4 2.6.9-55.0.6.EL (RHEL 4),
5 2.6.16.46-0.14 (SLES 10),
6 2.6.18-8.1.14.EL5 (RHEL 5).
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
15 Severity : enhancement
17 Description: self-adjustable client's lru lists
18 Details : use adaptive algorithm for managing client cached locks lru
19 lists according to current server load, other client's work
20 pattern, memory activities, etc. Both, server and client
21 side namespaces provide number of proc tunables for controlling
24 Severity : enhancement
26 Description: light-weight GSS support
27 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
28 and reduce performance overhead.
30 Severity : enhancement
32 Description: Linux keyring support
33 Details : Support using service of Linux keyring for Lustre GSS internal
34 context refresh/cache mechanism.
38 Description: Fix errors in lfs documentation
39 Details : Fixes man pages
43 Description: don't use GFP_* in generic Lustre code.
44 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
48 Description: obdclass is limited by single OBD_ALLOC(idarray)
49 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
53 Description: warnings with build patchless client with vanila 2.6.19 and up
54 Details : change old ctl_table style and replace ctl_table/ctl_table_header
55 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
59 Frequency : Only for SLES
60 Description: sanity_quota fail test_1
61 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
62 file, which makes TSTID[2] inunique.
66 Frequency : Only for SLES9
67 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
68 Details : Update target/ChangeLog/which_patch .
72 Frequency : Only for test_10 in sanity-quota.sh
73 Description: a bug in quota test code
74 Details : checking a test flag in a wrong place
78 Frequency : Only for test_13 in sanity-quota.sh
79 Description: a bug in quota test script
80 Details : shouldn't check overfull allocated quota
84 Description: jbd statistics
85 Details : Port older jbd statistics patch for sles10
87 Severity : enhancement
89 Description: Add printing inode info into message about error in writepage.
94 Description: reply_lock_interpret crash due to race with it and lock cancel.
95 Details : Do not replay locks that are being cancelled. Do not reference
96 locks by their address during replay, just by their handle.
100 Frequency : When flocks are used.
101 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
102 Details : Flock locks might destroy just granted lock if it could be merged
103 with another existing flock, this is done in completion handler,
104 so teach ldlm_cli_enquque_fini that this is a valid case for
110 Description: Oops in read and write path when failing to allocate lock.
111 Details : Check if lock allocation failed and return error back.
115 Description: lstripe command fails for valid OST index
116 Details : The stripe offset is compared to lov->desc.ld_tgt_count
117 instead of lov->desc.ld_active_tgt_count.
121 Description: sanity.sh failed test 103
122 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
127 Description: df doesn't work properly if diskfs blocksize != 4K
128 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
132 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
133 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
137 Description: obd_health_check_timeout too short
138 Details : set obd_health_check_timeout as 1.5x of obd_timeout
142 Description: llapi_file_create() does not allow some changes
143 Details : add llapi_file_open() that allows specifying the mode and
144 open flags, and also returns an open file handle.
148 Description: merge and cleanup kernel patches.
149 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
153 Description: Add journal checksum support.(Kernel part)
154 Details : The journal checksum feature adds two new flags i.e
155 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
156 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
157 indicates that the commit block contains the checksum for
158 the blocks described by the descriptor blocks. Now commit
159 record can be sent to disk without waiting for descriptor
160 blocks to be written to disk. This behavior is controlled
161 using JBD2_FEATURE_ASYNC_COMMIT flag.
165 Description: OSS needs mutliple precreate threads
166 Details : Add ability to start more than one create thread per OSS.
170 Description: Sanity.sh test_65k failed: setstripe should have succeeded
171 Details : execute lfs setstripe on client
175 Description: mds_obd_create error creating tmp object
176 Details : When the user sets quota on root, llog will be affected and can't
177 create files and write files.
180 Frequency : Always on ia64 patchless client, and possibly others.
182 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
183 Details : This allows the patchless client to be loaded on architectures
188 Description: RedHat Update kernel for RHEL5
189 Details : Add the kernel config file for RHEL5.
193 Description: RedHat Update kernel for RHEL5
194 Details : Modify the kernel config file more closer RHEL5.
198 Description: Build failure against Centos5 (RHEL5)
199 Details : Define PAGE_SIZE when it isn't present.
203 Description: test_7 on sanity-quota.sh can't be used on separate machines
204 Details : add facet to handle it
208 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
209 Details : with NFS, the anon dentry's parent was set to itself in
210 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
211 corresponding dentry other than use rec->ur_name.
213 Severity : enhancement
215 Description: lfs setstripe enhancement
216 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
220 Description: testing performance impact of enabling checksumming
221 Details : enable checksum by default, allow --disable-checksum
222 configure option and "-o nochecksum" mount option
226 Description: lustre support for RHEL5
227 Details : Add support for RHEL5.
231 Description: Client eviction due to failover config
232 Details : after a connection loss, the lustre client should attempt to
233 reconnect to the last active server first before trying the
234 other potential connections.
238 Description: when mds and osts use different quota unit(32bit and 64bit),
239 quota will be released repeatly.
240 Details : void sending multiple quota reqs to mds, which will keep the
241 status between the reqs.
245 Description: osts not allocated evenly to files
246 Details : change the condition to increase offset_idx
249 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
251 Description: Data corruption for OSTs that are formatted larger than 2TB
253 Details : When generating the bio request for lustre file writes the
254 sector number would overflow a temporary variable before being
255 used for the IO. The data reads correctly from Lustre (which
256 will overflow in a similar manner) but other file data or
257 filesystem metadata may be corrupted in some cases.
261 Description: Tune the kernel for good SCSI performance.
262 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
263 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
268 Description: rewrite ext2-derived code in obdclass/uuid.c
269 Details : rewrite inherited code (uuid parsing code from ext2 utils)
270 from scratch preserving functionality.
274 Description: "lfs find -obd UUID" prints directories
275 Details : "lfs find -obd UUID" will return all directory names instead
276 of just file names. It is incorrect because the directories
277 do not reside on the OSTs.
281 Description: use i_size_read and i_size_write in 2.6 port
282 Details : replace inode->i_size access with i_size_read/write()
285 Frequency : when removing large files
287 Description: scheduling issue during removal of large Lustre files
288 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
289 It causes scheduling issues when removing large files (17TB in the
293 Frequency : only with liblustre clients on XT3
295 Description: evictions taking too long
296 Details : allow llrd to evict clients directly on OSTs
299 Frequency : only on ppc
301 Description: /proc/fs/lustre/devices broken on ppc
302 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
303 We should make sure the type of 'index' is loff_t and then cast
304 explicitly as needed below. Do not assign an explicitly cast
308 Frequency : only for rhel5
310 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
311 Details : Modify the target file & which_kernel.
313 Severity : enhancement
315 Description: omit set fsid for export NFS
316 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
317 as FS_REQUIRES_DEV for avoid problems with generate fsid.
321 Frequency : Always, for kernels after 2.6.16
322 Description: Fix warning idr_remove called for id=.. which is not
324 Details : Last kernels save old s_dev before kill super and not allow
325 to restore from callback - restore it before call kill_anon_super.
329 Description: buffer overruns could theoretically occur
330 Details : llapi_semantic_traverse() modifies the "path" argument by
331 appending values to the end of the origin string, and a
332 overrun may occur. Adding buffer overrun check in liblustreapi.
336 Description: client sometimes tried revalidate root dentry.
337 Details : not revalidate dentry if it root dentry.
341 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
342 Details : Modify vm-tunables-rhel4.patch.
346 Description: Kernel config for 2.6.18-vanilla.
347 Details : Modify targets/2.6-vanilla.target.in.
348 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
349 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
350 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
351 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
355 Description: improve handling recoverable errors
356 Details : if request processig with error which can be recoverable on server
357 request should be resend, otherwise page released from cache and
362 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
363 Details : Modify target file & which_patch.
364 A flaw was found in the IA32 system call emulation provided
365 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
366 value could be stored in the %RAX register, which could trigger an
367 out-of-bounds system call table access. An untrusted local user
368 could exploit this flaw to run code in the kernel
369 (ie a root privilege escalation). (CVE-2007-4573).
373 Description: Remove 2.4 kernel checks lustre 1.8
374 Details : Remove 2.5.0 check from quota_check.c, quota_ctl.c,
375 quota_interface.c in quota subsystem.
376 Remove 2.4 kernel check from obd_mount.c, linux/linux-sysctl.c in
378 Remove lock_24kernel from fsfilt_ext3.c fsfilt_reiserfs.c
379 lvfs_linux.c upcall_cache.c in lvfs subsystem.
382 --------------------------------------------------------------------------------
384 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
386 * Support for kernels:
387 2.6.5-7.283 (SLES 9),
388 2.6.9-55.EL (RHEL 4),
389 2.6.16.46-0.14 (SLES 10),
390 2.6.18.8 vanilla (kernel.org)
391 * Client support for unpatched kernels:
392 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
393 2.6.16 - 2.6.22 vanilla (kernel.org)
394 * Due to recently discovered recovery problems, we do not recommend
395 using patchless RHEL 4 clients with this or any earlier release.
396 * Recommended e2fsprogs version: 1.39.cfs8
397 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
398 * Starting with this release, the ldiskfs backing filesystem required
399 by Lustre is now in its own package, lustre-ldiskfs. This package
400 should be installed. It is versioned separately from Lustre and
401 may be released separately in future.
405 Description: block reactivating mgc import until all deactivates complete
406 Details : Fix race when failing back MDT/MGS to itself (testing)
408 Severity : enhancement
410 Description: add optional extra BUILD_VERSION info
411 Details : add a new environment variable (namely LUSTRE_VERS) which allows
412 to override the lustre version.
415 Frequency : 2.6.18 servers only
417 Description: ll_kern_mount() doesn't release the module reference
418 Details : The ldiskfs module reference count never drops down to 0
419 because ll_kern_mount() doesn't release the module reference.
424 Description: server LBUG when using old ost_num_threads parameter
425 Details : Accept the old ost_num_threads parameter but warn that it
426 is deprecated, and fix an off-by-one error that caused an LBUG.
431 Description: Transient SCSI error results in persistent IO issue
432 Details : iobuf->dr_error is not reinitialized to 0 between two
436 Frequency : sometimes when underlying device returns I/O errors
438 Description: OSTs not going read-only during write failures
439 Details : OSTs are not remounted read-only when the journal commit threads
440 get I/O errors because fsfilt_ext3 calls journal_start/stop()
441 instead of the ext3 wrappers.
445 Description: poor connect scaling with increasing client count
446 Details : Don't run filter_grant_sanity_check for more than 100 exports
447 to improve scaling for large numbers of clients.
450 Frequency : SLES10 only
452 Description: sanity-quota.sh quotacheck failed: rc = -22
453 Details : Quotas cannot be enabled on SLES10.
456 Frequency : liblustre clients only
458 Description: getdirentries does not give error when run on compute nodes
459 Details : getdirentries does not fail when the size specified as an argument
460 is too small to contain at least one entry
462 Severity : enhancement
464 Description: Add LNET router traceability for debug purposes
465 Details : If a checksum failure occurs with a router as part of the
466 IO path, the NID of the last router that forwarded the bulk data
467 is printed so it can be identified.
472 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
473 Details : Due to a race condition, liblustre clients were occasionally
476 Severity : enhancement
478 Description: lfs setstripe use optional parameters instead of postional
481 Severity : enhancement
483 Description: Nanosecond timestamp support for ldiskfs
484 Details : The on-disk ldiskfs filesystem has added support for nanosecond
485 resolution timestamps. There is not yet support for this at
486 the Lustre filesystem level.
489 Frequency : during server recovery
491 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
492 Details : request with rq_no_resend flag not awake l_wait_event if they get a
496 Frequency : nfs export on patchless client
498 Description: connectathon hang when test nfs export over patchless client
499 Details : Disconnected dentry cannot be found with lookup, so we do not need
500 to unhash it or make it invalid
503 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
504 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
505 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
510 Description: open req refcounting wrong on reconnect
511 Details : If reconnect happened between getting open reply from server and
512 call to mdc_set_replay_data in ll_file_open, we will schedule
513 replay for unreferenced request that we are about to free.
514 Subsequent close will crash in variety of ways.
515 Check that request is still eligible for replay in
516 mdc_set_replay_data().
521 Description: disable writes to filesystem when reading health_check file
522 Details : the default for reading the health_check proc file has changed
523 to NOT do a journal transaction and write to disk, because this
524 can cause reads of the /proc file to hang and block HA state
525 checking on a healthy but otherwise heavily loaded system. It
526 is possible to return to the previous behaviour during configure
527 with --enable-health-write.
529 Severity : enhancement
531 Description: 64-bit inode version
532 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
533 to the inode. This will be required for version-based recovery.
538 Description: MDS fails to start if a duplicate client export is detected
539 Details : in some rare cases it was possible for a client to connect to
540 an MDS multiple times. Upon recovery the MDS would detect this
541 and fail during startup. Handle this more gracefully.
543 Severity : enhancement
545 Description: Add -o localflock option to simulate old noflock
547 Details : This will achieve local-only flock/fcntl locks
553 Description: log_commit_thread vs filter_destroy race leads to crash
554 Details : Take import reference before releasing llog record semaphore
559 Description: Wrong request locking in request set processing
560 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
561 imp_delayed_list, in this place should be used imp_lock.
564 Frequency : when reconnection
566 Description: Grant Leak when osc reconnect to OST
567 Details : When osc reconnect ost, OST(filter) should check whether it
568 should grant more space to client by comparing fed_grant and
569 cl_avail_grant, and return the granted space to client instead
570 of "new granted" space, because client will call osc_init_grant
571 to update the client grant space info.
574 Frequency : when client reconnect to OST
576 Description: Grant Leak when osc do resend and replay bulk write
577 Details : When osc reconnect to OST, OST(filter)should clear grant info of
578 bulk write request, because the grant info while be sync between
579 OSC and OST when reconnect, and we should ignore the grant info
580 these of resend/replay write req.
585 Description: Grant space more than avaiable left space sometimes.
586 Details : When then OST is about to be full, if two bulk writing from
587 different clients came to OST. Accord the avaliable space of the
588 OST, the first req should be permitted, and the second one
589 should be denied by ENOSPC. But if the seconde arrived before
590 the first one is commited. The OST might wrongly permit second
591 writing, which will cause grant space > avaiable space.
594 Frequency : when client is evicted
596 Description: Grant might be wrongly erased when osc is evicted by OST
597 Details : when the import is evicted by server, it will fork another
598 thread ptlrpc_invalidate_import_thread to invalidate the
599 import, where the grant will be set to 0. While the original
600 thread will update the grant it got when connecting. So if
601 the former happened latter, the grant will be wrongly errased
602 because of this race.
607 Description: Checking Stale with correct fid
608 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
609 to check whether it is stale, because sometimes, we want the
610 enqueue happened anyway, and op_data.fid2 will not be initialized.
612 Severity : enhancement
614 Description: update patchless client
615 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
618 Frequency : only with 2.4 kernel
620 Description: random memory corruption
621 Details : size of struct ll_inode_info is to big for union inode.u and this
622 can be cause of random memory corruption.
627 Description: Memory leak in recovery
628 Details : Lov_mds_md was not free in an error handler in mds_create_object.
629 It should also check obd_fail before fsfilt_start, otherwise if
630 fsfilt_start return -EROFS,(failover mds during mds recovery).
631 then the req will return with repmsg->transno = 0 and rc = EROFS.
632 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
633 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
634 be freed no matter whether fsfilt_commit success or not.
637 Frequency : only with huge count clients
639 Description: Prevents from taking the superblock lock in llap_from_page for
641 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
642 for llap_from_page call in ll_removepage prevents from taking the
643 superblock lock for a soon died page.
648 Description: Not check open intent error before release open handle
649 Details : in some rare cases, the open intent error is not checked before
650 release open handle, which may cause
651 ASSERTION(open_req->rq_transno != 0), because it tries to release
652 the failed open handle.
657 Description: Set cat log bitmap only after create log success.
658 Details : in some rare cases, the cat log bitmap is set too early. and it
659 should be set only after create log success.
663 Description: Accessing a block bevice can re-enable I/O when Lustre is
664 tearing down a device.
665 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
670 Description: service threads may hog cpus when there are a lot of requests
672 Details : Insert cond_resched to give other threads a chance to use some of
678 Description: the cat log was not initialized in recovery
679 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
680 unlink log on mds will not be initialized until mds post
681 recovery. And also in mds post recovery, the unlink log will
682 initialization will be done asynchronausly, so there will be race
683 between add unlink log and unlink log initialization.
687 Description: brw_stats were being printed incorrectly
688 Details : brw_stats were being printed as log2 but all of them were not
689 recorded as log2. Also remove some code duplication arising from
690 filter_tally_{read,write}.
694 Frequency : rare, only in recovery.
695 Description: ASSERTION(req->rq_type != LI_POISON) failed
696 Details : imp_lock should be held while iterating over imp_sending_list for
697 prevent destroy request after get timeout in ptlrpc_queue_wait.
701 Description: replay-single.sh test 52 fails
702 Details : A lock's skiplist need to be cleanup when it being unlinked
703 from its resource list.
707 Description: Short directio read returns full requested size rather than
709 Details : Direct I/O operations should return actual amount of bytes
710 transferred rather than requested size.
712 Severity : enhancement
714 Description: metadata RPC reduction (e.g. for rm performance)
715 Details : decrease the amount of synchronous RPC between clients and servers
716 by canceling conflicing lock before the operation on the client side
717 and packing thier handles into the main operation RPC to server.
719 Severity : enhancement
721 Description: Async OSC create to avoid the blocking unnecessarily.
722 Details : If a OST has no remain object, system will block on the creating
723 when need to create a new object on this OST. Now, ways use
724 pre-created objects when available, instead of blocking on an
725 empty osc while others are not empty. If we must block, we block
726 for the shortest possible period of time.
730 Description: improve handling recoverable errors
731 Details : if request processig with error which can be recoverable on server
732 request should be resend, otherwise page released from cache and
735 Severity : enhancement
737 Description: refine locking for avoid write wrong info into lov_objid file
738 Details : fix possible races with add new target and write/update data in
741 --------------------------------------------------------------------------------
743 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
748 Frequency : on some architectures
750 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
751 Details : Uninitialized flags sometimes cause configuration commands to
755 Frequency : patchless clients only
757 Description: missing __iget() symbol export
758 Details : The __iget() symbol export is missing. To avoid the need for
759 this on patchless clients the deathrow inode reaper is turned
760 off, and we depend on the VM to clean up old inodes. This
761 dependency was during via the fix for bug 12181.
763 --------------------------------------------------------------------------------
765 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
767 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
768 INTEROPERATE with older versions automatically. In many cases a
769 special upgrade step is needed. Please read the
770 user documentation before upgrading any part of a live system.
771 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
772 version will not interoperate with 1.6 betas before beta5 (1.5.95).
773 * WARNING: Lustre configuration and startup changes are required with
774 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
780 Severity : enhancement
782 Description: MountConf
783 Details : Lustre configuration is now managed via mkfs and mount
784 commands instead of lmc and lconf. New obd types (MGS, MGC)
785 are added for dynamic configuration management. See
786 https://mail.clusterfs.com/wikis/lustre/MountConf for
789 Severity : enhancement
791 Description: dynamic OST addition
792 Details : OSTs can now be added to a live filesystem
794 Severity : enhancement
796 Description: startup order invariance
797 Details : MDTs and OSTs can be started in any order. Clients only
798 require the MDT to complete startup.
800 Severity : enhancement
802 Description: parallel, asynchronous orphan cleanup
803 Details : orphan cleanup is now performed in separate threads for each
804 OST, allowing parallel non-blocking operation.
806 Severity : enhancement
808 Description: optimized stripe assignment
809 Details : stripe assignments are now made based on ost space available,
810 ost previous usage, and OSS previous usage, in order to try
811 to optimize storage space and networking resources.
813 Severity : enhancement
815 Description: Permanently set tunables
816 Details : All writable /proc/fs/lustre tunables can now be permanently
817 set on a per-server basis, at mkfs time or on a live
820 Severity : enhancement
822 Description: Lustre message v2
823 Details : Add lustre message format v2.
825 Severity : enhancement
827 Description: client OST exclusion list
828 Details : Clients can be started with a list of OSTs that should be
829 declared "inactive" for known non-responsive OSTs.
833 Description: SPEC SFS validation failure on NFS v2 over lustre.
834 Details : Changes the blocksize for regular files to be 2x RPC size,
835 and not depend on stripe size.
837 Severity : enhancement
839 Description: Multiple MD RPCs in flight.
840 Details : Further unserialise some read-only MDS RPCs - learn about intents.
841 To avoid overly-overloading MDS, introduce a limit on number of
842 MDS RPCs in flight for a single client and add /proc controls
843 to adjust this limit.
845 Severity : enhancement
847 Description: client read/write statistics
848 Details : Add client read/write call usage stats for performance
849 analysis of user processes.
850 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
851 file access. extents_stats shows chunk size distribution.
852 extents_stats_per_process show chunk size distribution per
855 Severity : enhancement
857 Description: mds statistics
858 Details : Add detailed mds operations statistics in
859 /proc/fs/lustre/mds/*/stats.
863 Description: Failure of copying files with lustre special EAs.
864 Details : Client side always return success for setxattr call for lustre
865 special xattr (currently only "trusted.lov").
870 Description: Refcount LNET uuids
871 Details : The global LNET uuid list grew linearly with every startup;
872 refcount repeated list entries instead of always adding to
875 Severity : enhancement
877 Description: Dynamic service threads
878 Details : Within a small range, start extra service threads
879 automatically when the request queue builds up.
882 Frequency : mixed-endian client/server environments
884 Description: mixed-endian crashes
885 Details : The new msg_v2 system had some failures in mixed-endian
888 Severity : enhancement
890 Description: Easy OST removal
891 Details : OSTs can be permanently deactivated with e.g. 'lctl
892 conf_param lustre-OST0001.osc.active=0'
894 Severity : enhancement
896 Description: MGS proc entries
897 Details : Added basic proc entries for the MGS showing what filesystems
900 Severity : enhancement
902 Description: provide MGS failover
903 Details : Added config lock reacquisition after MGS server failover.
905 Severity : enhancement
907 Description: add Linux 2.4 support
908 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
912 Description: a large application tries to do I/O to the same resource and dies
914 Details : Check the req->rq_arrival time after the call to
915 ost_brw_lock_get(), but before we do anything about
916 processing it & sending the BULK transfer request. This
917 should help move old stale pending locks off the queue as
918 quickly as obd_timeout.
921 Frequency : when an incorrect nid is specified during startup
923 Description: ptlrpc connect to non-existant node causes kernel crash
924 Details : LNET can't be re-entered from an event callback, which
925 happened when we expire a message after the export has been
926 cleaned up. Instead, hand the zombie cleanup off to another
929 Severity : enhancement
931 Description: plain/inodebits lock performance improvement
932 Details : Grouping plain/inodebits in granted list by their request modes
933 and bits policy, thus improving the performance of search through
937 Frequency : only if OST filesystem is corrupted
939 Description: client incorrectly hits assertion in ptlrpc_replay_req()
940 Details : for a short time RPCs with bulk IO are in the replay list,
941 but replay of bulk IOs is unimplemented. If the OST filesystem
942 is corrupted due to disk cache incoherency and then replay is
943 started it is possible to trip an assertion. Avoid putting
944 committed RPCs into the replay list at all to avoid this issue.
947 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
949 Description: System hang on startup
950 Details : This bug allowed the liblustre (e.g. catamount) client to
951 return to the app before handling all startup RPCs. This
952 could leave the node unresponsive to lustre network traffic
953 and manifested as a server ptllnd timeout.
955 Severity : enhancement
957 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
958 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
959 boolean to control whether to print peer debug info when a
960 client's RPC times out.
963 Frequency : only for kernels with patches from Lustre below 1.4.3
965 Description: Remove old rdonly API
966 Details : Remove old rdonly API which unused from at least lustre 1.4.3
969 Frequency : only for devices with external journals
971 Description: Set external device read-only also
972 Details : During a commanded failover stop, we set the disk device
973 read-only while the server shuts down. We now also set any
974 external journal device read-only at the same time.
977 Frequency : when upgrading from 1.4 while trying to change parameters
979 Description: The wrong (new) MDC name was used when setting parameters for
980 upgraded MDT's. Also allows changing of OSC (and MDC)
981 parameters if --writeconf is specified at tunefs upgrade time.
984 Frequency : when setting specific ost indicies
986 Description: QOS code breaks on skipped indicies
987 Details : Add checks for missing OST indicies in the QOS code, so OSTs
988 created with --index need not be sequential.
990 Severity : enhancement
992 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
993 Details : The uninit_groups feature works in conjunction with the kernel
994 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
995 up the pass1 processing of e2fsck. This is a read-only feature
996 in ldiskfs2 only, so older kernels and current ldiskfs cannot
997 mount filesystems that have had this feature enabled.
999 Severity : enhancement
1001 Description: Improve multi-block allocation algorithm to avoid fragmentation
1002 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1003 allocation locality and avoid filesystem fragmentation.
1005 ------------------------------------------------------------------------------
1007 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1009 * Support for kernels:
1010 2.4.21-47.0.1.EL (RHEL 3)
1011 2.6.5-7.283 (SLES 9)
1012 2.6.9-42.0.10.EL (RHEL 4)
1013 2.6.12.6 vanilla (kernel.org)
1014 2.6.16.27-0.9 (SLES 10)
1015 * Recommended e2fsprogs version: 1.39.cfs5
1017 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1021 Frequency : occasional, depends on client load and configuration
1022 Bugzilla : 12181, 12203
1023 Description: data loss for recently-modified files
1025 Details : In some cases it is possible that recently written or created
1026 files may not be written to disk in a timely manner (this should
1027 normally be within 30s unless client IO load is very high).
1028 The problem appears as zero-length files or files that are a
1029 multiple of 1MB in size after a client crash or client eviction
1030 that are missing data at the end of the file.
1032 This problem is more likely to be hit on clients where files are
1033 repeatedly created and unlinked in the same directory, clients
1034 have a large amount of RAM, have many CPUs, the filesystem has
1035 many OSTs, the clients are rebooted frequently, and/or the files
1036 are not accessed by other nodes after being written.
1038 The presence of the problem can be detected by looking at
1039 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1040 smaller than the second (nr_unused) then dirty files will not
1041 be flushed automatically to disk. "sync; sleep 10" should be
1042 run several times on the node before unmounting it to update
1043 Lustre (this is also safe to run on nodes without this problem).
1045 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1046 that can cause this same problem, so customers using that kernel
1047 also need to update the kernel in addition to Lustre. In order
1048 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1050 It is normal that files written just before a client crash (less
1051 than 30s) may not yet have been flushed to disk, even for local
1055 Frequency : frequent on thin XT3 nodes
1057 Description: UUID collision on thin XT3 Linux nodes
1058 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1059 randomly, since we relied on an insufficiently-seeded PRNG.
1064 Description: OSS hangs after "All ost request buffers busy"
1065 Details : A deadlock between quota and journal operations caused OSS
1066 hangs after printing "All ost request buffers busy."
1069 Frequency : always on liblustre builds
1071 Description: Cleanup compiler warnings on liblustre
1074 Frequency : always on liblustre builds on XT3
1076 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1080 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1081 > 32000 subdirectories
1082 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1083 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1084 subdirectory count crosses 32000. This will aid e2fsck to
1085 correctly handle more than 32000 subdirectories.
1088 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1090 Description: System hang on startup
1091 Details : This bug allowed the liblustre (e.g. catamount) client to
1092 return to the app before handling all startup RPCs. This
1093 could leave the node unresponsive to lustre network traffic
1094 and manifested as a server ptllnd timeout.
1096 Severity : enhancement
1098 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1099 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1100 boolean to control whether to print peer debug info when a
1101 client's RPC times out.
1106 Description: make O_SYNC working on 2.6 kernels
1107 Details : 2.6 kernels use different method for mark pages for write,
1108 so need add a code to lustre for O_SYNC work.
1113 Description: Failure to close file and release space on NFS
1114 Details : Put inode details into lock acquired in ll_intent_file_open.
1115 Use mdc_intent_lock in ll_intent_open to properly
1116 detect all kind of errors unhandled by mdc_enqueue.
1121 Description: proc file read during shutdown sometimes raced obd removal,
1123 Details : Add lock to prevent obd access after proc file removal.
1126 Frequency : Only for files larger than 4GB on 32-bit clients.
1128 Description: improperly doing page alignment of locks
1129 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1130 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1135 Description: RPCs being resent when they shouldn't be
1136 Details : Some RPCs that should not be resent are being resent. This
1137 can cause inconsistencies in the RPC state machine. Do not
1138 resend such requests.
1141 Frequency : rare, only with NFS export
1143 Description: Crash on NFS re-export node
1144 Details : under very unusual load conditions an assertion is hit in
1145 ll_intent_file_open()
1148 Frequency : only if OST filesystem is corrupted
1150 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1151 Details : for a short time RPCs with bulk IO are in the replay list,
1152 but replay of bulk IOs is unimplemented. If the OST filesystem
1153 is corrupted due to disk cache incoherency and then replay is
1154 started it is possible to trip an assertion. Avoid putting
1155 committed RPCs into the replay list at all to avoid this issue.
1160 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1161 Details : Large single O_DIRECT read and write calls can fail to allocate
1162 a sufficiently large buffer to process the request. In case of
1163 allocation failure the allocation is retried with a smaller
1164 buffer and broken into smaller requests.
1166 Severity : enhancement
1168 Description: Add -o localflock option to simulate old noflock behaviour.
1169 Details : This will achieve local-only flock/fcntl locks coherentness.
1174 Description: versioning check is incomplete
1175 Details : Checking the version difference of client vs. server, report
1176 error if the gap is too big.
1181 Description: add support PG_writeback bit
1182 Details : add support for PG_writeback bit for Lustre, for more carefull
1183 work with page cache in 2.6 kernel. This also fix some deadlocks
1184 and remove hack for work O_SYNC with 2.6 kernel.
1186 Severity : enhancement
1188 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1189 Details : The uninit_groups feature works in conjunction with the kernel
1190 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1191 up the pass1 processing of e2fsck. This is a read-only feature
1192 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1193 mount filesystems that have had this feature enabled.
1195 Severity : enhancement
1197 Description: Improve multi-block allocation algorithm to avoid fragmentation
1198 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1199 allocation locality and avoid filesystem fragmentation.
1201 ------------------------------------------------------------------------------
1203 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1205 * Support for kernels:
1206 2.6.9-42.0.3.EL (RHEL 4)
1207 2.6.5-7.276 (SLES 9)
1208 2.4.21-47.0.1.EL (RHEL 3)
1209 2.6.12.6 vanilla (kernel.org)
1210 2.6.16.21-0.8 (SLES10)
1211 * Recommended e2fsprogs version: 1.39.cfs2-0
1213 * The backwards-compatible /proc/sys/portals symlink has been removed
1214 in this release. Before upgrading, please ensure that you change
1215 any configuration scripts or /etc/sysctl.conf files that access
1216 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1217 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1218 in advance of the upgrade on any system running Lustre 1.4.6 or
1219 newer, since /proc/sys/lnet was added in that version.
1220 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1224 Frequency : only when quota is used
1226 Description: avoid scanning export list for quota master
1227 Details : Change the algorithms to avoid scanning export list in order
1228 to improve the efficiency.
1231 Frequency : MDS failover only, very rarely
1233 Description: "went back in time" messages on mds failover
1234 Details : The greatest transno may be lost when the current operation
1235 finishes with an error (transno==0) and the client's last_rcvd
1236 record is over-written. Save the greatest transno in the
1237 mds_last_transno for this case.
1240 Frequency : always for specific kernels and striping counts
1242 Description: client may get "Matching packet too big" without ACL support
1243 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1244 when trying to access files in certain configurations. The
1245 clients should in fact be denied when mounting because they do
1246 not understand ACLs.
1249 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1251 Description: many clients connecting with IO in progress causes connect timeouts
1252 Details : Avoid synchronous journal commits to avoid delays caused by many
1253 clients connecting/disconnecting when bulk IO is in progress.
1254 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1255 OST_IO_PORTAL to avoid delays behind potentially many pending
1259 Frequency : occasionally with multiple writers to a single file
1261 Description: shared writes to file may result in wrong size reported by stat()
1262 Details : Allow growing of kms when extent lock is cancelled
1265 Frequency : always with random mmap IO to multi-striped file
1267 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1268 Details : Only if the hole is at the end of OST object so that kms is too
1269 small. Fix is to increase kms accordingly in ll_nopage.
1272 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1274 Description: writes to a missing object would leak memory on the OST
1275 Details : If there is an inconsistency between the MDS and OST filesystems,
1276 such that the MDS references an object that doesn't exist, writes
1277 to that object will leak memory due to incorrect cleanup in the
1278 error handling path, eventually running out of memory on the OST.
1283 Description: Creating too long symlink causes lustre errors
1284 Details : Check symlink and name lengths before sending requests to MDS.
1287 Frequency : only if flock is enabled (not on by default)
1289 Description: posix locks not released on fd closure on 2.6.9+
1290 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1291 kernels, this caused such locks not to be released on fd close and
1292 then assertions on fs unmount about still used locks.
1295 Frequency : MDS failover only, very rarely
1297 Description: clients may get ASSERTION(granted_lock != NULL)
1298 Details : When request was taking a long time, and a client was resending
1299 a getattr by name lock request. The were multiple lock requests
1300 with the same client lock handle and
1301 mds_getattr_name->fixup_handle_for_resent_request found one of the
1302 lock handles but later failed with ASSERTION(granted_lock != NULL).
1307 Description: handle->h_buffer_credits > 0, assertion failure
1308 Details : h_buffer_credits is zero after truncate, causing assertion
1309 failure. This patch extends the transaction or creates a new
1313 Frequency : NFS re-export or patchless client
1314 Bugzilla : 11179, 10796
1315 Description: Crash on NFS re-export node (__d_move)
1316 Details : We do not want to hash the dentry if we don't have a lock.
1317 But if this dentry is later used in d_move, we'd hit uninitialised
1318 list head d_hash, so we just do this to init d_hash field but
1319 leave dentry unhashed.
1322 Frequency : NFS re-export or patchless client
1324 Description: NFS exports has problem with symbolic link
1325 Details : lustre client didn't properly install dentry when re-exported
1326 to NFS or running patchless client.
1329 Frequency : NFS re-export or patchless client
1331 Description: Various nfs/patchless fixes.
1332 Details : fixes reuse disconected alias for lookup process - this fixes
1333 warning "find_exported_dentry: npd != pd",
1334 fix permission error with open files at nfs.
1335 fix apply umask when do revalidate.
1338 Frequency : occasional
1340 Description: Crash on NFS re-export node
1341 Details : calling clear_page() on the wrong pointer triggered oops in
1342 generic_mapping_read().
1345 Frequency : rarely, using O_DIRECT IO
1347 Description: unaligned directio crashes client with LASSERT
1348 Details : check for unaligned buffers before trying any requests.
1351 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1353 Description: stale data returned from RAID cache
1354 Details : If only a small amount of IO is done to the RAID device before
1355 reading it again it is possible to get stale data from the RAID
1356 cache instead of reading it from disk.
1359 Frequency : always for sles10 kernel
1361 Description: sles10 support
1362 Details : ll_follow_link: compile fixes and using of nd_set_link
1363 under newer kernels.
1366 Frequency : depends on arch, kernel and compiler version, always on sles10
1369 Description: recursive or deep enough symlinks cause stack overflow
1370 Details : getting rid of large stack-allocated variable in
1374 Frequency : depends on hardware
1376 Description: lustre write performance loss in the SLES10 kernel
1377 Details : the performance loss is caused by using of write barriers in the
1378 ext3 code. The SLES10 kernel turns barrier support on by
1379 default. The fix is to undo that change for ldiskfs.
1381 ------------------------------------------------------------------------------
1383 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1385 * Support for kernels:
1386 2.6.9-42.0.3EL (RHEL 4)
1387 2.6.5-7.276 (SLES 9)
1388 2.4.21-47.0.1.EL (RHEL 3)
1389 2.6.12.6 vanilla (kernel.org)
1393 Frequency : quota enabled and large files being deleted
1395 Description: releasing more than 4GB of quota at once hangs OST
1396 Details : If a user deletes more than 4GB of files on a single OST it
1397 will cause the OST to spin in an infinite loop. Release
1398 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1403 Description: statfs data retrieved from /proc may be stale or zero
1404 Details : When reading per-device statfs data from /proc, in the
1405 {kbytes,files}_{total,free,avail} files, it may appear
1406 as zero or be out of date.
1409 Frequency : systems with MD RAID1 external journal devices
1411 Description: lconf's call to blkid is confused by RAID1 journal devices
1412 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1413 of returning all block devices that match the journal UUID.
1416 Frequency : always, for aggregate stripe size over 4GB
1418 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1419 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1420 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1423 Frequency : always if "lfs find" used on a local file/directory
1425 Description: "lfs find" segfaults if used on a local file/directory
1426 Details : The case where a directory component was not specified wasn't
1427 handled correctly. Handle this properly.
1430 Frequency : always on ppc64
1432 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1433 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1434 perform endianness conversion.
1437 Frequency : rarely (truncate to non-zero file size after write under load)
1438 Bugzilla : 10730, 10687
1439 Description: Files padded with zeros to next 4K multiple
1440 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1441 it is possible that files that are truncated to a non-zero size
1442 immediately after being written are filled with zero bytes beyond
1443 the truncated size. No file data is lost.
1445 Severity : enhancement
1447 Description: Allow recovery/failover for liblustre clients.
1448 Details : liblustre clients were unaware of failover configurations until
1451 Severity : enhancement
1453 Description: user file locks should fail when not mounting with flock option
1454 Details : Set up an error-returning stub in ll_file_operations.lock field
1455 to prevent incorrect behaviour when client is mounted without
1456 flock option. Also, set up properly f_op->flock field for
1460 Frequency : always on ia64
1462 Description: "lfs df" loops on printing out MDS statfs information
1463 Details : The obd_ioctl_data was not initialized and in some systems
1464 this caused a failure during the ioctl that did not return
1465 an error. Initialize the struct and return an error on failure.
1468 Frequency : SLES 9 only
1470 Description: Error of copying files with lustre special EAs as root
1471 Details : Client side always return success for setxattr call for lustre
1472 special xattr (currently only "trusted.lov").
1475 Frequency : rarely on clusters with both ia64+i386 clients
1477 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1478 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1479 clients are doing concurrent writes to the same file it is
1480 possible that the ia64 clients may LASSERT because the OST
1481 extent locks are not PAGE_SIZE aligned. Ensure that grown
1482 locks are always aligned on the request boundary.
1485 Frequency : specific use, occasional
1487 Description: Overwriting in use executable truncates on-disk binary image
1488 Details : If one node attempts to overwrite an executable in use by
1489 another node, we now correctly return ETXTBSY instead of
1490 truncating the file.
1492 Severity : enhancement
1494 Description: Async OSC create to avoid the blocking unnecessarily.
1495 Details : If a OST has no remain object, system will block on the creating
1496 when need to create a new object on this OST. Now, ways use
1497 pre-created objects when available, instead of blocking on an
1498 empty osc while others are not empty. If we must block, we block
1499 for the shortest possible period of time.
1504 Description: chmod on Lustre root is propagated to other clients
1505 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1506 invalid by the follow_mount time.
1509 Frequency : liblustre clients only
1511 Description: Race in 'instant cancel' lock handling could lead to such locks
1512 never to be granted in case of SMP MDS
1513 Details : Do not destroy not yet granted but cbpending locks in
1517 Frequency : replay/resend of open
1519 Description: non null lock assetion failure in mds_intent_policy
1520 Details : Trying to replay/resend lockless open requests resulted in
1521 mds_open() returning 0 with no lock. Now it sets a flag if
1522 a lock is going to be returned.
1524 Severity : enhancement
1526 Description: Checksum enhancements
1527 Details : New checksum enhancements allow for resending RPCs that failed
1530 Severity : enhancement
1532 Description: Tunables on number of dirty pages in cacche
1533 Details : Allow to set limit on number of dirty pages cached.
1538 Description: client crash on unmount - lock still has references
1539 Details : In some error handling cases it was possible to leak a lock
1540 reference on a client while accessing a file. This was not
1541 harmful to the client during operation, but would cause the
1542 client to crash when the filesystem is unmounted.
1545 Frequency : specific case, rare
1547 Description: ETXTBSY on mds though file not in use by client
1548 Details : ETXTBSY is no longer incorrectly returned when attempting to
1549 chmod or chown a directory that the user previously tried to
1550 execute or a currently-executing binary.
1553 Frequency : extremely rare except on liblustre-based clients
1555 Description: Lustre space not freed when files are deleted
1556 Details : Clean up open-unlinked files after client eviction. Previously
1557 the unlink was skipped and the files remained as orphans.
1562 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1563 Details : In some cases it was possible to send a blocking callback to
1564 a client doing a glimpse, even though that client didn't get
1565 a lock granted. When the glimpse lock is cancelled on the OST
1566 the freed lock is left on the waiting list and corrupted the list.
1569 Frequency : all core dumps
1571 Description: Broke core dumps to lustre
1572 Details : Negative dentry may be unhashed if parent does not have UPDATE
1573 lock, but some callers, e.g. do_coredump, expect dentry to be
1574 hashed after successful create, hash it in ll_create_it.
1576 ------------------------------------------------------------------------------
1578 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1580 * Support for kernels:
1581 2.6.9-42.0.2.EL (RHEL 4)
1582 2.6.5-7.276 (SLES 9)
1583 2.4.21-40.EL (RHEL 3)
1584 2.6.12.6 vanilla (kernel.org)
1588 Frequency : always on RHEL 3
1590 Description: Number of open files grows over time
1591 Details : The number of open files grows over time, whether or not
1592 Lustre is started. This was due to a filp leak introduced
1593 by one of our kernel patches.
1595 ------------------------------------------------------------------------------
1597 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1599 * Support for kernels:
1600 2.6.9-42.EL (RHEL 4)
1601 2.6.5-7.276 (SLES 9)
1602 2.4.21-40.EL (RHEL 3)
1603 2.6.12.6 vanilla (kernel.org)
1608 Bugzilla : 5719, 9635, 9792, 9684
1609 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1610 Details : If a server is under heavy load and cannot reply to new
1611 connection requests before the client resends the (re)connect,
1612 the connection handling code can behave badly if two service
1613 threads are concurrently handing separate (re)connections from
1614 the same client. Add better locking to the connection handling
1615 code, and ensure that only a single connection will be processed
1616 for a given client UUID, even if the lock is dropped.
1618 Severity : enhancement
1620 Description: add TCP zero-copy support to kernel
1621 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1622 sends if the hardware supports scatter-gather and checksumming.
1623 This allows socklnd to do client-write and server-read more
1624 efficiently and reduce CPU utilization from skbuf copying.
1627 Frequency : only if NFS exporting from client
1629 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1630 Details : If setting attributes on a file created under NFS that had
1631 never been opened it would be possible to oops the client
1632 if the file had no objects.
1635 Frequency : always for liblustre
1637 Description: liblustre client does MDS+OSTs setattr RPC for each write
1638 Details : When doing a write from a liblustre client, the client
1639 incorrectly issued an RPC to the MDS and each OST the file was
1640 striped over in order to update the timestamps. When writing
1641 with small chunks and many clients this could overwhelm the MDS
1642 with RPCs. In all cases it would slow down the write because
1643 these RPCs are unnecessary.
1645 Severity : enhancement
1647 Description: allow number of MDS service threads to be changed at module load
1648 Details : It is now possible to change the number of MDS service threads
1649 running. Adding "options mds mds_num_threads={N}" to the MDS's
1650 /etc/modprobe.conf will set the number of threads for the next
1651 time Lustre is restarted (assuming the "mds" module is also
1652 reloaded at that time). The default number of threads will
1653 stay the same, 32 for most systems.
1658 Description: OST crash if filesystem is unformatted or corrupt
1659 Details : If an OST is started on a device that has never been formatted
1660 or if the filesystem is corrupt and cannot even mount then the
1661 error handling cleanup routines would dereference a NULL pointer.
1666 Description: NULL pointer deref in llap_from_page.
1667 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1668 incorrect) mapping if the page was truncated/reclaimed while it was
1669 searched for. Check for this condition and skip such pages when
1670 doing readahead. Introduce extra check to llap_from_page() to
1671 verify page->mapping->host is non-NULL (so page is not anonymous).
1674 Frequency : Sometimes when using sys_sendfile
1676 Description: "page not covered by a lock" warnings from ll_readpage
1677 Details : sendfile called ll_readpage without right page locks present.
1678 Now we introduced ll_file_sendfile that does necessary locking
1679 around call to generic_file_sendfile() much like we do in
1683 Frequency : with certain MDS communication failures at client mount time
1685 Description: NULL pointer deref after failed client mount
1686 Details : a client connection request may delayed by the network layer
1687 and not be sent until after the PTLRPC layer has timed out the
1688 request. If the client fails the mount immediately it will try
1689 to clean up before the network times out the request. Add a
1690 reference from the request import to the obd device and delay
1691 the cleanup until the network drops the request.
1694 Frequency : occasionally during client (re)connect
1696 Description: assertion failure during client (re)connect
1697 Details : processing a client connection request may be delayed by the
1698 client or server longer than the client connect timeout. This
1699 causes the client to resend the connection request. If the
1700 original connection request is replied in this interval, the
1701 client may trip an assertion failure in ptlrpc_connect_interpret()
1702 which thought it would be the only running connect process.
1705 Frequency : only with obd_echo servers and clients that are rebooted
1707 Description: kernel BUG accessing uninitialized data structure
1708 Details : When running an obd_echo server it did not start the ping_evictor
1709 thread, and when a client was evicted an uninitialized data
1710 structure was accessed. Start the ping_evictor in the RPC
1711 service startup instead of the OBD startup.
1713 Severity : enhancement
1714 Bugzilla : 10193 (patchless)
1715 Description: Remove dependency on various unexported kernel interfaces.
1716 Details : No longer need reparent_to_init, exit_mm, exit_files,
1717 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1720 Frequency : rare (only users of deprecated and unsupported LDAP config)
1722 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1723 Details : LDAP apparently contains 'lustreName' attributes instead of
1724 'name'. A simple remapping of the name is sufficient.
1727 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1729 Description: waiting_locks_callback trips kernel BUG if client is evicted
1730 Details : Running with the dump_on_timeout debug flag turned on makes
1731 it possible that the waiting_locks_callback() can try to dump
1732 the Lustre kernel debug logs from an interrupt handler. Defer
1733 this log dumping to the expired_lock_main() thread.
1735 Severity : enhancement
1737 Description: Support NFS exporting on 2.6 kernels.
1738 Details : Implement non-rawops metadata methods for NFS server to use without
1739 changing NFS server code.
1742 Frequency : very rare (synthetic metadata workload only)
1744 Description: two racing renames might cause an MDS thread to deadlock
1745 Details : Running the "racer" program may cause one MDS thread to rename
1746 a file from being the source of a rename to being the target of
1747 a rename at exactly the same time that another thread is doing
1748 so, and the second thread has already enqueued these locks after
1749 doing a lookup of the target and is trying to relock them in
1750 order. Ensure that we don't try to re-lock the same resource.
1753 Frequency : only very large systems with liblustre clients
1755 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1756 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1757 all OSTs in parallel. This allows the eviction of stale liblustre
1758 clients to proceed much faster than if they were done in series,
1759 and also offers similar improvements for other set_info RPCs.
1764 Description: excessive CPU usage during initial read phase on client
1765 Details : During the initial read phase on a client, it would agressively
1766 retry readahead on the file, consuming too much CPU and impacting
1767 performance (since 1.4.5.8). Improve the readahead algorithm
1768 to avoid this, and also improve some other common cases (read
1769 of small files in particular, where "small" is files smaller than
1770 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1775 Description: MDS crash when receiving packet with unknown intent.
1776 Details : Do not LBUG in unknown intent case, just return -EFAULT
1778 Severity : enhancement
1779 Bugzilla : 9293, 9385
1780 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1781 Details : Do not serialize getattr (non-intent version) and statfs.
1784 Frequency : occasional, when OST network is overloaded/intermittent
1786 Description: client evicted by OST after bulk IO timeout
1787 Details : If a client sends a bulk IO request (read or write) the OST
1788 may evict the client if it is unresposive to its data GET/PUT
1789 request. This is incorrect if the network is overloaded (takes
1790 too long to transfer the RPC data) or dropped the OST GET/PUT
1791 request. There is no need to evict the client at all, since
1792 the pinger and/or lock callbacks will handle this, and the
1793 client can restart the bulk request.
1796 Frequency : Always when mmapping file with no objects
1798 Description: client crashes when mmapping file with no objects
1799 Details : Check that we actually have objects in a file before doing any
1800 operations on objects in ll_vm_open, ll_vm_close and
1806 Description: Request leak when working with deleted CWD
1807 Details : Introduce advanced request refcount tracking for requests
1808 referenced from lustre intent.
1810 Severity : Enhancement
1812 Description: Cache open file handles on client.
1813 Details : MDS now will return special lock along with openhandle, if
1814 requested and client is allowed to hold openhandle, even if unused,
1815 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1816 closing files for every read/write openration.
1818 Severity : Enhancement
1820 Description: Cache open negative dentries on client when possible.
1821 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1822 negative dentries on lock revocation.
1827 Description: Remounting a client read-only wasn't possible with a zconf mount
1828 Details : It wasn't possible to remount a client read-only with llmount.
1830 Severity : enhancement
1831 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1832 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1833 MPICH is included to add Lustre-specific ADIO interfaces.
1834 This is based closely on the UFS ADIO layer and only differs
1835 in file creation, in order to allow the OST striping to be set.
1836 This is user-contributed code and not supported by CFS.
1841 Description: extended inode attributes (immutable, append-only) work improperly
1842 when 2.4 and 2.6 kernels are used on client/server or vice versa
1843 Details : Introduce kernel-independent values for these flags.
1845 Severity : enhancement
1848 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1849 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1850 granularity. Now, Fractional MB granularity is be supported,
1851 this is very useful for low memory system.
1853 Severity : enhancement
1855 Description: Getattr by fid
1856 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1857 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1858 have locks and avoids getattr rpc requests for it on every stat.
1861 Frequency : Always, for filesystems larger than 2TB
1863 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1864 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1865 filesystems larger than 2TB to be mounted with mballoc enabled.
1868 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1870 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1871 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1872 filesystems larger than 2TB is mounted then the kernel will
1873 silently corrupt the start of the filesystem. CONFIG_LBD is
1874 enabled for all CFS-supported kernels, but the possibility of
1875 this happening with a modified kernel config exists.
1877 Severity : enhancement
1879 Description: add client O_DIRECT support for 2.6 kernels
1880 Details : It is now possible to do O_DIRECT reads and writes to files
1881 in the Lustre client mountpoint on 2.6 kernel clients.
1883 Severity : enhancement
1885 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1886 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1887 parallel, not waiting for response from every OST before sending
1888 a rpc to the next OST.
1893 Description: setattr vs write race when updating file timestamps
1894 Details : Client processes that update a file timestamp into the past
1895 right after writing to the file (e.g. tar) it is possible that
1896 the updated file modification time can be reset to the current
1897 time due to a race between processing the setattr and write RPC.
1899 Severity : enhancement
1901 Description: Bring 'lfs find' closer in line with regular Linux find.
1902 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1903 -print0 options and obtains all the needed info through the lustre
1906 Severity : enhancement
1908 Description: support up to 1024 configured devices on one node
1909 Details : change obd_dev array from statically allocated to dynamically
1910 allocated structs as they are first used to reduce memory usage
1915 Description: Flush dirty partially truncated pages during truncate
1916 Details : Immediatelly flush partially truncated pages in filter_setattr,
1917 this way we completely avoid having any pages in page cache on OST
1918 and can retire ugly workarounds during writes to flush such pages.
1923 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1924 Details : protect inode from truncation within vfs_unlink() context
1925 just take a reference before calling vfs_unlink() and release it
1926 when parent's i_sem is free.
1931 Description: last_id value checked outside lock on OST caused LASSERT failure
1932 Details : If there were multiple MDS->OST object precreate requests in
1933 flight, it was possible that the OST's last object id was checked
1934 outside a lock and incorrectly tripped an assertion. Move checks
1935 inside locks, and discard old precreate requests.
1938 Frequency : always, if extents are used on OSTs
1940 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1941 Details : OSTs using the extents format would not zero the high 16 bits of
1942 the index physical block number. This is not a problem for any
1943 OST filesystems smaller than 16TB, and no kernels support ext3
1944 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1945 new/modified files) and can be fixed for existing filesystems
1946 with e2fsprogs-1.39-cfs1.
1951 Description: import connection selection may be incorrect if timer wraps
1952 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1953 import connections to be ignored if the 32-bit jiffies counter
1954 wraps. Use a 64-bit jiffies counter.
1957 Frequency : very large clusters immediately after boot
1959 Description: LNET request buffers exhausted under heavy short-term load
1960 Details : If a large number of client requests are generated on a service
1961 that has previously never seen so many requests it is possible
1962 that the request buffer growth cannot keep up with the spike in
1963 demand. Instead of dropping incoming requests, they are held in
1964 the LND until the RPC service can accept more requests.
1967 Frequency : Sometimes during replay
1969 Description: Assertion failure in ll_local_open after replay.
1970 Details : If replay happened on an open request reply before we were able
1971 to set replay handler, reply will become not swabbed tripping the
1972 assertion in ll_local_open. Now we set the handler right after
1973 recognising of open request
1976 Frequency : very rare
1978 Description: kernel reports "badness in vsnprintf"
1979 Details : Reading from the "recovery_status" /proc file in small chunks
1980 may cause a negative length in lprocfs_obd_rd_recovery_status()
1981 call to vsnprintf() (which is otherwise harmless). Exit early
1982 if there is no more space in the output buffer.
1984 Severity : enhancement
1986 Description: clear OBD RPC statistics by writing to them
1987 Details : It is now possible to clear the OBD RPC statistics by writing
1988 to the "stats" file.
1993 Description: Client mtime is not the same on different clients after utimes
1994 Details : In some cases, the client was using the utimes() syscall on
1995 a file cached on another node. The clients now validate the
1996 ctime from the MDS + OSTs to determine which one is right.
2001 Description: Inability to activate failout mode
2002 Details : lconf script incorrectly assumed that in pythong string's numeric
2003 value is used in comparisons.
2006 Frequency : always with multiple stripes per file
2008 Description: Inefficient object allocation for mutli-stripe files
2009 Details : When selecting which OSTs to stripe files over, for files with
2010 a stripe count that divides evenly into the number of OSTs,
2011 the MDS is always picking the same starting OST for each file.
2012 Return the OST selection heuristic to the original design.
2017 Description: mount failures may take full timeout to return an error
2018 Details : Under some heavy load conditions it is possible that a
2019 failed mount can wait for the full obd_timeout interval,
2020 possibly several minutes, before reporting an error.
2021 Instead return an error as soon as the status is known.
2023 Frequency : quota enabled and large files being deleted
2025 Description: releasing more than 4GB of quota at once hangs OST
2026 Details : If a user deletes more than 4GB of files on a single OST it
2027 will cause the OST to spin in an infinite loop. Release
2028 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2033 Description: statfs data retrieved from /proc may be stale or zero
2034 Details : When reading per-device statfs data from /proc, in the
2035 {kbytes,files}_{total,free,avail} files, it may appear
2036 as zero or be out of date.
2039 Frequency : systems with MD RAID1 external journal devices
2041 Description: lconf's call to blkid is confused by RAID1 journal devices
2042 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2043 of returning all block devices that match the journal UUID.
2046 Frequency : always, for aggregate stripe size over 4GB
2048 Description: assertion fails when trying to use 4GB stripe size
2049 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2050 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2053 Frequency : always on ppc64
2055 Description: the first write on an ext3 filesystem with mballoc got stuck
2056 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2057 perform endianness conversion.
2059 ------------------------------------------------------------------------------
2061 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2063 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2064 INTEROPERATE with older versions automatically. Please read the
2065 user documentation before upgrading any part of a live system.
2066 * WARNING: Lustre networking configuration changes are required with
2067 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2070 * Support for kernels:
2071 2.6.9-22.0.2.EL (RHEL 4)
2072 2.6.5-7.244 (SLES 9)
2073 2.6.12.6 vanilla (kernel.org)
2076 Severity : enhancement
2077 Bugzilla : 7981/8208
2078 Description: Introduced Lustre Networking (LNET)
2079 Details : LNET is new networking infrastructure for Lustre, it includes
2080 a reorganized network configuration mode (see the user
2081 documentation for full details) as well as support for routing
2082 between different network fabrics. Lustre Networking Devices
2083 (LNDs) for the supported network fabrics have also been
2084 created for this new infrastructure.
2086 Severity : enhancement
2087 Description: Introduced Access control lists
2088 Details : clients can set ACLs on files and directories in order to have
2089 more fine-grained permissions than the standard Unix UGO+RWX.
2090 The MDS must be started with the "-o acl" mount option.
2092 Severity : enhancement
2093 Description: Introduced filesystem quotas
2094 Details : Administrators may now establish per-user quotas on the
2097 Severity : enhancement
2099 Description: Configuration change for the XT3
2100 The PTLLND is now used to run Lustre over Portals on the XT3
2101 The configure option(s) --with-cray-portals are no longer used.
2102 Rather --with-portals=<path-to-portals-includes> is used to
2103 enable building on the XT3. In addition to enable XT3 specific
2104 features the option --enable-cray-xt3 must be used.
2109 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2110 Details : A race between allocating a new llcd and re-getting the llcd_lock
2111 allowed another thread to grab newly-allocated llcd.
2113 Severity : enhancement
2115 Description: 2.6 OST async journal commit and locking fix to improve performance
2116 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2117 2.6 kernels are now async as they already were in 2.4 kernels so
2118 that they can commit concurrently with the network bulk transfer.
2119 For block-allocated files the filter allocation semaphore is held
2120 to avoid filesystem fragmentation during allocation. BKL lock
2121 removed for 2.6 xattr operations where it is no longer needed.
2126 Description: lconf incorrectly determined whether two IP networks could talk
2127 Details : In some more complicated routing and multiple-network
2128 configurations, lconf will avoid trying to make a network
2129 connection to a disjoint part of the IP space. It was doing the
2130 math incorrectly for one set of cases.
2135 Description: Fix for potential infinite loop processing records in an llog.
2136 Details : If an llog record is corrupted/zeroed, it is possible to loop
2137 forever in llog_process(). Validate the llog record length
2138 and skip the remainder of the block on error.
2141 Frequency : occasional (liblustre only)
2143 Description: liblustre could not open files whose last component is a symlink
2144 Details : sysio_path_walk() would incorrectly pass the open intent to
2145 intermediate path components.
2148 Frequency : rare (liblustre only with non-standard tuning)
2149 Bugzilla : 7201 (7350)
2150 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2151 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2152 after it has been released, only do this for referenced locks
2154 Severity : enhancement
2156 Description: specifying an (invalid) directory default stripe_size of -1
2157 would reset the directory default striping
2158 Details : stripe_size -1 was used internally to signal directory stripe
2159 removal, now use "all default" to signal dir stripe removal
2160 as a directory striping of "all default" is not useful
2163 Frequency : common for large clusters running liblustre clients
2165 Description: doing an ls when liblustre clients are running is slow
2166 Details : sending a glimpse AST to a liblustre client waits for every AST
2167 to time out, as liblustre clients will not respond. Since they
2168 cannot cache data we refresh the OST lock LVB from disk instead.
2170 Severity : enhancement
2172 Description: doing an ls at the same time as file IO can be slow
2173 Details : enqueue and other "small" requests can be blocked behind many
2174 large IO requests. Create a new OST IO portal for non-IO
2175 requests so they can be processed faster.
2178 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2180 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2181 created file with mode 000
2182 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2183 allow an ftruncate() as a non-root user to a file with mode 000.
2184 Lustre now respects this flag to disable mode checks when
2185 truncating a file owned by the user
2188 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2191 Description: Revoking locks from clients that went dead or catatonic might take
2193 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2194 cancellation of such locks instant on servers without waiting for
2195 any reply from clients. Clients drops these locks when cancel
2196 notification from server is received without replying.
2199 Frequency : liblustre-only, when liblustre client dies or becomes busy
2201 Description: Doing ls on Linux clients can take a long time with active
2203 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2204 granting such locks to it in the first place if possible. Locks
2205 are taken by proxy on the OST during the read or write and
2206 dropped immediately afterward. Add connect flags handling, do
2207 not grant locks to liblustre clients for glimpse ASTs.
2209 Severity : enhancement
2211 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2212 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2213 sequential reads and behaves badly with applications doing random
2214 reads. Improve it by reading ahead at least read region, and
2215 avoiding excessive large RPC for small reads.
2217 Severity : enhancement
2219 Description: Creating more than 1000 files for a single job may cause a load
2220 imbalance on the OSTs if there are also a large number of OSTs.
2221 Details : qos_prep_create() uses an OST index reseed value that is an
2222 even multiple of the number of available OSTs so that if the
2223 reseed happens in the middle of the object allocation it will
2224 still utilize the OSTs as uniformly as possible.
2229 Description: OST or MDS may oops in ping_evictor_main()
2230 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2231 but doesn't restart at beginning of obd_exports_timed list
2234 Severity : enhancement
2236 Description: improve by-nid export eviction on the MDS and OST
2237 Details : allow multiple exports with the same NID to be evicted at one
2238 time without re-searching the exports list.
2241 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2243 Description: MDS may oops in groups_free()
2244 Details : in rare race conditions a newly allocated group_info struct is
2245 freed again, and this can be NULL. The 2.4 compatibility code
2246 for groups_free() checked for a NULL pointer, but 2.6 did not.
2249 Frequency : common for liblustre clients doing little filesystem IO
2250 Bugzilla : 9352, 7313
2251 Description: server may evict liblustre clients accessing contended locks
2252 Details : if a client is granted a lock or receives a completion AST
2253 with a blocking AST already set it would not reply to the AST
2254 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2257 Frequency : lfs setstripe, only systems with more than 160 OSTs
2259 Description: unable to set striping with a starting offset beyond OST 160
2260 Details : llapi_create_file() incorrectly limited the starting stripe
2261 index to the maximum single-file stripe count.
2264 Frequency : LDAP users only
2266 Description: lconf did not handle in-kernel recovery with LDAP properly
2267 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2269 Severity : enhancement
2271 Description: bind OST threads to NUMA nodes to improve performance
2272 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2273 node and do their allocations there to localize memory access
2275 Severity : enhancement
2277 Description: llmount can determine client NID directly from Myrinet (GM)
2278 Details : the client NID code from gmnalnid was moved directly into
2279 llmount, removing the need to use this or specifying the
2280 client NID explicitly when mounting GM clients with zeroconf
2283 Frequency : if client is started with down MDS
2285 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2286 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2287 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2288 ptlrpc_import_delayed_req() because we didn't check if the
2289 request was interrupted, and we also didn't break out of the
2290 event loop if there was no timeout
2295 Description: data loss during non-page-aligned writes to a single file from
2296 both multiple nodes and multiple threads on one node at same time
2297 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2298 inconsistency led to false short-reads, that were cached and later
2299 used by ->prepare_write() to fill in partially written page,
2300 leading to data loss.
2303 Frequency : always, if lconf --abort_recovery used
2305 Description: lconf --abort_recovery fails with 'Operation not supported'
2306 Details : lconf was attempting to abort recovery on the MDT device and not
2309 Severity : enhancement
2311 Description: remove cleanup logs
2312 Details : replace lconf-generated cleanup logs with lustre internal
2313 cleanup routines. Eliminates the need for client-cleanup and
2316 Severity : enhancement
2318 Description: add support for EAs (user and system) on lustre filesystems
2319 Details : it is now possible to store extended attributes in the Lustre
2320 client filesystem, and with the user_xattr mount option it
2321 is possible to allow users to store EAs on their files also
2323 Severity : enhancement
2325 Description: Add possibility (config option) to show minimal available OST free
2327 Details : When compiled with --enable-mindf configure option, statfs(2)
2328 (and so, df) will return least minimal free space available from
2329 all OSTs as amount of free space on FS, instead of summary of
2330 free spaces of all OSTs.
2332 Severity : enhancement
2334 Description: do not expand extent locks acquired on OST-side
2335 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2336 by server: they are not cached anyway.
2339 Frequency : when mmap is used/binaries executed from Lustre
2341 Description: Unmmap pages before throwing them away from read cache.
2342 Details : llap_shrink cache now attempts to unmap pages before discarding
2343 them (if unmapping failed - do not discard). SLES9 kernel has
2344 extra checks that trigger if this unmapping is not done first.
2349 Description: lconf didn't resolve symlinks before checking to see whether a
2350 given mountpoint was already in use
2353 Frequency : when migrating failover services
2354 Bugzilla : 6395, 9514
2355 Description: When migrating a subset of services from a node (e.g. failback
2356 from a failover service node) the remaining services would
2357 time out and evict clients.
2358 Details : lconf --force (implied by --failover) sets the global obd_timeout
2359 to 5 seconds in order to quickly disconnect, but this caused
2360 other RPCs to time out too quickly. Do not change the global
2361 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2363 Severity : enhancement
2364 Frequency : if MDS is started with down OST
2365 Bugzilla : 9439,5706
2366 Description: Allow startup/shutdown of an MDS without depending on the
2367 availability of the OSTs.
2368 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2369 Add appropriate locking and lov-osc refcounts for safe
2370 cleaning. Add osc abort_inflight calls in case the
2371 synchronize never started.
2374 Frequency : occasional (Cray XT3 only)
2376 Description: root not authorized to access files in CRAY_PORTALS environment
2377 Details : The client process capabilities were not honoured on the MDS in
2378 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2379 been accessed by an authorized user then root was able to access
2380 the file on the local client also. The root user capabilities
2381 are now allowed on the MDS, as this environment has secure UID.
2384 Frequency : occasional
2386 Description: ldiskfs "too long searching" message happens too often
2387 Details : A debugging message (otherwise harmless) prints too often on
2388 the OST console. This has been reduced to only happen when
2389 there are fragmentation problems on the filesystem.
2394 Description: Division by zero in statfs when all OSCs are inactive
2395 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2396 lov_statfs divides by value returned by lov_get_stripecnt().
2400 Bugzilla : 9489, 3273
2401 Description: First write from each client to each OST was only 4kB in size,
2402 to initialize client writeback cache, which caused sub-optimal
2403 RPCs and poor layout on disk for the first writen file.
2404 Details : Clients now request an initial cache grant at (re)connect time
2405 and so that they can start streaming writes to the cache right
2406 away and always do full-sized RPCs if there is enough data.
2407 If the OST is rebooted the client also re-establishes its grant
2408 so that client cached writes will be honoured under the grant.
2413 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2414 Details : Now I/O RPCs go to different portal number and (presumably) fast
2415 lock requests (and glimses) and other RPCs get their own service
2416 threads pool that should be able to service those RPCs
2419 Severity : enhancement
2421 Description: Ability to exchange lustre version between client and servers and
2422 issue warnings at client side if client is too old. Also for
2423 liblustre clients there is ability to refuse connection of too old
2425 Details : New 'version' field is added to connect data structure that is
2426 filled with version info. That info is later checked by server and
2430 Frequency : rare, liblustre only.
2431 Bugzilla : 9296, 9581
2432 Description: Two simultaneous writes from liblustre at offset within same page
2433 might proceed at the same time overwriting eachother with stale
2435 Details : I/O lock withing llu_file_prwv was released too early, before data
2436 actually was hitting the wire. Extended lock-holding time until
2437 server acknowledges receiving data.
2440 Frequency : extremely rare. Never observed in practice.
2442 Description: avoid generating lustre_handle cookie of 0.
2443 Details : class_handle_hash() generates handle cookies by incrementing
2444 global counter, and can hit 0 occasionaly (this is unlikely, but
2445 not impossible, because initial value of cookie counter is
2446 selected randonly). Value of 0 is used as a sentinel meaning
2447 "unassigned handle" --- avoid it. Also coalesce two critical
2448 sections in this function into one.
2450 Severity : enhancement
2452 Description: allow liblustre clients to delegate truncate locking to OST
2453 Details : To avoid overhead of locking, liblustre client instructs OST to
2454 take extent lock in ost_punch() on client's behalf. New connection
2455 flag is added to handle backward compatibility.
2457 Severity : enhancement
2458 Bugzilla : 4928, 7341, 9758
2459 Description: allow number of OST service threads to be specified
2460 Details : a module parameter allows the number of OST service threads
2461 to be specified via "options ost ost_num_threads={N}" in the
2462 OSS's /etc/modules.conf or /etc/modprobe.conf.
2466 Bugzilla : 6146, 9635, 9895
2467 Description: servers crash with bad pointer in target_handle_connect()
2468 Details : In rare cases when a client is reconnecting it was possible that
2469 the connection request was the last reference for that export.
2470 We would temporarily drop the export reference and get a new
2471 one, but this may have been the last reference and the export
2472 was just destroyed. Get new reference before dropping old one.
2474 Severity : enhancement
2475 Frequency : if client is started with failover MDS
2477 Description: Allow multiple MDS hostnames in the mount command
2478 Details : Try to read the configuration from all specified MDS
2479 hostnames during a client mount in case the "primary"
2482 Severity : enhancement
2484 Description: Stop sending data to evicted clients as soon as possible.
2485 Details : Check if the client we are about to send or are sending data to
2486 was evicted already. (Check is done every second of waiting,
2487 for which l_wait_event interface was extended to allow checking
2488 of exit condition at specified intervals).
2491 Frequency : rare, normally only when NFS exporting is done from client
2493 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2495 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2498 Frequency : always, if srand() is called before liblustre initialization
2500 Description: Liblustre uses system PRNG disturbing its usage by user application
2501 Details : Introduce internal to lustre fast and high-quality PRNG for
2502 lustre usage and make liblustre and some other places in generic
2503 lustre code to use it.
2505 Severity : enhancement
2506 Bugzilla : 9477, 9557, 9870
2507 Description: Verify that the MDS configuration logs are updated when xml is
2508 Details : Check if the .xml configuration logs are newer than the config
2509 logs stored on the MDS and report an error if this is the case.
2510 Request --write-conf, or allow starting with --old_conf.
2512 Severity : enhancement
2514 Description: Handle symlinks in the path when checking if Lustre is mounted.
2515 Details : Resolve intermediate symlinks when checking if a client has
2516 mounted a filesystem to avoid duplicate client mounts.
2521 Description: lconf can hit an error exception but still return success.
2522 Details : The lconf command catches the Command error exception at the top
2523 level script context and will exit with the associated exit
2524 status, but doesn't ensure that this exit status is non-zero.
2529 Description: failure of ptlrpc thread startup can cause oops
2530 Details : Starting a ptlrpc service thread can fail if there are a large
2531 number of threads or the server memory is very fragmented.
2532 Handle this without oopsing.
2535 Frequency : always, only if liblustre and non-default acceptor port was used
2537 Description: liblustre cannot connect to servers with non-default acceptor port
2538 Details : tcpnal_set_default_params() was not called and was therefore
2539 ignoring the environment varaible TCPNAL_PORT, as well as other
2540 TCPNAL_ environment variables
2545 Description: two objects could be created on the same OST for a single file
2546 Details : If an OST is down, in some cases it was possible to create two
2547 objects on a single OST for a single file. No problems other
2548 than potential performance impact and spurious error messages.
2552 Bugzilla : 5681, 9562
2553 Description: Client may oops in ll_unhash_aliases
2554 Details : Client dcache may become inconsistent in race condition.
2555 In some cases "getcwd" can fail if the current directory is
2561 Description: Inode refcounting problems in NFS export code
2562 Details : link_raw functions used to call d_instantiate without obtaining
2563 extra inode reference first.
2567 Bugzilla : 9942, 9903
2568 Description: Referencing freed requests leading to crash, memleaks with NFS.
2569 Details : We used to require that call to ll_revalidate_it was always
2570 followed by ll_lookup_it. Also with revalidate_special() it is
2571 possible to call ll_revalidate_it() twice for the same dentry
2572 even if first occurence returned success. This fix changes semantic
2573 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2574 reference on a request referred from the intent.
2575 ll_intent_release() then releases such a request.
2578 Frequency : rare, normally benchmark loads only
2580 Description: unlinked inodes were kept in memory on the client
2581 Details : If a client is repeatedly creating and unlinking files it
2582 can accumulate a lot of stale inodes in the inode slab cache.
2583 If there is no other client load running this can cause the
2584 client node to run out of memory. Instead flush old inodes
2585 from client cache that have the same inode number as a new inode.
2588 Frequency : SLES9 2.6.5 kernel and long filenames only
2589 Bugzilla : 9969, 10379
2590 Description: utime reports stale NFS file handle
2591 Details : SLES9 uses out-of-dentry names in some cases, which confused
2592 the lustre dentry revalidation. Change it to always use the
2596 Frequency : rare, unless heavy write-truncate concurrency is continuous
2597 Bugzilla : 4180, 6984, 7171, 9963, 9331
2598 Description: OST becomes very slow and/or deadlocked during object unlink
2599 Details : filter_destroy() was holding onto the parent directory lock
2600 while truncating+unlinking objects. For very large objects this
2601 may block other threads for a long time and slow overall OST
2602 responsiveness. It may also be possible to get a lock ordering
2603 deadlock in this case, or run out of journal credits because of
2604 the combined truncate+unlink. Solution is to do object truncate
2605 first in one transaction without parent lock, and then do the
2606 final unlink in a new transaction with the parent lock. This
2607 reduces the lock hold time dramatically.
2610 Frequency : rare, 2.4 kernels only
2612 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2613 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2614 before unmounting their filesystem in order to avoid deadlock.
2615 The kernel_locked() function in 2.4 kernels only checks whether
2616 the kernel lock is held, not whether it is this process that is
2617 holding it as 2.6 kernels do.
2622 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2623 Details : The client ptlrpc code may be trying to reconnect to a down
2624 server before a previous connection attempt has timed out.
2625 Increase the reconnect interval to be longer than the connection
2626 timeout interval to avoid sending duplicate connections to
2630 Frequency : echo_client brw_test command
2632 Description: fix echo_client to work with OST preallocated code
2633 Details : OST preallocation code (5137) didn't take echo_client IO path
2634 into account: echo_client calls filter methods outside of any
2635 OST thread and, hence, there is no per-thread preallocated
2636 pages and buffers to use. Solution: hijack pga pages for IO. As
2637 a byproduct, this avoids unnecessary data copying.
2641 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2642 Description: Client can oops in mdc_commit_close() after open replay
2643 Details : It was possible for the MDS to return an open request with no
2644 transaction number in mds_finish_transno() if the client was
2645 evicted, but without actually returning an error. Clients
2646 would later try to replay that open and may trip an assertion
2647 Simplify the client close codepath, and always return an error
2648 from the MDS in case the open is not successful.
2651 Frequency : rare, 2.6 OSTs only
2653 Description: OST may deadlock under high load on fragmented files
2654 Details : If there was a heavy load and highly-fragmented OST filesystems
2655 it was possible to have all the OST threads deadlock waiting on
2656 allocation of biovecs, because the biovecs were not released
2657 until the entire RPC IO was completed. Instead, release biovecs
2658 as soon as they are complete to ensure forward IO progress.
2660 Severity : enhancement
2662 Description: Support for specifying external journal device at mount
2663 Details : If an OST or MDS device is formatted with an external journal
2664 device, this device major/minor is stored in the ext3 superblock
2665 and may not be valid for failover. Allow detecting and
2666 specifying the external journal at mount time.
2671 Description: Mounting an MDS with pending unlinked files may cause oops
2672 Details : target_finish_recovery() calls mds_postrecov() which returned
2673 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2674 considers this an error and immediately begins cleaning up the
2675 lov, just after starting the mds_lov process
2677 Severity : enhancement
2679 Description: Implement 'lfs df' to report actual free space on per-OST basis
2680 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2681 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2682 usage of objects; '-h' to report in human readable format.
2684 ------------------------------------------------------------------------------
2686 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2693 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2694 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2695 at startup during orphan unlinking. Instead, skip buddy update
2696 before setup, as it will be regenerated after recovery is complete.
2701 Description: If an OST is inactive, its locks might reference stale inodes.
2702 Details : lov_change_cbdata() must iterate over all namespaces, even if
2703 they are inactive to clear inode references from the lock.
2705 Severity : enhancement
2706 Frequency : occasional, if non-standard max_dirty_mb used
2708 Description: Client will block write RPCs if not enough grant
2709 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2710 then the client will block writes while waiting for another RPC
2711 to complete instead of consuming its dirty limit. With change
2712 we get improved performance when max_dirty_mb is small.
2714 Severity : enhancement
2715 Bugzilla : 3389, 6253
2716 Description: Add support for supplementary groups on the MDS.
2717 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2718 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2719 lookups for user supplementary groups into a cache.
2723 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2724 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2725 for liblustre O_EXCL is enforced by MDS.
2729 Description: readdir never returns NULL in liblustre.
2730 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2731 dentry in struct dirent.
2735 Frequency : liblustre only, depends on application IO pattern
2736 Description: liblustre clients evicted if not contacting servers
2737 Details : Don't put liblustre clients into the ping_evictor list, so
2738 they will not be evicted by the pinger ever.
2740 Severity : enhancement
2742 Description: Add ability to evict clients by NID from MDS.
2743 Details : By echoing "nid:$NID" string into
2744 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2745 $NID would be instantly evicted from this MDS and from all active
2746 OSTs connected to it.
2750 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2751 Details : lookup_it_finish() used to query file size from OSTs that was not
2756 Description: service threads change working directory to that of init
2757 Details : Starting lustre service threads may pin the working directory
2758 of the parent thread, making that filesystem busy. Threads
2759 now change to the working directory of init to avoid this.
2763 Frequency : during shutdown only
2764 Description: shutdown with a failed MDS or OST can cause unmount to hang
2765 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2770 Frequency : 2.6 only
2771 Description: chmod/chown may include an extra supplementary group
2772 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2773 supplementary group and if none is specified this is used.
2776 Bugzilla : 5479 (6816)
2778 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2779 Details : If lookup is in progress on a file that is unlinked we might try
2780 to revalidate the inode and fail in revalidate after lookup is
2781 complete and ll_file_open() enqueues the open again but
2782 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2785 Frequency : always, if lconf --abort_recovery used
2787 Description: lconf --abort_recovery fails with 'Operation not supported'
2788 Details : lconf was attempting to abort recovery on the MDT device and not
2791 ------------------------------------------------------------------------------
2793 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2798 Frequency : rare (only unsupported configurations with a node running as an
2800 Bugzilla : 6514, 5137
2801 Description: Mounting a Lustre file system on a node running as an OST could
2803 Details : OSTs now preallocates memory needed to write out data at
2804 startup, instead of when needed, to avoid having to
2805 allocate memory in possibly low memory situations.
2806 Specifically, if the file system is mounted on on OST,
2807 memory pressure could force it to try to write out data,
2808 which it needed to allocate memory to do. Due to the low
2809 memory, it would be unable to do so and the node would
2810 become unresponsive.
2812 Severity : enhancement
2814 Description: Addition of lconf --service command line option
2815 Details : lconf now accepts a '--service <arg>' option, which is
2816 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2818 Severity : enhancement
2820 Description: Failover mode is now the default for OSTs.
2821 Details : By default, OSTs will now run in failover mode. To return to
2822 the old behaviour, add '--failout' to the lmc line for OSTs.
2824 Severity : enhancement
2826 Description: Health checks are now provided for MDS and OSTs
2827 Details : Additional detailed health check information on MSD and OSTs
2828 is now provided through the procfs health_check value.
2831 Frequency : occasional, depends on IO load
2833 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2834 after numerous create/delete cycles
2835 Details : The ext3 inode allocation policy would not allocate new inodes
2836 very well on the OSTs because there are no new directories
2837 being created. Instead we look for groups with free space if
2838 the parent directories are nearly full.
2843 Description: Network or server problems during mount may cause partially
2844 mounted clients instead of returning an error.
2845 Details : The config llog parsing code may overwrite the error return
2846 code during mount error handling, returning success instead
2852 Description: MDS can fail to allocate large reply buffers
2853 Details : After long uptimes the MDS can fail to allocate large reply
2854 buffers (e.g. zconf client mount config records) due to memory
2855 fragmentation or consumption by the buffer cache. Preallocate
2856 some large reply buffers so that these replies can be sent even
2857 under memory pressure.
2861 Frequency : rare (liblustre)
2862 Description: fsx running with liblustre complained that using truncate() to
2863 extend the file doesn't work. This patch corrects that issue.
2864 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2865 ATTR_SIZE and lsm use in llu_setattr_raw.
2869 Frequency : rare, only 2.6 kernels
2870 Description: Unusual file access patterns on the MDS may result in inode
2871 data being lost in very rare circumstances.
2872 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2873 code in the 2.6 kernel caused the inode and/or EA data not to
2874 be read from disk, causing single-file corruption.
2878 Frequency : rare, only 2.6 filesystems using extents
2879 Description: Heavy concurrent write and delete load may cause data corruption.
2880 Details : It was possible under high-load situations to have an extent
2881 metadata block in the block device cache from a just-unlinked
2882 file overwrite a newly-allocated data block. We now unmap any
2883 metadata buffers that alias just-allocated data blocks.
2887 Frequency : filesystems with default stripe_count larger than 77
2888 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2889 Details : lconf specifies an inode size of 4096 bytes when the default
2890 stripe_count is larger than 77. This conflicts with the default
2891 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2892 this case to avoid pinning too much memory for large EAs.
2894 ------------------------------------------------------------------------------
2896 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2901 Frequency : rare (extremely heavy IO load with hundreds of clients)
2903 Description: Client is evicted, gets IO error writing to file
2904 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2905 caused clients to be evicted for AST timeouts. The fixes in
2906 bug 5192 mean we no longer need to have such short AST timeouts
2907 so ldlm_timeout has been increased.
2910 Frequency : occasional during --force or --failover shutdown under load
2911 Bugzilla : 5949, 4834
2912 Description: Server oops/LBUG if stopped with --force or --failover under load
2913 Details : a collection of import/export refcount and cleanup ordering
2914 issues fixed for safer force cleanup
2917 Frequency : only filesystems larger than 120 OSTs
2918 Bugzilla : 5990, 6223
2919 Description: lfs getstripe would oops on a very large filesystem
2920 Details : lov_getconfig used kfree on vmalloc'd memory
2923 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2924 Bugzilla : 6242, 6243
2925 Description: reading from files that had been truncated to a non-zero size
2926 but never opened returned no data
2927 Details : ll_file_read() reads zeros from no-object files to EOF
2932 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2933 Details : There's little harm in aborting MDS/OSS recovery and letting it
2934 try again, so I removed the LASSERT and return an error instead.
2936 Severity : enhancement
2938 Description: New debugging infrastructure for tracking down data corruption
2939 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2940 (b) cover more of the client-side code path, and (c) try to narrow
2941 down where problems occurred
2945 Bugzilla : 3819, 4364, 4397, 6313
2946 Description: Racing close and eviction MDS could cause assertion in mds_close
2947 Details : It was possible to get multiple mfd references during close and
2948 client eviction, leading to one thread referencing a freed mfd.
2950 Severity: : enhancement
2951 Bugzilla : 3262, 6359
2952 Description: Attempts to reconnect to servers are now more aggressive.
2953 Details : This builds on the enhanced upcall-less recovery that was added
2954 in 1.4.2. When trying to reconnect to servers, clients will
2955 now try each server in the failover group every 10 seconds. By
2956 default, clients would previously try one server every 25 seconds.
2961 Description: After recovery, certain operations trigger a failed
2962 assertion on a client.
2963 Details : Failing over an mds, using lconf -d --failover, while a
2964 client was doing a readdir() call would cause the client to
2965 LBUG after recovery completed and the readdir() was resent.
2967 Severity : enhancement
2969 Description: Default groups are now added by lconf
2970 Details : You can now run lconf --group <servicename> without having to
2971 manually add groups with lmc.
2974 Frequency : occasional
2976 Description: Nodes with an elan id of 0 trigger a failed assertion
2979 Frequency : always when accessing e.g. tty/console device nodes
2981 Description: tty and some other devices nodes cannot be used on lustre
2982 Details : file's private_data field is used by device data and lustre
2983 values in there got lost. New field was added to struct file to
2984 store fs-specific private data.
2987 Frequency : when exporting Lustre via NFS
2989 Description: NFSD failed occasionally when looking up a path component
2990 Details : NFSD is looking up ".." which was broken in ext3 directories
2991 that had grown large enough to become hashed.
2994 Frequency : Clusters with multiple interfaces not on the same subnet
2996 Description: Nodes will repeatedly try to reconnect to an interface which it
2997 cannot reach and report an error to the log.
2998 Details : Extra peer list entries will be created by lconf with some peers
2999 unreachable. lconf now validates the peer before adding it.
3002 Frequency : Only if a default stripe is set on the filesystem root.
3004 Description: Setting a default stripe on the filesystem root prevented the
3005 filesystem from being remounted.
3006 Details : The client was sending extra request flags in the root getattr
3007 request and did not allocate a reply buffer for the dir EA.
3010 Frequency : occasional, higher if lots of files are accessed by one client
3011 Bugzilla : 6159, 6097
3012 Description: Client trips assertion regarding lsm mismatch/magic
3013 Details : While revalidating inodes the VFS looks up inodes with ifind()
3014 and in rare cases can find an inode that is being freed.
3015 The ll_test_inode() code will free the lsm during ifind()
3016 when it finds an existing inode and then the VFS later attaches
3017 this free lsm to a new inode.
3021 Bugzilla : 6422, 7030
3022 Description: MDS deadlock between mkdir and client eviction
3023 Details : Creating a new file via mkdir or mknod (starting a transaction
3024 and getting the ns lock) can deadlock with client eviction
3025 (gets ns lock and trying to finish a synchronous transaction).
3028 Frequency : occasional
3029 Description: While starting a server, the fsfilt_ext3 module could not be
3031 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3032 kernels. Previously, lconf would still use the ext3 name
3033 when trying to load modules. Now, it will correctly use
3034 ext3 on 2.4 and ldiskfs on 2.6.
3036 Severity : enhancement
3037 Description: The default stripe count has been changed to 1
3038 Details : The interpretation of the default stripe count (0, to lfs
3039 or lmc) has been changed to mean striping across a single
3040 OST, rather than all available. For general usage we have
3041 found a stripe count of 1 or 2 works best.
3043 Severity : enhancement
3044 Description: Add support for compiling against Cray portals.
3045 Details : Conditional compiling for some areas that are different
3049 Frequency : occasional
3050 Bugzilla : 6409, 6834
3051 Description: Creating files with an explicit stripe count may lead to
3052 a failed assertion on the MDS
3053 Details : If some OSTs are full or unavailable, creating files may
3054 trigger a failed assertion on the MDS. Now, Lustre will
3055 try to use other servers or return an error to the
3059 Frequency : occasional
3061 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3062 caused lock timeout during readahead (since 1.4.2).
3063 Details : Processes doing readahead might match a lock that hasn't been
3064 granted yet if there are overlapping and conflicting lock
3065 requests. The readahead process waits on ungranted lock
3066 (original lock is CBPENDING), while OST waits for that process
3067 to cancel CBPENDING read lock and eventually evicts client.
3069 Severity : enhancement
3071 Description: Initial enabling of flock support for clients
3072 Details : Implements fcntl advisory locking and file status functions.
3073 This feature is provided as an optional mount flag (default
3074 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3075 locking are implemented yet, and those that are are not guaranteed
3076 to be completely correct in production environments.
3077 mount -t lustre -o [flock|noflock] ...
3080 Frequency : occasional
3082 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3083 assertion in the ext3 JBD (journaling) layer.
3084 Details : The b_committed_data struct is protected by the big kernel lock
3085 in 2.4 kernels, serializing journal_commit_transaction() and
3086 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3087 access to this struct. In 2.6 kernels there is finer grained
3088 locking to improve SMP performance of the JBD layer.
3092 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3097 Description: merge and cleanup kernel patches.
3100 -----------------------------------------------------------------------------
3102 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3104 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3105 versions of Lustre. Please update all servers and clients to
3106 version 1.4.2 or later at the same time. You must also run
3107 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3108 to update the configuration logs.
3110 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3111 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3112 - swab llog records of type '0' so we get proper header size/idx (5861)
3113 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3114 - fix rename of one directory over another leaking an inode (5953)
3115 - avoid SetPageDirty on 2.6 (5981)
3116 - don't re-add just-being-destroyed locks to the waiting list (5653)
3117 - when creating new directories, inherit the parent's custom
3118 striping settings if present parent (3048)
3119 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3120 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3121 - don't hold client locks on temporary worklist from l_lru (5666)
3122 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3123 - automatically evict dead clients (5921)
3124 - Update file size properly in create+truncate+fstat case (6196)
3125 - Do not unhash mountpoint dentries, do not allow removal of
3127 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3128 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3129 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3130 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3131 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3132 - Correctly handle reads of files with no objects (6243)
3133 - lctl recover will also mark a device active if deactivate used (5933)
3135 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3136 - allow --write-conf on an MDS with different nettype than client (5619)
3137 - don't write config llogs to MDS for mounts not from that MDS (5617)
3138 - lconf should create multiple TCP connections from a client (5201)
3139 - init scripts are now turned off by default; run chkconfig --on
3140 lustre and chkconfig --on lustrefs to use them
3141 - upcalls are no longer needed for clients to recover to failover
3143 - add --abort-recovery option to lconf to abort recovery on device
3145 - add support for an arbitrary number of OSTs (3026)
3146 - Quota support protocol changes.
3147 - forward compatibility changes to wire structs (6007)
3148 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3149 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3150 - improved "lustre status" script
3151 - initialize blocksize for non-regular files (6062)
3152 - added --disable-server and --disable-client configure options (5782)
3153 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3154 - Vanilla 2.4.29 support
3155 - increase maximum number of obd devices to 520 (6242)
3156 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3157 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3158 SLES 9 2.6.5 kernels
3159 - limit stripes per file to 160 (the maximum EA size) (6093)
3161 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3164 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3165 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3166 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3167 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3168 - small ext3 extents cleanups and fixes (5733)
3169 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3170 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3171 - avoid needless client->OST connect, fix handle mismatch (5317)
3172 - fix DLM error path that led to out-of-sync client, long delays (5779)
3173 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3174 - fix several locking issues related to i_size (5492,5624,5654,5672)
3175 - don't move pending lock onto export if it is already evicted (5683)
3176 - fix kernel oops when creating .foo in unlinked directory (5548)
3177 - fix deadlock in obdfilter statistics vs. object create (5811)
3178 - use time_{before,after} to avoid timer jiffies wrap (5882)
3179 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3180 - Do not leak request if server was not able to process it (5154)
3181 - If mds_open unable to find parent dir, make that negative lookup(5154)
3182 - don't create new directories with extent-mapping (5909, 5936)
3184 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3185 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3186 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3187 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3188 - support for RHEL4 kernel on i686 (5773)
3189 - provide error messages when incompatible logs are encountered (5898)
3191 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3192 * version 1.4.0.10 (1.4.1 release candidate 1)
3194 - don't keep a lock reference when lock is not granted (4238)
3195 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3196 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3197 - drop import inflight refcount on signal_completed_replay error (5255)
3198 - unlock page after async write error during send (3677)
3199 - handle missing objects in filter_preprw_read properly (5265)
3200 - no transno return for symlink open, don't save no-trasno open (3440)
3201 - don't try to complete elan receive that already failed (4012)
3202 - free RPC server reply state on error (5406)
3203 - clean up thread from ptlrpc_start_thread() on error (5160)
3204 - readahead could read extra page into cache that wasn't ejected (5388)
3205 - prevent races in class_attach/setup/cleanup/detach (5260)
3206 - don't dereference de->d_inode after l_dput of de (5458)
3207 - use "int" for stripe value returned from lock_to_stripe (5544)
3208 - mballoc allocation and error-checking fixes in 2.6 (5504)
3209 - block device patches to fix I/O request sizes in 2.6 (5482)
3210 - look up hostnames for IB nals (5602)
3211 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3212 - don't start multiple acceptors for the same port (5277)
3213 - fix incorrect LASSERT in mds_getattr_name (5635)
3214 - export a proc file for general "ping" checking (5628)
3215 - fix "lfs check" to not block when the MDS is down (5628)
3217 - service request history (4965)
3218 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3219 - create an "evict_client" /proc entry on OSTs, like the MDS has
3220 - fix mount usage message, return errors per mount(8) (5168)
3221 - change grep [] to grep "[]" in tests so they work in more UMLs
3222 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3223 - remove ancient LOV_MAGIC_V0 EA support (5047)
3224 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3225 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3226 - allow readahead allocations to fail when low on memory (5383)
3227 - mmap locking landed again, after considerable improvement (2828)
3228 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3230 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3233 - send OST transaction number in read/write reply to free req (4966)
3234 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3235 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3236 - the watchdog thread now runs as interruptible (5246)
3237 - flock/lockf fixes (but it's still disabled, pending 5135)
3238 - don't use EXT3 constants in llite code (5094)
3239 - memory shortage at startup could cause assertion (5176)
3241 - reorganization of lov code
3242 - single portals codebase
3244 - add extents/mballoc support (5025)
3245 - direct I/O reads in the obdfilter (4048)
3246 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3248 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3251 - send OST transaction number in read/write reply to free req (4966)
3252 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3253 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3254 - the watchdog thread now runs as interruptible (5246)
3255 - handle missing objects in filter_preprw_read properly (5265)
3256 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3257 - drop import inflight refcount on signal_completed_replay error (5255)
3258 - unlock page after async write error during send (3677)
3259 - return original error code on reconstructed replies (3761)
3260 - no transno return for symlink open, don't save no-trasno open (3440)
3262 - add pid to ldlm debugging output (4922)
3263 - bump the watchdog timeouts -- we can't handle 30sec yet
3264 - extra debugging for orphan dentry/inode bug (5259)
3266 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3269 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3270 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3271 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3272 - fix NULL dereference in /proc/sys/portals/routes (4827)
3273 - allow failed mdc_close() operations to be interrupted (4561)
3274 - stop precreate on OST before MDS would time out on it (4778)
3275 - don't send partial-page writes before EOF from client (4410)
3276 - discard client grant for sub-page writes on large-page clients (4520)
3277 - don't free dentries not owned by NFS code, check generation (4806)
3278 - fix lsm leak if mds_create_objects() fails (4801)
3279 - limit debug_daemon file size, always print CERROR messages (4789)
3280 - use transno after validating reply (3892)
3281 - process timed out requests if import state changes (3754)
3282 - update mtime on OST during writes, return in glimpse (4829)
3283 - add mkfsoptions to LDAP (4679)
3284 - use ->max_readahead method instead of zapping global ra (5039)
3285 - don't interrupt __l_wait_event() during strace
3287 - add software watchdogs to catch hung threads quickly (4941)
3288 - make lustrefs init script start after nfs is mounted
3289 - fix CWARN/ERROR duplication (4930)
3290 - return async write errors to application if possible (2248)
3291 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3292 - print NAL number in %x format (4645)
3293 - update barely-supported suse-2.4.21-171 series (4842)
3294 - support for sles 9 %post scripts
3295 - support for building 2.6 kernel-source packages
3296 - support for sles km_* packages
3298 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3301 - ignore -ENOENT errors in osc_destroy (3639)
3302 - notify osc create thread that OSC is being cleaned up (4600)
3303 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3304 - reconstruct ost_handle() like mds_handle() (4657)
3305 - create a new thread to do import eviction to avoid deadlock (3969)
3306 - let lconf resolve symlinked-to devices (4629)
3307 - don't unlink "objects" from directory with default EA (4554)
3308 - hold socknal file ref over connect in case target is down (4394)
3309 - allow more than 32000 subdirectories in a single directory (3244)
3310 - fix blocks count for O_DIRECT writes (3751)
3311 - OST returns ENOSPC from object create when no space left (4539)
3312 - don't send truncate RPC if file size isn't changing (4410)
3313 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3314 - bind to privileged port in socknal and tcpnal (3689)
3316 - rate limit CERROR/CWARN console message to avoid overload (4519)
3317 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3318 - basic mmap support (3918)
3319 - kernel patch series update from b1_4 (4711)
3321 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3324 - avoid crash during MDS cleanup with OST shut down (2775)
3325 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3326 - don't use bad inodes on the MDS (3744)
3327 - dynamic object preallocation to improve recovery speed (4236)
3328 - don't hold spinlock over lock dumping or change debug flags (4401)
3329 - don't zero obd_dev when it is force cleaned (3651)
3330 - print grants to console if they go negative (4431)
3331 - "lctl deactivate" will stop automatic recovery attempts (3406)
3332 - look for existing locks in ldlm_handle_enqueue() (3764)
3333 - don't resolve lock handle twice in recovery avoiding race (4401)
3334 - revalidate should check working dir is a directory (4134)
3336 - don't always mark "slow" obdfilter messages as errors (4418)
3338 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3341 - don't close LustreDB during write_conf until it is done (3860)
3342 - fix typo in lconf for_each_profile (3821)
3343 - allow dumping logs from multiple threads at one time (3820)
3344 - don't allow multiple threads in OSC recovery (3812)
3345 - fix debug_size parameters (3864)
3346 - fix mds_postrecov to initialize import for llog ctxt (3121)
3347 - replace config semaphore with spinlock (3306)
3348 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3349 - don't allow enqueue to complete on a destroyed export (3822)
3350 - down write_lock before checking llog header bitmap (3825)
3351 - recover from lock replay timeout (3764)
3352 - up llog sem before sending rpc (3652)
3353 - reduce ns lock hold times when setting kms (3267)
3354 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3355 - fix NULL deref and obd_dev leak on setup error (3312)
3356 - replace some LBUG about llog ops with error handling (3841)
3357 - don't match INVALID dentries from d_lookup and spin (3784)
3358 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3359 - fix invalid assertion in ptlrpc_set_wait (3880)
3361 - add libwrap support for the TCP acceptor (3996)
3362 - add /proc/sys/portals/routes for non-root route listing (3994)
3363 - allow setting MDS UUID in .xml (2580)
3364 - print the stack of a process that LBUGs (4228)
3366 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3369 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3370 - make sure to unset replay flag from failed open requests (3440)
3371 - if default stripe count is 0, use OST count for inode size (3636)
3372 - update parent mtime/ctime on client for create/unlink (2611)
3373 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3374 - free recovery state on server during a forced cleanup (3571)
3375 - unregister_reply for resent reqs (3063)
3376 - loop back devices mounting and status check on 2.6 (3563)
3377 - fix resource-creation race that can provoke i_size == 0 (3513)
3378 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3379 - more debugging for page-accounting assertion (3746)
3380 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3381 - avoid deadlock after precreation failure (3758)
3382 - fix race and lock order deadlock in orphan handling (3450, 3750)
3383 - add validity checks when grabbing inodes from l_ast_data (3599)
3385 - add /proc/.../recovery_status to obdfilter (3428)
3386 - lightweight CDEBUG infrastructure, debug daemon (3668)
3387 - change default OSC RPC parameters to be better on small clusters
3388 - turn off OST read cache for files smaller than 32MB
3389 - install man pages and include them in rpms (3100)
3390 - add new init script for (un)mounting lustre filesystems (2593)
3391 - run chkconfig in %post for init scripts (3701)
3392 - drop scimac NAL (unmaintained)
3394 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3397 - clean kiobufs before and after use (3485)
3398 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3399 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3400 - change init script to fail more clearly if not run as root (1528)
3401 - allow clients to reconnect during replay (1742)
3402 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3403 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3404 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3405 - fix cond_resched (3554)
3406 - only evict unfinished clients after recovery (3515)
3407 - allow bulk resend, prevent data loss (3570)
3408 - dynamic ptlrpc request buffer allocation (2102)
3409 - don't allow unlinking open directory if it isn't empty (2904)
3410 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3411 - remove extraneous obd dereference causing LASSERT failure (3334)
3412 - don't use get_cycles() when creating temp. files on the mds (3156)
3413 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3414 - handle EEXIST for set-stripe, set proper directory name (3336)
3416 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3417 - fix ksocknal_fmb_callback() error messages (2918)
3419 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3422 - don't copy lvb into (possibly NULL) reply on error (2983)
3423 - don't deref dentry after dput, don't free lvb on error (2922)
3424 - use the kms to determine writeback rpc length (2947)
3425 - increment oti_logcookies when osc is inactive (2948)
3426 - update client's i_blocks count via lvb messages (2543)
3427 - handle intent open/close of special files properly (1557)
3428 - mount MDS with errors=remount-ro, like obdfilter (2009)
3429 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3430 - don't use cancelling-locks' kms values (2947)
3431 - use highest lock extent for kms, not last one (2925)
3432 - don't dereference ERR_PTR() dentry in error handling path (3107)
3433 - fix thread race in portals_debug_dumplog() (3122)
3434 - create lprocfs device entries at setup instead of at attach (1519)
3435 - common AST error handler, don't evict client on completion race (3145)
3436 - zero nameidata in detach_mnt in 2.6 (3118)
3437 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3438 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3439 - initialize RPC timeout timer earlier for 2.6 (3219)
3440 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3441 - print nal/nid for unknown nid (3258)
3442 - additional checks for oscc recovery before doing precreate (3284)
3443 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3444 - don't crash in mdc_close for bad permissions on open (3285)
3445 - zero i_rdev for non-device files (3147)
3446 - clear page->private before handing to FS, better assertion (3119)
3447 - tune the read pipeline (3236)
3448 - fix incorrect decref of invalidated dentry (2350)
3449 - provide read-ahead stats and refine rpc in flight stats (3328)
3450 - don't hold journal transaction open across create RPC (3313)
3451 - update atime on MDS at close time (3265)
3452 - close LDAP connection when recovering to avoid server load (3315)
3453 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3454 - don't leak open file on MDS after open resend (3325)
3455 - serialize filter_precreate and filter_destroy_precreated (3329)
3456 - loop device shouldn't call sync_dev() for nul device (3092)
3457 - clear page cache after eviction (2766)
3458 - resynchronize MDS->OST in background (2824)
3459 - refuse to mount the same filesystem twice on same mountpoint (3394)
3460 - allow llmount to create routes for mounting behind routers (3320)
3461 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3462 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3463 - fix rare problem with rename on htree directories (3417)
3465 - allow default OST striping configuration per directory (1414)
3466 - fix compilation for qswnal for 2.6 kernels (3125)
3467 - increase maximum number of MDS request buffers for large systems
3468 - change liblustreapi to be useful for external progs like lfsck (3098)
3469 - increase local configuration timeout for slow disks (3353)
3470 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3472 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3475 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3476 - don't overwrite extent policy data in reply if lock was blocked (2901)
3477 - drop filter export grants atomically with removal from device (2663)
3478 - del obd_self_export from work_list in class_disconnect_exports (2908)
3479 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3480 - swab reply message in mdc_close, other PPC fixes (2464)
3481 - fix destroying of named logs (2325)
3482 - overwrite old logs when running lconf --write_conf (2264)
3483 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3484 - fix race in target_handle_connect (2898)
3485 - mds_reint_create() should take same inode create lock (2926)
3486 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3487 - don't close files for self_export to avoid uninitialized obd (2936)
3488 - allow MDS with the same name as client node (2939)
3489 - hold dentry reference for closed log files for unlink (2325)
3490 - reserve space for all logs during transactions (2059)
3491 - don't evict page beyond end of stripe extent (2925)
3492 - don't oops on a deleted current working directory (2399)
3493 - handle hard links to targets without a parent properly (2517)
3494 - don't dereference NULL lock when racing during eviction (2867)
3495 - don't grow lock extents when lots of conflicting locks (2919)
3497 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3500 - account for cache space usage on clients to avoid data loss (974)
3501 - lfsck support in lustre kernel code (2349)
3502 - reduce journal credits needed for BRW writes (2370)
3503 - orphan handling to avoid losing space on client/server crashes
3504 - ptlrpcd can be blocked, stopping ALL progress (2477)
3505 - use lock value blocks to assist in proper KMS, faster stat (1021)
3506 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3507 - recovery for initial connections (2355)
3508 - fixes for mds_cleanup_orphans (1934)
3509 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3510 - block all file creations until orphan recovery completes (1901)
3511 - client remove rq_connection from request struct (2423)
3512 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3513 - recovery timer race (2670)
3514 - mdc_close recovey bug (2532)
3515 - ptlrpc cleanup bug (2710)
3516 - mds timeout on local locks (2588)
3517 - namespace lock held during RPCs (2431)
3518 - handle interrupted sync write properly (2503)
3519 - don't try to handle a message that hasn't been replied to (2699)
3520 - client assert failure during cleanup after abort recovery (2701)
3521 - leak mdc device after failed mount (2712)
3522 - ptlrpc_check_set allows timedout requests to complete (2714)
3523 - wait for inflight reqs when ptlrpcd finishes (2710)
3524 - make sure unregistered services are removed from the srv_list
3525 - reset bulk XID's when resending them (caught by 1138 test)
3526 - unregister_bulk after timeout
3527 - fix lconf error (2694)
3528 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3529 - readahead locks pages, leaves pending causing memory pressure (2673)
3530 - increase OST request buffers to 4096 on large machines (2729)
3531 - fix up permission of existing directories in simple_mkdir (2661)
3532 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3533 - don't assign transno to errored transactions (2742)
3534 - don't delete objects on OST if given a bogus objid from MDS (2751)
3535 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3536 - if rq_no_resend, then timeout request after recovery (2432)
3537 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3538 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3539 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3540 - MDS recovery shouldn't delete objects that it has given out (2730)
3541 - if enqueue arrives after completion, don't clobber LVB (2819)
3542 - don't unlock pages twice when trigger_group_io returns error (2814)
3543 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3544 - don't write pages to disk if there was an error (1450)
3545 - don't ping imports that have recovery disabled (2676)
3546 - take buffered bytes into account when balancing socknal conn (2817)
3547 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3548 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3549 - remove little-endian swabbing of llog records (1987)
3550 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3551 - retry reposting request buffers if they fail (1191)
3552 - grow extent at grant time to avoid granting a revoked lock (2809)
3553 - lock revoke doesn't evict page if covered by a second lock (2765)
3554 - disable VM readahead to avoid reading outside lock extents (2805)
3556 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3557 - updated kernel patches for hp-2.4.20 kernel (2681)
3559 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3562 - fix truncated write corruption (2366)
3563 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3565 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3566 - ensure len(uuid) < 37 in lmc (1171)
3567 - fix ia64 OOPS in llog_test (2255)
3568 - zero end of page at obdfilter for partial page writes (2648)
3569 - don't leave stale dentries around after renames (bug 2428)
3570 - fix timeouts when evicting a client with a single lock held (2642)
3571 - set deadline for the initial HELLO message to drain (2634)
3572 - print out dotted-quad IP addresses in the socknal (2302)
3574 - additional debugging for MDS client eviction problem (2443)
3575 - fix mkfsoptions support for osts (2603, 2604)
3577 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3580 - add series for the vanilla 2.6.0 kernel
3581 - add series for the vanilla 2.4.24 kernel
3582 - add series for a cray x86/64 UL kernel drop
3583 - fix xattr patches for the vanilla 2.4.19 series
3585 - generate true UUIDs in lmc (1171)
3586 - have portals stack dumping break in UML (2466)
3587 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3588 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3589 - wake callback waiting threads on client eviction (2460)
3590 - Add --ptldebug and --subsystem to lmc (1719)
3591 - update assertion to allow safe interrupt allocation
3592 - set rq_no_resend for cancel requests (2432)
3593 - recalculate ptlrpcd timeout after resend (2494)
3594 - call vfs_rmdir when removing pending directories (2368)
3595 - fix renaming a file to itself (2429)
3596 - lmc creates a default one-stripe lov (2454)
3597 - expand procfs space to handle large clusters (2326)
3598 - increase UML stack to avoid overflow
3599 - update lconf's list of debug and subsystem masks
3600 - fix lfs find --obd (2510)
3601 - /proc tunable for disabling filter read caching (2591)
3602 - stop rpm packages from altering slapd.conf (2301)
3603 - disable nagle in the socknal under 0conf (2578)
3604 - choose mds inode size based on stripe count (2572)
3605 - fix kernel-source rpm problems (2516)
3607 - add --disable-doc to avoid pdf generation (2421)
3608 - update documentation, tests, type-os, comments
3609 - avoid format warnings on ia64
3610 - remove the TOE NAL
3611 - tiny code cleanups by removing unused fields
3613 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3616 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3617 - remove the most blatant lies from BUILDING file (2371)
3618 - change default debug level to reasonable production setting
3619 - reduce client side cache size to reduce cache flush time
3620 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3621 - make TCP zerocopy and pinger support enabled by default (2476)
3622 - sync writes completed after process exits caused crashes (2319)
3623 - maintain correct mount count on the MDS (2356)
3624 - backout 1557, because 2316 wasn't really fixed
3625 - better file I/O statistics gathering in /proc
3626 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3627 - another kernel patch to fix zero-copy TCP function export
3628 - don't take duplicate lock when processing re-sent getattr (2420)
3629 - lctl uses obd_self_export instead of creating new conn (2353)
3630 - MDS/OST recovery case which requires object creation asserted (2425)
3631 - move lfs from /usr/sbin to /usr/bin in packages
3632 - fix race between mds_client_add and mds_client_free (2417)
3633 - use kmalloc instead of slabs in portals (2430)
3634 - don't create duplicate records when a failover MDS is present (2442)
3635 - remove unnecessary mount age check (2332)
3636 - don't remove directory inodes from locks prematurely (2451)
3637 - don't break if MDS service name is the same as hostname (2103)
3638 - fix races in client write RPC generation when cache full (2482)
3640 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3643 - remove now-unused request->rq_obd (278)
3644 - if an allocation fails, print out how much memory we've used (1933)
3645 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3646 - add the "configurable stack size" patch to most series files (1256)
3647 - ability to write large log records, for 100+ OST configs (2306)
3648 - fix NULL deref when filter_prep fails (2314)
3649 - fix operator precedence error in filter_sync
3650 - dynamic allocation of socknal TX descriptors (2315)
3651 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3652 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3653 - ensure that CWARN messages in Portals always get to the syslog
3654 - __init/__exit are not for prototype decls (ldlm_init/exit)
3655 - x86-64 compile warning fixes
3656 - fix gateway LMC keyword conflict (2318)
3657 - fix MDS lock inversions in getattr/reint paths (1844)
3658 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3659 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3660 - DLM race condition prevented some lock evictions (2328)
3661 - ENOMEM detection and retry on socknal sends (2230)
3662 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3663 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3664 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3665 - better error messages when a client is rejected during recovery (1505)
3666 - avoid cancelling locks which were never granted, after failure (2330)
3667 - fix i_sem/journal inversion in mds_client_add (2333)
3668 - fix truncate/getattr lock cycle deadlock (2334)
3669 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3670 - fix two rare exit paths which could leak an l_lock() ref (2321)
3671 - fencepost error in MDS/OST orphan recovery (2226)
3672 - make log record alignment 8 bytes (1988)
3673 - lstripe now fails when requested offset > ost_count (2237)
3674 - ensure that all kernel series have a complete list.h (1607)
3675 - fix crashes in special-file operations (2316)
3676 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3678 - allow configurable automake binary, for testing new versions
3679 - small update to the lfs documentation
3681 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3683 * fix negative export reference count in fsfilt_sync (2312)
3685 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3686 * release candidate 0.9.1
3688 - orphans are moved into the PENDING directory for possible recovery
3689 - replayed opens now open by fid for orphan/rename safety (1042)
3690 - last close of an orphan inode generates a transno (683)
3691 - chdir() and mount() now pin the directory entry (1020)
3692 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3693 - discard very old requests without processing them (1502)
3694 - remove some common, well-understood CERRORs (1505)
3695 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3696 - clear "grant" flags in OST replies until OST grant code lands (1644)
3697 - fix read performance by not clobbering i_blksize on client (1598)
3698 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3699 - make LRU size a /proc tunable, clears locks when reduced (707)
3700 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3701 - server locks take references on exports now (1558)
3702 - build fixes for 2.4.20-rh trees (1663)
3703 - return an error from lov_create if all OSCs are inactive (1751)
3704 - fix import levels when a reconnect happens without a timeout (1597)
3705 - exit early from mds_open if we get a lookup error (1749)
3706 - partial page read at EOF wouldn't wait for disk before sending (1642)
3707 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3708 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3709 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3710 - write/truncate lock inversion (1639)
3711 - Don't auto-load obdclass, portals modules during cleanup (1495)
3712 - fix timestamps from jumping to "now" (1763)
3713 - extra journal assertions (1648)
3714 - add an extra multiunlink test (1771)
3715 - fix read_record/write_record API (1776)
3716 - fix leak of offset_extent, possible incorrect i_size later (1772)
3717 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3718 - Debugging for the kqswnal_get_idle_tx problems (1820)
3719 - Allow recovery to be attempted multiple times (1536)
3720 - Write out MDS last_rcvd file after it is first created (1600)
3721 - Fix tx_descriptor leak in failed transmit situations (1827)
3722 - ext3 journaling fixes for assertion failure after IO error (1871)
3723 - class_export_put() on freed export after completion AST error (1896)
3724 - Fix revalidate looping in VFS (1322)
3725 - Don't access a freed export during MDS_REINT timeout (1521)
3726 - Add open-unlink recovery support on the MDS (1673,1764)
3727 - Return an error if no MDS data was read from last_rcvd (1946)
3728 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3729 - Don't setup a disk device that doesn't match exported UUID (317)
3730 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3731 - avoid committing NULL handle in force close
3732 - local.sh is now a one-stripe LOV configuration
3733 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3734 - fix ext3 htree duplicate directory entry corruption (1516)
3735 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3736 - update to new LOV EA format (2097)
3737 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3739 2003-06-15 Phil Schwan <phil@clusterfs.com>
3742 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3743 - per-import/export recovery handling (958, 931, 959)
3744 - multiple last-rcvd slots, for serving multiple FSes (949)
3745 - connections are again shared between multiple imp/exports (963, 964)
3746 - "umount -f" would hang if any requests needed to be sent (393, 978)
3747 - avoid pinning large req buffer by copying for queued messages (989)
3748 - add "uuid" to "lctl device" command to help upcalls (991)
3749 - "open" RPCs with transnos would confuse recovery counters (1037)
3750 - do proper endian conversion of all wire messages (288, 340, 891)
3751 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3752 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3753 - fix ptlrpc_connection leak in target_handle_connect (1174)
3754 - fix import refcounting bug in OST and MDS cleanup (1134)
3755 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3756 - fix very unlikely obd_types race condition (501)
3757 - remove osc_open hack for echo_client (1187)
3758 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3759 - a failure in read_inode2 leads to deadlock (1139)
3760 - cancel ack-locks as soon as transaction is committed (1072)
3761 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3762 - make sure to commitrw after any preprw to avoid deadlock (1162)
3763 - failing to execute a file in a lustre FS would lock inode (1203)
3764 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3765 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3766 - obd_uuid2tgt would walk off the end of the list (1255)
3767 - on IA64 the osc would give portals incorrect bulk size (1258)
3768 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3769 - fix lock inversion caused by new llite matching code (1282)
3770 - limit the number of dirty pages on a client to 10MB (1286)
3771 - timed out locks were not being corrected cancelled (1289)
3772 - fix O_DIRECT above 4GB on IA-32 (1292)
3773 * major user-visible changes
3774 - fail out/fail over policy now controlled by the upcall (993)
3776 - add OBD_PING to check server availability and failure (954)
3777 - lustre messages are now sent in sending host order (288, 340, 891)
3778 - add eadatalen to MDS getattr reply (340)
3779 - OST read replies may contain second buffer, with per-page status (593)
3781 2003-03-11 Phil Schwan <phil@clusterfs.com>
3784 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3785 - failed open()s could cause deadlock; fixed (867, 869)
3786 - stop cancelling OST locks when files are closed (481)
3787 - overlapping XID spaces caused network corruption (851, 853)
3788 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3789 - setattr_raw added, to do single-RPC, server-side setattrs
3790 - lmc/lconf syntax change for OST UUIDs
3791 - fix crashy race condition between ptlrpc_free_req and osc_close
3792 - don't use request in mdc_enqueue if we hit a timeout (889)
3793 - don't set the inode i_size for regular files from the MDS (896)
3794 - handle out of order completion AST (842)
3795 - don't LBUG if a lock request times out after receiving AST (913)
3796 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3797 - if a bad lock AST arrives, send an error instead of dropping entirely
3798 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3799 - fix leak in bulk IO when only partially completed (899, 900, 926)
3800 - fix O_DIRECT for ia64 (55)
3801 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3802 - C-z after timeout could hang a process forever; fixed (977)
3804 - client-side I/O cache (678, 924, 929, 941, 970)
3806 - READPAGE and SETATTRs which don't take server-side locks get
3809 2003-02-11 Phil Schwan <phil@clusterfs.com>
3812 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3813 - Fix fsfilt_extN_readpage() to read a full page of directory
3814 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3815 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3816 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3817 - Most intent operations were removed, in favour of a new RPC mode
3818 that does a single RPC to the server and bypasses most of the VFS
3819 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3820 - Aggressively cancel local locks on DLM servers
3821 - mds_reint_unlink sends EA to the client if it's the last nlink.
3822 client uses that EA to unlink OST objects.
3823 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3824 - recursive symlinks were fixed (439)
3825 - fixed NULL deref in DEBUG_REQ
3826 - filter_update_lastobjid no longer calls sync, which annoyed extN
3827 - fixed multi-client small-writes to a single file problem (445)
3828 - fixed mtime updates during file writes (607)
3829 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3830 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3831 - fixed Linux OST size reporting problem (444, 656)
3832 - OST now updates object mtime with writes or setattr (607, 619)
3833 - client verifies file size before zeroing page past EOF (445)
3834 - OST now writes last allocated objid to disk with allocation (108)
3835 - LOV on echo now works (409)
3837 - mds_reint_unlink sends a new buffer, with the EA included. this
3838 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3839 set if a regular file was being unlinked, and it was the last link
3840 - use PtlGet from the target for bulk writes (315)
3841 - OST now updates object mtime with writes or setattr (607, 619)
3842 - LDLM now has a grant-time callback to revalidate locked items, if
3844 - Many MDS operations were reorganized to combat race conditions
3846 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3847 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3849 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3852 - Fully reactivate OST imports after reconnection (512, others)
3853 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3854 - More graceful error handling for truncating on dead OST (515)
3855 - Don't error out unless we're actually accessing dead stripes (474)
3856 - Fix garbage sizes when stripes are missing (410)
3857 - LRU counters were broken, causing constant lock purge (433, 432)
3858 - garbage on read from stripes with failed OSTs (441)
3859 - mark OSCs as active before reconnecting during recovery (438)
3860 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3861 - lfind did not preserve OST order in output (443)
3862 - symlinks cause hung clients, incorrect data (439)
3863 - stop dereferencing request after dropping refcount (457)
3864 - don't LASSERT(spin_is_locked) on non-SMP (455)
3865 - fixes for many rename() bugs
3866 - fstat didn't correctly synchronize attributes (399)
3867 - server must handle lock cancellation during blocking AST prep (487)
3868 - bulk descriptors were free()d too soon (511)
3869 - fix paths in lconf, which would load incorrect modules (451, 507)
3870 - fix confusing lconf 'host not found' error message (386)
3871 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3872 - fix race condition in mdc_blocking_ast() for inode access (526)
3873 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3874 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3875 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3876 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3877 - fix lstripe in several ways (kernel side) (527)
3878 - fix request leak in ldlm_cli_enqueue (262)
3879 - incorrect OSC was marked inactive after OST failure
3880 - call mds_fs_cleanup before unmounting filesystem (524)
3881 - fix races between taking ns_lock and ldlm_lock_change_resource
3882 - fix races updating LOV export open file list
3883 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3884 - fix recovery NULL deref in ldlm_cli_cancel_unused
3885 - fix some DLM races by using new hash table for lock handles (419)
3886 - permit the client to specify desired inodes, at replay
3887 - duplicate requests when we queue them for replay reintegration
3888 - fix last_rcvd offset calculation
3889 - sync after each recovered transaction, so we always make progress
3890 - never, not always, ERESTART requests without transnos
3891 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3893 - skip replay if the MDS says that the client is already connected
3894 - don't check for a recovery-enabled export to match lctl's UUID
3895 - don't INC_USE_COUNT for phantom exports
3896 - don't crash when cleaning up phantom exports (567)
3897 - don't double-finish or set replay data for errored mdc_open requests
3898 - abort requests when they time out, so we don't get old replies
3899 - send/receive replies for AST messages again
3900 - if the client says that it doesn't have the lock, cancel it on the
3902 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3903 mark it as destroyed, it will all work out when decref is called
3904 - fix module use counts (22, 581)
3906 - ASTs now expect a reply (server cancels lock on error reply)
3908 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3911 - fix many simultaneous client startup (392)
3912 - fix dentry->d_it clobbering
3913 - credentials weren't being shipped for readdir/getattr operations
3914 - remove invalid assertions triggered during some concurrent MD
3916 - proper Lustre versions added (336, 389)
3917 - fix memory leak for create error case (398)
3918 - fix LOV locking bug that would get cli/srv out of sync
3919 - fix echo client over LOV (409)
3920 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3921 - fix double-O_EXCL intent crash (424)
3922 - avoid sending multiple lock CANCELs (352)
3924 - MDS can do multi-client recovery (modulo bugs in new code)
3926 - many updates, edits, cleanups
3928 2002-11-18 Phil Schwan <phil@clusterfs.com>
3931 - fix null d_it dereference (346)
3932 - fix full OST/dbench hang (333)
3933 - fix permission problem with file removal (286)
3934 - fix removal of OSCs from LOV when they fail
3935 - fix NULL deref during bulk timeout (214)
3936 - fix problems related to multiple filesystems on one MDS (241)
3937 - fixed serious subtle metadata locking bugs
3938 - free locks on clients when inodes are removed due to memory
3940 - fix inode pointer in lock data (285)
3941 - partial support for multiple MDS on a single host (241)
3942 - data locks weren't cancelled at clear_inode time (290, 311)
3943 - intent locks could lead to unbounded lock growth (205)
3944 - added a maximum lock count, an LRU list, and a flusher
3945 - fix multiple rename (365)
3946 - properly abstracted the echo client
3947 - OSC locked 1 byte too many; fixed
3948 - rewrote brw callback code:
3949 - fixed recovery bugs related to LOVs (306)
3950 - fixed too-many-pages-in-one-write crash (191)
3951 - fixed (again) crash in sync_io_timeout (214)
3952 - probably fixed callback-related race (385)
3954 - Add capability to MDS protocol
3955 - LDLM cancellations and callbacks on different portals
3957 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3960 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3961 - defer open object destruction to close time (601981, 138)
3962 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3963 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3964 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3965 - don't call obd_setattr in ll_file_release for destroyed objects
3966 * protocol change to lustre_msg: move |version| and add |flags|
3967 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3968 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3969 * added replay of create, unlink, link and rename operations during
3970 MDS failover; recovery should be much more robust now
3971 * remove failed OSCs from LOVs (only lov_create uses this so far)
3972 * the lustre-HOWTO was brought (more) up to date (582544)
3974 2002-10-23 Phil Schwan <phil@clusterfs.com>
3977 - in-use dentries weren't being reused properly (617851)
3978 - prevent multiple LDLM setup (599178)
3979 - fix LOV size calculations for truncate (617853)
3980 - fix client handling of MDS intent errors (POSIX)
3981 - fix permission bug in lovstripe.c test (624321)
3982 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3983 - truncate past end of file could corrupt data
3984 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3985 - a race in recovery could return ETIMEDOUT to apps (623947)
3986 - building outside the source directory was fixed
3987 * the lustre-HOWTO was brought (more) up to date (582544)
3988 * major progress was made on recovery functionality
3990 2002-10-10 Phil Schwan <phil@clusterfs.com>
3993 - recovery deadlock fix
3994 - rm -rf causes LBUG fix (617817)
3995 - file open by multiple tasks fix (618962)
3996 - directory permissions bugs (602707 and 620007)
3997 - journal_stop fixed with locking (611313)
3998 - O_APPEND failures resolved (618273, perhaps 614459)
3999 - lconf PATH fix (619770)
4000 - IA64 build fix (621450)
4001 - RPC buffer sizes scale with amount of memory
4003 2002-10-01 Phil Schwan <phil@clusterfs.com>
4006 - locks would be cancelled without throwing away data pages,
4007 resulting in inconsistent data (605627)
4008 - inode attributes were not always being refreshed (605627, 612449)
4009 - lconf now continues to cleanup after lctl reports an error
4010 - MDS now enforces user permissions (602707)
4011 - lprocfs cleanup fixed, but not yet enabled (614157)
4012 - fixed infinite server hang, should a client not respond to an AST
4013 - avoid going into recovery if user calls readlink() with a buffer
4014 that's too small (613941)
4015 - AST RPCs no longer require replies (614867) -- this may be changed
4016 - don't crash server if client sends an IOV that's too big (611336)
4017 - fixed lock conversion deadlock (611892)
4018 - fixed the following of symlinks (614622)
4019 * recovery: the server can remove locks from a client that dies, other
4020 clients can make progress
4021 * more extN patch fixes
4022 * compile-time configurable ptlrpc buffer allocations
4024 - collaborative read cache document
4025 - Lustre Lite Performance CDR document-in-progress
4027 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4030 - fix typo in patch-2.4.18
4032 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4035 - clear ptlrpc request each time in handle_incoming_request()
4036 - unlink of files now destroys the object on the OST
4038 2002-09-19 Peter Braam <braam@clusterfs.com>
4040 * add hard link support
4041 * change obdfile creation method
4042 * kernel patch changed
4044 2002-09-19 Peter Braam <braam@clusterfs.com>
4047 - stack overflow bug in extN fixed
4049 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4051 * documentation updates
4052 - add man pages for config tools
4053 - update tests/README to describe testing with new config tools
4054 - finish metadata API descriptions
4055 * bug fixes and cleanups
4056 - statfs workaround for 16TB limit
4057 - LOV stripe allocation improved, can stripe on subset of OSTs
4058 - LOV file size/IO offset was wrong for files > 4GB in size
4059 - object EA data was being dropped, caused files to be unreadable
4060 - memory overflow with non-LOV OST caused memory corruption
4061 - fixed regression tests to work with new config tools, obdfilter
4062 - fixed bug when directory size became larger than 1 block
4063 - fixed bug (for single client case) when PWD was deleted
4064 - invalidate local directory pages when doing intent-based ops
4065 - avoid LDLM oops when lock callback contained bad data
4067 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4069 * documentation updates
4070 * bug fixes and cleanups
4071 - configuration tools
4074 - 64-bit compile warnings
4075 - 64-bit internal statfs data
4077 * test_brw on persistent OST devices
4079 * lprocfs (disabled)
4081 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4083 * documentation updates
4084 * bug fixes and cleanups
4085 * configuration tools
4087 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4090 * many small fixes to 0_5_4
4091 * io/network handling
4092 * thinkos in MDS operations
4094 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4097 * crucial basic fixes to 0.5.3
4098 * IOR, Iozone work over Elan
4101 2002-08-07 Phil Schwan <phil@clusterfs.com>
4102 * version 0_5_3, our first alpha
4103 * we use the new Portals iovs
4104 * documentation updates
4105 * bug fixes and cleanups
4106 * small changes in the DLM wire protocol
4108 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4109 * version 0_5_1 with some initial stability,
4110 * locking on MD and file I/O.
4111 * documentation updates
4112 * several bug fixes since 0.5.0
4113 * small changes in wire protocol
4115 2002-07-18 Phil Schwan <phil@clusterfs.com>
4117 * delivered as Lustre Light Alpha
4118 * fixed a crash after handling invalid MDS requests
4119 * fixed directory pages for architectures with non-4k pages sizes
4121 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4122 * release version v0_4_4
4123 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4125 * Fixes a number of interruption problems with OST operations.
4126 * Update documentation for portals header changes
4127 * Move all wire protocol structs/defines to lustre_idl.h
4128 * Fixes symlink length bug.
4129 * Add tcpdump to repository.
4131 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4132 * release version v0_4_3
4133 * Fixes statfs for inodes on extN.
4134 * Fixes bug in runtests which would delete /etc/hosts.
4135 * Use 64-bit object IDs wherever possible (not into VFS though)
4136 Remove ost_get_info, which is unused by lustre, and out of date.
4138 2002-07-03 Peter Braam <braam@clusterfs.com>
4139 * release version v0_4_2 Fixes a lookup error (type not passed)
4140 * move forward to head of Portals
4141 * move forward to latest Lustre kernel
4143 2002-06-25 Peter Braam <braam@clusterfs.com>
4144 * release version v0_4_1. Hopefully stable on single node use.