check for this condition and skip such pages when
1593 doing readahead. Introduce extra check to llap_from_page() to
1594 verify page->mapping->host is non-NULL (so page is not anonymous).
1597 Frequency : Sometimes when using sys_sendfile
1599 Description: "page not covered by a lock" warnings from ll_readpage
1600 Details : sendfile called ll_readpage without right page locks present.
1601 Now we introduced ll_file_sendfile that does necessary locking
1602 around call to generic_file_sendfile() much like we do in
1606 Frequency : with certain MDS communication failures at client mount time
1608 Description: NULL pointer deref after failed client mount
1609 Details : a client connection request may delayed by the network layer
1610 and not be sent until after the PTLRPC layer has timed out the
1611 request. If the client fails the mount immediately it will try
1612 to clean up before the network times out the request. Add a
1613 reference from the request import to the obd device and delay
1614 the cleanup until the network drops the request.
1617 Frequency : occasionally during client (re)connect
1619 Description: assertion failure during client (re)connect
1620 Details : processing a client connection request may be delayed by the
1621 client or server longer than the client connect timeout. This
1622 causes the client to resend the connection request. If the
1623 original connection request is replied in this interval, the
1624 client may trip an assertion failure in ptlrpc_connect_interpret()
1625 which thought it would be the only running connect process.
1628 Frequency : only with obd_echo servers and clients that are rebooted
1630 Description: kernel BUG accessing uninitialized data structure
1631 Details : When running an obd_echo server it did not start the ping_evictor
1632 thread, and when a client was evicted an uninitialized data
1633 structure was accessed. Start the ping_evictor in the RPC
1634 service startup instead of the OBD startup.
1636 Severity : enhancement
1637 Bugzilla : 10193 (patchless)
1638 Description: Remove dependency on various unexported kernel interfaces.
1639 Details : No longer need reparent_to_init, exit_mm, exit_files,
1640 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1643 Frequency : rare (only users of deprecated and unsupported LDAP config)
1645 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1646 Details : LDAP apparently contains 'lustreName' attributes instead of
1647 'name'. A simple remapping of the name is sufficient.
1650 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1652 Description: waiting_locks_callback trips kernel BUG if client is evicted
1653 Details : Running with the dump_on_timeout debug flag turned on makes
1654 it possible that the waiting_locks_callback() can try to dump
1655 the Lustre kernel debug logs from an interrupt handler. Defer
1656 this log dumping to the expired_lock_main() thread.
1658 Severity : enhancement
1660 Description: Support NFS exporting on 2.6 kernels.
1661 Details : Implement non-rawops metadata methods for NFS server to use without
1662 changing NFS server code.
1665 Frequency : very rare (synthetic metadata workload only)
1667 Description: two racing renames might cause an MDS thread to deadlock
1668 Details : Running the "racer" program may cause one MDS thread to rename
1669 a file from being the source of a rename to being the target of
1670 a rename at exactly the same time that another thread is doing
1671 so, and the second thread has already enqueued these locks after
1672 doing a lookup of the target and is trying to relock them in
1673 order. Ensure that we don't try to re-lock the same resource.
1676 Frequency : only very large systems with liblustre clients
1678 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1679 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1680 all OSTs in parallel. This allows the eviction of stale liblustre
1681 clients to proceed much faster than if they were done in series,
1682 and also offers similar improvements for other set_info RPCs.
1687 Description: excessive CPU usage during initial read phase on client
1688 Details : During the initial read phase on a client, it would agressively
1689 retry readahead on the file, consuming too much CPU and impacting
1690 performance (since 1.4.5.8). Improve the readahead algorithm
1691 to avoid this, and also improve some other common cases (read
1692 of small files in particular, where "small" is files smaller than
1693 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1698 Description: MDS crash when receiving packet with unknown intent.
1699 Details : Do not LBUG in unknown intent case, just return -EFAULT
1701 Severity : enhancement
1702 Bugzilla : 9293, 9385
1703 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1704 Details : Do not serialize getattr (non-intent version) and statfs.
1707 Frequency : occasional, when OST network is overloaded/intermittent
1709 Description: client evicted by OST after bulk IO timeout
1710 Details : If a client sends a bulk IO request (read or write) the OST
1711 may evict the client if it is unresposive to its data GET/PUT
1712 request. This is incorrect if the network is overloaded (takes
1713 too long to transfer the RPC data) or dropped the OST GET/PUT
1714 request. There is no need to evict the client at all, since
1715 the pinger and/or lock callbacks will handle this, and the
1716 client can restart the bulk request.
1719 Frequency : Always when mmapping file with no objects
1721 Description: client crashes when mmapping file with no objects
1722 Details : Check that we actually have objects in a file before doing any
1723 operations on objects in ll_vm_open, ll_vm_close and
1729 Description: Request leak when working with deleted CWD
1730 Details : Introduce advanced request refcount tracking for requests
1731 referenced from lustre intent.
1733 Severity : Enhancement
1735 Description: Cache open file handles on client.
1736 Details : MDS now will return special lock along with openhandle, if
1737 requested and client is allowed to hold openhandle, even if unused,
1738 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1739 closing files for every read/write openration.
1741 Severity : Enhancement
1743 Description: Cache open negative dentries on client when possible.
1744 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1745 negative dentries on lock revocation.
1750 Description: Remounting a client read-only wasn't possible with a zconf mount
1751 Details : It wasn't possible to remount a client read-only with llmount.
1753 Severity : enhancement
1754 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1755 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1756 MPICH is included to add Lustre-specific ADIO interfaces.
1757 This is based closely on the UFS ADIO layer and only differs
1758 in file creation, in order to allow the OST striping to be set.
1759 This is user-contributed code and not supported by CFS.
1764 Description: extended inode attributes (immutable, append-only) work improperly
1765 when 2.4 and 2.6 kernels are used on client/server or vice versa
1766 Details : Introduce kernel-independent values for these flags.
1768 Severity : enhancement
1771 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1772 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1773 granularity. Now, Fractional MB granularity is be supported,
1774 this is very useful for low memory system.
1776 Severity : enhancement
1778 Description: Getattr by fid
1779 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1780 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1781 have locks and avoids getattr rpc requests for it on every stat.
1784 Frequency : Always, for filesystems larger than 2TB
1786 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1787 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1788 filesystems larger than 2TB to be mounted with mballoc enabled.
1791 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1793 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1794 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1795 filesystems larger than 2TB is mounted then the kernel will
1796 silently corrupt the start of the filesystem. CONFIG_LBD is
1797 enabled for all CFS-supported kernels, but the possibility of
1798 this happening with a modified kernel config exists.
1800 Severity : enhancement
1802 Description: add client O_DIRECT support for 2.6 kernels
1803 Details : It is now possible to do O_DIRECT reads and writes to files
1804 in the Lustre client mountpoint on 2.6 kernel clients.
1806 Severity : enhancement
1808 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1809 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1810 parallel, not waiting for response from every OST before sending
1811 a rpc to the next OST.
1816 Description: setattr vs write race when updating file timestamps
1817 Details : Client processes that update a file timestamp into the past
1818 right after writing to the file (e.g. tar) it is possible that
1819 the updated file modification time can be reset to the current
1820 time due to a race between processing the setattr and write RPC.
1822 Severity : enhancement
1824 Description: Bring 'lfs find' closer in line with regular Linux find.
1825 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1826 -print0 options and obtains all the needed info through the lustre
1829 Severity : enhancement
1831 Description: support up to 1024 configured devices on one node
1832 Details : change obd_dev array from statically allocated to dynamically
1833 allocated structs as they are first used to reduce memory usage
1838 Description: Flush dirty partially truncated pages during truncate
1839 Details : Immediatelly flush partially truncated pages in filter_setattr,
1840 this way we completely avoid having any pages in page cache on OST
1841 and can retire ugly workarounds during writes to flush such pages.
1846 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1847 Details : protect inode from truncation within vfs_unlink() context
1848 just take a reference before calling vfs_unlink() and release it
1849 when parent's i_sem is free.
1854 Description: last_id value checked outside lock on OST caused LASSERT failure
1855 Details : If there were multiple MDS->OST object precreate requests in
1856 flight, it was possible that the OST's last object id was checked
1857 outside a lock and incorrectly tripped an assertion. Move checks
1858 inside locks, and discard old precreate requests.
1861 Frequency : always, if extents are used on OSTs
1863 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1864 Details : OSTs using the extents format would not zero the high 16 bits of
1865 the index physical block number. This is not a problem for any
1866 OST filesystems smaller than 16TB, and no kernels support ext3
1867 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1868 new/modified files) and can be fixed for existing filesystems
1869 with e2fsprogs-1.39-cfs1.
1874 Description: import connection selection may be incorrect if timer wraps
1875 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1876 import connections to be ignored if the 32-bit jiffies counter
1877 wraps. Use a 64-bit jiffies counter.
1880 Frequency : very large clusters immediately after boot
1882 Description: LNET request buffers exhausted under heavy short-term load
1883 Details : If a large number of client requests are generated on a service
1884 that has previously never seen so many requests it is possible
1885 that the request buffer growth cannot keep up with the spike in
1886 demand. Instead of dropping incoming requests, they are held in
1887 the LND until the RPC service can accept more requests.
1890 Frequency : Sometimes during replay
1892 Description: Assertion failure in ll_local_open after replay.
1893 Details : If replay happened on an open request reply before we were able
1894 to set replay handler, reply will become not swabbed tripping the
1895 assertion in ll_local_open. Now we set the handler right after
1896 recognising of open request
1899 Frequency : very rare
1901 Description: kernel reports "badness in vsnprintf"
1902 Details : Reading from the "recovery_status" /proc file in small chunks
1903 may cause a negative length in lprocfs_obd_rd_recovery_status()
1904 call to vsnprintf() (which is otherwise harmless). Exit early
1905 if there is no more space in the output buffer.
1907 Severity : enhancement
1909 Description: clear OBD RPC statistics by writing to them
1910 Details : It is now possible to clear the OBD RPC statistics by writing
1911 to the "stats" file.
1916 Description: Client mtime is not the same on different clients after utimes
1917 Details : In some cases, the client was using the utimes() syscall on
1918 a file cached on another node. The clients now validate the
1919 ctime from the MDS + OSTs to determine which one is right.
1924 Description: Inability to activate failout mode
1925 Details : lconf script incorrectly assumed that in pythong string's numeric
1926 value is used in comparisons.
1929 Frequency : always with multiple stripes per file
1931 Description: Inefficient object allocation for mutli-stripe files
1932 Details : When selecting which OSTs to stripe files over, for files with
1933 a stripe count that divides evenly into the number of OSTs,
1934 the MDS is always picking the same starting OST for each file.
1935 Return the OST selection heuristic to the original design.
1940 Description: mount failures may take full timeout to return an error
1941 Details : Under some heavy load conditions it is possible that a
1942 failed mount can wait for the full obd_timeout interval,
1943 possibly several minutes, before reporting an error.
1944 Instead return an error as soon as the status is known.
1946 Frequency : quota enabled and large files being deleted
1948 Description: releasing more than 4GB of quota at once hangs OST
1949 Details : If a user deletes more than 4GB of files on a single OST it
1950 will cause the OST to spin in an infinite loop. Release
1951 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1956 Description: statfs data retrieved from /proc may be stale or zero
1957 Details : When reading per-device statfs data from /proc, in the
1958 {kbytes,files}_{total,free,avail} files, it may appear
1959 as zero or be out of date.
1962 Frequency : systems with MD RAID1 external journal devices
1964 Description: lconf's call to blkid is confused by RAID1 journal devices
1965 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1966 of returning all block devices that match the journal UUID.
1969 Frequency : always, for aggregate stripe size over 4GB
1971 Description: assertion fails when trying to use 4GB stripe size
1972 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1973 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1976 Frequency : always on ppc64
1978 Description: the first write on an ext3 filesystem with mballoc got stuck
1979 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1980 perform endianness conversion.
1982 ------------------------------------------------------------------------------
1984 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1986 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1987 INTEROPERATE with older versions automatically. Please read the
1988 user documentation before upgrading any part of a live system.
1989 * WARNING: Lustre networking configuration changes are required with
1990 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1993 * Support for kernels:
1994 2.6.9-22.0.2.EL (RHEL 4)
1995 2.6.5-7.244 (SLES 9)
1996 2.6.12.6 vanilla (kernel.org)
1999 Severity : enhancement
2000 Bugzilla : 7981/8208
2001 Description: Introduced Lustre Networking (LNET)
2002 Details : LNET is new networking infrastructure for Lustre, it includes
2003 a reorganized network configuration mode (see the user
2004 documentation for full details) as well as support for routing
2005 between different network fabrics. Lustre Networking Devices
2006 (LNDs) for the supported network fabrics have also been
2007 created for this new infrastructure.
2009 Severity : enhancement
2010 Description: Introduced Access control lists
2011 Details : clients can set ACLs on files and directories in order to have
2012 more fine-grained permissions than the standard Unix UGO+RWX.
2013 The MDS must be started with the "-o acl" mount option.
2015 Severity : enhancement
2016 Description: Introduced filesystem quotas
2017 Details : Administrators may now establish per-user quotas on the
2020 Severity : enhancement
2022 Description: Configuration change for the XT3
2023 The PTLLND is now used to run Lustre over Portals on the XT3
2024 The configure option(s) --with-cray-portals are no longer used.
2025 Rather --with-portals=<path-to-portals-includes> is used to
2026 enable building on the XT3. In addition to enable XT3 specific
2027 features the option --enable-cray-xt3 must be used.
2032 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2033 Details : A race between allocating a new llcd and re-getting the llcd_lock
2034 allowed another thread to grab newly-allocated llcd.
2036 Severity : enhancement
2038 Description: 2.6 OST async journal commit and locking fix to improve performance
2039 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2040 2.6 kernels are now async as they already were in 2.4 kernels so
2041 that they can commit concurrently with the network bulk transfer.
2042 For block-allocated files the filter allocation semaphore is held
2043 to avoid filesystem fragmentation during allocation. BKL lock
2044 removed for 2.6 xattr operations where it is no longer needed.
2049 Description: lconf incorrectly determined whether two IP networks could talk
2050 Details : In some more complicated routing and multiple-network
2051 configurations, lconf will avoid trying to make a network
2052 connection to a disjoint part of the IP space. It was doing the
2053 math incorrectly for one set of cases.
2058 Description: Fix for potential infinite loop processing records in an llog.
2059 Details : If an llog record is corrupted/zeroed, it is possible to loop
2060 forever in llog_process(). Validate the llog record length
2061 and skip the remainder of the block on error.
2064 Frequency : occasional (liblustre only)
2066 Description: liblustre could not open files whose last component is a symlink
2067 Details : sysio_path_walk() would incorrectly pass the open intent to
2068 intermediate path components.
2071 Frequency : rare (liblustre only with non-standard tuning)
2072 Bugzilla : 7201 (7350)
2073 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2074 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2075 after it has been released, only do this for referenced locks
2077 Severity : enhancement
2079 Description: specifying an (invalid) directory default stripe_size of -1
2080 would reset the directory default striping
2081 Details : stripe_size -1 was used internally to signal directory stripe
2082 removal, now use "all default" to signal dir stripe removal
2083 as a directory striping of "all default" is not useful
2086 Frequency : common for large clusters running liblustre clients
2088 Description: doing an ls when liblustre clients are running is slow
2089 Details : sending a glimpse AST to a liblustre client waits for every AST
2090 to time out, as liblustre clients will not respond. Since they
2091 cannot cache data we refresh the OST lock LVB from disk instead.
2093 Severity : enhancement
2095 Description: doing an ls at the same time as file IO can be slow
2096 Details : enqueue and other "small" requests can be blocked behind many
2097 large IO requests. Create a new OST IO portal for non-IO
2098 requests so they can be processed faster.
2101 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2103 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2104 created file with mode 000
2105 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2106 allow an ftruncate() as a non-root user to a file with mode 000.
2107 Lustre now respects this flag to disable mode checks when
2108 truncating a file owned by the user
2111 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2114 Description: Revoking locks from clients that went dead or catatonic might take
2116 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2117 cancellation of such locks instant on servers without waiting for
2118 any reply from clients. Clients drops these locks when cancel
2119 notification from server is received without replying.
2122 Frequency : liblustre-only, when liblustre client dies or becomes busy
2124 Description: Doing ls on Linux clients can take a long time with active
2126 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2127 granting such locks to it in the first place if possible. Locks
2128 are taken by proxy on the OST during the read or write and
2129 dropped immediately afterward. Add connect flags handling, do
2130 not grant locks to liblustre clients for glimpse ASTs.
2132 Severity : enhancement
2134 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2135 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2136 sequential reads and behaves badly with applications doing random
2137 reads. Improve it by reading ahead at least read region, and
2138 avoiding excessive large RPC for small reads.
2140 Severity : enhancement
2142 Description: Creating more than 1000 files for a single job may cause a load
2143 imbalance on the OSTs if there are also a large number of OSTs.
2144 Details : qos_prep_create() uses an OST index reseed value that is an
2145 even multiple of the number of available OSTs so that if the
2146 reseed happens in the middle of the object allocation it will
2147 still utilize the OSTs as uniformly as possible.
2152 Description: OST or MDS may oops in ping_evictor_main()
2153 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2154 but doesn't restart at beginning of obd_exports_timed list
2157 Severity : enhancement
2159 Description: improve by-nid export eviction on the MDS and OST
2160 Details : allow multiple exports with the same NID to be evicted at one
2161 time without re-searching the exports list.
2164 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2166 Description: MDS may oops in groups_free()
2167 Details : in rare race conditions a newly allocated group_info struct is
2168 freed again, and this can be NULL. The 2.4 compatibility code
2169 for groups_free() checked for a NULL pointer, but 2.6 did not.
2172 Frequency : common for liblustre clients doing little filesystem IO
2173 Bugzilla : 9352, 7313
2174 Description: server may evict liblustre clients accessing contended locks
2175 Details : if a client is granted a lock or receives a completion AST
2176 with a blocking AST already set it would not reply to the AST
2177 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2180 Frequency : lfs setstripe, only systems with more than 160 OSTs
2182 Description: unable to set striping with a starting offset beyond OST 160
2183 Details : llapi_create_file() incorrectly limited the starting stripe
2184 index to the maximum single-file stripe count.
2187 Frequency : LDAP users only
2189 Description: lconf did not handle in-kernel recovery with LDAP properly
2190 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2192 Severity : enhancement
2194 Description: bind OST threads to NUMA nodes to improve performance
2195 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2196 node and do their allocations there to localize memory access
2198 Severity : enhancement
2200 Description: llmount can determine client NID directly from Myrinet (GM)
2201 Details : the client NID code from gmnalnid was moved directly into
2202 llmount, removing the need to use this or specifying the
2203 client NID explicitly when mounting GM clients with zeroconf
2206 Frequency : if client is started with down MDS
2208 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2209 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2210 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2211 ptlrpc_import_delayed_req() because we didn't check if the
2212 request was interrupted, and we also didn't break out of the
2213 event loop if there was no timeout
2218 Description: data loss during non-page-aligned writes to a single file from
2219 both multiple nodes and multiple threads on one node at same time
2220 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2221 inconsistency led to false short-reads, that were cached and later
2222 used by ->prepare_write() to fill in partially written page,
2223 leading to data loss.
2226 Frequency : always, if lconf --abort_recovery used
2228 Description: lconf --abort_recovery fails with 'Operation not supported'
2229 Details : lconf was attempting to abort recovery on the MDT device and not
2232 Severity : enhancement
2234 Description: remove cleanup logs
2235 Details : replace lconf-generated cleanup logs with lustre internal
2236 cleanup routines. Eliminates the need for client-cleanup and
2239 Severity : enhancement
2241 Description: add support for EAs (user and system) on lustre filesystems
2242 Details : it is now possible to store extended attributes in the Lustre
2243 client filesystem, and with the user_xattr mount option it
2244 is possible to allow users to store EAs on their files also
2246 Severity : enhancement
2248 Description: Add possibility (config option) to show minimal available OST free
2250 Details : When compiled with --enable-mindf configure option, statfs(2)
2251 (and so, df) will return least minimal free space available from
2252 all OSTs as amount of free space on FS, instead of summary of
2253 free spaces of all OSTs.
2255 Severity : enhancement
2257 Description: do not expand extent locks acquired on OST-side
2258 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2259 by server: they are not cached anyway.
2262 Frequency : when mmap is used/binaries executed from Lustre
2264 Description: Unmmap pages before throwing them away from read cache.
2265 Details : llap_shrink cache now attempts to unmap pages before discarding
2266 them (if unmapping failed - do not discard). SLES9 kernel has
2267 extra checks that trigger if this unmapping is not done first.
2272 Description: lconf didn't resolve symlinks before checking to see whether a
2273 given mountpoint was already in use
2276 Frequency : when migrating failover services
2277 Bugzilla : 6395, 9514
2278 Description: When migrating a subset of services from a node (e.g. failback
2279 from a failover service node) the remaining services would
2280 time out and evict clients.
2281 Details : lconf --force (implied by --failover) sets the global obd_timeout
2282 to 5 seconds in order to quickly disconnect, but this caused
2283 other RPCs to time out too quickly. Do not change the global
2284 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2286 Severity : enhancement
2287 Frequency : if MDS is started with down OST
2288 Bugzilla : 9439,5706
2289 Description: Allow startup/shutdown of an MDS without depending on the
2290 availability of the OSTs.
2291 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2292 Add appropriate locking and lov-osc refcounts for safe
2293 cleaning. Add osc abort_inflight calls in case the
2294 synchronize never started.
2297 Frequency : occasional (Cray XT3 only)
2299 Description: root not authorized to access files in CRAY_PORTALS environment
2300 Details : The client process capabilities were not honoured on the MDS in
2301 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2302 been accessed by an authorized user then root was able to access
2303 the file on the local client also. The root user capabilities
2304 are now allowed on the MDS, as this environment has secure UID.
2307 Frequency : occasional
2309 Description: ldiskfs "too long searching" message happens too often
2310 Details : A debugging message (otherwise harmless) prints too often on
2311 the OST console. This has been reduced to only happen when
2312 there are fragmentation problems on the filesystem.
2317 Description: Division by zero in statfs when all OSCs are inactive
2318 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2319 lov_statfs divides by value returned by lov_get_stripecnt().
2323 Bugzilla : 9489, 3273
2324 Description: First write from each client to each OST was only 4kB in size,
2325 to initialize client writeback cache, which caused sub-optimal
2326 RPCs and poor layout on disk for the first writen file.
2327 Details : Clients now request an initial cache grant at (re)connect time
2328 and so that they can start streaming writes to the cache right
2329 away and always do full-sized RPCs if there is enough data.
2330 If the OST is rebooted the client also re-establishes its grant
2331 so that client cached writes will be honoured under the grant.
2336 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2337 Details : Now I/O RPCs go to different portal number and (presumably) fast
2338 lock requests (and glimses) and other RPCs get their own service
2339 threads pool that should be able to service those RPCs
2342 Severity : enhancement
2344 Description: Ability to exchange lustre version between client and servers and
2345 issue warnings at client side if client is too old. Also for
2346 liblustre clients there is ability to refuse connection of too old
2348 Details : New 'version' field is added to connect data structure that is
2349 filled with version info. That info is later checked by server and
2353 Frequency : rare, liblustre only.
2354 Bugzilla : 9296, 9581
2355 Description: Two simultaneous writes from liblustre at offset within same page
2356 might proceed at the same time overwriting eachother with stale
2358 Details : I/O lock withing llu_file_prwv was released too early, before data
2359 actually was hitting the wire. Extended lock-holding time until
2360 server acknowledges receiving data.
2363 Frequency : extremely rare. Never observed in practice.
2365 Description: avoid generating lustre_handle cookie of 0.
2366 Details : class_handle_hash() generates handle cookies by incrementing
2367 global counter, and can hit 0 occasionaly (this is unlikely, but
2368 not impossible, because initial value of cookie counter is
2369 selected randonly). Value of 0 is used as a sentinel meaning
2370 "unassigned handle" --- avoid it. Also coalesce two critical
2371 sections in this function into one.
2373 Severity : enhancement
2375 Description: allow liblustre clients to delegate truncate locking to OST
2376 Details : To avoid overhead of locking, liblustre client instructs OST to
2377 take extent lock in ost_punch() on client's behalf. New connection
2378 flag is added to handle backward compatibility.
2380 Severity : enhancement
2381 Bugzilla : 4928, 7341, 9758
2382 Description: allow number of OST service threads to be specified
2383 Details : a module parameter allows the number of OST service threads
2384 to be specified via "options ost ost_num_threads={N}" in the
2385 OSS's /etc/modules.conf or /etc/modprobe.conf.
2389 Bugzilla : 6146, 9635, 9895
2390 Description: servers crash with bad pointer in target_handle_connect()
2391 Details : In rare cases when a client is reconnecting it was possible that
2392 the connection request was the last reference for that export.
2393 We would temporarily drop the export reference and get a new
2394 one, but this may have been the last reference and the export
2395 was just destroyed. Get new reference before dropping old one.
2397 Severity : enhancement
2398 Frequency : if client is started with failover MDS
2400 Description: Allow multiple MDS hostnames in the mount command
2401 Details : Try to read the configuration from all specified MDS
2402 hostnames during a client mount in case the "primary"
2405 Severity : enhancement
2407 Description: Stop sending data to evicted clients as soon as possible.
2408 Details : Check if the client we are about to send or are sending data to
2409 was evicted already. (Check is done every second of waiting,
2410 for which l_wait_event interface was extended to allow checking
2411 of exit condition at specified intervals).
2414 Frequency : rare, normally only when NFS exporting is done from client
2416 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2418 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2421 Frequency : always, if srand() is called before liblustre initialization
2423 Description: Liblustre uses system PRNG disturbing its usage by user application
2424 Details : Introduce internal to lustre fast and high-quality PRNG for
2425 lustre usage and make liblustre and some other places in generic
2426 lustre code to use it.
2428 Severity : enhancement
2429 Bugzilla : 9477, 9557, 9870
2430 Description: Verify that the MDS configuration logs are updated when xml is
2431 Details : Check if the .xml configuration logs are newer than the config
2432 logs stored on the MDS and report an error if this is the case.
2433 Request --write-conf, or allow starting with --old_conf.
2435 Severity : enhancement
2437 Description: Handle symlinks in the path when checking if Lustre is mounted.
2438 Details : Resolve intermediate symlinks when checking if a client has
2439 mounted a filesystem to avoid duplicate client mounts.
2444 Description: lconf can hit an error exception but still return success.
2445 Details : The lconf command catches the Command error exception at the top
2446 level script context and will exit with the associated exit
2447 status, but doesn't ensure that this exit status is non-zero.
2452 Description: failure of ptlrpc thread startup can cause oops
2453 Details : Starting a ptlrpc service thread can fail if there are a large
2454 number of threads or the server memory is very fragmented.
2455 Handle this without oopsing.
2458 Frequency : always, only if liblustre and non-default acceptor port was used
2460 Description: liblustre cannot connect to servers with non-default acceptor port
2461 Details : tcpnal_set_default_params() was not called and was therefore
2462 ignoring the environment varaible TCPNAL_PORT, as well as other
2463 TCPNAL_ environment variables
2468 Description: two objects could be created on the same OST for a single file
2469 Details : If an OST is down, in some cases it was possible to create two
2470 objects on a single OST for a single file. No problems other
2471 than potential performance impact and spurious error messages.
2475 Bugzilla : 5681, 9562
2476 Description: Client may oops in ll_unhash_aliases
2477 Details : Client dcache may become inconsistent in race condition.
2478 In some cases "getcwd" can fail if the current directory is
2484 Description: Inode refcounting problems in NFS export code
2485 Details : link_raw functions used to call d_instantiate without obtaining
2486 extra inode reference first.
2490 Bugzilla : 9942, 9903
2491 Description: Referencing freed requests leading to crash, memleaks with NFS.
2492 Details : We used to require that call to ll_revalidate_it was always
2493 followed by ll_lookup_it. Also with revalidate_special() it is
2494 possible to call ll_revalidate_it() twice for the same dentry
2495 even if first occurence returned success. This fix changes semantic
2496 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2497 reference on a request referred from the intent.
2498 ll_intent_release() then releases such a request.
2501 Frequency : rare, normally benchmark loads only
2503 Description: unlinked inodes were kept in memory on the client
2504 Details : If a client is repeatedly creating and unlinking files it
2505 can accumulate a lot of stale inodes in the inode slab cache.
2506 If there is no other client load running this can cause the
2507 client node to run out of memory. Instead flush old inodes
2508 from client cache that have the same inode number as a new inode.
2511 Frequency : SLES9 2.6.5 kernel and long filenames only
2512 Bugzilla : 9969, 10379
2513 Description: utime reports stale NFS file handle
2514 Details : SLES9 uses out-of-dentry names in some cases, which confused
2515 the lustre dentry revalidation. Change it to always use the
2519 Frequency : rare, unless heavy write-truncate concurrency is continuous
2520 Bugzilla : 4180, 6984, 7171, 9963, 9331
2521 Description: OST becomes very slow and/or deadlocked during object unlink
2522 Details : filter_destroy() was holding onto the parent directory lock
2523 while truncating+unlinking objects. For very large objects this
2524 may block other threads for a long time and slow overall OST
2525 responsiveness. It may also be possible to get a lock ordering
2526 deadlock in this case, or run out of journal credits because of
2527 the combined truncate+unlink. Solution is to do object truncate
2528 first in one transaction without parent lock, and then do the
2529 final unlink in a new transaction with the parent lock. This
2530 reduces the lock hold time dramatically.
2533 Frequency : rare, 2.4 kernels only
2535 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2536 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2537 before unmounting their filesystem in order to avoid deadlock.
2538 The kernel_locked() function in 2.4 kernels only checks whether
2539 the kernel lock is held, not whether it is this process that is
2540 holding it as 2.6 kernels do.
2545 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2546 Details : The client ptlrpc code may be trying to reconnect to a down
2547 server before a previous connection attempt has timed out.
2548 Increase the reconnect interval to be longer than the connection
2549 timeout interval to avoid sending duplicate connections to
2553 Frequency : echo_client brw_test command
2555 Description: fix echo_client to work with OST preallocated code
2556 Details : OST preallocation code (5137) didn't take echo_client IO path
2557 into account: echo_client calls filter methods outside of any
2558 OST thread and, hence, there is no per-thread preallocated
2559 pages and buffers to use. Solution: hijack pga pages for IO. As
2560 a byproduct, this avoids unnecessary data copying.
2564 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2565 Description: Client can oops in mdc_commit_close() after open replay
2566 Details : It was possible for the MDS to return an open request with no
2567 transaction number in mds_finish_transno() if the client was
2568 evicted, but without actually returning an error. Clients
2569 would later try to replay that open and may trip an assertion
2570 Simplify the client close codepath, and always return an error
2571 from the MDS in case the open is not successful.
2574 Frequency : rare, 2.6 OSTs only
2576 Description: OST may deadlock under high load on fragmented files
2577 Details : If there was a heavy load and highly-fragmented OST filesystems
2578 it was possible to have all the OST threads deadlock waiting on
2579 allocation of biovecs, because the biovecs were not released
2580 until the entire RPC IO was completed. Instead, release biovecs
2581 as soon as they are complete to ensure forward IO progress.
2583 Severity : enhancement
2585 Description: Support for specifying external journal device at mount
2586 Details : If an OST or MDS device is formatted with an external journal
2587 device, this device major/minor is stored in the ext3 superblock
2588 and may not be valid for failover. Allow detecting and
2589 specifying the external journal at mount time.
2594 Description: Mounting an MDS with pending unlinked files may cause oops
2595 Details : target_finish_recovery() calls mds_postrecov() which returned
2596 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2597 considers this an error and immediately begins cleaning up the
2598 lov, just after starting the mds_lov process
2600 Severity : enhancement
2602 Description: Implement 'lfs df' to report actual free space on per-OST basis
2603 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2604 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2605 usage of objects; '-h' to report in human readable format.
2607 ------------------------------------------------------------------------------
2609 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2616 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2617 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2618 at startup during orphan unlinking. Instead, skip buddy update
2619 before setup, as it will be regenerated after recovery is complete.
2624 Description: If an OST is inactive, its locks might reference stale inodes.
2625 Details : lov_change_cbdata() must iterate over all namespaces, even if
2626 they are inactive to clear inode references from the lock.
2628 Severity : enhancement
2629 Frequency : occasional, if non-standard max_dirty_mb used
2631 Description: Client will block write RPCs if not enough grant
2632 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2633 then the client will block writes while waiting for another RPC
2634 to complete instead of consuming its dirty limit. With change
2635 we get improved performance when max_dirty_mb is small.
2637 Severity : enhancement
2638 Bugzilla : 3389, 6253
2639 Description: Add support for supplementary groups on the MDS.
2640 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2641 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2642 lookups for user supplementary groups into a cache.
2646 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2647 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2648 for liblustre O_EXCL is enforced by MDS.
2652 Description: readdir never returns NULL in liblustre.
2653 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2654 dentry in struct dirent.
2658 Frequency : liblustre only, depends on application IO pattern
2659 Description: liblustre clients evicted if not contacting servers
2660 Details : Don't put liblustre clients into the ping_evictor list, so
2661 they will not be evicted by the pinger ever.
2663 Severity : enhancement
2665 Description: Add ability to evict clients by NID from MDS.
2666 Details : By echoing "nid:$NID" string into
2667 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2668 $NID would be instantly evicted from this MDS and from all active
2669 OSTs connected to it.
2673 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2674 Details : lookup_it_finish() used to query file size from OSTs that was not
2679 Description: service threads change working directory to that of init
2680 Details : Starting lustre service threads may pin the working directory
2681 of the parent thread, making that filesystem busy. Threads
2682 now change to the working directory of init to avoid this.
2686 Frequency : during shutdown only
2687 Description: shutdown with a failed MDS or OST can cause unmount to hang
2688 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2693 Frequency : 2.6 only
2694 Description: chmod/chown may include an extra supplementary group
2695 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2696 supplementary group and if none is specified this is used.
2699 Bugzilla : 5479 (6816)
2701 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2702 Details : If lookup is in progress on a file that is unlinked we might try
2703 to revalidate the inode and fail in revalidate after lookup is
2704 complete and ll_file_open() enqueues the open again but
2705 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2708 Frequency : always, if lconf --abort_recovery used
2710 Description: lconf --abort_recovery fails with 'Operation not supported'
2711 Details : lconf was attempting to abort recovery on the MDT device and not
2714 ------------------------------------------------------------------------------
2716 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2721 Frequency : rare (only unsupported configurations with a node running as an
2723 Bugzilla : 6514, 5137
2724 Description: Mounting a Lustre file system on a node running as an OST could
2726 Details : OSTs now preallocates memory needed to write out data at
2727 startup, instead of when needed, to avoid having to
2728 allocate memory in possibly low memory situations.
2729 Specifically, if the file system is mounted on on OST,
2730 memory pressure could force it to try to write out data,
2731 which it needed to allocate memory to do. Due to the low
2732 memory, it would be unable to do so and the node would
2733 become unresponsive.
2735 Severity : enhancement
2737 Description: Addition of lconf --service command line option
2738 Details : lconf now accepts a '--service <arg>' option, which is
2739 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2741 Severity : enhancement
2743 Description: Failover mode is now the default for OSTs.
2744 Details : By default, OSTs will now run in failover mode. To return to
2745 the old behaviour, add '--failout' to the lmc line for OSTs.
2747 Severity : enhancement
2749 Description: Health checks are now provided for MDS and OSTs
2750 Details : Additional detailed health check information on MSD and OSTs
2751 is now provided through the procfs health_check value.
2754 Frequency : occasional, depends on IO load
2756 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2757 after numerous create/delete cycles
2758 Details : The ext3 inode allocation policy would not allocate new inodes
2759 very well on the OSTs because there are no new directories
2760 being created. Instead we look for groups with free space if
2761 the parent directories are nearly full.
2766 Description: Network or server problems during mount may cause partially
2767 mounted clients instead of returning an error.
2768 Details : The config llog parsing code may overwrite the error return
2769 code during mount error handling, returning success instead
2775 Description: MDS can fail to allocate large reply buffers
2776 Details : After long uptimes the MDS can fail to allocate large reply
2777 buffers (e.g. zconf client mount config records) due to memory
2778 fragmentation or consumption by the buffer cache. Preallocate
2779 some large reply buffers so that these replies can be sent even
2780 under memory pressure.
2784 Frequency : rare (liblustre)
2785 Description: fsx running with liblustre complained that using truncate() to
2786 extend the file doesn't work. This patch corrects that issue.
2787 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2788 ATTR_SIZE and lsm use in llu_setattr_raw.
2792 Frequency : rare, only 2.6 kernels
2793 Description: Unusual file access patterns on the MDS may result in inode
2794 data being lost in very rare circumstances.
2795 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2796 code in the 2.6 kernel caused the inode and/or EA data not to
2797 be read from disk, causing single-file corruption.
2801 Frequency : rare, only 2.6 filesystems using extents
2802 Description: Heavy concurrent write and delete load may cause data corruption.
2803 Details : It was possible under high-load situations to have an extent
2804 metadata block in the block device cache from a just-unlinked
2805 file overwrite a newly-allocated data block. We now unmap any
2806 metadata buffers that alias just-allocated data blocks.
2810 Frequency : filesystems with default stripe_count larger than 77
2811 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2812 Details : lconf specifies an inode size of 4096 bytes when the default
2813 stripe_count is larger than 77. This conflicts with the default
2814 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2815 this case to avoid pinning too much memory for large EAs.
2817 ------------------------------------------------------------------------------
2819 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2824 Frequency : rare (extremely heavy IO load with hundreds of clients)
2826 Description: Client is evicted, gets IO error writing to file
2827 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2828 caused clients to be evicted for AST timeouts. The fixes in
2829 bug 5192 mean we no longer need to have such short AST timeouts
2830 so ldlm_timeout has been increased.
2833 Frequency : occasional during --force or --failover shutdown under load
2834 Bugzilla : 5949, 4834
2835 Description: Server oops/LBUG if stopped with --force or --failover under load
2836 Details : a collection of import/export refcount and cleanup ordering
2837 issues fixed for safer force cleanup
2840 Frequency : only filesystems larger than 120 OSTs
2841 Bugzilla : 5990, 6223
2842 Description: lfs getstripe would oops on a very large filesystem
2843 Details : lov_getconfig used kfree on vmalloc'd memory
2846 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2847 Bugzilla : 6242, 6243
2848 Description: reading from files that had been truncated to a non-zero size
2849 but never opened returned no data
2850 Details : ll_file_read() reads zeros from no-object files to EOF
2855 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2856 Details : There's little harm in aborting MDS/OSS recovery and letting it
2857 try again, so I removed the LASSERT and return an error instead.
2859 Severity : enhancement
2861 Description: New debugging infrastructure for tracking down data corruption
2862 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2863 (b) cover more of the client-side code path, and (c) try to narrow
2864 down where problems occurred
2868 Bugzilla : 3819, 4364, 4397, 6313
2869 Description: Racing close and eviction MDS could cause assertion in mds_close
2870 Details : It was possible to get multiple mfd references during close and
2871 client eviction, leading to one thread referencing a freed mfd.
2873 Severity: : enhancement
2874 Bugzilla : 3262, 6359
2875 Description: Attempts to reconnect to servers are now more aggressive.
2876 Details : This builds on the enhanced upcall-less recovery that was added
2877 in 1.4.2. When trying to reconnect to servers, clients will
2878 now try each server in the failover group every 10 seconds. By
2879 default, clients would previously try one server every 25 seconds.
2884 Description: After recovery, certain operations trigger a failed
2885 assertion on a client.
2886 Details : Failing over an mds, using lconf -d --failover, while a
2887 client was doing a readdir() call would cause the client to
2888 LBUG after recovery completed and the readdir() was resent.
2890 Severity : enhancement
2892 Description: Default groups are now added by lconf
2893 Details : You can now run lconf --group <servicename> without having to
2894 manually add groups with lmc.
2897 Frequency : occasional
2899 Description: Nodes with an elan id of 0 trigger a failed assertion
2902 Frequency : always when accessing e.g. tty/console device nodes
2904 Description: tty and some other devices nodes cannot be used on lustre
2905 Details : file's private_data field is used by device data and lustre
2906 values in there got lost. New field was added to struct file to
2907 store fs-specific private data.
2910 Frequency : when exporting Lustre via NFS
2912 Description: NFSD failed occasionally when looking up a path component
2913 Details : NFSD is looking up ".." which was broken in ext3 directories
2914 that had grown large enough to become hashed.
2917 Frequency : Clusters with multiple interfaces not on the same subnet
2919 Description: Nodes will repeatedly try to reconnect to an interface which it
2920 cannot reach and report an error to the log.
2921 Details : Extra peer list entries will be created by lconf with some peers
2922 unreachable. lconf now validates the peer before adding it.
2925 Frequency : Only if a default stripe is set on the filesystem root.
2927 Description: Setting a default stripe on the filesystem root prevented the
2928 filesystem from being remounted.
2929 Details : The client was sending extra request flags in the root getattr
2930 request and did not allocate a reply buffer for the dir EA.
2933 Frequency : occasional, higher if lots of files are accessed by one client
2934 Bugzilla : 6159, 6097
2935 Description: Client trips assertion regarding lsm mismatch/magic
2936 Details : While revalidating inodes the VFS looks up inodes with ifind()
2937 and in rare cases can find an inode that is being freed.
2938 The ll_test_inode() code will free the lsm during ifind()
2939 when it finds an existing inode and then the VFS later attaches
2940 this free lsm to a new inode.
2944 Bugzilla : 6422, 7030
2945 Description: MDS deadlock between mkdir and client eviction
2946 Details : Creating a new file via mkdir or mknod (starting a transaction
2947 and getting the ns lock) can deadlock with client eviction
2948 (gets ns lock and trying to finish a synchronous transaction).
2951 Frequency : occasional
2952 Description: While starting a server, the fsfilt_ext3 module could not be
2954 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2955 kernels. Previously, lconf would still use the ext3 name
2956 when trying to load modules. Now, it will correctly use
2957 ext3 on 2.4 and ldiskfs on 2.6.
2959 Severity : enhancement
2960 Description: The default stripe count has been changed to 1
2961 Details : The interpretation of the default stripe count (0, to lfs
2962 or lmc) has been changed to mean striping across a single
2963 OST, rather than all available. For general usage we have
2964 found a stripe count of 1 or 2 works best.
2966 Severity : enhancement
2967 Description: Add support for compiling against Cray portals.
2968 Details : Conditional compiling for some areas that are different
2972 Frequency : occasional
2973 Bugzilla : 6409, 6834
2974 Description: Creating files with an explicit stripe count may lead to
2975 a failed assertion on the MDS
2976 Details : If some OSTs are full or unavailable, creating files may
2977 trigger a failed assertion on the MDS. Now, Lustre will
2978 try to use other servers or return an error to the
2982 Frequency : occasional
2984 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2985 caused lock timeout during readahead (since 1.4.2).
2986 Details : Processes doing readahead might match a lock that hasn't been
2987 granted yet if there are overlapping and conflicting lock
2988 requests. The readahead process waits on ungranted lock
2989 (original lock is CBPENDING), while OST waits for that process
2990 to cancel CBPENDING read lock and eventually evicts client.
2992 Severity : enhancement
2994 Description: Initial enabling of flock support for clients
2995 Details : Implements fcntl advisory locking and file status functions.
2996 This feature is provided as an optional mount flag (default
2997 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2998 locking are implemented yet, and those that are are not guaranteed
2999 to be completely correct in production environments.
3000 mount -t lustre -o [flock|noflock] ...
3003 Frequency : occasional
3005 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3006 assertion in the ext3 JBD (journaling) layer.
3007 Details : The b_committed_data struct is protected by the big kernel lock
3008 in 2.4 kernels, serializing journal_commit_transaction() and
3009 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3010 access to this struct. In 2.6 kernels there is finer grained
3011 locking to improve SMP performance of the JBD layer.
3015 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3020 Description: merge and cleanup kernel patches.
3023 -----------------------------------------------------------------------------
3025 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3027 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3028 versions of Lustre. Please update all servers and clients to
3029 version 1.4.2 or later at the same time. You must also run
3030 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3031 to update the configuration logs.
3033 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3034 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3035 - swab llog records of type '0' so we get proper header size/idx (5861)
3036 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3037 - fix rename of one directory over another leaking an inode (5953)
3038 - avoid SetPageDirty on 2.6 (5981)
3039 - don't re-add just-being-destroyed locks to the waiting list (5653)
3040 - when creating new directories, inherit the parent's custom
3041 striping settings if present parent (3048)
3042 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3043 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3044 - don't hold client locks on temporary worklist from l_lru (5666)
3045 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3046 - automatically evict dead clients (5921)
3047 - Update file size properly in create+truncate+fstat case (6196)
3048 - Do not unhash mountpoint dentries, do not allow removal of
3050 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3051 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3052 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3053 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3054 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3055 - Correctly handle reads of files with no objects (6243)
3056 - lctl recover will also mark a device active if deactivate used (5933)
3058 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3059 - allow --write-conf on an MDS with different nettype than client (5619)
3060 - don't write config llogs to MDS for mounts not from that MDS (5617)
3061 - lconf should create multiple TCP connections from a client (5201)
3062 - init scripts are now turned off by default; run chkconfig --on
3063 lustre and chkconfig --on lustrefs to use them
3064 - upcalls are no longer needed for clients to recover to failover
3066 - add --abort-recovery option to lconf to abort recovery on device
3068 - add support for an arbitrary number of OSTs (3026)
3069 - Quota support protocol changes.
3070 - forward compatibility changes to wire structs (6007)
3071 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3072 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3073 - improved "lustre status" script
3074 - initialize blocksize for non-regular files (6062)
3075 - added --disable-server and --disable-client configure options (5782)
3076 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3077 - Vanilla 2.4.29 support
3078 - increase maximum number of obd devices to 520 (6242)
3079 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3080 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3081 SLES 9 2.6.5 kernels
3082 - limit stripes per file to 160 (the maximum EA size) (6093)
3084 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3087 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3088 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3089 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3090 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3091 - small ext3 extents cleanups and fixes (5733)
3092 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3093 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3094 - avoid needless client->OST connect, fix handle mismatch (5317)
3095 - fix DLM error path that led to out-of-sync client, long delays (5779)
3096 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3097 - fix several locking issues related to i_size (5492,5624,5654,5672)
3098 - don't move pending lock onto export if it is already evicted (5683)
3099 - fix kernel oops when creating .foo in unlinked directory (5548)
3100 - fix deadlock in obdfilter statistics vs. object create (5811)
3101 - use time_{before,after} to avoid timer jiffies wrap (5882)
3102 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3103 - Do not leak request if server was not able to process it (5154)
3104 - If mds_open unable to find parent dir, make that negative lookup(5154)
3105 - don't create new directories with extent-mapping (5909, 5936)
3107 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3108 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3109 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3110 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3111 - support for RHEL4 kernel on i686 (5773)
3112 - provide error messages when incompatible logs are encountered (5898)
3114 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3115 * version 1.4.0.10 (1.4.1 release candidate 1)
3117 - don't keep a lock reference when lock is not granted (4238)
3118 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3119 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3120 - drop import inflight refcount on signal_completed_replay error (5255)
3121 - unlock page after async write error during send (3677)
3122 - handle missing objects in filter_preprw_read properly (5265)
3123 - no transno return for symlink open, don't save no-trasno open (3440)
3124 - don't try to complete elan receive that already failed (4012)
3125 - free RPC server reply state on error (5406)
3126 - clean up thread from ptlrpc_start_thread() on error (5160)
3127 - readahead could read extra page into cache that wasn't ejected (5388)
3128 - prevent races in class_attach/setup/cleanup/detach (5260)
3129 - don't dereference de->d_inode after l_dput of de (5458)
3130 - use "int" for stripe value returned from lock_to_stripe (5544)
3131 - mballoc allocation and error-checking fixes in 2.6 (5504)
3132 - block device patches to fix I/O request sizes in 2.6 (5482)
3133 - look up hostnames for IB nals (5602)
3134 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3135 - don't start multiple acceptors for the same port (5277)
3136 - fix incorrect LASSERT in mds_getattr_name (5635)
3137 - export a proc file for general "ping" checking (5628)
3138 - fix "lfs check" to not block when the MDS is down (5628)
3140 - service request history (4965)
3141 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3142 - create an "evict_client" /proc entry on OSTs, like the MDS has
3143 - fix mount usage message, return errors per mount(8) (5168)
3144 - change grep [] to grep "[]" in tests so they work in more UMLs
3145 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3146 - remove ancient LOV_MAGIC_V0 EA support (5047)
3147 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3148 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3149 - allow readahead allocations to fail when low on memory (5383)
3150 - mmap locking landed again, after considerable improvement (2828)
3151 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3153 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3156 - send OST transaction number in read/write reply to free req (4966)
3157 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3158 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3159 - the watchdog thread now runs as interruptible (5246)
3160 - flock/lockf fixes (but it's still disabled, pending 5135)
3161 - don't use EXT3 constants in llite code (5094)
3162 - memory shortage at startup could cause assertion (5176)
3164 - reorganization of lov code
3165 - single portals codebase
3167 - add extents/mballoc support (5025)
3168 - direct I/O reads in the obdfilter (4048)
3169 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3171 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3174 - send OST transaction number in read/write reply to free req (4966)
3175 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3176 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3177 - the watchdog thread now runs as interruptible (5246)
3178 - handle missing objects in filter_preprw_read properly (5265)
3179 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3180 - drop import inflight refcount on signal_completed_replay error (5255)
3181 - unlock page after async write error during send (3677)
3182 - return original error code on reconstructed replies (3761)
3183 - no transno return for symlink open, don't save no-trasno open (3440)
3185 - add pid to ldlm debugging output (4922)
3186 - bump the watchdog timeouts -- we can't handle 30sec yet
3187 - extra debugging for orphan dentry/inode bug (5259)
3189 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3192 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3193 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3194 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3195 - fix NULL dereference in /proc/sys/portals/routes (4827)
3196 - allow failed mdc_close() operations to be interrupted (4561)
3197 - stop precreate on OST before MDS would time out on it (4778)
3198 - don't send partial-page writes before EOF from client (4410)
3199 - discard client grant for sub-page writes on large-page clients (4520)
3200 - don't free dentries not owned by NFS code, check generation (4806)
3201 - fix lsm leak if mds_create_objects() fails (4801)
3202 - limit debug_daemon file size, always print CERROR messages (4789)
3203 - use transno after validating reply (3892)
3204 - process timed out requests if import state changes (3754)
3205 - update mtime on OST during writes, return in glimpse (4829)
3206 - add mkfsoptions to LDAP (4679)
3207 - use ->max_readahead method instead of zapping global ra (5039)
3208 - don't interrupt __l_wait_event() during strace
3210 - add software watchdogs to catch hung threads quickly (4941)
3211 - make lustrefs init script start after nfs is mounted
3212 - fix CWARN/ERROR duplication (4930)
3213 - return async write errors to application if possible (2248)
3214 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3215 - print NAL number in %x format (4645)
3216 - update barely-supported suse-2.4.21-171 series (4842)
3217 - support for sles 9 %post scripts
3218 - support for building 2.6 kernel-source packages
3219 - support for sles km_* packages
3221 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3224 - ignore -ENOENT errors in osc_destroy (3639)
3225 - notify osc create thread that OSC is being cleaned up (4600)
3226 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3227 - reconstruct ost_handle() like mds_handle() (4657)
3228 - create a new thread to do import eviction to avoid deadlock (3969)
3229 - let lconf resolve symlinked-to devices (4629)
3230 - don't unlink "objects" from directory with default EA (4554)
3231 - hold socknal file ref over connect in case target is down (4394)
3232 - allow more than 32000 subdirectories in a single directory (3244)
3233 - fix blocks count for O_DIRECT writes (3751)
3234 - OST returns ENOSPC from object create when no space left (4539)
3235 - don't send truncate RPC if file size isn't changing (4410)
3236 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3237 - bind to privileged port in socknal and tcpnal (3689)
3239 - rate limit CERROR/CWARN console message to avoid overload (4519)
3240 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3241 - basic mmap support (3918)
3242 - kernel patch series update from b1_4 (4711)
3244 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3247 - avoid crash during MDS cleanup with OST shut down (2775)
3248 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3249 - don't use bad inodes on the MDS (3744)
3250 - dynamic object preallocation to improve recovery speed (4236)
3251 - don't hold spinlock over lock dumping or change debug flags (4401)
3252 - don't zero obd_dev when it is force cleaned (3651)
3253 - print grants to console if they go negative (4431)
3254 - "lctl deactivate" will stop automatic recovery attempts (3406)
3255 - look for existing locks in ldlm_handle_enqueue() (3764)
3256 - don't resolve lock handle twice in recovery avoiding race (4401)
3257 - revalidate should check working dir is a directory (4134)
3259 - don't always mark "slow" obdfilter messages as errors (4418)
3261 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3264 - don't close LustreDB during write_conf until it is done (3860)
3265 - fix typo in lconf for_each_profile (3821)
3266 - allow dumping logs from multiple threads at one time (3820)
3267 - don't allow multiple threads in OSC recovery (3812)
3268 - fix debug_size parameters (3864)
3269 - fix mds_postrecov to initialize import for llog ctxt (3121)
3270 - replace config semaphore with spinlock (3306)
3271 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3272 - don't allow enqueue to complete on a destroyed export (3822)
3273 - down write_lock before checking llog header bitmap (3825)
3274 - recover from lock replay timeout (3764)
3275 - up llog sem before sending rpc (3652)
3276 - reduce ns lock hold times when setting kms (3267)
3277 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3278 - fix NULL deref and obd_dev leak on setup error (3312)
3279 - replace some LBUG about llog ops with error handling (3841)
3280 - don't match INVALID dentries from d_lookup and spin (3784)
3281 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3282 - fix invalid assertion in ptlrpc_set_wait (3880)
3284 - add libwrap support for the TCP acceptor (3996)
3285 - add /proc/sys/portals/routes for non-root route listing (3994)
3286 - allow setting MDS UUID in .xml (2580)
3287 - print the stack of a process that LBUGs (4228)
3289 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3292 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3293 - make sure to unset replay flag from failed open requests (3440)
3294 - if default stripe count is 0, use OST count for inode size (3636)
3295 - update parent mtime/ctime on client for create/unlink (2611)
3296 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3297 - free recovery state on server during a forced cleanup (3571)
3298 - unregister_reply for resent reqs (3063)
3299 - loop back devices mounting and status check on 2.6 (3563)
3300 - fix resource-creation race that can provoke i_size == 0 (3513)
3301 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3302 - more debugging for page-accounting assertion (3746)
3303 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3304 - avoid deadlock after precreation failure (3758)
3305 - fix race and lock order deadlock in orphan handling (3450, 3750)
3306 - add validity checks when grabbing inodes from l_ast_data (3599)
3308 - add /proc/.../recovery_status to obdfilter (3428)
3309 - lightweight CDEBUG infrastructure, debug daemon (3668)
3310 - change default OSC RPC parameters to be better on small clusters
3311 - turn off OST read cache for files smaller than 32MB
3312 - install man pages and include them in rpms (3100)
3313 - add new init script for (un)mounting lustre filesystems (2593)
3314 - run chkconfig in %post for init scripts (3701)
3315 - drop scimac NAL (unmaintained)
3317 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3320 - clean kiobufs before and after use (3485)
3321 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3322 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3323 - change init script to fail more clearly if not run as root (1528)
3324 - allow clients to reconnect during replay (1742)
3325 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3326 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3327 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3328 - fix cond_resched (3554)
3329 - only evict unfinished clients after recovery (3515)
3330 - allow bulk resend, prevent data loss (3570)
3331 - dynamic ptlrpc request buffer allocation (2102)
3332 - don't allow unlinking open directory if it isn't empty (2904)
3333 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3334 - remove extraneous obd dereference causing LASSERT failure (3334)
3335 - don't use get_cycles() when creating temp. files on the mds (3156)
3336 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3337 - handle EEXIST for set-stripe, set proper directory name (3336)
3339 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3340 - fix ksocknal_fmb_callback() error messages (2918)
3342 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3345 - don't copy lvb into (possibly NULL) reply on error (2983)
3346 - don't deref dentry after dput, don't free lvb on error (2922)
3347 - use the kms to determine writeback rpc length (2947)
3348 - increment oti_logcookies when osc is inactive (2948)
3349 - update client's i_blocks count via lvb messages (2543)
3350 - handle intent open/close of special files properly (1557)
3351 - mount MDS with errors=remount-ro, like obdfilter (2009)
3352 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3353 - don't use cancelling-locks' kms values (2947)
3354 - use highest lock extent for kms, not last one (2925)
3355 - don't dereference ERR_PTR() dentry in error handling path (3107)
3356 - fix thread race in portals_debug_dumplog() (3122)
3357 - create lprocfs device entries at setup instead of at attach (1519)
3358 - common AST error handler, don't evict client on completion race (3145)
3359 - zero nameidata in detach_mnt in 2.6 (3118)
3360 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3361 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3362 - initialize RPC timeout timer earlier for 2.6 (3219)
3363 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3364 - print nal/nid for unknown nid (3258)
3365 - additional checks for oscc recovery before doing precreate (3284)
3366 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3367 - don't crash in mdc_close for bad permissions on open (3285)
3368 - zero i_rdev for non-device files (3147)
3369 - clear page->private before handing to FS, better assertion (3119)
3370 - tune the read pipeline (3236)
3371 - fix incorrect decref of invalidated dentry (2350)
3372 - provide read-ahead stats and refine rpc in flight stats (3328)
3373 - don't hold journal transaction open across create RPC (3313)
3374 - update atime on MDS at close time (3265)
3375 - close LDAP connection when recovering to avoid server load (3315)
3376 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3377 - don't leak open file on MDS after open resend (3325)
3378 - serialize filter_precreate and filter_destroy_precreated (3329)
3379 - loop device shouldn't call sync_dev() for nul device (3092)
3380 - clear page cache after eviction (2766)
3381 - resynchronize MDS->OST in background (2824)
3382 - refuse to mount the same filesystem twice on same mountpoint (3394)
3383 - allow llmount to create routes for mounting behind routers (3320)
3384 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3385 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3386 - fix rare problem with rename on htree directories (3417)
3388 - allow default OST striping configuration per directory (1414)
3389 - fix compilation for qswnal for 2.6 kernels (3125)
3390 - increase maximum number of MDS request buffers for large systems
3391 - change liblustreapi to be useful for external progs like lfsck (3098)
3392 - increase local configuration timeout for slow disks (3353)
3393 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3395 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3398 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3399 - don't overwrite extent policy data in reply if lock was blocked (2901)
3400 - drop filter export grants atomically with removal from device (2663)
3401 - del obd_self_export from work_list in class_disconnect_exports (2908)
3402 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3403 - swab reply message in mdc_close, other PPC fixes (2464)
3404 - fix destroying of named logs (2325)
3405 - overwrite old logs when running lconf --write_conf (2264)
3406 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3407 - fix race in target_handle_connect (2898)
3408 - mds_reint_create() should take same inode create lock (2926)
3409 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3410 - don't close files for self_export to avoid uninitialized obd (2936)
3411 - allow MDS with the same name as client node (2939)
3412 - hold dentry reference for closed log files for unlink (2325)
3413 - reserve space for all logs during transactions (2059)
3414 - don't evict page beyond end of stripe extent (2925)
3415 - don't oops on a deleted current working directory (2399)
3416 - handle hard links to targets without a parent properly (2517)
3417 - don't dereference NULL lock when racing during eviction (2867)
3418 - don't grow lock extents when lots of conflicting locks (2919)
3420 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3423 - account for cache space usage on clients to avoid data loss (974)
3424 - lfsck support in lustre kernel code (2349)
3425 - reduce journal credits needed for BRW writes (2370)
3426 - orphan handling to avoid losing space on client/server crashes
3427 - ptlrpcd can be blocked, stopping ALL progress (2477)
3428 - use lock value blocks to assist in proper KMS, faster stat (1021)
3429 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3430 - recovery for initial connections (2355)
3431 - fixes for mds_cleanup_orphans (1934)
3432 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3433 - block all file creations until orphan recovery completes (1901)
3434 - client remove rq_connection from request struct (2423)
3435 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3436 - recovery timer race (2670)
3437 - mdc_close recovey bug (2532)
3438 - ptlrpc cleanup bug (2710)
3439 - mds timeout on local locks (2588)
3440 - namespace lock held during RPCs (2431)
3441 - handle interrupted sync write properly (2503)
3442 - don't try to handle a message that hasn't been replied to (2699)
3443 - client assert failure during cleanup after abort recovery (2701)
3444 - leak mdc device after failed mount (2712)
3445 - ptlrpc_check_set allows timedout requests to complete (2714)
3446 - wait for inflight reqs when ptlrpcd finishes (2710)
3447 - make sure unregistered services are removed from the srv_list
3448 - reset bulk XID's when resending them (caught by 1138 test)
3449 - unregister_bulk after timeout
3450 - fix lconf error (2694)
3451 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3452 - readahead locks pages, leaves pending causing memory pressure (2673)
3453 - increase OST request buffers to 4096 on large machines (2729)
3454 - fix up permission of existing directories in simple_mkdir (2661)
3455 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3456 - don't assign transno to errored transactions (2742)
3457 - don't delete objects on OST if given a bogus objid from MDS (2751)
3458 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3459 - if rq_no_resend, then timeout request after recovery (2432)
3460 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3461 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3462 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3463 - MDS recovery shouldn't delete objects that it has given out (2730)
3464 - if enqueue arrives after completion, don't clobber LVB (2819)
3465 - don't unlock pages twice when trigger_group_io returns error (2814)
3466 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3467 - don't write pages to disk if there was an error (1450)
3468 - don't ping imports that have recovery disabled (2676)
3469 - take buffered bytes into account when balancing socknal conn (2817)
3470 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3471 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3472 - remove little-endian swabbing of llog records (1987)
3473 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3474 - retry reposting request buffers if they fail (1191)
3475 - grow extent at grant time to avoid granting a revoked lock (2809)
3476 - lock revoke doesn't evict page if covered by a second lock (2765)
3477 - disable VM readahead to avoid reading outside lock extents (2805)
3479 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3480 - updated kernel patches for hp-2.4.20 kernel (2681)
3482 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3485 - fix truncated write corruption (2366)
3486 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3488 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3489 - ensure len(uuid) < 37 in lmc (1171)
3490 - fix ia64 OOPS in llog_test (2255)
3491 - zero end of page at obdfilter for partial page writes (2648)
3492 - don't leave stale dentries around after renames (bug 2428)
3493 - fix timeouts when evicting a client with a single lock held (2642)
3494 - set deadline for the initial HELLO message to drain (2634)
3495 - print out dotted-quad IP addresses in the socknal (2302)
3497 - additional debugging for MDS client eviction problem (2443)
3498 - fix mkfsoptions support for osts (2603, 2604)
3500 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3503 - add series for the vanilla 2.6.0 kernel
3504 - add series for the vanilla 2.4.24 kernel
3505 - add series for a cray x86/64 UL kernel drop
3506 - fix xattr patches for the vanilla 2.4.19 series
3508 - generate true UUIDs in lmc (1171)
3509 - have portals stack dumping break in UML (2466)
3510 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3511 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3512 - wake callback waiting threads on client eviction (2460)
3513 - Add --ptldebug and --subsystem to lmc (1719)
3514 - update assertion to allow safe interrupt allocation
3515 - set rq_no_resend for cancel requests (2432)
3516 - recalculate ptlrpcd timeout after resend (2494)
3517 - call vfs_rmdir when removing pending directories (2368)
3518 - fix renaming a file to itself (2429)
3519 - lmc creates a default one-stripe lov (2454)
3520 - expand procfs space to handle large clusters (2326)
3521 - increase UML stack to avoid overflow
3522 - update lconf's list of debug and subsystem masks
3523 - fix lfs find --obd (2510)
3524 - /proc tunable for disabling filter read caching (2591)
3525 - stop rpm packages from altering slapd.conf (2301)
3526 - disable nagle in the socknal under 0conf (2578)
3527 - choose mds inode size based on stripe count (2572)
3528 - fix kernel-source rpm problems (2516)
3530 - add --disable-doc to avoid pdf generation (2421)
3531 - update documentation, tests, type-os, comments
3532 - avoid format warnings on ia64
3533 - remove the TOE NAL
3534 - tiny code cleanups by removing unused fields
3536 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3539 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3540 - remove the most blatant lies from BUILDING file (2371)
3541 - change default debug level to reasonable production setting
3542 - reduce client side cache size to reduce cache flush time
3543 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3544 - make TCP zerocopy and pinger support enabled by default (2476)
3545 - sync writes completed after process exits caused crashes (2319)
3546 - maintain correct mount count on the MDS (2356)
3547 - backout 1557, because 2316 wasn't really fixed
3548 - better file I/O statistics gathering in /proc
3549 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3550 - another kernel patch to fix zero-copy TCP function export
3551 - don't take duplicate lock when processing re-sent getattr (2420)
3552 - lctl uses obd_self_export instead of creating new conn (2353)
3553 - MDS/OST recovery case which requires object creation asserted (2425)
3554 - move lfs from /usr/sbin to /usr/bin in packages
3555 - fix race between mds_client_add and mds_client_free (2417)
3556 - use kmalloc instead of slabs in portals (2430)
3557 - don't create duplicate records when a failover MDS is present (2442)
3558 - remove unnecessary mount age check (2332)
3559 - don't remove directory inodes from locks prematurely (2451)
3560 - don't break if MDS service name is the same as hostname (2103)
3561 - fix races in client write RPC generation when cache full (2482)
3563 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3566 - remove now-unused request->rq_obd (278)
3567 - if an allocation fails, print out how much memory we've used (1933)
3568 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3569 - add the "configurable stack size" patch to most series files (1256)
3570 - ability to write large log records, for 100+ OST configs (2306)
3571 - fix NULL deref when filter_prep fails (2314)
3572 - fix operator precedence error in filter_sync
3573 - dynamic allocation of socknal TX descriptors (2315)
3574 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3575 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3576 - ensure that CWARN messages in Portals always get to the syslog
3577 - __init/__exit are not for prototype decls (ldlm_init/exit)
3578 - x86-64 compile warning fixes
3579 - fix gateway LMC keyword conflict (2318)
3580 - fix MDS lock inversions in getattr/reint paths (1844)
3581 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3582 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3583 - DLM race condition prevented some lock evictions (2328)
3584 - ENOMEM detection and retry on socknal sends (2230)
3585 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3586 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3587 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3588 - better error messages when a client is rejected during recovery (1505)
3589 - avoid cancelling locks which were never granted, after failure (2330)
3590 - fix i_sem/journal inversion in mds_client_add (2333)
3591 - fix truncate/getattr lock cycle deadlock (2334)
3592 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3593 - fix two rare exit paths which could leak an l_lock() ref (2321)
3594 - fencepost error in MDS/OST orphan recovery (2226)
3595 - make log record alignment 8 bytes (1988)
3596 - lstripe now fails when requested offset > ost_count (2237)
3597 - ensure that all kernel series have a complete list.h (1607)
3598 - fix crashes in special-file operations (2316)
3599 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3601 - allow configurable automake binary, for testing new versions
3602 - small update to the lfs documentation
3604 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3606 * fix negative export reference count in fsfilt_sync (2312)
3608 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3609 * release candidate 0.9.1
3611 - orphans are moved into the PENDING directory for possible recovery
3612 - replayed opens now open by fid for orphan/rename safety (1042)
3613 - last close of an orphan inode generates a transno (683)
3614 - chdir() and mount() now pin the directory entry (1020)
3615 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3616 - discard very old requests without processing them (1502)
3617 - remove some common, well-understood CERRORs (1505)
3618 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3619 - clear "grant" flags in OST replies until OST grant code lands (1644)
3620 - fix read performance by not clobbering i_blksize on client (1598)
3621 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3622 - make LRU size a /proc tunable, clears locks when reduced (707)
3623 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3624 - server locks take references on exports now (1558)
3625 - build fixes for 2.4.20-rh trees (1663)
3626 - return an error from lov_create if all OSCs are inactive (1751)
3627 - fix import levels when a reconnect happens without a timeout (1597)
3628 - exit early from mds_open if we get a lookup error (1749)
3629 - partial page read at EOF wouldn't wait for disk before sending (1642)
3630 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3631 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3632 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3633 - write/truncate lock inversion (1639)
3634 - Don't auto-load obdclass, portals modules during cleanup (1495)
3635 - fix timestamps from jumping to "now" (1763)
3636 - extra journal assertions (1648)
3637 - add an extra multiunlink test (1771)
3638 - fix read_record/write_record API (1776)
3639 - fix leak of offset_extent, possible incorrect i_size later (1772)
3640 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3641 - Debugging for the kqswnal_get_idle_tx problems (1820)
3642 - Allow recovery to be attempted multiple times (1536)
3643 - Write out MDS last_rcvd file after it is first created (1600)
3644 - Fix tx_descriptor leak in failed transmit situations (1827)
3645 - ext3 journaling fixes for assertion failure after IO error (1871)
3646 - class_export_put() on freed export after completion AST error (1896)
3647 - Fix revalidate looping in VFS (1322)
3648 - Don't access a freed export during MDS_REINT timeout (1521)
3649 - Add open-unlink recovery support on the MDS (1673,1764)
3650 - Return an error if no MDS data was read from last_rcvd (1946)
3651 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3652 - Don't setup a disk device that doesn't match exported UUID (317)
3653 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3654 - avoid committing NULL handle in force close
3655 - local.sh is now a one-stripe LOV configuration
3656 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3657 - fix ext3 htree duplicate directory entry corruption (1516)
3658 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3659 - update to new LOV EA format (2097)
3660 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3662 2003-06-15 Phil Schwan <phil@clusterfs.com>
3665 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3666 - per-import/export recovery handling (958, 931, 959)
3667 - multiple last-rcvd slots, for serving multiple FSes (949)
3668 - connections are again shared between multiple imp/exports (963, 964)
3669 - "umount -f" would hang if any requests needed to be sent (393, 978)
3670 - avoid pinning large req buffer by copying for queued messages (989)
3671 - add "uuid" to "lctl device" command to help upcalls (991)
3672 - "open" RPCs with transnos would confuse recovery counters (1037)
3673 - do proper endian conversion of all wire messages (288, 340, 891)
3674 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3675 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3676 - fix ptlrpc_connection leak in target_handle_connect (1174)
3677 - fix import refcounting bug in OST and MDS cleanup (1134)
3678 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3679 - fix very unlikely obd_types race condition (501)
3680 - remove osc_open hack for echo_client (1187)
3681 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3682 - a failure in read_inode2 leads to deadlock (1139)
3683 - cancel ack-locks as soon as transaction is committed (1072)
3684 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3685 - make sure to commitrw after any preprw to avoid deadlock (1162)
3686 - failing to execute a file in a lustre FS would lock inode (1203)
3687 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3688 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3689 - obd_uuid2tgt would walk off the end of the list (1255)
3690 - on IA64 the osc would give portals incorrect bulk size (1258)
3691 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3692 - fix lock inversion caused by new llite matching code (1282)
3693 - limit the number of dirty pages on a client to 10MB (1286)
3694 - timed out locks were not being corrected cancelled (1289)
3695 - fix O_DIRECT above 4GB on IA-32 (1292)
3696 * major user-visible changes
3697 - fail out/fail over policy now controlled by the upcall (993)
3699 - add OBD_PING to check server availability and failure (954)
3700 - lustre messages are now sent in sending host order (288, 340, 891)
3701 - add eadatalen to MDS getattr reply (340)
3702 - OST read replies may contain second buffer, with per-page status (593)
3704 2003-03-11 Phil Schwan <phil@clusterfs.com>
3707 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3708 - failed open()s could cause deadlock; fixed (867, 869)
3709 - stop cancelling OST locks when files are closed (481)
3710 - overlapping XID spaces caused network corruption (851, 853)
3711 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3712 - setattr_raw added, to do single-RPC, server-side setattrs
3713 - lmc/lconf syntax change for OST UUIDs
3714 - fix crashy race condition between ptlrpc_free_req and osc_close
3715 - don't use request in mdc_enqueue if we hit a timeout (889)
3716 - don't set the inode i_size for regular files from the MDS (896)
3717 - handle out of order completion AST (842)
3718 - don't LBUG if a lock request times out after receiving AST (913)
3719 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3720 - if a bad lock AST arrives, send an error instead of dropping entirely
3721 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3722 - fix leak in bulk IO when only partially completed (899, 900, 926)
3723 - fix O_DIRECT for ia64 (55)
3724 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3725 - C-z after timeout could hang a process forever; fixed (977)
3727 - client-side I/O cache (678, 924, 929, 941, 970)
3729 - READPAGE and SETATTRs which don't take server-side locks get
3732 2003-02-11 Phil Schwan <phil@clusterfs.com>
3735 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3736 - Fix fsfilt_extN_readpage() to read a full page of directory
3737 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3738 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3739 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3740 - Most intent operations were removed, in favour of a new RPC mode
3741 that does a single RPC to the server and bypasses most of the VFS
3742 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3743 - Aggressively cancel local locks on DLM servers
3744 - mds_reint_unlink sends EA to the client if it's the last nlink.
3745 client uses that EA to unlink OST objects.
3746 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3747 - recursive symlinks were fixed (439)
3748 - fixed NULL deref in DEBUG_REQ
3749 - filter_update_lastobjid no longer calls sync, which annoyed extN
3750 - fixed multi-client small-writes to a single file problem (445)
3751 - fixed mtime updates during file writes (607)
3752 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3753 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3754 - fixed Linux OST size reporting problem (444, 656)
3755 - OST now updates object mtime with writes or setattr (607, 619)
3756 - client verifies file size before zeroing page past EOF (445)
3757 - OST now writes last allocated objid to disk with allocation (108)
3758 - LOV on echo now works (409)
3760 - mds_reint_unlink sends a new buffer, with the EA included. this
3761 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3762 set if a regular file was being unlinked, and it was the last link
3763 - use PtlGet from the target for bulk writes (315)
3764 - OST now updates object mtime with writes or setattr (607, 619)
3765 - LDLM now has a grant-time callback to revalidate locked items, if
3767 - Many MDS operations were reorganized to combat race conditions
3769 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3770 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3772 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3775 - Fully reactivate OST imports after reconnection (512, others)
3776 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3777 - More graceful error handling for truncating on dead OST (515)
3778 - Don't error out unless we're actually accessing dead stripes (474)
3779 - Fix garbage sizes when stripes are missing (410)
3780 - LRU counters were broken, causing constant lock purge (433, 432)
3781 - garbage on read from stripes with failed OSTs (441)
3782 - mark OSCs as active before reconnecting during recovery (438)
3783 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3784 - lfind did not preserve OST order in output (443)
3785 - symlinks cause hung clients, incorrect data (439)
3786 - stop dereferencing request after dropping refcount (457)
3787 - don't LASSERT(spin_is_locked) on non-SMP (455)
3788 - fixes for many rename() bugs
3789 - fstat didn't correctly synchronize attributes (399)
3790 - server must handle lock cancellation during blocking AST prep (487)
3791 - bulk descriptors were free()d too soon (511)
3792 - fix paths in lconf, which would load incorrect modules (451, 507)
3793 - fix confusing lconf 'host not found' error message (386)
3794 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3795 - fix race condition in mdc_blocking_ast() for inode access (526)
3796 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3797 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3798 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3799 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3800 - fix lstripe in several ways (kernel side) (527)
3801 - fix request leak in ldlm_cli_enqueue (262)
3802 - incorrect OSC was marked inactive after OST failure
3803 - call mds_fs_cleanup before unmounting filesystem (524)
3804 - fix races between taking ns_lock and ldlm_lock_change_resource
3805 - fix races updating LOV export open file list
3806 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3807 - fix recovery NULL deref in ldlm_cli_cancel_unused
3808 - fix some DLM races by using new hash table for lock handles (419)
3809 - permit the client to specify desired inodes, at replay
3810 - duplicate requests when we queue them for replay reintegration
3811 - fix last_rcvd offset calculation
3812 - sync after each recovered transaction, so we always make progress
3813 - never, not always, ERESTART requests without transnos
3814 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3816 - skip replay if the MDS says that the client is already connected
3817 - don't check for a recovery-enabled export to match lctl's UUID
3818 - don't INC_USE_COUNT for phantom exports
3819 - don't crash when cleaning up phantom exports (567)
3820 - don't double-finish or set replay data for errored mdc_open requests
3821 - abort requests when they time out, so we don't get old replies
3822 - send/receive replies for AST messages again
3823 - if the client says that it doesn't have the lock, cancel it on the
3825 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3826 mark it as destroyed, it will all work out when decref is called
3827 - fix module use counts (22, 581)
3829 - ASTs now expect a reply (server cancels lock on error reply)
3831 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3834 - fix many simultaneous client startup (392)
3835 - fix dentry->d_it clobbering
3836 - credentials weren't being shipped for readdir/getattr operations
3837 - remove invalid assertions triggered during some concurrent MD
3839 - proper Lustre versions added (336, 389)
3840 - fix memory leak for create error case (398)
3841 - fix LOV locking bug that would get cli/srv out of sync
3842 - fix echo client over LOV (409)
3843 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3844 - fix double-O_EXCL intent crash (424)
3845 - avoid sending multiple lock CANCELs (352)
3847 - MDS can do multi-client recovery (modulo bugs in new code)
3849 - many updates, edits, cleanups
3851 2002-11-18 Phil Schwan <phil@clusterfs.com>
3854 - fix null d_it dereference (346)
3855 - fix full OST/dbench hang (333)
3856 - fix permission problem with file removal (286)
3857 - fix removal of OSCs from LOV when they fail
3858 - fix NULL deref during bulk timeout (214)
3859 - fix problems related to multiple filesystems on one MDS (241)
3860 - fixed serious subtle metadata locking bugs
3861 - free locks on clients when inodes are removed due to memory
3863 - fix inode pointer in lock data (285)
3864 - partial support for multiple MDS on a single host (241)
3865 - data locks weren't cancelled at clear_inode time (290, 311)
3866 - intent locks could lead to unbounded lock growth (205)
3867 - added a maximum lock count, an LRU list, and a flusher
3868 - fix multiple rename (365)
3869 - properly abstracted the echo client
3870 - OSC locked 1 byte too many; fixed
3871 - rewrote brw callback code:
3872 - fixed recovery bugs related to LOVs (306)
3873 - fixed too-many-pages-in-one-write crash (191)
3874 - fixed (again) crash in sync_io_timeout (214)
3875 - probably fixed callback-related race (385)
3877 - Add capability to MDS protocol
3878 - LDLM cancellations and callbacks on different portals
3880 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3883 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3884 - defer open object destruction to close time (601981, 138)
3885 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3886 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3887 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3888 - don't call obd_setattr in ll_file_release for destroyed objects
3889 * protocol change to lustre_msg: move |version| and add |flags|
3890 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3891 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3892 * added replay of create, unlink, link and rename operations during
3893 MDS failover; recovery should be much more robust now
3894 * remove failed OSCs from LOVs (only lov_create uses this so far)
3895 * the lustre-HOWTO was brought (more) up to date (582544)
3897 2002-10-23 Phil Schwan <phil@clusterfs.com>
3900 - in-use dentries weren't being reused properly (617851)
3901 - prevent multiple LDLM setup (599178)
3902 - fix LOV size calculations for truncate (617853)
3903 - fix client handling of MDS intent errors (POSIX)
3904 - fix permission bug in lovstripe.c test (624321)
3905 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3906 - truncate past end of file could corrupt data
3907 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3908 - a race in recovery could return ETIMEDOUT to apps (623947)
3909 - building outside the source directory was fixed
3910 * the lustre-HOWTO was brought (more) up to date (582544)
3911 * major progress was made on recovery functionality
3913 2002-10-10 Phil Schwan <phil@clusterfs.com>
3916 - recovery deadlock fix
3917 - rm -rf causes LBUG fix (617817)
3918 - file open by multiple tasks fix (618962)
3919 - directory permissions bugs (602707 and 620007)
3920 - journal_stop fixed with locking (611313)
3921 - O_APPEND failures resolved (618273, perhaps 614459)
3922 - lconf PATH fix (619770)
3923 - IA64 build fix (621450)
3924 - RPC buffer sizes scale with amount of memory
3926 2002-10-01 Phil Schwan <phil@clusterfs.com>
3929 - locks would be cancelled without throwing away data pages,
3930 resulting in inconsistent data (605627)
3931 - inode attributes were not always being refreshed (605627, 612449)
3932 - lconf now continues to cleanup after lctl reports an error
3933 - MDS now enforces user permissions (602707)
3934 - lprocfs cleanup fixed, but not yet enabled (614157)
3935 - fixed infinite server hang, should a client not respond to an AST
3936 - avoid going into recovery if user calls readlink() with a buffer
3937 that's too small (613941)
3938 - AST RPCs no longer require replies (614867) -- this may be changed
3939 - don't crash server if client sends an IOV that's too big (611336)
3940 - fixed lock conversion deadlock (611892)
3941 - fixed the following of symlinks (614622)
3942 * recovery: the server can remove locks from a client that dies, other
3943 clients can make progress
3944 * more extN patch fixes
3945 * compile-time configurable ptlrpc buffer allocations
3947 - collaborative read cache document
3948 - Lustre Lite Performance CDR document-in-progress
3950 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3953 - fix typo in patch-2.4.18
3955 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3958 - clear ptlrpc request each time in handle_incoming_request()
3959 - unlink of files now destroys the object on the OST
3961 2002-09-19 Peter Braam <braam@clusterfs.com>
3963 * add hard link support
3964 * change obdfile creation method
3965 * kernel patch changed
3967 2002-09-19 Peter Braam <braam@clusterfs.com>
3970 - stack overflow bug in extN fixed
3972 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3974 * documentation updates
3975 - add man pages for config tools
3976 - update tests/README to describe testing with new config tools
3977 - finish metadata API descriptions
3978 * bug fixes and cleanups
3979 - statfs workaround for 16TB limit
3980 - LOV stripe allocation improved, can stripe on subset of OSTs
3981 - LOV file size/IO offset was wrong for files > 4GB in size
3982 - object EA data was being dropped, caused files to be unreadable
3983 - memory overflow with non-LOV OST caused memory corruption
3984 - fixed regression tests to work with new config tools, obdfilter
3985 - fixed bug when directory size became larger than 1 block
3986 - fixed bug (for single client case) when PWD was deleted
3987 - invalidate local directory pages when doing intent-based ops
3988 - avoid LDLM oops when lock callback contained bad data
3990 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3992 * documentation updates
3993 * bug fixes and cleanups
3994 - configuration tools
3997 - 64-bit compile warnings
3998 - 64-bit internal statfs data
4000 * test_brw on persistent OST devices
4002 * lprocfs (disabled)
4004 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4006 * documentation updates
4007 * bug fixes and cleanups
4008 * configuration tools
4010 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4013 * many small fixes to 0_5_4
4014 * io/network handling
4015 * thinkos in MDS operations
4017 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4020 * crucial basic fixes to 0.5.3
4021 * IOR, Iozone work over Elan
4024 2002-08-07 Phil Schwan <phil@clusterfs.com>
4025 * version 0_5_3, our first alpha
4026 * we use the new Portals iovs
4027 * documentation updates
4028 * bug fixes and cleanups
4029 * small changes in the DLM wire protocol
4031 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4032 * version 0_5_1 with some initial stability,
4033 * locking on MD and file I/O.
4034 * documentation updates
4035 * several bug fixes since 0.5.0
4036 * small changes in wire protocol
4038 2002-07-18 Phil Schwan <phil@clusterfs.com>
4040 * delivered as Lustre Light Alpha
4041 * fixed a crash after handling invalid MDS requests
4042 * fixed directory pages for architectures with non-4k pages sizes
4044 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4045 * release version v0_4_4
4046 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4048 * Fixes a number of interruption problems with OST operations.
4049 * Update documentation for portals header changes
4050 * Move all wire protocol structs/defines to lustre_idl.h
4051 * Fixes symlink length bug.
4052 * Add tcpdump to repository.
4054 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4055 * release version v0_4_3
4056 * Fixes statfs for inodes on extN.
4057 * Fixes bug in runtests which would delete /etc/hosts.
4058 * Use 64-bit object IDs wherever possible (not into VFS though)
4059 Remove ost_get_info, which is unused by lustre, and out of date.
4061 2002-07-03 Peter Braam <braam@clusterfs.com>
4062 * release version v0_4_2 Fixes a lookup error (type not passed)
4063 * move forward to head of Portals
4064 * move forward to latest Lustre kernel
4066 2002-06-25 Peter Braam <braam@clusterfs.com>
4067 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob