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.
16 Description: Remove client patches from SLES 10 kernel.
17 Details : This causes SLES 10 clients to behave as patchless clients
18 even on a Lustre-patched (server) kernel.
20 Severity : enhancement
22 Description: self-adjustable client's lru lists
23 Details : use adaptive algorithm for managing client cached locks lru
24 lists according to current server load, other client's work
25 pattern, memory activities, etc. Both, server and client
26 side namespaces provide number of proc tunables for controlling
29 Severity : enhancement
31 Description: light-weight GSS support
32 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
33 and reduce performance overhead.
35 Severity : enhancement
37 Description: Linux keyring support
38 Details : Support using service of Linux keyring for Lustre GSS internal
39 context refresh/cache mechanism.
43 Description: Fix errors in lfs documentation
44 Details : Fixes man pages
48 Description: don't use GFP_* in generic Lustre code.
49 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
53 Description: obdclass is limited by single OBD_ALLOC(idarray)
54 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
58 Description: warnings with build patchless client with vanila 2.6.19 and up
59 Details : change old ctl_table style and replace ctl_table/ctl_table_header
60 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
64 Frequency : Only for SLES
65 Description: sanity_quota fail test_1
66 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
67 file, which makes TSTID[2] inunique.
71 Frequency : Only for SLES9
72 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
73 Details : Update target/ChangeLog/which_patch .
77 Frequency : Only for test_10 in sanity-quota.sh
78 Description: a bug in quota test code
79 Details : checking a test flag in a wrong place
83 Frequency : Only for test_13 in sanity-quota.sh
84 Description: a bug in quota test script
85 Details : shouldn't check overfull allocated quota
89 Description: jbd statistics
90 Details : Port older jbd statistics patch for sles10
92 Severity : enhancement
94 Description: Add printing inode info into message about error in writepage.
99 Description: reply_lock_interpret crash due to race with it and lock cancel.
100 Details : Do not replay locks that are being cancelled. Do not reference
101 locks by their address during replay, just by their handle.
105 Frequency : When flocks are used.
106 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
107 Details : Flock locks might destroy just granted lock if it could be merged
108 with another existing flock, this is done in completion handler,
109 so teach ldlm_cli_enquque_fini that this is a valid case for
115 Description: Oops in read and write path when failing to allocate lock.
116 Details : Check if lock allocation failed and return error back.
120 Description: lstripe command fails for valid OST index
121 Details : The stripe offset is compared to lov->desc.ld_tgt_count
122 instead of lov->desc.ld_active_tgt_count.
126 Description: sanity.sh failed test 103
127 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
132 Description: df doesn't work properly if diskfs blocksize != 4K
133 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
137 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
138 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
142 Description: obd_health_check_timeout too short
143 Details : set obd_health_check_timeout as 1.5x of obd_timeout
147 Description: llapi_file_create() does not allow some changes
148 Details : add llapi_file_open() that allows specifying the mode and
149 open flags, and also returns an open file handle.
153 Description: merge and cleanup kernel patches.
154 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
158 Description: Add journal checksum support.(Kernel part)
159 Details : The journal checksum feature adds two new flags i.e
160 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
161 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
162 indicates that the commit block contains the checksum for
163 the blocks described by the descriptor blocks. Now commit
164 record can be sent to disk without waiting for descriptor
165 blocks to be written to disk. This behavior is controlled
166 using JBD2_FEATURE_ASYNC_COMMIT flag.
170 Description: OSS needs mutliple precreate threads
171 Details : Add ability to start more than one create thread per OSS.
175 Description: Sanity.sh test_65k failed: setstripe should have succeeded
176 Details : execute lfs setstripe on client
180 Description: mds_obd_create error creating tmp object
181 Details : When the user sets quota on root, llog will be affected and can't
182 create files and write files.
185 Frequency : Always on ia64 patchless client, and possibly others.
187 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
188 Details : This allows the patchless client to be loaded on architectures
193 Description: RedHat Update kernel for RHEL5
194 Details : Add the kernel config file for RHEL5.
198 Description: RedHat Update kernel for RHEL5
199 Details : Modify the kernel config file more closer RHEL5.
203 Description: Build failure against Centos5 (RHEL5)
204 Details : Define PAGE_SIZE when it isn't present.
208 Description: test_7 on sanity-quota.sh can't be used on separate machines
209 Details : add facet to handle it
213 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
214 Details : with NFS, the anon dentry's parent was set to itself in
215 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
216 corresponding dentry other than use rec->ur_name.
218 Severity : enhancement
220 Description: lfs setstripe enhancement
221 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
225 Description: testing performance impact of enabling checksumming
226 Details : enable checksum by default, allow --disable-checksum
227 configure option and "-o nochecksum" mount option
231 Description: lustre support for RHEL5
232 Details : Add support for RHEL5.
236 Description: Client eviction due to failover config
237 Details : after a connection loss, the lustre client should attempt to
238 reconnect to the last active server first before trying the
239 other potential connections.
243 Description: when mds and osts use different quota unit(32bit and 64bit),
244 quota will be released repeatly.
245 Details : void sending multiple quota reqs to mds, which will keep the
246 status between the reqs.
250 Description: osts not allocated evenly to files
251 Details : change the condition to increase offset_idx
254 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
256 Description: Data corruption for OSTs that are formatted larger than 2TB
258 Details : When generating the bio request for lustre file writes the
259 sector number would overflow a temporary variable before being
260 used for the IO. The data reads correctly from Lustre (which
261 will overflow in a similar manner) but other file data or
262 filesystem metadata may be corrupted in some cases.
266 Description: Tune the kernel for good SCSI performance.
267 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
268 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
273 Description: rewrite ext2-derived code in obdclass/uuid.c
274 Details : rewrite inherited code (uuid parsing code from ext2 utils)
275 from scratch preserving functionality.
279 Description: "lfs find -obd UUID" prints directories
280 Details : "lfs find -obd UUID" will return all directory names instead
281 of just file names. It is incorrect because the directories
282 do not reside on the OSTs.
286 Description: use i_size_read and i_size_write in 2.6 port
287 Details : replace inode->i_size access with i_size_read/write()
290 Frequency : when removing large files
292 Description: scheduling issue during removal of large Lustre files
293 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
294 It causes scheduling issues when removing large files (17TB in the
298 Frequency : only with liblustre clients on XT3
300 Description: evictions taking too long
301 Details : allow llrd to evict clients directly on OSTs
304 Frequency : only on ppc
306 Description: /proc/fs/lustre/devices broken on ppc
307 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
308 We should make sure the type of 'index' is loff_t and then cast
309 explicitly as needed below. Do not assign an explicitly cast
313 Frequency : only for rhel5
315 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
316 Details : Modify the target file & which_kernel.
318 Severity : enhancement
320 Description: omit set fsid for export NFS
321 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
322 as FS_REQUIRES_DEV for avoid problems with generate fsid.
326 Frequency : Always, for kernels after 2.6.16
327 Description: Fix warning idr_remove called for id=.. which is not
329 Details : Last kernels save old s_dev before kill super and not allow
330 to restore from callback - restore it before call kill_anon_super.
334 Description: buffer overruns could theoretically occur
335 Details : llapi_semantic_traverse() modifies the "path" argument by
336 appending values to the end of the origin string, and a
337 overrun may occur. Adding buffer overrun check in liblustreapi.
341 Description: client sometimes tried revalidate root dentry.
342 Details : not revalidate dentry if it root dentry.
346 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
347 Details : Modify vm-tunables-rhel4.patch.
351 Description: Kernel config for 2.6.18-vanilla.
352 Details : Modify targets/2.6-vanilla.target.in.
353 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
354 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
355 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
356 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
360 Description: improve handling recoverable errors
361 Details : if request processig with error which can be recoverable on server
362 request should be resend, otherwise page released from cache and
367 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
368 Details : Modify target file & which_patch.
369 A flaw was found in the IA32 system call emulation provided
370 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
371 value could be stored in the %RAX register, which could trigger an
372 out-of-bounds system call table access. An untrusted local user
373 could exploit this flaw to run code in the kernel
374 (ie a root privilege escalation). (CVE-2007-4573).
378 Description: Remove 2.4 kernel checks lustre 1.8
379 Details : Remove 2.5.0 check from quota_check.c, quota_ctl.c,
380 quota_interface.c in quota subsystem.
381 Remove 2.4 kernel check from obd_mount.c, linux/linux-sysctl.c in
383 Remove lock_24kernel from fsfilt_ext3.c fsfilt_reiserfs.c
384 lvfs_linux.c upcall_cache.c in lvfs subsystem.
388 Description: O_DIRECT bypasses client statistics.
389 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
390 read_ahead_stats were updated. copied the stats section from
391 osc_send_oap_rpc() into async_internal().
395 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
399 Description: change order of libsysio includes
400 Details : '#include sysio.h' should always come before '#include xtio.h'
402 --------------------------------------------------------------------------------
404 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
406 * Support for kernels:
407 2.6.5-7.283 (SLES 9),
408 2.6.9-55.EL (RHEL 4),
409 2.6.16.46-0.14 (SLES 10),
410 2.6.18.8 vanilla (kernel.org)
411 * Client support for unpatched kernels:
412 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
413 2.6.16 - 2.6.22 vanilla (kernel.org)
414 * Due to recently discovered recovery problems, we do not recommend
415 using patchless RHEL 4 clients with this or any earlier release.
416 * Recommended e2fsprogs version: 1.39.cfs8
417 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
418 * Starting with this release, the ldiskfs backing filesystem required
419 by Lustre is now in its own package, lustre-ldiskfs. This package
420 should be installed. It is versioned separately from Lustre and
421 may be released separately in future.
425 Description: block reactivating mgc import until all deactivates complete
426 Details : Fix race when failing back MDT/MGS to itself (testing)
428 Severity : enhancement
430 Description: add optional extra BUILD_VERSION info
431 Details : add a new environment variable (namely LUSTRE_VERS) which allows
432 to override the lustre version.
435 Frequency : 2.6.18 servers only
437 Description: ll_kern_mount() doesn't release the module reference
438 Details : The ldiskfs module reference count never drops down to 0
439 because ll_kern_mount() doesn't release the module reference.
444 Description: server LBUG when using old ost_num_threads parameter
445 Details : Accept the old ost_num_threads parameter but warn that it
446 is deprecated, and fix an off-by-one error that caused an LBUG.
451 Description: Transient SCSI error results in persistent IO issue
452 Details : iobuf->dr_error is not reinitialized to 0 between two
456 Frequency : sometimes when underlying device returns I/O errors
458 Description: OSTs not going read-only during write failures
459 Details : OSTs are not remounted read-only when the journal commit threads
460 get I/O errors because fsfilt_ext3 calls journal_start/stop()
461 instead of the ext3 wrappers.
465 Description: poor connect scaling with increasing client count
466 Details : Don't run filter_grant_sanity_check for more than 100 exports
467 to improve scaling for large numbers of clients.
470 Frequency : SLES10 only
472 Description: sanity-quota.sh quotacheck failed: rc = -22
473 Details : Quotas cannot be enabled on SLES10.
476 Frequency : liblustre clients only
478 Description: getdirentries does not give error when run on compute nodes
479 Details : getdirentries does not fail when the size specified as an argument
480 is too small to contain at least one entry
482 Severity : enhancement
484 Description: Add LNET router traceability for debug purposes
485 Details : If a checksum failure occurs with a router as part of the
486 IO path, the NID of the last router that forwarded the bulk data
487 is printed so it can be identified.
492 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
493 Details : Due to a race condition, liblustre clients were occasionally
496 Severity : enhancement
498 Description: lfs setstripe use optional parameters instead of postional
501 Severity : enhancement
503 Description: Nanosecond timestamp support for ldiskfs
504 Details : The on-disk ldiskfs filesystem has added support for nanosecond
505 resolution timestamps. There is not yet support for this at
506 the Lustre filesystem level.
509 Frequency : during server recovery
511 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
512 Details : request with rq_no_resend flag not awake l_wait_event if they get a
516 Frequency : nfs export on patchless client
518 Description: connectathon hang when test nfs export over patchless client
519 Details : Disconnected dentry cannot be found with lookup, so we do not need
520 to unhash it or make it invalid
523 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
524 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
525 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
530 Description: open req refcounting wrong on reconnect
531 Details : If reconnect happened between getting open reply from server and
532 call to mdc_set_replay_data in ll_file_open, we will schedule
533 replay for unreferenced request that we are about to free.
534 Subsequent close will crash in variety of ways.
535 Check that request is still eligible for replay in
536 mdc_set_replay_data().
541 Description: disable writes to filesystem when reading health_check file
542 Details : the default for reading the health_check proc file has changed
543 to NOT do a journal transaction and write to disk, because this
544 can cause reads of the /proc file to hang and block HA state
545 checking on a healthy but otherwise heavily loaded system. It
546 is possible to return to the previous behaviour during configure
547 with --enable-health-write.
549 Severity : enhancement
551 Description: 64-bit inode version
552 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
553 to the inode. This will be required for version-based recovery.
558 Description: MDS fails to start if a duplicate client export is detected
559 Details : in some rare cases it was possible for a client to connect to
560 an MDS multiple times. Upon recovery the MDS would detect this
561 and fail during startup. Handle this more gracefully.
563 Severity : enhancement
565 Description: Add -o localflock option to simulate old noflock
567 Details : This will achieve local-only flock/fcntl locks
573 Description: log_commit_thread vs filter_destroy race leads to crash
574 Details : Take import reference before releasing llog record semaphore
579 Description: Wrong request locking in request set processing
580 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
581 imp_delayed_list, in this place should be used imp_lock.
584 Frequency : when reconnection
586 Description: Grant Leak when osc reconnect to OST
587 Details : When osc reconnect ost, OST(filter) should check whether it
588 should grant more space to client by comparing fed_grant and
589 cl_avail_grant, and return the granted space to client instead
590 of "new granted" space, because client will call osc_init_grant
591 to update the client grant space info.
594 Frequency : when client reconnect to OST
596 Description: Grant Leak when osc do resend and replay bulk write
597 Details : When osc reconnect to OST, OST(filter)should clear grant info of
598 bulk write request, because the grant info while be sync between
599 OSC and OST when reconnect, and we should ignore the grant info
600 these of resend/replay write req.
605 Description: Grant space more than avaiable left space sometimes.
606 Details : When then OST is about to be full, if two bulk writing from
607 different clients came to OST. Accord the avaliable space of the
608 OST, the first req should be permitted, and the second one
609 should be denied by ENOSPC. But if the seconde arrived before
610 the first one is commited. The OST might wrongly permit second
611 writing, which will cause grant space > avaiable space.
614 Frequency : when client is evicted
616 Description: Grant might be wrongly erased when osc is evicted by OST
617 Details : when the import is evicted by server, it will fork another
618 thread ptlrpc_invalidate_import_thread to invalidate the
619 import, where the grant will be set to 0. While the original
620 thread will update the grant it got when connecting. So if
621 the former happened latter, the grant will be wrongly errased
622 because of this race.
627 Description: Checking Stale with correct fid
628 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
629 to check whether it is stale, because sometimes, we want the
630 enqueue happened anyway, and op_data.fid2 will not be initialized.
632 Severity : enhancement
634 Description: update patchless client
635 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
638 Frequency : only with 2.4 kernel
640 Description: random memory corruption
641 Details : size of struct ll_inode_info is to big for union inode.u and this
642 can be cause of random memory corruption.
647 Description: Memory leak in recovery
648 Details : Lov_mds_md was not free in an error handler in mds_create_object.
649 It should also check obd_fail before fsfilt_start, otherwise if
650 fsfilt_start return -EROFS,(failover mds during mds recovery).
651 then the req will return with repmsg->transno = 0 and rc = EROFS.
652 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
653 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
654 be freed no matter whether fsfilt_commit success or not.
657 Frequency : only with huge count clients
659 Description: Prevents from taking the superblock lock in llap_from_page for
661 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
662 for llap_from_page call in ll_removepage prevents from taking the
663 superblock lock for a soon died page.
668 Description: Not check open intent error before release open handle
669 Details : in some rare cases, the open intent error is not checked before
670 release open handle, which may cause
671 ASSERTION(open_req->rq_transno != 0), because it tries to release
672 the failed open handle.
677 Description: Set cat log bitmap only after create log success.
678 Details : in some rare cases, the cat log bitmap is set too early. and it
679 should be set only after create log success.
683 Description: Accessing a block bevice can re-enable I/O when Lustre is
684 tearing down a device.
685 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
690 Description: service threads may hog cpus when there are a lot of requests
692 Details : Insert cond_resched to give other threads a chance to use some of
698 Description: the cat log was not initialized in recovery
699 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
700 unlink log on mds will not be initialized until mds post
701 recovery. And also in mds post recovery, the unlink log will
702 initialization will be done asynchronausly, so there will be race
703 between add unlink log and unlink log initialization.
707 Description: brw_stats were being printed incorrectly
708 Details : brw_stats were being printed as log2 but all of them were not
709 recorded as log2. Also remove some code duplication arising from
710 filter_tally_{read,write}.
714 Frequency : rare, only in recovery.
715 Description: ASSERTION(req->rq_type != LI_POISON) failed
716 Details : imp_lock should be held while iterating over imp_sending_list for
717 prevent destroy request after get timeout in ptlrpc_queue_wait.
721 Description: replay-single.sh test 52 fails
722 Details : A lock's skiplist need to be cleanup when it being unlinked
723 from its resource list.
727 Description: Short directio read returns full requested size rather than
729 Details : Direct I/O operations should return actual amount of bytes
730 transferred rather than requested size.
732 Severity : enhancement
734 Description: metadata RPC reduction (e.g. for rm performance)
735 Details : decrease the amount of synchronous RPC between clients and servers
736 by canceling conflicing lock before the operation on the client side
737 and packing thier handles into the main operation RPC to server.
739 Severity : enhancement
741 Description: Async OSC create to avoid the blocking unnecessarily.
742 Details : If a OST has no remain object, system will block on the creating
743 when need to create a new object on this OST. Now, ways use
744 pre-created objects when available, instead of blocking on an
745 empty osc while others are not empty. If we must block, we block
746 for the shortest possible period of time.
750 Description: improve handling recoverable errors
751 Details : if request processig with error which can be recoverable on server
752 request should be resend, otherwise page released from cache and
755 Severity : enhancement
757 Description: refine locking for avoid write wrong info into lov_objid file
758 Details : fix possible races with add new target and write/update data in
761 --------------------------------------------------------------------------------
763 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
768 Frequency : on some architectures
770 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
771 Details : Uninitialized flags sometimes cause configuration commands to
775 Frequency : patchless clients only
777 Description: missing __iget() symbol export
778 Details : The __iget() symbol export is missing. To avoid the need for
779 this on patchless clients the deathrow inode reaper is turned
780 off, and we depend on the VM to clean up old inodes. This
781 dependency was during via the fix for bug 12181.
783 --------------------------------------------------------------------------------
785 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
787 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
788 INTEROPERATE with older versions automatically. In many cases a
789 special upgrade step is needed. Please read the
790 user documentation before upgrading any part of a live system.
791 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
792 version will not interoperate with 1.6 betas before beta5 (1.5.95).
793 * WARNING: Lustre configuration and startup changes are required with
794 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
800 Severity : enhancement
802 Description: MountConf
803 Details : Lustre configuration is now managed via mkfs and mount
804 commands instead of lmc and lconf. New obd types (MGS, MGC)
805 are added for dynamic configuration management. See
806 https://mail.clusterfs.com/wikis/lustre/MountConf for
809 Severity : enhancement
811 Description: dynamic OST addition
812 Details : OSTs can now be added to a live filesystem
814 Severity : enhancement
816 Description: startup order invariance
817 Details : MDTs and OSTs can be started in any order. Clients only
818 require the MDT to complete startup.
820 Severity : enhancement
822 Description: parallel, asynchronous orphan cleanup
823 Details : orphan cleanup is now performed in separate threads for each
824 OST, allowing parallel non-blocking operation.
826 Severity : enhancement
828 Description: optimized stripe assignment
829 Details : stripe assignments are now made based on ost space available,
830 ost previous usage, and OSS previous usage, in order to try
831 to optimize storage space and networking resources.
833 Severity : enhancement
835 Description: Permanently set tunables
836 Details : All writable /proc/fs/lustre tunables can now be permanently
837 set on a per-server basis, at mkfs time or on a live
840 Severity : enhancement
842 Description: Lustre message v2
843 Details : Add lustre message format v2.
845 Severity : enhancement
847 Description: client OST exclusion list
848 Details : Clients can be started with a list of OSTs that should be
849 declared "inactive" for known non-responsive OSTs.
853 Description: SPEC SFS validation failure on NFS v2 over lustre.
854 Details : Changes the blocksize for regular files to be 2x RPC size,
855 and not depend on stripe size.
857 Severity : enhancement
859 Description: Multiple MD RPCs in flight.
860 Details : Further unserialise some read-only MDS RPCs - learn about intents.
861 To avoid overly-overloading MDS, introduce a limit on number of
862 MDS RPCs in flight for a single client and add /proc controls
863 to adjust this limit.
865 Severity : enhancement
867 Description: client read/write statistics
868 Details : Add client read/write call usage stats for performance
869 analysis of user processes.
870 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
871 file access. extents_stats shows chunk size distribution.
872 extents_stats_per_process show chunk size distribution per
875 Severity : enhancement
877 Description: mds statistics
878 Details : Add detailed mds operations statistics in
879 /proc/fs/lustre/mds/*/stats.
883 Description: Failure of copying files with lustre special EAs.
884 Details : Client side always return success for setxattr call for lustre
885 special xattr (currently only "trusted.lov").
890 Description: Refcount LNET uuids
891 Details : The global LNET uuid list grew linearly with every startup;
892 refcount repeated list entries instead of always adding to
895 Severity : enhancement
897 Description: Dynamic service threads
898 Details : Within a small range, start extra service threads
899 automatically when the request queue builds up.
902 Frequency : mixed-endian client/server environments
904 Description: mixed-endian crashes
905 Details : The new msg_v2 system had some failures in mixed-endian
908 Severity : enhancement
910 Description: Easy OST removal
911 Details : OSTs can be permanently deactivated with e.g. 'lctl
912 conf_param lustre-OST0001.osc.active=0'
914 Severity : enhancement
916 Description: MGS proc entries
917 Details : Added basic proc entries for the MGS showing what filesystems
920 Severity : enhancement
922 Description: provide MGS failover
923 Details : Added config lock reacquisition after MGS server failover.
925 Severity : enhancement
927 Description: add Linux 2.4 support
928 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
932 Description: a large application tries to do I/O to the same resource and dies
934 Details : Check the req->rq_arrival time after the call to
935 ost_brw_lock_get(), but before we do anything about
936 processing it & sending the BULK transfer request. This
937 should help move old stale pending locks off the queue as
938 quickly as obd_timeout.
941 Frequency : when an incorrect nid is specified during startup
943 Description: ptlrpc connect to non-existant node causes kernel crash
944 Details : LNET can't be re-entered from an event callback, which
945 happened when we expire a message after the export has been
946 cleaned up. Instead, hand the zombie cleanup off to another
949 Severity : enhancement
951 Description: plain/inodebits lock performance improvement
952 Details : Grouping plain/inodebits in granted list by their request modes
953 and bits policy, thus improving the performance of search through
957 Frequency : only if OST filesystem is corrupted
959 Description: client incorrectly hits assertion in ptlrpc_replay_req()
960 Details : for a short time RPCs with bulk IO are in the replay list,
961 but replay of bulk IOs is unimplemented. If the OST filesystem
962 is corrupted due to disk cache incoherency and then replay is
963 started it is possible to trip an assertion. Avoid putting
964 committed RPCs into the replay list at all to avoid this issue.
967 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
969 Description: System hang on startup
970 Details : This bug allowed the liblustre (e.g. catamount) client to
971 return to the app before handling all startup RPCs. This
972 could leave the node unresponsive to lustre network traffic
973 and manifested as a server ptllnd timeout.
975 Severity : enhancement
977 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
978 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
979 boolean to control whether to print peer debug info when a
980 client's RPC times out.
983 Frequency : only for kernels with patches from Lustre below 1.4.3
985 Description: Remove old rdonly API
986 Details : Remove old rdonly API which unused from at least lustre 1.4.3
989 Frequency : only for devices with external journals
991 Description: Set external device read-only also
992 Details : During a commanded failover stop, we set the disk device
993 read-only while the server shuts down. We now also set any
994 external journal device read-only at the same time.
997 Frequency : when upgrading from 1.4 while trying to change parameters
999 Description: The wrong (new) MDC name was used when setting parameters for
1000 upgraded MDT's. Also allows changing of OSC (and MDC)
1001 parameters if --writeconf is specified at tunefs upgrade time.
1004 Frequency : when setting specific ost indicies
1006 Description: QOS code breaks on skipped indicies
1007 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1008 created with --index need not be sequential.
1010 Severity : enhancement
1012 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1013 Details : The uninit_groups feature works in conjunction with the kernel
1014 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1015 up the pass1 processing of e2fsck. This is a read-only feature
1016 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1017 mount filesystems that have had this feature enabled.
1019 Severity : enhancement
1021 Description: Improve multi-block allocation algorithm to avoid fragmentation
1022 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1023 allocation locality and avoid filesystem fragmentation.
1025 ------------------------------------------------------------------------------
1027 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1029 * Support for kernels:
1030 2.4.21-47.0.1.EL (RHEL 3)
1031 2.6.5-7.283 (SLES 9)
1032 2.6.9-42.0.10.EL (RHEL 4)
1033 2.6.12.6 vanilla (kernel.org)
1034 2.6.16.27-0.9 (SLES 10)
1035 * Recommended e2fsprogs version: 1.39.cfs5
1037 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1041 Frequency : occasional, depends on client load and configuration
1042 Bugzilla : 12181, 12203
1043 Description: data loss for recently-modified files
1045 Details : In some cases it is possible that recently written or created
1046 files may not be written to disk in a timely manner (this should
1047 normally be within 30s unless client IO load is very high).
1048 The problem appears as zero-length files or files that are a
1049 multiple of 1MB in size after a client crash or client eviction
1050 that are missing data at the end of the file.
1052 This problem is more likely to be hit on clients where files are
1053 repeatedly created and unlinked in the same directory, clients
1054 have a large amount of RAM, have many CPUs, the filesystem has
1055 many OSTs, the clients are rebooted frequently, and/or the files
1056 are not accessed by other nodes after being written.
1058 The presence of the problem can be detected by looking at
1059 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1060 smaller than the second (nr_unused) then dirty files will not
1061 be flushed automatically to disk. "sync; sleep 10" should be
1062 run several times on the node before unmounting it to update
1063 Lustre (this is also safe to run on nodes without this problem).
1065 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1066 that can cause this same problem, so customers using that kernel
1067 also need to update the kernel in addition to Lustre. In order
1068 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1070 It is normal that files written just before a client crash (less
1071 than 30s) may not yet have been flushed to disk, even for local
1075 Frequency : frequent on thin XT3 nodes
1077 Description: UUID collision on thin XT3 Linux nodes
1078 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1079 randomly, since we relied on an insufficiently-seeded PRNG.
1084 Description: OSS hangs after "All ost request buffers busy"
1085 Details : A deadlock between quota and journal operations caused OSS
1086 hangs after printing "All ost request buffers busy."
1089 Frequency : always on liblustre builds
1091 Description: Cleanup compiler warnings on liblustre
1094 Frequency : always on liblustre builds on XT3
1096 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1100 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1101 > 32000 subdirectories
1102 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1103 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1104 subdirectory count crosses 32000. This will aid e2fsck to
1105 correctly handle more than 32000 subdirectories.
1108 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1110 Description: System hang on startup
1111 Details : This bug allowed the liblustre (e.g. catamount) client to
1112 return to the app before handling all startup RPCs. This
1113 could leave the node unresponsive to lustre network traffic
1114 and manifested as a server ptllnd timeout.
1116 Severity : enhancement
1118 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1119 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1120 boolean to control whether to print peer debug info when a
1121 client's RPC times out.
1126 Description: make O_SYNC working on 2.6 kernels
1127 Details : 2.6 kernels use different method for mark pages for write,
1128 so need add a code to lustre for O_SYNC work.
1133 Description: Failure to close file and release space on NFS
1134 Details : Put inode details into lock acquired in ll_intent_file_open.
1135 Use mdc_intent_lock in ll_intent_open to properly
1136 detect all kind of errors unhandled by mdc_enqueue.
1141 Description: proc file read during shutdown sometimes raced obd removal,
1143 Details : Add lock to prevent obd access after proc file removal.
1146 Frequency : Only for files larger than 4GB on 32-bit clients.
1148 Description: improperly doing page alignment of locks
1149 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1150 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1155 Description: RPCs being resent when they shouldn't be
1156 Details : Some RPCs that should not be resent are being resent. This
1157 can cause inconsistencies in the RPC state machine. Do not
1158 resend such requests.
1161 Frequency : rare, only with NFS export
1163 Description: Crash on NFS re-export node
1164 Details : under very unusual load conditions an assertion is hit in
1165 ll_intent_file_open()
1168 Frequency : only if OST filesystem is corrupted
1170 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1171 Details : for a short time RPCs with bulk IO are in the replay list,
1172 but replay of bulk IOs is unimplemented. If the OST filesystem
1173 is corrupted due to disk cache incoherency and then replay is
1174 started it is possible to trip an assertion. Avoid putting
1175 committed RPCs into the replay list at all to avoid this issue.
1180 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1181 Details : Large single O_DIRECT read and write calls can fail to allocate
1182 a sufficiently large buffer to process the request. In case of
1183 allocation failure the allocation is retried with a smaller
1184 buffer and broken into smaller requests.
1186 Severity : enhancement
1188 Description: Add -o localflock option to simulate old noflock behaviour.
1189 Details : This will achieve local-only flock/fcntl locks coherentness.
1194 Description: versioning check is incomplete
1195 Details : Checking the version difference of client vs. server, report
1196 error if the gap is too big.
1201 Description: add support PG_writeback bit
1202 Details : add support for PG_writeback bit for Lustre, for more carefull
1203 work with page cache in 2.6 kernel. This also fix some deadlocks
1204 and remove hack for work O_SYNC with 2.6 kernel.
1206 Severity : enhancement
1208 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1209 Details : The uninit_groups feature works in conjunction with the kernel
1210 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1211 up the pass1 processing of e2fsck. This is a read-only feature
1212 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1213 mount filesystems that have had this feature enabled.
1215 Severity : enhancement
1217 Description: Improve multi-block allocation algorithm to avoid fragmentation
1218 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1219 allocation locality and avoid filesystem fragmentation.
1221 ------------------------------------------------------------------------------
1223 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1225 * Support for kernels:
1226 2.6.9-42.0.3.EL (RHEL 4)
1227 2.6.5-7.276 (SLES 9)
1228 2.4.21-47.0.1.EL (RHEL 3)
1229 2.6.12.6 vanilla (kernel.org)
1230 2.6.16.21-0.8 (SLES10)
1231 * Recommended e2fsprogs version: 1.39.cfs2-0
1233 * The backwards-compatible /proc/sys/portals symlink has been removed
1234 in this release. Before upgrading, please ensure that you change
1235 any configuration scripts or /etc/sysctl.conf files that access
1236 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1237 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1238 in advance of the upgrade on any system running Lustre 1.4.6 or
1239 newer, since /proc/sys/lnet was added in that version.
1240 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1244 Frequency : only when quota is used
1246 Description: avoid scanning export list for quota master
1247 Details : Change the algorithms to avoid scanning export list in order
1248 to improve the efficiency.
1251 Frequency : MDS failover only, very rarely
1253 Description: "went back in time" messages on mds failover
1254 Details : The greatest transno may be lost when the current operation
1255 finishes with an error (transno==0) and the client's last_rcvd
1256 record is over-written. Save the greatest transno in the
1257 mds_last_transno for this case.
1260 Frequency : always for specific kernels and striping counts
1262 Description: client may get "Matching packet too big" without ACL support
1263 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1264 when trying to access files in certain configurations. The
1265 clients should in fact be denied when mounting because they do
1266 not understand ACLs.
1269 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1271 Description: many clients connecting with IO in progress causes connect timeouts
1272 Details : Avoid synchronous journal commits to avoid delays caused by many
1273 clients connecting/disconnecting when bulk IO is in progress.
1274 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1275 OST_IO_PORTAL to avoid delays behind potentially many pending
1279 Frequency : occasionally with multiple writers to a single file
1281 Description: shared writes to file may result in wrong size reported by stat()
1282 Details : Allow growing of kms when extent lock is cancelled
1285 Frequency : always with random mmap IO to multi-striped file
1287 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1288 Details : Only if the hole is at the end of OST object so that kms is too
1289 small. Fix is to increase kms accordingly in ll_nopage.
1292 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1294 Description: writes to a missing object would leak memory on the OST
1295 Details : If there is an inconsistency between the MDS and OST filesystems,
1296 such that the MDS references an object that doesn't exist, writes
1297 to that object will leak memory due to incorrect cleanup in the
1298 error handling path, eventually running out of memory on the OST.
1303 Description: Creating too long symlink causes lustre errors
1304 Details : Check symlink and name lengths before sending requests to MDS.
1307 Frequency : only if flock is enabled (not on by default)
1309 Description: posix locks not released on fd closure on 2.6.9+
1310 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1311 kernels, this caused such locks not to be released on fd close and
1312 then assertions on fs unmount about still used locks.
1315 Frequency : MDS failover only, very rarely
1317 Description: clients may get ASSERTION(granted_lock != NULL)
1318 Details : When request was taking a long time, and a client was resending
1319 a getattr by name lock request. The were multiple lock requests
1320 with the same client lock handle and
1321 mds_getattr_name->fixup_handle_for_resent_request found one of the
1322 lock handles but later failed with ASSERTION(granted_lock != NULL).
1327 Description: handle->h_buffer_credits > 0, assertion failure
1328 Details : h_buffer_credits is zero after truncate, causing assertion
1329 failure. This patch extends the transaction or creates a new
1333 Frequency : NFS re-export or patchless client
1334 Bugzilla : 11179, 10796
1335 Description: Crash on NFS re-export node (__d_move)
1336 Details : We do not want to hash the dentry if we don't have a lock.
1337 But if this dentry is later used in d_move, we'd hit uninitialised
1338 list head d_hash, so we just do this to init d_hash field but
1339 leave dentry unhashed.
1342 Frequency : NFS re-export or patchless client
1344 Description: NFS exports has problem with symbolic link
1345 Details : lustre client didn't properly install dentry when re-exported
1346 to NFS or running patchless client.
1349 Frequency : NFS re-export or patchless client
1351 Description: Various nfs/patchless fixes.
1352 Details : fixes reuse disconected alias for lookup process - this fixes
1353 warning "find_exported_dentry: npd != pd",
1354 fix permission error with open files at nfs.
1355 fix apply umask when do revalidate.
1358 Frequency : occasional
1360 Description: Crash on NFS re-export node
1361 Details : calling clear_page() on the wrong pointer triggered oops in
1362 generic_mapping_read().
1365 Frequency : rarely, using O_DIRECT IO
1367 Description: unaligned directio crashes client with LASSERT
1368 Details : check for unaligned buffers before trying any requests.
1371 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1373 Description: stale data returned from RAID cache
1374 Details : If only a small amount of IO is done to the RAID device before
1375 reading it again it is possible to get stale data from the RAID
1376 cache instead of reading it from disk.
1379 Frequency : always for sles10 kernel
1381 Description: sles10 support
1382 Details : ll_follow_link: compile fixes and using of nd_set_link
1383 under newer kernels.
1386 Frequency : depends on arch, kernel and compiler version, always on sles10
1389 Description: recursive or deep enough symlinks cause stack overflow
1390 Details : getting rid of large stack-allocated variable in
1394 Frequency : depends on hardware
1396 Description: lustre write performance loss in the SLES10 kernel
1397 Details : the performance loss is caused by using of write barriers in the
1398 ext3 code. The SLES10 kernel turns barrier support on by
1399 default. The fix is to undo that change for ldiskfs.
1401 ------------------------------------------------------------------------------
1403 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1405 * Support for kernels:
1406 2.6.9-42.0.3EL (RHEL 4)
1407 2.6.5-7.276 (SLES 9)
1408 2.4.21-47.0.1.EL (RHEL 3)
1409 2.6.12.6 vanilla (kernel.org)
1413 Frequency : quota enabled and large files being deleted
1415 Description: releasing more than 4GB of quota at once hangs OST
1416 Details : If a user deletes more than 4GB of files on a single OST it
1417 will cause the OST to spin in an infinite loop. Release
1418 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1423 Description: statfs data retrieved from /proc may be stale or zero
1424 Details : When reading per-device statfs data from /proc, in the
1425 {kbytes,files}_{total,free,avail} files, it may appear
1426 as zero or be out of date.
1429 Frequency : systems with MD RAID1 external journal devices
1431 Description: lconf's call to blkid is confused by RAID1 journal devices
1432 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1433 of returning all block devices that match the journal UUID.
1436 Frequency : always, for aggregate stripe size over 4GB
1438 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1439 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1440 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1443 Frequency : always if "lfs find" used on a local file/directory
1445 Description: "lfs find" segfaults if used on a local file/directory
1446 Details : The case where a directory component was not specified wasn't
1447 handled correctly. Handle this properly.
1450 Frequency : always on ppc64
1452 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1453 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1454 perform endianness conversion.
1457 Frequency : rarely (truncate to non-zero file size after write under load)
1458 Bugzilla : 10730, 10687
1459 Description: Files padded with zeros to next 4K multiple
1460 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1461 it is possible that files that are truncated to a non-zero size
1462 immediately after being written are filled with zero bytes beyond
1463 the truncated size. No file data is lost.
1465 Severity : enhancement
1467 Description: Allow recovery/failover for liblustre clients.
1468 Details : liblustre clients were unaware of failover configurations until
1471 Severity : enhancement
1473 Description: user file locks should fail when not mounting with flock option
1474 Details : Set up an error-returning stub in ll_file_operations.lock field
1475 to prevent incorrect behaviour when client is mounted without
1476 flock option. Also, set up properly f_op->flock field for
1480 Frequency : always on ia64
1482 Description: "lfs df" loops on printing out MDS statfs information
1483 Details : The obd_ioctl_data was not initialized and in some systems
1484 this caused a failure during the ioctl that did not return
1485 an error. Initialize the struct and return an error on failure.
1488 Frequency : SLES 9 only
1490 Description: Error of copying files with lustre special EAs as root
1491 Details : Client side always return success for setxattr call for lustre
1492 special xattr (currently only "trusted.lov").
1495 Frequency : rarely on clusters with both ia64+i386 clients
1497 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1498 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1499 clients are doing concurrent writes to the same file it is
1500 possible that the ia64 clients may LASSERT because the OST
1501 extent locks are not PAGE_SIZE aligned. Ensure that grown
1502 locks are always aligned on the request boundary.
1505 Frequency : specific use, occasional
1507 Description: Overwriting in use executable truncates on-disk binary image
1508 Details : If one node attempts to overwrite an executable in use by
1509 another node, we now correctly return ETXTBSY instead of
1510 truncating the file.
1512 Severity : enhancement
1514 Description: Async OSC create to avoid the blocking unnecessarily.
1515 Details : If a OST has no remain object, system will block on the creating
1516 when need to create a new object on this OST. Now, ways use
1517 pre-created objects when available, instead of blocking on an
1518 empty osc while others are not empty. If we must block, we block
1519 for the shortest possible period of time.
1524 Description: chmod on Lustre root is propagated to other clients
1525 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1526 invalid by the follow_mount time.
1529 Frequency : liblustre clients only
1531 Description: Race in 'instant cancel' lock handling could lead to such locks
1532 never to be granted in case of SMP MDS
1533 Details : Do not destroy not yet granted but cbpending locks in
1537 Frequency : replay/resend of open
1539 Description: non null lock assetion failure in mds_intent_policy
1540 Details : Trying to replay/resend lockless open requests resulted in
1541 mds_open() returning 0 with no lock. Now it sets a flag if
1542 a lock is going to be returned.
1544 Severity : enhancement
1546 Description: Checksum enhancements
1547 Details : New checksum enhancements allow for resending RPCs that failed
1550 Severity : enhancement
1552 Description: Tunables on number of dirty pages in cacche
1553 Details : Allow to set limit on number of dirty pages cached.
1558 Description: client crash on unmount - lock still has references
1559 Details : In some error handling cases it was possible to leak a lock
1560 reference on a client while accessing a file. This was not
1561 harmful to the client during operation, but would cause the
1562 client to crash when the filesystem is unmounted.
1565 Frequency : specific case, rare
1567 Description: ETXTBSY on mds though file not in use by client
1568 Details : ETXTBSY is no longer incorrectly returned when attempting to
1569 chmod or chown a directory that the user previously tried to
1570 execute or a currently-executing binary.
1573 Frequency : extremely rare except on liblustre-based clients
1575 Description: Lustre space not freed when files are deleted
1576 Details : Clean up open-unlinked files after client eviction. Previously
1577 the unlink was skipped and the files remained as orphans.
1582 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1583 Details : In some cases it was possible to send a blocking callback to
1584 a client doing a glimpse, even though that client didn't get
1585 a lock granted. When the glimpse lock is cancelled on the OST
1586 the freed lock is left on the waiting list and corrupted the list.
1589 Frequency : all core dumps
1591 Description: Broke core dumps to lustre
1592 Details : Negative dentry may be unhashed if parent does not have UPDATE
1593 lock, but some callers, e.g. do_coredump, expect dentry to be
1594 hashed after successful create, hash it in ll_create_it.
1596 ------------------------------------------------------------------------------
1598 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1600 * Support for kernels:
1601 2.6.9-42.0.2.EL (RHEL 4)
1602 2.6.5-7.276 (SLES 9)
1603 2.4.21-40.EL (RHEL 3)
1604 2.6.12.6 vanilla (kernel.org)
1608 Frequency : always on RHEL 3
1610 Description: Number of open files grows over time
1611 Details : The number of open files grows over time, whether or not
1612 Lustre is started. This was due to a filp leak introduced
1613 by one of our kernel patches.
1615 ------------------------------------------------------------------------------
1617 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1619 * Support for kernels:
1620 2.6.9-42.EL (RHEL 4)
1621 2.6.5-7.276 (SLES 9)
1622 2.4.21-40.EL (RHEL 3)
1623 2.6.12.6 vanilla (kernel.org)
1628 Bugzilla : 5719, 9635, 9792, 9684
1629 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1630 Details : If a server is under heavy load and cannot reply to new
1631 connection requests before the client resends the (re)connect,
1632 the connection handling code can behave badly if two service
1633 threads are concurrently handing separate (re)connections from
1634 the same client. Add better locking to the connection handling
1635 code, and ensure that only a single connection will be processed
1636 for a given client UUID, even if the lock is dropped.
1638 Severity : enhancement
1640 Description: add TCP zero-copy support to kernel
1641 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1642 sends if the hardware supports scatter-gather and checksumming.
1643 This allows socklnd to do client-write and server-read more
1644 efficiently and reduce CPU utilization from skbuf copying.
1647 Frequency : only if NFS exporting from client
1649 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1650 Details : If setting attributes on a file created under NFS that had
1651 never been opened it would be possible to oops the client
1652 if the file had no objects.
1655 Frequency : always for liblustre
1657 Description: liblustre client does MDS+OSTs setattr RPC for each write
1658 Details : When doing a write from a liblustre client, the client
1659 incorrectly issued an RPC to the MDS and each OST the file was
1660 striped over in order to update the timestamps. When writing
1661 with small chunks and many clients this could overwhelm the MDS
1662 with RPCs. In all cases it would slow down the write because
1663 these RPCs are unnecessary.
1665 Severity : enhancement
1667 Description: allow number of MDS service threads to be changed at module load
1668 Details : It is now possible to change the number of MDS service threads
1669 running. Adding "options mds mds_num_threads={N}" to the MDS's
1670 /etc/modprobe.conf will set the number of threads for the next
1671 time Lustre is restarted (assuming the "mds" module is also
1672 reloaded at that time). The default number of threads will
1673 stay the same, 32 for most systems.
1678 Description: OST crash if filesystem is unformatted or corrupt
1679 Details : If an OST is started on a device that has never been formatted
1680 or if the filesystem is corrupt and cannot even mount then the
1681 error handling cleanup routines would dereference a NULL pointer.
1686 Description: NULL pointer deref in llap_from_page.
1687 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1688 incorrect) mapping if the page was truncated/reclaimed while it was
1689 searched for. Check for this condition and skip such pages when
1690 doing readahead. Introduce extra check to llap_from_page() to
1691 verify page->mapping->host is non-NULL (so page is not anonymous).
1694 Frequency : Sometimes when using sys_sendfile
1696 Description: "page not covered by a lock" warnings from ll_readpage
1697 Details : sendfile called ll_readpage without right page locks present.
1698 Now we introduced ll_file_sendfile that does necessary locking
1699 around call to generic_file_sendfile() much like we do in
1703 Frequency : with certain MDS communication failures at client mount time
1705 Description: NULL pointer deref after failed client mount
1706 Details : a client connection request may delayed by the network layer
1707 and not be sent until after the PTLRPC layer has timed out the
1708 request. If the client fails the mount immediately it will try
1709 to clean up before the network times out the request. Add a
1710 reference from the request import to the obd device and delay
1711 the cleanup until the network drops the request.
1714 Frequency : occasionally during client (re)connect
1716 Description: assertion failure during client (re)connect
1717 Details : processing a client connection request may be delayed by the
1718 client or server longer than the client connect timeout. This
1719 causes the client to resend the connection request. If the
1720 original connection request is replied in this interval, the
1721 client may trip an assertion failure in ptlrpc_connect_interpret()
1722 which thought it would be the only running connect process.
1725 Frequency : only with obd_echo servers and clients that are rebooted
1727 Description: kernel BUG accessing uninitialized data structure
1728 Details : When running an obd_echo server it did not start the ping_evictor
1729 thread, and when a client was evicted an uninitialized data
1730 structure was accessed. Start the ping_evictor in the RPC
1731 service startup instead of the OBD startup.
1733 Severity : enhancement
1734 Bugzilla : 10193 (patchless)
1735 Description: Remove dependency on various unexported kernel interfaces.
1736 Details : No longer need reparent_to_init, exit_mm, exit_files,
1737 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1740 Frequency : rare (only users of deprecated and unsupported LDAP config)
1742 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1743 Details : LDAP apparently contains 'lustreName' attributes instead of
1744 'name'. A simple remapping of the name is sufficient.
1747 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1749 Description: waiting_locks_callback trips kernel BUG if client is evicted
1750 Details : Running with the dump_on_timeout debug flag turned on makes
1751 it possible that the waiting_locks_callback() can try to dump
1752 the Lustre kernel debug logs from an interrupt handler. Defer
1753 this log dumping to the expired_lock_main() thread.
1755 Severity : enhancement
1757 Description: Support NFS exporting on 2.6 kernels.
1758 Details : Implement non-rawops metadata methods for NFS server to use without
1759 changing NFS server code.
1762 Frequency : very rare (synthetic metadata workload only)
1764 Description: two racing renames might cause an MDS thread to deadlock
1765 Details : Running the "racer" program may cause one MDS thread to rename
1766 a file from being the source of a rename to being the target of
1767 a rename at exactly the same time that another thread is doing
1768 so, and the second thread has already enqueued these locks after
1769 doing a lookup of the target and is trying to relock them in
1770 order. Ensure that we don't try to re-lock the same resource.
1773 Frequency : only very large systems with liblustre clients
1775 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1776 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1777 all OSTs in parallel. This allows the eviction of stale liblustre
1778 clients to proceed much faster than if they were done in series,
1779 and also offers similar improvements for other set_info RPCs.
1784 Description: excessive CPU usage during initial read phase on client
1785 Details : During the initial read phase on a client, it would agressively
1786 retry readahead on the file, consuming too much CPU and impacting
1787 performance (since 1.4.5.8). Improve the readahead algorithm
1788 to avoid this, and also improve some other common cases (read
1789 of small files in particular, where "small" is files smaller than
1790 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1795 Description: MDS crash when receiving packet with unknown intent.
1796 Details : Do not LBUG in unknown intent case, just return -EFAULT
1798 Severity : enhancement
1799 Bugzilla : 9293, 9385
1800 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1801 Details : Do not serialize getattr (non-intent version) and statfs.
1804 Frequency : occasional, when OST network is overloaded/intermittent
1806 Description: client evicted by OST after bulk IO timeout
1807 Details : If a client sends a bulk IO request (read or write) the OST
1808 may evict the client if it is unresposive to its data GET/PUT
1809 request. This is incorrect if the network is overloaded (takes
1810 too long to transfer the RPC data) or dropped the OST GET/PUT
1811 request. There is no need to evict the client at all, since
1812 the pinger and/or lock callbacks will handle this, and the
1813 client can restart the bulk request.
1816 Frequency : Always when mmapping file with no objects
1818 Description: client crashes when mmapping file with no objects
1819 Details : Check that we actually have objects in a file before doing any
1820 operations on objects in ll_vm_open, ll_vm_close and
1826 Description: Request leak when working with deleted CWD
1827 Details : Introduce advanced request refcount tracking for requests
1828 referenced from lustre intent.
1830 Severity : Enhancement
1832 Description: Cache open file handles on client.
1833 Details : MDS now will return special lock along with openhandle, if
1834 requested and client is allowed to hold openhandle, even if unused,
1835 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1836 closing files for every read/write openration.
1838 Severity : Enhancement
1840 Description: Cache open negative dentries on client when possible.
1841 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1842 negative dentries on lock revocation.
1847 Description: Remounting a client read-only wasn't possible with a zconf mount
1848 Details : It wasn't possible to remount a client read-only with llmount.
1850 Severity : enhancement
1851 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1852 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1853 MPICH is included to add Lustre-specific ADIO interfaces.
1854 This is based closely on the UFS ADIO layer and only differs
1855 in file creation, in order to allow the OST striping to be set.
1856 This is user-contributed code and not supported by CFS.
1861 Description: extended inode attributes (immutable, append-only) work improperly
1862 when 2.4 and 2.6 kernels are used on client/server or vice versa
1863 Details : Introduce kernel-independent values for these flags.
1865 Severity : enhancement
1868 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1869 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1870 granularity. Now, Fractional MB granularity is be supported,
1871 this is very useful for low memory system.
1873 Severity : enhancement
1875 Description: Getattr by fid
1876 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1877 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1878 have locks and avoids getattr rpc requests for it on every stat.
1881 Frequency : Always, for filesystems larger than 2TB
1883 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1884 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1885 filesystems larger than 2TB to be mounted with mballoc enabled.
1888 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1890 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1891 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1892 filesystems larger than 2TB is mounted then the kernel will
1893 silently corrupt the start of the filesystem. CONFIG_LBD is
1894 enabled for all CFS-supported kernels, but the possibility of
1895 this happening with a modified kernel config exists.
1897 Severity : enhancement
1899 Description: add client O_DIRECT support for 2.6 kernels
1900 Details : It is now possible to do O_DIRECT reads and writes to files
1901 in the Lustre client mountpoint on 2.6 kernel clients.
1903 Severity : enhancement
1905 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1906 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1907 parallel, not waiting for response from every OST before sending
1908 a rpc to the next OST.
1913 Description: setattr vs write race when updating file timestamps
1914 Details : Client processes that update a file timestamp into the past
1915 right after writing to the file (e.g. tar) it is possible that
1916 the updated file modification time can be reset to the current
1917 time due to a race between processing the setattr and write RPC.
1919 Severity : enhancement
1921 Description: Bring 'lfs find' closer in line with regular Linux find.
1922 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1923 -print0 options and obtains all the needed info through the lustre
1926 Severity : enhancement
1928 Description: support up to 1024 configured devices on one node
1929 Details : change obd_dev array from statically allocated to dynamically
1930 allocated structs as they are first used to reduce memory usage
1935 Description: Flush dirty partially truncated pages during truncate
1936 Details : Immediatelly flush partially truncated pages in filter_setattr,
1937 this way we completely avoid having any pages in page cache on OST
1938 and can retire ugly workarounds during writes to flush such pages.
1943 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1944 Details : protect inode from truncation within vfs_unlink() context
1945 just take a reference before calling vfs_unlink() and release it
1946 when parent's i_sem is free.
1951 Description: last_id value checked outside lock on OST caused LASSERT failure
1952 Details : If there were multiple MDS->OST object precreate requests in
1953 flight, it was possible that the OST's last object id was checked
1954 outside a lock and incorrectly tripped an assertion. Move checks
1955 inside locks, and discard old precreate requests.
1958 Frequency : always, if extents are used on OSTs
1960 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1961 Details : OSTs using the extents format would not zero the high 16 bits of
1962 the index physical block number. This is not a problem for any
1963 OST filesystems smaller than 16TB, and no kernels support ext3
1964 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1965 new/modified files) and can be fixed for existing filesystems
1966 with e2fsprogs-1.39-cfs1.
1971 Description: import connection selection may be incorrect if timer wraps
1972 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1973 import connections to be ignored if the 32-bit jiffies counter
1974 wraps. Use a 64-bit jiffies counter.
1977 Frequency : very large clusters immediately after boot
1979 Description: LNET request buffers exhausted under heavy short-term load
1980 Details : If a large number of client requests are generated on a service
1981 that has previously never seen so many requests it is possible
1982 that the request buffer growth cannot keep up with the spike in
1983 demand. Instead of dropping incoming requests, they are held in
1984 the LND until the RPC service can accept more requests.
1987 Frequency : Sometimes during replay
1989 Description: Assertion failure in ll_local_open after replay.
1990 Details : If replay happened on an open request reply before we were able
1991 to set replay handler, reply will become not swabbed tripping the
1992 assertion in ll_local_open. Now we set the handler right after
1993 recognising of open request
1996 Frequency : very rare
1998 Description: kernel reports "badness in vsnprintf"
1999 Details : Reading from the "recovery_status" /proc file in small chunks
2000 may cause a negative length in lprocfs_obd_rd_recovery_status()
2001 call to vsnprintf() (which is otherwise harmless). Exit early
2002 if there is no more space in the output buffer.
2004 Severity : enhancement
2006 Description: clear OBD RPC statistics by writing to them
2007 Details : It is now possible to clear the OBD RPC statistics by writing
2008 to the "stats" file.
2013 Description: Client mtime is not the same on different clients after utimes
2014 Details : In some cases, the client was using the utimes() syscall on
2015 a file cached on another node. The clients now validate the
2016 ctime from the MDS + OSTs to determine which one is right.
2021 Description: Inability to activate failout mode
2022 Details : lconf script incorrectly assumed that in pythong string's numeric
2023 value is used in comparisons.
2026 Frequency : always with multiple stripes per file
2028 Description: Inefficient object allocation for mutli-stripe files
2029 Details : When selecting which OSTs to stripe files over, for files with
2030 a stripe count that divides evenly into the number of OSTs,
2031 the MDS is always picking the same starting OST for each file.
2032 Return the OST selection heuristic to the original design.
2037 Description: mount failures may take full timeout to return an error
2038 Details : Under some heavy load conditions it is possible that a
2039 failed mount can wait for the full obd_timeout interval,
2040 possibly several minutes, before reporting an error.
2041 Instead return an error as soon as the status is known.
2043 Frequency : quota enabled and large files being deleted
2045 Description: releasing more than 4GB of quota at once hangs OST
2046 Details : If a user deletes more than 4GB of files on a single OST it
2047 will cause the OST to spin in an infinite loop. Release
2048 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2053 Description: statfs data retrieved from /proc may be stale or zero
2054 Details : When reading per-device statfs data from /proc, in the
2055 {kbytes,files}_{total,free,avail} files, it may appear
2056 as zero or be out of date.
2059 Frequency : systems with MD RAID1 external journal devices
2061 Description: lconf's call to blkid is confused by RAID1 journal devices
2062 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2063 of returning all block devices that match the journal UUID.
2066 Frequency : always, for aggregate stripe size over 4GB
2068 Description: assertion fails when trying to use 4GB stripe size
2069 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2070 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2073 Frequency : always on ppc64
2075 Description: the first write on an ext3 filesystem with mballoc got stuck
2076 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2077 perform endianness conversion.
2079 ------------------------------------------------------------------------------
2081 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2083 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2084 INTEROPERATE with older versions automatically. Please read the
2085 user documentation before upgrading any part of a live system.
2086 * WARNING: Lustre networking configuration changes are required with
2087 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2090 * Support for kernels:
2091 2.6.9-22.0.2.EL (RHEL 4)
2092 2.6.5-7.244 (SLES 9)
2093 2.6.12.6 vanilla (kernel.org)
2096 Severity : enhancement
2097 Bugzilla : 7981/8208
2098 Description: Introduced Lustre Networking (LNET)
2099 Details : LNET is new networking infrastructure for Lustre, it includes
2100 a reorganized network configuration mode (see the user
2101 documentation for full details) as well as support for routing
2102 between different network fabrics. Lustre Networking Devices
2103 (LNDs) for the supported network fabrics have also been
2104 created for this new infrastructure.
2106 Severity : enhancement
2107 Description: Introduced Access control lists
2108 Details : clients can set ACLs on files and directories in order to have
2109 more fine-grained permissions than the standard Unix UGO+RWX.
2110 The MDS must be started with the "-o acl" mount option.
2112 Severity : enhancement
2113 Description: Introduced filesystem quotas
2114 Details : Administrators may now establish per-user quotas on the
2117 Severity : enhancement
2119 Description: Configuration change for the XT3
2120 The PTLLND is now used to run Lustre over Portals on the XT3
2121 The configure option(s) --with-cray-portals are no longer used.
2122 Rather --with-portals=<path-to-portals-includes> is used to
2123 enable building on the XT3. In addition to enable XT3 specific
2124 features the option --enable-cray-xt3 must be used.
2129 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2130 Details : A race between allocating a new llcd and re-getting the llcd_lock
2131 allowed another thread to grab newly-allocated llcd.
2133 Severity : enhancement
2135 Description: 2.6 OST async journal commit and locking fix to improve performance
2136 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2137 2.6 kernels are now async as they already were in 2.4 kernels so
2138 that they can commit concurrently with the network bulk transfer.
2139 For block-allocated files the filter allocation semaphore is held
2140 to avoid filesystem fragmentation during allocation. BKL lock
2141 removed for 2.6 xattr operations where it is no longer needed.
2146 Description: lconf incorrectly determined whether two IP networks could talk
2147 Details : In some more complicated routing and multiple-network
2148 configurations, lconf will avoid trying to make a network
2149 connection to a disjoint part of the IP space. It was doing the
2150 math incorrectly for one set of cases.
2155 Description: Fix for potential infinite loop processing records in an llog.
2156 Details : If an llog record is corrupted/zeroed, it is possible to loop
2157 forever in llog_process(). Validate the llog record length
2158 and skip the remainder of the block on error.
2161 Frequency : occasional (liblustre only)
2163 Description: liblustre could not open files whose last component is a symlink
2164 Details : sysio_path_walk() would incorrectly pass the open intent to
2165 intermediate path components.
2168 Frequency : rare (liblustre only with non-standard tuning)
2169 Bugzilla : 7201 (7350)
2170 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2171 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2172 after it has been released, only do this for referenced locks
2174 Severity : enhancement
2176 Description: specifying an (invalid) directory default stripe_size of -1
2177 would reset the directory default striping
2178 Details : stripe_size -1 was used internally to signal directory stripe
2179 removal, now use "all default" to signal dir stripe removal
2180 as a directory striping of "all default" is not useful
2183 Frequency : common for large clusters running liblustre clients
2185 Description: doing an ls when liblustre clients are running is slow
2186 Details : sending a glimpse AST to a liblustre client waits for every AST
2187 to time out, as liblustre clients will not respond. Since they
2188 cannot cache data we refresh the OST lock LVB from disk instead.
2190 Severity : enhancement
2192 Description: doing an ls at the same time as file IO can be slow
2193 Details : enqueue and other "small" requests can be blocked behind many
2194 large IO requests. Create a new OST IO portal for non-IO
2195 requests so they can be processed faster.
2198 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2200 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2201 created file with mode 000
2202 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2203 allow an ftruncate() as a non-root user to a file with mode 000.
2204 Lustre now respects this flag to disable mode checks when
2205 truncating a file owned by the user
2208 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2211 Description: Revoking locks from clients that went dead or catatonic might take
2213 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2214 cancellation of such locks instant on servers without waiting for
2215 any reply from clients. Clients drops these locks when cancel
2216 notification from server is received without replying.
2219 Frequency : liblustre-only, when liblustre client dies or becomes busy
2221 Description: Doing ls on Linux clients can take a long time with active
2223 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2224 granting such locks to it in the first place if possible. Locks
2225 are taken by proxy on the OST during the read or write and
2226 dropped immediately afterward. Add connect flags handling, do
2227 not grant locks to liblustre clients for glimpse ASTs.
2229 Severity : enhancement
2231 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2232 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2233 sequential reads and behaves badly with applications doing random
2234 reads. Improve it by reading ahead at least read region, and
2235 avoiding excessive large RPC for small reads.
2237 Severity : enhancement
2239 Description: Creating more than 1000 files for a single job may cause a load
2240 imbalance on the OSTs if there are also a large number of OSTs.
2241 Details : qos_prep_create() uses an OST index reseed value that is an
2242 even multiple of the number of available OSTs so that if the
2243 reseed happens in the middle of the object allocation it will
2244 still utilize the OSTs as uniformly as possible.
2249 Description: OST or MDS may oops in ping_evictor_main()
2250 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2251 but doesn't restart at beginning of obd_exports_timed list
2254 Severity : enhancement
2256 Description: improve by-nid export eviction on the MDS and OST
2257 Details : allow multiple exports with the same NID to be evicted at one
2258 time without re-searching the exports list.
2261 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2263 Description: MDS may oops in groups_free()
2264 Details : in rare race conditions a newly allocated group_info struct is
2265 freed again, and this can be NULL. The 2.4 compatibility code
2266 for groups_free() checked for a NULL pointer, but 2.6 did not.
2269 Frequency : common for liblustre clients doing little filesystem IO
2270 Bugzilla : 9352, 7313
2271 Description: server may evict liblustre clients accessing contended locks
2272 Details : if a client is granted a lock or receives a completion AST
2273 with a blocking AST already set it would not reply to the AST
2274 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2277 Frequency : lfs setstripe, only systems with more than 160 OSTs
2279 Description: unable to set striping with a starting offset beyond OST 160
2280 Details : llapi_create_file() incorrectly limited the starting stripe
2281 index to the maximum single-file stripe count.
2284 Frequency : LDAP users only
2286 Description: lconf did not handle in-kernel recovery with LDAP properly
2287 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2289 Severity : enhancement
2291 Description: bind OST threads to NUMA nodes to improve performance
2292 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2293 node and do their allocations there to localize memory access
2295 Severity : enhancement
2297 Description: llmount can determine client NID directly from Myrinet (GM)
2298 Details : the client NID code from gmnalnid was moved directly into
2299 llmount, removing the need to use this or specifying the
2300 client NID explicitly when mounting GM clients with zeroconf
2303 Frequency : if client is started with down MDS
2305 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2306 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2307 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2308 ptlrpc_import_delayed_req() because we didn't check if the
2309 request was interrupted, and we also didn't break out of the
2310 event loop if there was no timeout
2315 Description: data loss during non-page-aligned writes to a single file from
2316 both multiple nodes and multiple threads on one node at same time
2317 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2318 inconsistency led to false short-reads, that were cached and later
2319 used by ->prepare_write() to fill in partially written page,
2320 leading to data loss.
2323 Frequency : always, if lconf --abort_recovery used
2325 Description: lconf --abort_recovery fails with 'Operation not supported'
2326 Details : lconf was attempting to abort recovery on the MDT device and not
2329 Severity : enhancement
2331 Description: remove cleanup logs
2332 Details : replace lconf-generated cleanup logs with lustre internal
2333 cleanup routines. Eliminates the need for client-cleanup and
2336 Severity : enhancement
2338 Description: add support for EAs (user and system) on lustre filesystems
2339 Details : it is now possible to store extended attributes in the Lustre
2340 client filesystem, and with the user_xattr mount option it
2341 is possible to allow users to store EAs on their files also
2343 Severity : enhancement
2345 Description: Add possibility (config option) to show minimal available OST free
2347 Details : When compiled with --enable-mindf configure option, statfs(2)
2348 (and so, df) will return least minimal free space available from
2349 all OSTs as amount of free space on FS, instead of summary of
2350 free spaces of all OSTs.
2352 Severity : enhancement
2354 Description: do not expand extent locks acquired on OST-side
2355 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2356 by server: they are not cached anyway.
2359 Frequency : when mmap is used/binaries executed from Lustre
2361 Description: Unmmap pages before throwing them away from read cache.
2362 Details : llap_shrink cache now attempts to unmap pages before discarding
2363 them (if unmapping failed - do not discard). SLES9 kernel has
2364 extra checks that trigger if this unmapping is not done first.
2369 Description: lconf didn't resolve symlinks before checking to see whether a
2370 given mountpoint was already in use
2373 Frequency : when migrating failover services
2374 Bugzilla : 6395, 9514
2375 Description: When migrating a subset of services from a node (e.g. failback
2376 from a failover service node) the remaining services would
2377 time out and evict clients.
2378 Details : lconf --force (implied by --failover) sets the global obd_timeout
2379 to 5 seconds in order to quickly disconnect, but this caused
2380 other RPCs to time out too quickly. Do not change the global
2381 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2383 Severity : enhancement
2384 Frequency : if MDS is started with down OST
2385 Bugzilla : 9439,5706
2386 Description: Allow startup/shutdown of an MDS without depending on the
2387 availability of the OSTs.
2388 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2389 Add appropriate locking and lov-osc refcounts for safe
2390 cleaning. Add osc abort_inflight calls in case the
2391 synchronize never started.
2394 Frequency : occasional (Cray XT3 only)
2396 Description: root not authorized to access files in CRAY_PORTALS environment
2397 Details : The client process capabilities were not honoured on the MDS in
2398 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2399 been accessed by an authorized user then root was able to access
2400 the file on the local client also. The root user capabilities
2401 are now allowed on the MDS, as this environment has secure UID.
2404 Frequency : occasional
2406 Description: ldiskfs "too long searching" message happens too often
2407 Details : A debugging message (otherwise harmless) prints too often on
2408 the OST console. This has been reduced to only happen when
2409 there are fragmentation problems on the filesystem.
2414 Description: Division by zero in statfs when all OSCs are inactive
2415 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2416 lov_statfs divides by value returned by lov_get_stripecnt().
2420 Bugzilla : 9489, 3273
2421 Description: First write from each client to each OST was only 4kB in size,
2422 to initialize client writeback cache, which caused sub-optimal
2423 RPCs and poor layout on disk for the first writen file.
2424 Details : Clients now request an initial cache grant at (re)connect time
2425 and so that they can start streaming writes to the cache right
2426 away and always do full-sized RPCs if there is enough data.
2427 If the OST is rebooted the client also re-establishes its grant
2428 so that client cached writes will be honoured under the grant.
2433 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2434 Details : Now I/O RPCs go to different portal number and (presumably) fast
2435 lock requests (and glimses) and other RPCs get their own service
2436 threads pool that should be able to service those RPCs
2439 Severity : enhancement
2441 Description: Ability to exchange lustre version between client and servers and
2442 issue warnings at client side if client is too old. Also for
2443 liblustre clients there is ability to refuse connection of too old
2445 Details : New 'version' field is added to connect data structure that is
2446 filled with version info. That info is later checked by server and
2450 Frequency : rare, liblustre only.
2451 Bugzilla : 9296, 9581
2452 Description: Two simultaneous writes from liblustre at offset within same page
2453 might proceed at the same time overwriting eachother with stale
2455 Details : I/O lock withing llu_file_prwv was released too early, before data
2456 actually was hitting the wire. Extended lock-holding time until
2457 server acknowledges receiving data.
2460 Frequency : extremely rare. Never observed in practice.
2462 Description: avoid generating lustre_handle cookie of 0.
2463 Details : class_handle_hash() generates handle cookies by incrementing
2464 global counter, and can hit 0 occasionaly (this is unlikely, but
2465 not impossible, because initial value of cookie counter is
2466 selected randonly). Value of 0 is used as a sentinel meaning
2467 "unassigned handle" --- avoid it. Also coalesce two critical
2468 sections in this function into one.
2470 Severity : enhancement
2472 Description: allow liblustre clients to delegate truncate locking to OST
2473 Details : To avoid overhead of locking, liblustre client instructs OST to
2474 take extent lock in ost_punch() on client's behalf. New connection
2475 flag is added to handle backward compatibility.
2477 Severity : enhancement
2478 Bugzilla : 4928, 7341, 9758
2479 Description: allow number of OST service threads to be specified
2480 Details : a module parameter allows the number of OST service threads
2481 to be specified via "options ost ost_num_threads={N}" in the
2482 OSS's /etc/modules.conf or /etc/modprobe.conf.
2486 Bugzilla : 6146, 9635, 9895
2487 Description: servers crash with bad pointer in target_handle_connect()
2488 Details : In rare cases when a client is reconnecting it was possible that
2489 the connection request was the last reference for that export.
2490 We would temporarily drop the export reference and get a new
2491 one, but this may have been the last reference and the export
2492 was just destroyed. Get new reference before dropping old one.
2494 Severity : enhancement
2495 Frequency : if client is started with failover MDS
2497 Description: Allow multiple MDS hostnames in the mount command
2498 Details : Try to read the configuration from all specified MDS
2499 hostnames during a client mount in case the "primary"
2502 Severity : enhancement
2504 Description: Stop sending data to evicted clients as soon as possible.
2505 Details : Check if the client we are about to send or are sending data to
2506 was evicted already. (Check is done every second of waiting,
2507 for which l_wait_event interface was extended to allow checking
2508 of exit condition at specified intervals).
2511 Frequency : rare, normally only when NFS exporting is done from client
2513 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2515 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2518 Frequency : always, if srand() is called before liblustre initialization
2520 Description: Liblustre uses system PRNG disturbing its usage by user application
2521 Details : Introduce internal to lustre fast and high-quality PRNG for
2522 lustre usage and make liblustre and some other places in generic
2523 lustre code to use it.
2525 Severity : enhancement
2526 Bugzilla : 9477, 9557, 9870
2527 Description: Verify that the MDS configuration logs are updated when xml is
2528 Details : Check if the .xml configuration logs are newer than the config
2529 logs stored on the MDS and report an error if this is the case.
2530 Request --write-conf, or allow starting with --old_conf.
2532 Severity : enhancement
2534 Description: Handle symlinks in the path when checking if Lustre is mounted.
2535 Details : Resolve intermediate symlinks when checking if a client has
2536 mounted a filesystem to avoid duplicate client mounts.
2541 Description: lconf can hit an error exception but still return success.
2542 Details : The lconf command catches the Command error exception at the top
2543 level script context and will exit with the associated exit
2544 status, but doesn't ensure that this exit status is non-zero.
2549 Description: failure of ptlrpc thread startup can cause oops
2550 Details : Starting a ptlrpc service thread can fail if there are a large
2551 number of threads or the server memory is very fragmented.
2552 Handle this without oopsing.
2555 Frequency : always, only if liblustre and non-default acceptor port was used
2557 Description: liblustre cannot connect to servers with non-default acceptor port
2558 Details : tcpnal_set_default_params() was not called and was therefore
2559 ignoring the environment varaible TCPNAL_PORT, as well as other
2560 TCPNAL_ environment variables
2565 Description: two objects could be created on the same OST for a single file
2566 Details : If an OST is down, in some cases it was possible to create two
2567 objects on a single OST for a single file. No problems other
2568 than potential performance impact and spurious error messages.
2572 Bugzilla : 5681, 9562
2573 Description: Client may oops in ll_unhash_aliases
2574 Details : Client dcache may become inconsistent in race condition.
2575 In some cases "getcwd" can fail if the current directory is
2581 Description: Inode refcounting problems in NFS export code
2582 Details : link_raw functions used to call d_instantiate without obtaining
2583 extra inode reference first.
2587 Bugzilla : 9942, 9903
2588 Description: Referencing freed requests leading to crash, memleaks with NFS.
2589 Details : We used to require that call to ll_revalidate_it was always
2590 followed by ll_lookup_it. Also with revalidate_special() it is
2591 possible to call ll_revalidate_it() twice for the same dentry
2592 even if first occurence returned success. This fix changes semantic
2593 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2594 reference on a request referred from the intent.
2595 ll_intent_release() then releases such a request.
2598 Frequency : rare, normally benchmark loads only
2600 Description: unlinked inodes were kept in memory on the client
2601 Details : If a client is repeatedly creating and unlinking files it
2602 can accumulate a lot of stale inodes in the inode slab cache.
2603 If there is no other client load running this can cause the
2604 client node to run out of memory. Instead flush old inodes
2605 from client cache that have the same inode number as a new inode.
2608 Frequency : SLES9 2.6.5 kernel and long filenames only
2609 Bugzilla : 9969, 10379
2610 Description: utime reports stale NFS file handle
2611 Details : SLES9 uses out-of-dentry names in some cases, which confused
2612 the lustre dentry revalidation. Change it to always use the
2616 Frequency : rare, unless heavy write-truncate concurrency is continuous
2617 Bugzilla : 4180, 6984, 7171, 9963, 9331
2618 Description: OST becomes very slow and/or deadlocked during object unlink
2619 Details : filter_destroy() was holding onto the parent directory lock
2620 while truncating+unlinking objects. For very large objects this
2621 may block other threads for a long time and slow overall OST
2622 responsiveness. It may also be possible to get a lock ordering
2623 deadlock in this case, or run out of journal credits because of
2624 the combined truncate+unlink. Solution is to do object truncate
2625 first in one transaction without parent lock, and then do the
2626 final unlink in a new transaction with the parent lock. This
2627 reduces the lock hold time dramatically.
2630 Frequency : rare, 2.4 kernels only
2632 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2633 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2634 before unmounting their filesystem in order to avoid deadlock.
2635 The kernel_locked() function in 2.4 kernels only checks whether
2636 the kernel lock is held, not whether it is this process that is
2637 holding it as 2.6 kernels do.
2642 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2643 Details : The client ptlrpc code may be trying to reconnect to a down
2644 server before a previous connection attempt has timed out.
2645 Increase the reconnect interval to be longer than the connection
2646 timeout interval to avoid sending duplicate connections to
2650 Frequency : echo_client brw_test command
2652 Description: fix echo_client to work with OST preallocated code
2653 Details : OST preallocation code (5137) didn't take echo_client IO path
2654 into account: echo_client calls filter methods outside of any
2655 OST thread and, hence, there is no per-thread preallocated
2656 pages and buffers to use. Solution: hijack pga pages for IO. As
2657 a byproduct, this avoids unnecessary data copying.
2661 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2662 Description: Client can oops in mdc_commit_close() after open replay
2663 Details : It was possible for the MDS to return an open request with no
2664 transaction number in mds_finish_transno() if the client was
2665 evicted, but without actually returning an error. Clients
2666 would later try to replay that open and may trip an assertion
2667 Simplify the client close codepath, and always return an error
2668 from the MDS in case the open is not successful.
2671 Frequency : rare, 2.6 OSTs only
2673 Description: OST may deadlock under high load on fragmented files
2674 Details : If there was a heavy load and highly-fragmented OST filesystems
2675 it was possible to have all the OST threads deadlock waiting on
2676 allocation of biovecs, because the biovecs were not released
2677 until the entire RPC IO was completed. Instead, release biovecs
2678 as soon as they are complete to ensure forward IO progress.
2680 Severity : enhancement
2682 Description: Support for specifying external journal device at mount
2683 Details : If an OST or MDS device is formatted with an external journal
2684 device, this device major/minor is stored in the ext3 superblock
2685 and may not be valid for failover. Allow detecting and
2686 specifying the external journal at mount time.
2691 Description: Mounting an MDS with pending unlinked files may cause oops
2692 Details : target_finish_recovery() calls mds_postrecov() which returned
2693 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2694 considers this an error and immediately begins cleaning up the
2695 lov, just after starting the mds_lov process
2697 Severity : enhancement
2699 Description: Implement 'lfs df' to report actual free space on per-OST basis
2700 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2701 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2702 usage of objects; '-h' to report in human readable format.
2704 ------------------------------------------------------------------------------
2706 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2713 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2714 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2715 at startup during orphan unlinking. Instead, skip buddy update
2716 before setup, as it will be regenerated after recovery is complete.
2721 Description: If an OST is inactive, its locks might reference stale inodes.
2722 Details : lov_change_cbdata() must iterate over all namespaces, even if
2723 they are inactive to clear inode references from the lock.
2725 Severity : enhancement
2726 Frequency : occasional, if non-standard max_dirty_mb used
2728 Description: Client will block write RPCs if not enough grant
2729 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2730 then the client will block writes while waiting for another RPC
2731 to complete instead of consuming its dirty limit. With change
2732 we get improved performance when max_dirty_mb is small.
2734 Severity : enhancement
2735 Bugzilla : 3389, 6253
2736 Description: Add support for supplementary groups on the MDS.
2737 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2738 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2739 lookups for user supplementary groups into a cache.
2743 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2744 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2745 for liblustre O_EXCL is enforced by MDS.
2749 Description: readdir never returns NULL in liblustre.
2750 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2751 dentry in struct dirent.
2755 Frequency : liblustre only, depends on application IO pattern
2756 Description: liblustre clients evicted if not contacting servers
2757 Details : Don't put liblustre clients into the ping_evictor list, so
2758 they will not be evicted by the pinger ever.
2760 Severity : enhancement
2762 Description: Add ability to evict clients by NID from MDS.
2763 Details : By echoing "nid:$NID" string into
2764 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2765 $NID would be instantly evicted from this MDS and from all active
2766 OSTs connected to it.
2770 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2771 Details : lookup_it_finish() used to query file size from OSTs that was not
2776 Description: service threads change working directory to that of init
2777 Details : Starting lustre service threads may pin the working directory
2778 of the parent thread, making that filesystem busy. Threads
2779 now change to the working directory of init to avoid this.
2783 Frequency : during shutdown only
2784 Description: shutdown with a failed MDS or OST can cause unmount to hang
2785 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2790 Frequency : 2.6 only
2791 Description: chmod/chown may include an extra supplementary group
2792 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2793 supplementary group and if none is specified this is used.
2796 Bugzilla : 5479 (6816)
2798 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2799 Details : If lookup is in progress on a file that is unlinked we might try
2800 to revalidate the inode and fail in revalidate after lookup is
2801 complete and ll_file_open() enqueues the open again but
2802 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2805 Frequency : always, if lconf --abort_recovery used
2807 Description: lconf --abort_recovery fails with 'Operation not supported'
2808 Details : lconf was attempting to abort recovery on the MDT device and not
2811 ------------------------------------------------------------------------------
2813 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2818 Frequency : rare (only unsupported configurations with a node running as an
2820 Bugzilla : 6514, 5137
2821 Description: Mounting a Lustre file system on a node running as an OST could
2823 Details : OSTs now preallocates memory needed to write out data at
2824 startup, instead of when needed, to avoid having to
2825 allocate memory in possibly low memory situations.
2826 Specifically, if the file system is mounted on on OST,
2827 memory pressure could force it to try to write out data,
2828 which it needed to allocate memory to do. Due to the low
2829 memory, it would be unable to do so and the node would
2830 become unresponsive.
2832 Severity : enhancement
2834 Description: Addition of lconf --service command line option
2835 Details : lconf now accepts a '--service <arg>' option, which is
2836 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2838 Severity : enhancement
2840 Description: Failover mode is now the default for OSTs.
2841 Details : By default, OSTs will now run in failover mode. To return to
2842 the old behaviour, add '--failout' to the lmc line for OSTs.
2844 Severity : enhancement
2846 Description: Health checks are now provided for MDS and OSTs
2847 Details : Additional detailed health check information on MSD and OSTs
2848 is now provided through the procfs health_check value.
2851 Frequency : occasional, depends on IO load
2853 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2854 after numerous create/delete cycles
2855 Details : The ext3 inode allocation policy would not allocate new inodes
2856 very well on the OSTs because there are no new directories
2857 being created. Instead we look for groups with free space if
2858 the parent directories are nearly full.
2863 Description: Network or server problems during mount may cause partially
2864 mounted clients instead of returning an error.
2865 Details : The config llog parsing code may overwrite the error return
2866 code during mount error handling, returning success instead
2872 Description: MDS can fail to allocate large reply buffers
2873 Details : After long uptimes the MDS can fail to allocate large reply
2874 buffers (e.g. zconf client mount config records) due to memory
2875 fragmentation or consumption by the buffer cache. Preallocate
2876 some large reply buffers so that these replies can be sent even
2877 under memory pressure.
2881 Frequency : rare (liblustre)
2882 Description: fsx running with liblustre complained that using truncate() to
2883 extend the file doesn't work. This patch corrects that issue.
2884 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2885 ATTR_SIZE and lsm use in llu_setattr_raw.
2889 Frequency : rare, only 2.6 kernels
2890 Description: Unusual file access patterns on the MDS may result in inode
2891 data being lost in very rare circumstances.
2892 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2893 code in the 2.6 kernel caused the inode and/or EA data not to
2894 be read from disk, causing single-file corruption.
2898 Frequency : rare, only 2.6 filesystems using extents
2899 Description: Heavy concurrent write and delete load may cause data corruption.
2900 Details : It was possible under high-load situations to have an extent
2901 metadata block in the block device cache from a just-unlinked
2902 file overwrite a newly-allocated data block. We now unmap any
2903 metadata buffers that alias just-allocated data blocks.
2907 Frequency : filesystems with default stripe_count larger than 77
2908 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2909 Details : lconf specifies an inode size of 4096 bytes when the default
2910 stripe_count is larger than 77. This conflicts with the default
2911 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2912 this case to avoid pinning too much memory for large EAs.
2914 ------------------------------------------------------------------------------
2916 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2921 Frequency : rare (extremely heavy IO load with hundreds of clients)
2923 Description: Client is evicted, gets IO error writing to file
2924 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2925 caused clients to be evicted for AST timeouts. The fixes in
2926 bug 5192 mean we no longer need to have such short AST timeouts
2927 so ldlm_timeout has been increased.
2930 Frequency : occasional during --force or --failover shutdown under load
2931 Bugzilla : 5949, 4834
2932 Description: Server oops/LBUG if stopped with --force or --failover under load
2933 Details : a collection of import/export refcount and cleanup ordering
2934 issues fixed for safer force cleanup
2937 Frequency : only filesystems larger than 120 OSTs
2938 Bugzilla : 5990, 6223
2939 Description: lfs getstripe would oops on a very large filesystem
2940 Details : lov_getconfig used kfree on vmalloc'd memory
2943 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2944 Bugzilla : 6242, 6243
2945 Description: reading from files that had been truncated to a non-zero size
2946 but never opened returned no data
2947 Details : ll_file_read() reads zeros from no-object files to EOF
2952 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2953 Details : There's little harm in aborting MDS/OSS recovery and letting it
2954 try again, so I removed the LASSERT and return an error instead.
2956 Severity : enhancement
2958 Description: New debugging infrastructure for tracking down data corruption
2959 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2960 (b) cover more of the client-side code path, and (c) try to narrow
2961 down where problems occurred
2965 Bugzilla : 3819, 4364, 4397, 6313
2966 Description: Racing close and eviction MDS could cause assertion in mds_close
2967 Details : It was possible to get multiple mfd references during close and
2968 client eviction, leading to one thread referencing a freed mfd.
2970 Severity: : enhancement
2971 Bugzilla : 3262, 6359
2972 Description: Attempts to reconnect to servers are now more aggressive.
2973 Details : This builds on the enhanced upcall-less recovery that was added
2974 in 1.4.2. When trying to reconnect to servers, clients will
2975 now try each server in the failover group every 10 seconds. By
2976 default, clients would previously try one server every 25 seconds.
2981 Description: After recovery, certain operations trigger a failed
2982 assertion on a client.
2983 Details : Failing over an mds, using lconf -d --failover, while a
2984 client was doing a readdir() call would cause the client to
2985 LBUG after recovery completed and the readdir() was resent.
2987 Severity : enhancement
2989 Description: Default groups are now added by lconf
2990 Details : You can now run lconf --group <servicename> without having to
2991 manually add groups with lmc.
2994 Frequency : occasional
2996 Description: Nodes with an elan id of 0 trigger a failed assertion
2999 Frequency : always when accessing e.g. tty/console device nodes
3001 Description: tty and some other devices nodes cannot be used on lustre
3002 Details : file's private_data field is used by device data and lustre
3003 values in there got lost. New field was added to struct file to
3004 store fs-specific private data.
3007 Frequency : when exporting Lustre via NFS
3009 Description: NFSD failed occasionally when looking up a path component
3010 Details : NFSD is looking up ".." which was broken in ext3 directories
3011 that had grown large enough to become hashed.
3014 Frequency : Clusters with multiple interfaces not on the same subnet
3016 Description: Nodes will repeatedly try to reconnect to an interface which it
3017 cannot reach and report an error to the log.
3018 Details : Extra peer list entries will be created by lconf with some peers
3019 unreachable. lconf now validates the peer before adding it.
3022 Frequency : Only if a default stripe is set on the filesystem root.
3024 Description: Setting a default stripe on the filesystem root prevented the
3025 filesystem from being remounted.
3026 Details : The client was sending extra request flags in the root getattr
3027 request and did not allocate a reply buffer for the dir EA.
3030 Frequency : occasional, higher if lots of files are accessed by one client
3031 Bugzilla : 6159, 6097
3032 Description: Client trips assertion regarding lsm mismatch/magic
3033 Details : While revalidating inodes the VFS looks up inodes with ifind()
3034 and in rare cases can find an inode that is being freed.
3035 The ll_test_inode() code will free the lsm during ifind()
3036 when it finds an existing inode and then the VFS later attaches
3037 this free lsm to a new inode.
3041 Bugzilla : 6422, 7030
3042 Description: MDS deadlock between mkdir and client eviction
3043 Details : Creating a new file via mkdir or mknod (starting a transaction
3044 and getting the ns lock) can deadlock with client eviction
3045 (gets ns lock and trying to finish a synchronous transaction).
3048 Frequency : occasional
3049 Description: While starting a server, the fsfilt_ext3 module could not be
3051 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3052 kernels. Previously, lconf would still use the ext3 name
3053 when trying to load modules. Now, it will correctly use
3054 ext3 on 2.4 and ldiskfs on 2.6.
3056 Severity : enhancement
3057 Description: The default stripe count has been changed to 1
3058 Details : The interpretation of the default stripe count (0, to lfs
3059 or lmc) has been changed to mean striping across a single
3060 OST, rather than all available. For general usage we have
3061 found a stripe count of 1 or 2 works best.
3063 Severity : enhancement
3064 Description: Add support for compiling against Cray portals.
3065 Details : Conditional compiling for some areas that are different
3069 Frequency : occasional
3070 Bugzilla : 6409, 6834
3071 Description: Creating files with an explicit stripe count may lead to
3072 a failed assertion on the MDS
3073 Details : If some OSTs are full or unavailable, creating files may
3074 trigger a failed assertion on the MDS. Now, Lustre will
3075 try to use other servers or return an error to the
3079 Frequency : occasional
3081 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3082 caused lock timeout during readahead (since 1.4.2).
3083 Details : Processes doing readahead might match a lock that hasn't been
3084 granted yet if there are overlapping and conflicting lock
3085 requests. The readahead process waits on ungranted lock
3086 (original lock is CBPENDING), while OST waits for that process
3087 to cancel CBPENDING read lock and eventually evicts client.
3089 Severity : enhancement
3091 Description: Initial enabling of flock support for clients
3092 Details : Implements fcntl advisory locking and file status functions.
3093 This feature is provided as an optional mount flag (default
3094 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3095 locking are implemented yet, and those that are are not guaranteed
3096 to be completely correct in production environments.
3097 mount -t lustre -o [flock|noflock] ...
3100 Frequency : occasional
3102 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3103 assertion in the ext3 JBD (journaling) layer.
3104 Details : The b_committed_data struct is protected by the big kernel lock
3105 in 2.4 kernels, serializing journal_commit_transaction() and
3106 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3107 access to this struct. In 2.6 kernels there is finer grained
3108 locking to improve SMP performance of the JBD layer.
3112 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3117 Description: merge and cleanup kernel patches.
3120 -----------------------------------------------------------------------------
3122 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3124 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3125 versions of Lustre. Please update all servers and clients to
3126 version 1.4.2 or later at the same time. You must also run
3127 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3128 to update the configuration logs.
3130 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3131 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3132 - swab llog records of type '0' so we get proper header size/idx (5861)
3133 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3134 - fix rename of one directory over another leaking an inode (5953)
3135 - avoid SetPageDirty on 2.6 (5981)
3136 - don't re-add just-being-destroyed locks to the waiting list (5653)
3137 - when creating new directories, inherit the parent's custom
3138 striping settings if present parent (3048)
3139 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3140 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3141 - don't hold client locks on temporary worklist from l_lru (5666)
3142 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3143 - automatically evict dead clients (5921)
3144 - Update file size properly in create+truncate+fstat case (6196)
3145 - Do not unhash mountpoint dentries, do not allow removal of
3147 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3148 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3149 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3150 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3151 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3152 - Correctly handle reads of files with no objects (6243)
3153 - lctl recover will also mark a device active if deactivate used (5933)
3155 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3156 - allow --write-conf on an MDS with different nettype than client (5619)
3157 - don't write config llogs to MDS for mounts not from that MDS (5617)
3158 - lconf should create multiple TCP connections from a client (5201)
3159 - init scripts are now turned off by default; run chkconfig --on
3160 lustre and chkconfig --on lustrefs to use them
3161 - upcalls are no longer needed for clients to recover to failover
3163 - add --abort-recovery option to lconf to abort recovery on device
3165 - add support for an arbitrary number of OSTs (3026)
3166 - Quota support protocol changes.
3167 - forward compatibility changes to wire structs (6007)
3168 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3169 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3170 - improved "lustre status" script
3171 - initialize blocksize for non-regular files (6062)
3172 - added --disable-server and --disable-client configure options (5782)
3173 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3174 - Vanilla 2.4.29 support
3175 - increase maximum number of obd devices to 520 (6242)
3176 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3177 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3178 SLES 9 2.6.5 kernels
3179 - limit stripes per file to 160 (the maximum EA size) (6093)
3181 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3184 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3185 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3186 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3187 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3188 - small ext3 extents cleanups and fixes (5733)
3189 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3190 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3191 - avoid needless client->OST connect, fix handle mismatch (5317)
3192 - fix DLM error path that led to out-of-sync client, long delays (5779)
3193 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3194 - fix several locking issues related to i_size (5492,5624,5654,5672)
3195 - don't move pending lock onto export if it is already evicted (5683)
3196 - fix kernel oops when creating .foo in unlinked directory (5548)
3197 - fix deadlock in obdfilter statistics vs. object create (5811)
3198 - use time_{before,after} to avoid timer jiffies wrap (5882)
3199 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3200 - Do not leak request if server was not able to process it (5154)
3201 - If mds_open unable to find parent dir, make that negative lookup(5154)
3202 - don't create new directories with extent-mapping (5909, 5936)
3204 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3205 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3206 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3207 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3208 - support for RHEL4 kernel on i686 (5773)
3209 - provide error messages when incompatible logs are encountered (5898)
3211 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3212 * version 1.4.0.10 (1.4.1 release candidate 1)
3214 - don't keep a lock reference when lock is not granted (4238)
3215 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3216 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3217 - drop import inflight refcount on signal_completed_replay error (5255)
3218 - unlock page after async write error during send (3677)
3219 - handle missing objects in filter_preprw_read properly (5265)
3220 - no transno return for symlink open, don't save no-trasno open (3440)
3221 - don't try to complete elan receive that already failed (4012)
3222 - free RPC server reply state on error (5406)
3223 - clean up thread from ptlrpc_start_thread() on error (5160)
3224 - readahead could read extra page into cache that wasn't ejected (5388)
3225 - prevent races in class_attach/setup/cleanup/detach (5260)
3226 - don't dereference de->d_inode after l_dput of de (5458)
3227 - use "int" for stripe value returned from lock_to_stripe (5544)
3228 - mballoc allocation and error-checking fixes in 2.6 (5504)
3229 - block device patches to fix I/O request sizes in 2.6 (5482)
3230 - look up hostnames for IB nals (5602)
3231 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3232 - don't start multiple acceptors for the same port (5277)
3233 - fix incorrect LASSERT in mds_getattr_name (5635)
3234 - export a proc file for general "ping" checking (5628)
3235 - fix "lfs check" to not block when the MDS is down (5628)
3237 - service request history (4965)
3238 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3239 - create an "evict_client" /proc entry on OSTs, like the MDS has
3240 - fix mount usage message, return errors per mount(8) (5168)
3241 - change grep [] to grep "[]" in tests so they work in more UMLs
3242 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3243 - remove ancient LOV_MAGIC_V0 EA support (5047)
3244 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3245 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3246 - allow readahead allocations to fail when low on memory (5383)
3247 - mmap locking landed again, after considerable improvement (2828)
3248 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3250 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3253 - send OST transaction number in read/write reply to free req (4966)
3254 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3255 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3256 - the watchdog thread now runs as interruptible (5246)
3257 - flock/lockf fixes (but it's still disabled, pending 5135)
3258 - don't use EXT3 constants in llite code (5094)
3259 - memory shortage at startup could cause assertion (5176)
3261 - reorganization of lov code
3262 - single portals codebase
3264 - add extents/mballoc support (5025)
3265 - direct I/O reads in the obdfilter (4048)
3266 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3268 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3271 - send OST transaction number in read/write reply to free req (4966)
3272 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3273 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3274 - the watchdog thread now runs as interruptible (5246)
3275 - handle missing objects in filter_preprw_read properly (5265)
3276 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3277 - drop import inflight refcount on signal_completed_replay error (5255)
3278 - unlock page after async write error during send (3677)
3279 - return original error code on reconstructed replies (3761)
3280 - no transno return for symlink open, don't save no-trasno open (3440)
3282 - add pid to ldlm debugging output (4922)
3283 - bump the watchdog timeouts -- we can't handle 30sec yet
3284 - extra debugging for orphan dentry/inode bug (5259)
3286 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3289 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3290 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3291 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3292 - fix NULL dereference in /proc/sys/portals/routes (4827)
3293 - allow failed mdc_close() operations to be interrupted (4561)
3294 - stop precreate on OST before MDS would time out on it (4778)
3295 - don't send partial-page writes before EOF from client (4410)
3296 - discard client grant for sub-page writes on large-page clients (4520)
3297 - don't free dentries not owned by NFS code, check generation (4806)
3298 - fix lsm leak if mds_create_objects() fails (4801)
3299 - limit debug_daemon file size, always print CERROR messages (4789)
3300 - use transno after validating reply (3892)
3301 - process timed out requests if import state changes (3754)
3302 - update mtime on OST during writes, return in glimpse (4829)
3303 - add mkfsoptions to LDAP (4679)
3304 - use ->max_readahead method instead of zapping global ra (5039)
3305 - don't interrupt __l_wait_event() during strace
3307 - add software watchdogs to catch hung threads quickly (4941)
3308 - make lustrefs init script start after nfs is mounted
3309 - fix CWARN/ERROR duplication (4930)
3310 - return async write errors to application if possible (2248)
3311 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3312 - print NAL number in %x format (4645)
3313 - update barely-supported suse-2.4.21-171 series (4842)
3314 - support for sles 9 %post scripts
3315 - support for building 2.6 kernel-source packages
3316 - support for sles km_* packages
3318 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3321 - ignore -ENOENT errors in osc_destroy (3639)
3322 - notify osc create thread that OSC is being cleaned up (4600)
3323 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3324 - reconstruct ost_handle() like mds_handle() (4657)
3325 - create a new thread to do import eviction to avoid deadlock (3969)
3326 - let lconf resolve symlinked-to devices (4629)
3327 - don't unlink "objects" from directory with default EA (4554)
3328 - hold socknal file ref over connect in case target is down (4394)
3329 - allow more than 32000 subdirectories in a single directory (3244)
3330 - fix blocks count for O_DIRECT writes (3751)
3331 - OST returns ENOSPC from object create when no space left (4539)
3332 - don't send truncate RPC if file size isn't changing (4410)
3333 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3334 - bind to privileged port in socknal and tcpnal (3689)
3336 - rate limit CERROR/CWARN console message to avoid overload (4519)
3337 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3338 - basic mmap support (3918)
3339 - kernel patch series update from b1_4 (4711)
3341 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3344 - avoid crash during MDS cleanup with OST shut down (2775)
3345 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3346 - don't use bad inodes on the MDS (3744)
3347 - dynamic object preallocation to improve recovery speed (4236)
3348 - don't hold spinlock over lock dumping or change debug flags (4401)
3349 - don't zero obd_dev when it is force cleaned (3651)
3350 - print grants to console if they go negative (4431)
3351 - "lctl deactivate" will stop automatic recovery attempts (3406)
3352 - look for existing locks in ldlm_handle_enqueue() (3764)
3353 - don't resolve lock handle twice in recovery avoiding race (4401)
3354 - revalidate should check working dir is a directory (4134)
3356 - don't always mark "slow" obdfilter messages as errors (4418)
3358 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3361 - don't close LustreDB during write_conf until it is done (3860)
3362 - fix typo in lconf for_each_profile (3821)
3363 - allow dumping logs from multiple threads at one time (3820)
3364 - don't allow multiple threads in OSC recovery (3812)
3365 - fix debug_size parameters (3864)
3366 - fix mds_postrecov to initialize import for llog ctxt (3121)
3367 - replace config semaphore with spinlock (3306)
3368 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3369 - don't allow enqueue to complete on a destroyed export (3822)
3370 - down write_lock before checking llog header bitmap (3825)
3371 - recover from lock replay timeout (3764)
3372 - up llog sem before sending rpc (3652)
3373 - reduce ns lock hold times when setting kms (3267)
3374 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3375 - fix NULL deref and obd_dev leak on setup error (3312)
3376 - replace some LBUG about llog ops with error handling (3841)
3377 - don't match INVALID dentries from d_lookup and spin (3784)
3378 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3379 - fix invalid assertion in ptlrpc_set_wait (3880)
3381 - add libwrap support for the TCP acceptor (3996)
3382 - add /proc/sys/portals/routes for non-root route listing (3994)
3383 - allow setting MDS UUID in .xml (2580)
3384 - print the stack of a process that LBUGs (4228)
3386 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3389 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3390 - make sure to unset replay flag from failed open requests (3440)
3391 - if default stripe count is 0, use OST count for inode size (3636)
3392 - update parent mtime/ctime on client for create/unlink (2611)
3393 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3394 - free recovery state on server during a forced cleanup (3571)
3395 - unregister_reply for resent reqs (3063)
3396 - loop back devices mounting and status check on 2.6 (3563)
3397 - fix resource-creation race that can provoke i_size == 0 (3513)
3398 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3399 - more debugging for page-accounting assertion (3746)
3400 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3401 - avoid deadlock after precreation failure (3758)
3402 - fix race and lock order deadlock in orphan handling (3450, 3750)
3403 - add validity checks when grabbing inodes from l_ast_data (3599)
3405 - add /proc/.../recovery_status to obdfilter (3428)
3406 - lightweight CDEBUG infrastructure, debug daemon (3668)
3407 - change default OSC RPC parameters to be better on small clusters
3408 - turn off OST read cache for files smaller than 32MB
3409 - install man pages and include them in rpms (3100)
3410 - add new init script for (un)mounting lustre filesystems (2593)
3411 - run chkconfig in %post for init scripts (3701)
3412 - drop scimac NAL (unmaintained)
3414 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3417 - clean kiobufs before and after use (3485)
3418 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3419 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3420 - change init script to fail more clearly if not run as root (1528)
3421 - allow clients to reconnect during replay (1742)
3422 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3423 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3424 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3425 - fix cond_resched (3554)
3426 - only evict unfinished clients after recovery (3515)
3427 - allow bulk resend, prevent data loss (3570)
3428 - dynamic ptlrpc request buffer allocation (2102)
3429 - don't allow unlinking open directory if it isn't empty (2904)
3430 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3431 - remove extraneous obd dereference causing LASSERT failure (3334)
3432 - don't use get_cycles() when creating temp. files on the mds (3156)
3433 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3434 - handle EEXIST for set-stripe, set proper directory name (3336)
3436 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3437 - fix ksocknal_fmb_callback() error messages (2918)
3439 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3442 - don't copy lvb into (possibly NULL) reply on error (2983)
3443 - don't deref dentry after dput, don't free lvb on error (2922)
3444 - use the kms to determine writeback rpc length (2947)
3445 - increment oti_logcookies when osc is inactive (2948)
3446 - update client's i_blocks count via lvb messages (2543)
3447 - handle intent open/close of special files properly (1557)
3448 - mount MDS with errors=remount-ro, like obdfilter (2009)
3449 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3450 - don't use cancelling-locks' kms values (2947)
3451 - use highest lock extent for kms, not last one (2925)
3452 - don't dereference ERR_PTR() dentry in error handling path (3107)
3453 - fix thread race in portals_debug_dumplog() (3122)
3454 - create lprocfs device entries at setup instead of at attach (1519)
3455 - common AST error handler, don't evict client on completion race (3145)
3456 - zero nameidata in detach_mnt in 2.6 (3118)
3457 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3458 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3459 - initialize RPC timeout timer earlier for 2.6 (3219)
3460 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3461 - print nal/nid for unknown nid (3258)
3462 - additional checks for oscc recovery before doing precreate (3284)
3463 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3464 - don't crash in mdc_close for bad permissions on open (3285)
3465 - zero i_rdev for non-device files (3147)
3466 - clear page->private before handing to FS, better assertion (3119)
3467 - tune the read pipeline (3236)
3468 - fix incorrect decref of invalidated dentry (2350)
3469 - provide read-ahead stats and refine rpc in flight stats (3328)
3470 - don't hold journal transaction open across create RPC (3313)
3471 - update atime on MDS at close time (3265)
3472 - close LDAP connection when recovering to avoid server load (3315)
3473 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3474 - don't leak open file on MDS after open resend (3325)
3475 - serialize filter_precreate and filter_destroy_precreated (3329)
3476 - loop device shouldn't call sync_dev() for nul device (3092)
3477 - clear page cache after eviction (2766)
3478 - resynchronize MDS->OST in background (2824)
3479 - refuse to mount the same filesystem twice on same mountpoint (3394)
3480 - allow llmount to create routes for mounting behind routers (3320)
3481 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3482 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3483 - fix rare problem with rename on htree directories (3417)
3485 - allow default OST striping configuration per directory (1414)
3486 - fix compilation for qswnal for 2.6 kernels (3125)
3487 - increase maximum number of MDS request buffers for large systems
3488 - change liblustreapi to be useful for external progs like lfsck (3098)
3489 - increase local configuration timeout for slow disks (3353)
3490 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3492 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3495 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3496 - don't overwrite extent policy data in reply if lock was blocked (2901)
3497 - drop filter export grants atomically with removal from device (2663)
3498 - del obd_self_export from work_list in class_disconnect_exports (2908)
3499 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3500 - swab reply message in mdc_close, other PPC fixes (2464)
3501 - fix destroying of named logs (2325)
3502 - overwrite old logs when running lconf --write_conf (2264)
3503 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3504 - fix race in target_handle_connect (2898)
3505 - mds_reint_create() should take same inode create lock (2926)
3506 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3507 - don't close files for self_export to avoid uninitialized obd (2936)
3508 - allow MDS with the same name as client node (2939)
3509 - hold dentry reference for closed log files for unlink (2325)
3510 - reserve space for all logs during transactions (2059)
3511 - don't evict page beyond end of stripe extent (2925)
3512 - don't oops on a deleted current working directory (2399)
3513 - handle hard links to targets without a parent properly (2517)
3514 - don't dereference NULL lock when racing during eviction (2867)
3515 - don't grow lock extents when lots of conflicting locks (2919)
3517 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3520 - account for cache space usage on clients to avoid data loss (974)
3521 - lfsck support in lustre kernel code (2349)
3522 - reduce journal credits needed for BRW writes (2370)
3523 - orphan handling to avoid losing space on client/server crashes
3524 - ptlrpcd can be blocked, stopping ALL progress (2477)
3525 - use lock value blocks to assist in proper KMS, faster stat (1021)
3526 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3527 - recovery for initial connections (2355)
3528 - fixes for mds_cleanup_orphans (1934)
3529 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3530 - block all file creations until orphan recovery completes (1901)
3531 - client remove rq_connection from request struct (2423)
3532 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3533 - recovery timer race (2670)
3534 - mdc_close recovey bug (2532)
3535 - ptlrpc cleanup bug (2710)
3536 - mds timeout on local locks (2588)
3537 - namespace lock held during RPCs (2431)
3538 - handle interrupted sync write properly (2503)
3539 - don't try to handle a message that hasn't been replied to (2699)
3540 - client assert failure during cleanup after abort recovery (2701)
3541 - leak mdc device after failed mount (2712)
3542 - ptlrpc_check_set allows timedout requests to complete (2714)
3543 - wait for inflight reqs when ptlrpcd finishes (2710)
3544 - make sure unregistered services are removed from the srv_list
3545 - reset bulk XID's when resending them (caught by 1138 test)
3546 - unregister_bulk after timeout
3547 - fix lconf error (2694)
3548 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3549 - readahead locks pages, leaves pending causing memory pressure (2673)
3550 - increase OST request buffers to 4096 on large machines (2729)
3551 - fix up permission of existing directories in simple_mkdir (2661)
3552 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3553 - don't assign transno to errored transactions (2742)
3554 - don't delete objects on OST if given a bogus objid from MDS (2751)
3555 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3556 - if rq_no_resend, then timeout request after recovery (2432)
3557 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3558 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3559 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3560 - MDS recovery shouldn't delete objects that it has given out (2730)
3561 - if enqueue arrives after completion, don't clobber LVB (2819)
3562 - don't unlock pages twice when trigger_group_io returns error (2814)
3563 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3564 - don't write pages to disk if there was an error (1450)
3565 - don't ping imports that have recovery disabled (2676)
3566 - take buffered bytes into account when balancing socknal conn (2817)
3567 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3568 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3569 - remove little-endian swabbing of llog records (1987)
3570 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3571 - retry reposting request buffers if they fail (1191)
3572 - grow extent at grant time to avoid granting a revoked lock (2809)
3573 - lock revoke doesn't evict page if covered by a second lock (2765)
3574 - disable VM readahead to avoid reading outside lock extents (2805)
3576 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3577 - updated kernel patches for hp-2.4.20 kernel (2681)
3579 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3582 - fix truncated write corruption (2366)
3583 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3585 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3586 - ensure len(uuid) < 37 in lmc (1171)
3587 - fix ia64 OOPS in llog_test (2255)
3588 - zero end of page at obdfilter for partial page writes (2648)
3589 - don't leave stale dentries around after renames (bug 2428)
3590 - fix timeouts when evicting a client with a single lock held (2642)
3591 - set deadline for the initial HELLO message to drain (2634)
3592 - print out dotted-quad IP addresses in the socknal (2302)
3594 - additional debugging for MDS client eviction problem (2443)
3595 - fix mkfsoptions support for osts (2603, 2604)
3597 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3600 - add series for the vanilla 2.6.0 kernel
3601 - add series for the vanilla 2.4.24 kernel
3602 - add series for a cray x86/64 UL kernel drop
3603 - fix xattr patches for the vanilla 2.4.19 series
3605 - generate true UUIDs in lmc (1171)
3606 - have portals stack dumping break in UML (2466)
3607 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3608 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3609 - wake callback waiting threads on client eviction (2460)
3610 - Add --ptldebug and --subsystem to lmc (1719)
3611 - update assertion to allow safe interrupt allocation
3612 - set rq_no_resend for cancel requests (2432)
3613 - recalculate ptlrpcd timeout after resend (2494)
3614 - call vfs_rmdir when removing pending directories (2368)
3615 - fix renaming a file to itself (2429)
3616 - lmc creates a default one-stripe lov (2454)
3617 - expand procfs space to handle large clusters (2326)
3618 - increase UML stack to avoid overflow
3619 - update lconf's list of debug and subsystem masks
3620 - fix lfs find --obd (2510)
3621 - /proc tunable for disabling filter read caching (2591)
3622 - stop rpm packages from altering slapd.conf (2301)
3623 - disable nagle in the socknal under 0conf (2578)
3624 - choose mds inode size based on stripe count (2572)
3625 - fix kernel-source rpm problems (2516)
3627 - add --disable-doc to avoid pdf generation (2421)
3628 - update documentation, tests, type-os, comments
3629 - avoid format warnings on ia64
3630 - remove the TOE NAL
3631 - tiny code cleanups by removing unused fields
3633 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3636 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3637 - remove the most blatant lies from BUILDING file (2371)
3638 - change default debug level to reasonable production setting
3639 - reduce client side cache size to reduce cache flush time
3640 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3641 - make TCP zerocopy and pinger support enabled by default (2476)
3642 - sync writes completed after process exits caused crashes (2319)
3643 - maintain correct mount count on the MDS (2356)
3644 - backout 1557, because 2316 wasn't really fixed
3645 - better file I/O statistics gathering in /proc
3646 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3647 - another kernel patch to fix zero-copy TCP function export
3648 - don't take duplicate lock when processing re-sent getattr (2420)
3649 - lctl uses obd_self_export instead of creating new conn (2353)
3650 - MDS/OST recovery case which requires object creation asserted (2425)
3651 - move lfs from /usr/sbin to /usr/bin in packages
3652 - fix race between mds_client_add and mds_client_free (2417)
3653 - use kmalloc instead of slabs in portals (2430)
3654 - don't create duplicate records when a failover MDS is present (2442)
3655 - remove unnecessary mount age check (2332)
3656 - don't remove directory inodes from locks prematurely (2451)
3657 - don't break if MDS service name is the same as hostname (2103)
3658 - fix races in client write RPC generation when cache full (2482)
3660 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3663 - remove now-unused request->rq_obd (278)
3664 - if an allocation fails, print out how much memory we've used (1933)
3665 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3666 - add the "configurable stack size" patch to most series files (1256)
3667 - ability to write large log records, for 100+ OST configs (2306)
3668 - fix NULL deref when filter_prep fails (2314)
3669 - fix operator precedence error in filter_sync
3670 - dynamic allocation of socknal TX descriptors (2315)
3671 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3672 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3673 - ensure that CWARN messages in Portals always get to the syslog
3674 - __init/__exit are not for prototype decls (ldlm_init/exit)
3675 - x86-64 compile warning fixes
3676 - fix gateway LMC keyword conflict (2318)
3677 - fix MDS lock inversions in getattr/reint paths (1844)
3678 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3679 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3680 - DLM race condition prevented some lock evictions (2328)
3681 - ENOMEM detection and retry on socknal sends (2230)
3682 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3683 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3684 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3685 - better error messages when a client is rejected during recovery (1505)
3686 - avoid cancelling locks which were never granted, after failure (2330)
3687 - fix i_sem/journal inversion in mds_client_add (2333)
3688 - fix truncate/getattr lock cycle deadlock (2334)
3689 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3690 - fix two rare exit paths which could leak an l_lock() ref (2321)
3691 - fencepost error in MDS/OST orphan recovery (2226)
3692 - make log record alignment 8 bytes (1988)
3693 - lstripe now fails when requested offset > ost_count (2237)
3694 - ensure that all kernel series have a complete list.h (1607)
3695 - fix crashes in special-file operations (2316)
3696 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3698 - allow configurable automake binary, for testing new versions
3699 - small update to the lfs documentation
3701 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3703 * fix negative export reference count in fsfilt_sync (2312)
3705 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3706 * release candidate 0.9.1
3708 - orphans are moved into the PENDING directory for possible recovery
3709 - replayed opens now open by fid for orphan/rename safety (1042)
3710 - last close of an orphan inode generates a transno (683)
3711 - chdir() and mount() now pin the directory entry (1020)
3712 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3713 - discard very old requests without processing them (1502)
3714 - remove some common, well-understood CERRORs (1505)
3715 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3716 - clear "grant" flags in OST replies until OST grant code lands (1644)
3717 - fix read performance by not clobbering i_blksize on client (1598)
3718 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3719 - make LRU size a /proc tunable, clears locks when reduced (707)
3720 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3721 - server locks take references on exports now (1558)
3722 - build fixes for 2.4.20-rh trees (1663)
3723 - return an error from lov_create if all OSCs are inactive (1751)
3724 - fix import levels when a reconnect happens without a timeout (1597)
3725 - exit early from mds_open if we get a lookup error (1749)
3726 - partial page read at EOF wouldn't wait for disk before sending (1642)
3727 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3728 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3729 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3730 - write/truncate lock inversion (1639)
3731 - Don't auto-load obdclass, portals modules during cleanup (1495)
3732 - fix timestamps from jumping to "now" (1763)
3733 - extra journal assertions (1648)
3734 - add an extra multiunlink test (1771)
3735 - fix read_record/write_record API (1776)
3736 - fix leak of offset_extent, possible incorrect i_size later (1772)
3737 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3738 - Debugging for the kqswnal_get_idle_tx problems (1820)
3739 - Allow recovery to be attempted multiple times (1536)
3740 - Write out MDS last_rcvd file after it is first created (1600)
3741 - Fix tx_descriptor leak in failed transmit situations (1827)
3742 - ext3 journaling fixes for assertion failure after IO error (1871)
3743 - class_export_put() on freed export after completion AST error (1896)
3744 - Fix revalidate looping in VFS (1322)
3745 - Don't access a freed export during MDS_REINT timeout (1521)
3746 - Add open-unlink recovery support on the MDS (1673,1764)
3747 - Return an error if no MDS data was read from last_rcvd (1946)
3748 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3749 - Don't setup a disk device that doesn't match exported UUID (317)
3750 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3751 - avoid committing NULL handle in force close
3752 - local.sh is now a one-stripe LOV configuration
3753 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3754 - fix ext3 htree duplicate directory entry corruption (1516)
3755 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3756 - update to new LOV EA format (2097)
3757 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3759 2003-06-15 Phil Schwan <phil@clusterfs.com>
3762 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3763 - per-import/export recovery handling (958, 931, 959)
3764 - multiple last-rcvd slots, for serving multiple FSes (949)
3765 - connections are again shared between multiple imp/exports (963, 964)
3766 - "umount -f" would hang if any requests needed to be sent (393, 978)
3767 - avoid pinning large req buffer by copying for queued messages (989)
3768 - add "uuid" to "lctl device" command to help upcalls (991)
3769 - "open" RPCs with transnos would confuse recovery counters (1037)
3770 - do proper endian conversion of all wire messages (288, 340, 891)
3771 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3772 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3773 - fix ptlrpc_connection leak in target_handle_connect (1174)
3774 - fix import refcounting bug in OST and MDS cleanup (1134)
3775 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3776 - fix very unlikely obd_types race condition (501)
3777 - remove osc_open hack for echo_client (1187)
3778 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3779 - a failure in read_inode2 leads to deadlock (1139)
3780 - cancel ack-locks as soon as transaction is committed (1072)
3781 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3782 - make sure to commitrw after any preprw to avoid deadlock (1162)
3783 - failing to execute a file in a lustre FS would lock inode (1203)
3784 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3785 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3786 - obd_uuid2tgt would walk off the end of the list (1255)
3787 - on IA64 the osc would give portals incorrect bulk size (1258)
3788 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3789 - fix lock inversion caused by new llite matching code (1282)
3790 - limit the number of dirty pages on a client to 10MB (1286)
3791 - timed out locks were not being corrected cancelled (1289)
3792 - fix O_DIRECT above 4GB on IA-32 (1292)
3793 * major user-visible changes
3794 - fail out/fail over policy now controlled by the upcall (993)
3796 - add OBD_PING to check server availability and failure (954)
3797 - lustre messages are now sent in sending host order (288, 340, 891)
3798 - add eadatalen to MDS getattr reply (340)
3799 - OST read replies may contain second buffer, with per-page status (593)
3801 2003-03-11 Phil Schwan <phil@clusterfs.com>
3804 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3805 - failed open()s could cause deadlock; fixed (867, 869)
3806 - stop cancelling OST locks when files are closed (481)
3807 - overlapping XID spaces caused network corruption (851, 853)
3808 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3809 - setattr_raw added, to do single-RPC, server-side setattrs
3810 - lmc/lconf syntax change for OST UUIDs
3811 - fix crashy race condition between ptlrpc_free_req and osc_close
3812 - don't use request in mdc_enqueue if we hit a timeout (889)
3813 - don't set the inode i_size for regular files from the MDS (896)
3814 - handle out of order completion AST (842)
3815 - don't LBUG if a lock request times out after receiving AST (913)
3816 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3817 - if a bad lock AST arrives, send an error instead of dropping entirely
3818 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3819 - fix leak in bulk IO when only partially completed (899, 900, 926)
3820 - fix O_DIRECT for ia64 (55)
3821 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3822 - C-z after timeout could hang a process forever; fixed (977)
3824 - client-side I/O cache (678, 924, 929, 941, 970)
3826 - READPAGE and SETATTRs which don't take server-side locks get
3829 2003-02-11 Phil Schwan <phil@clusterfs.com>
3832 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3833 - Fix fsfilt_extN_readpage() to read a full page of directory
3834 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3835 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3836 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3837 - Most intent operations were removed, in favour of a new RPC mode
3838 that does a single RPC to the server and bypasses most of the VFS
3839 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3840 - Aggressively cancel local locks on DLM servers
3841 - mds_reint_unlink sends EA to the client if it's the last nlink.
3842 client uses that EA to unlink OST objects.
3843 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3844 - recursive symlinks were fixed (439)
3845 - fixed NULL deref in DEBUG_REQ
3846 - filter_update_lastobjid no longer calls sync, which annoyed extN
3847 - fixed multi-client small-writes to a single file problem (445)
3848 - fixed mtime updates during file writes (607)
3849 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3850 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3851 - fixed Linux OST size reporting problem (444, 656)
3852 - OST now updates object mtime with writes or setattr (607, 619)
3853 - client verifies file size before zeroing page past EOF (445)
3854 - OST now writes last allocated objid to disk with allocation (108)
3855 - LOV on echo now works (409)
3857 - mds_reint_unlink sends a new buffer, with the EA included. this
3858 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3859 set if a regular file was being unlinked, and it was the last link
3860 - use PtlGet from the target for bulk writes (315)
3861 - OST now updates object mtime with writes or setattr (607, 619)
3862 - LDLM now has a grant-time callback to revalidate locked items, if
3864 - Many MDS operations were reorganized to combat race conditions
3866 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3867 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3869 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3872 - Fully reactivate OST imports after reconnection (512, others)
3873 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3874 - More graceful error handling for truncating on dead OST (515)
3875 - Don't error out unless we're actually accessing dead stripes (474)
3876 - Fix garbage sizes when stripes are missing (410)
3877 - LRU counters were broken, causing constant lock purge (433, 432)
3878 - garbage on read from stripes with failed OSTs (441)
3879 - mark OSCs as active before reconnecting during recovery (438)
3880 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3881 - lfind did not preserve OST order in output (443)
3882 - symlinks cause hung clients, incorrect data (439)
3883 - stop dereferencing request after dropping refcount (457)
3884 - don't LASSERT(spin_is_locked) on non-SMP (455)
3885 - fixes for many rename() bugs
3886 - fstat didn't correctly synchronize attributes (399)
3887 - server must handle lock cancellation during blocking AST prep (487)
3888 - bulk descriptors were free()d too soon (511)
3889 - fix paths in lconf, which would load incorrect modules (451, 507)
3890 - fix confusing lconf 'host not found' error message (386)
3891 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3892 - fix race condition in mdc_blocking_ast() for inode access (526)
3893 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3894 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3895 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3896 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3897 - fix lstripe in several ways (kernel side) (527)
3898 - fix request leak in ldlm_cli_enqueue (262)
3899 - incorrect OSC was marked inactive after OST failure
3900 - call mds_fs_cleanup before unmounting filesystem (524)
3901 - fix races between taking ns_lock and ldlm_lock_change_resource
3902 - fix races updating LOV export open file list
3903 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3904 - fix recovery NULL deref in ldlm_cli_cancel_unused
3905 - fix some DLM races by using new hash table for lock handles (419)
3906 - permit the client to specify desired inodes, at replay
3907 - duplicate requests when we queue them for replay reintegration
3908 - fix last_rcvd offset calculation
3909 - sync after each recovered transaction, so we always make progress
3910 - never, not always, ERESTART requests without transnos
3911 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3913 - skip replay if the MDS says that the client is already connected
3914 - don't check for a recovery-enabled export to match lctl's UUID
3915 - don't INC_USE_COUNT for phantom exports
3916 - don't crash when cleaning up phantom exports (567)
3917 - don't double-finish or set replay data for errored mdc_open requests
3918 - abort requests when they time out, so we don't get old replies
3919 - send/receive replies for AST messages again
3920 - if the client says that it doesn't have the lock, cancel it on the
3922 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3923 mark it as destroyed, it will all work out when decref is called
3924 - fix module use counts (22, 581)
3926 - ASTs now expect a reply (server cancels lock on error reply)
3928 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3931 - fix many simultaneous client startup (392)
3932 - fix dentry->d_it clobbering
3933 - credentials weren't being shipped for readdir/getattr operations
3934 - remove invalid assertions triggered during some concurrent MD
3936 - proper Lustre versions added (336, 389)
3937 - fix memory leak for create error case (398)
3938 - fix LOV locking bug that would get cli/srv out of sync
3939 - fix echo client over LOV (409)
3940 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3941 - fix double-O_EXCL intent crash (424)
3942 - avoid sending multiple lock CANCELs (352)
3944 - MDS can do multi-client recovery (modulo bugs in new code)
3946 - many updates, edits, cleanups
3948 2002-11-18 Phil Schwan <phil@clusterfs.com>
3951 - fix null d_it dereference (346)
3952 - fix full OST/dbench hang (333)
3953 - fix permission problem with file removal (286)
3954 - fix removal of OSCs from LOV when they fail
3955 - fix NULL deref during bulk timeout (214)
3956 - fix problems related to multiple filesystems on one MDS (241)
3957 - fixed serious subtle metadata locking bugs
3958 - free locks on clients when inodes are removed due to memory
3960 - fix inode pointer in lock data (285)
3961 - partial support for multiple MDS on a single host (241)
3962 - data locks weren't cancelled at clear_inode time (290, 311)
3963 - intent locks could lead to unbounded lock growth (205)
3964 - added a maximum lock count, an LRU list, and a flusher
3965 - fix multiple rename (365)
3966 - properly abstracted the echo client
3967 - OSC locked 1 byte too many; fixed
3968 - rewrote brw callback code:
3969 - fixed recovery bugs related to LOVs (306)
3970 - fixed too-many-pages-in-one-write crash (191)
3971 - fixed (again) crash in sync_io_timeout (214)
3972 - probably fixed callback-related race (385)
3974 - Add capability to MDS protocol
3975 - LDLM cancellations and callbacks on different portals
3977 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3980 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3981 - defer open object destruction to close time (601981, 138)
3982 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3983 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3984 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3985 - don't call obd_setattr in ll_file_release for destroyed objects
3986 * protocol change to lustre_msg: move |version| and add |flags|
3987 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3988 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3989 * added replay of create, unlink, link and rename operations during
3990 MDS failover; recovery should be much more robust now
3991 * remove failed OSCs from LOVs (only lov_create uses this so far)
3992 * the lustre-HOWTO was brought (more) up to date (582544)
3994 2002-10-23 Phil Schwan <phil@clusterfs.com>
3997 - in-use dentries weren't being reused properly (617851)
3998 - prevent multiple LDLM setup (599178)
3999 - fix LOV size calculations for truncate (617853)
4000 - fix client handling of MDS intent errors (POSIX)
4001 - fix permission bug in lovstripe.c test (624321)
4002 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4003 - truncate past end of file could corrupt data
4004 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4005 - a race in recovery could return ETIMEDOUT to apps (623947)
4006 - building outside the source directory was fixed
4007 * the lustre-HOWTO was brought (more) up to date (582544)
4008 * major progress was made on recovery functionality
4010 2002-10-10 Phil Schwan <phil@clusterfs.com>
4013 - recovery deadlock fix
4014 - rm -rf causes LBUG fix (617817)
4015 - file open by multiple tasks fix (618962)
4016 - directory permissions bugs (602707 and 620007)
4017 - journal_stop fixed with locking (611313)
4018 - O_APPEND failures resolved (618273, perhaps 614459)
4019 - lconf PATH fix (619770)
4020 - IA64 build fix (621450)
4021 - RPC buffer sizes scale with amount of memory
4023 2002-10-01 Phil Schwan <phil@clusterfs.com>
4026 - locks would be cancelled without throwing away data pages,
4027 resulting in inconsistent data (605627)
4028 - inode attributes were not always being refreshed (605627, 612449)
4029 - lconf now continues to cleanup after lctl reports an error
4030 - MDS now enforces user permissions (602707)
4031 - lprocfs cleanup fixed, but not yet enabled (614157)
4032 - fixed infinite server hang, should a client not respond to an AST
4033 - avoid going into recovery if user calls readlink() with a buffer
4034 that's too small (613941)
4035 - AST RPCs no longer require replies (614867) -- this may be changed
4036 - don't crash server if client sends an IOV that's too big (611336)
4037 - fixed lock conversion deadlock (611892)
4038 - fixed the following of symlinks (614622)
4039 * recovery: the server can remove locks from a client that dies, other
4040 clients can make progress
4041 * more extN patch fixes
4042 * compile-time configurable ptlrpc buffer allocations
4044 - collaborative read cache document
4045 - Lustre Lite Performance CDR document-in-progress
4047 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4050 - fix typo in patch-2.4.18
4052 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4055 - clear ptlrpc request each time in handle_incoming_request()
4056 - unlink of files now destroys the object on the OST
4058 2002-09-19 Peter Braam <braam@clusterfs.com>
4060 * add hard link support
4061 * change obdfile creation method
4062 * kernel patch changed
4064 2002-09-19 Peter Braam <braam@clusterfs.com>
4067 - stack overflow bug in extN fixed
4069 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4071 * documentation updates
4072 - add man pages for config tools
4073 - update tests/README to describe testing with new config tools
4074 - finish metadata API descriptions
4075 * bug fixes and cleanups
4076 - statfs workaround for 16TB limit
4077 - LOV stripe allocation improved, can stripe on subset of OSTs
4078 - LOV file size/IO offset was wrong for files > 4GB in size
4079 - object EA data was being dropped, caused files to be unreadable
4080 - memory overflow with non-LOV OST caused memory corruption
4081 - fixed regression tests to work with new config tools, obdfilter
4082 - fixed bug when directory size became larger than 1 block
4083 - fixed bug (for single client case) when PWD was deleted
4084 - invalidate local directory pages when doing intent-based ops
4085 - avoid LDLM oops when lock callback contained bad data
4087 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4089 * documentation updates
4090 * bug fixes and cleanups
4091 - configuration tools
4094 - 64-bit compile warnings
4095 - 64-bit internal statfs data
4097 * test_brw on persistent OST devices
4099 * lprocfs (disabled)
4101 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4103 * documentation updates
4104 * bug fixes and cleanups
4105 * configuration tools
4107 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4110 * many small fixes to 0_5_4
4111 * io/network handling
4112 * thinkos in MDS operations
4114 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4117 * crucial basic fixes to 0.5.3
4118 * IOR, Iozone work over Elan
4121 2002-08-07 Phil Schwan <phil@clusterfs.com>
4122 * version 0_5_3, our first alpha
4123 * we use the new Portals iovs
4124 * documentation updates
4125 * bug fixes and cleanups
4126 * small changes in the DLM wire protocol
4128 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4129 * version 0_5_1 with some initial stability,
4130 * locking on MD and file I/O.
4131 * documentation updates
4132 * several bug fixes since 0.5.0
4133 * small changes in wire protocol
4135 2002-07-18 Phil Schwan <phil@clusterfs.com>
4137 * delivered as Lustre Light Alpha
4138 * fixed a crash after handling invalid MDS requests
4139 * fixed directory pages for architectures with non-4k pages sizes
4141 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4142 * release version v0_4_4
4143 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4145 * Fixes a number of interruption problems with OST operations.
4146 * Update documentation for portals header changes
4147 * Move all wire protocol structs/defines to lustre_idl.h
4148 * Fixes symlink length bug.
4149 * Add tcpdump to repository.
4151 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4152 * release version v0_4_3
4153 * Fixes statfs for inodes on extN.
4154 * Fixes bug in runtests which would delete /etc/hosts.
4155 * Use 64-bit object IDs wherever possible (not into VFS though)
4156 Remove ost_get_info, which is unused by lustre, and out of date.
4158 2002-07-03 Peter Braam <braam@clusterfs.com>
4159 * release version v0_4_2 Fixes a lookup error (type not passed)
4160 * move forward to head of Portals
4161 * move forward to latest Lustre kernel
4163 2002-06-25 Peter Braam <braam@clusterfs.com>
4164 * release version v0_4_1. Hopefully stable on single node use.