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'
404 Description: handle "serious error: objid * already exists" more gracefully
405 Details : If LAST_ID value on disk is smaller than the objects existing in
406 the O/0/d* directories, it indicates disk corruption and causes an
407 LBUG(). If the object is 0-length, then we should use the existing
408 object. This will help to avoid a full fsck in most cases.
410 --------------------------------------------------------------------------------
412 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
414 * Support for kernels:
415 2.6.5-7.283 (SLES 9),
416 2.6.9-55.EL (RHEL 4),
417 2.6.16.46-0.14 (SLES 10),
418 2.6.18.8 vanilla (kernel.org)
419 * Client support for unpatched kernels:
420 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
421 2.6.16 - 2.6.22 vanilla (kernel.org)
422 * Due to recently discovered recovery problems, we do not recommend
423 using patchless RHEL 4 clients with this or any earlier release.
424 * Recommended e2fsprogs version: 1.39.cfs8
425 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
426 * Starting with this release, the ldiskfs backing filesystem required
427 by Lustre is now in its own package, lustre-ldiskfs. This package
428 should be installed. It is versioned separately from Lustre and
429 may be released separately in future.
433 Description: block reactivating mgc import until all deactivates complete
434 Details : Fix race when failing back MDT/MGS to itself (testing)
436 Severity : enhancement
438 Description: add optional extra BUILD_VERSION info
439 Details : add a new environment variable (namely LUSTRE_VERS) which allows
440 to override the lustre version.
443 Frequency : 2.6.18 servers only
445 Description: ll_kern_mount() doesn't release the module reference
446 Details : The ldiskfs module reference count never drops down to 0
447 because ll_kern_mount() doesn't release the module reference.
452 Description: server LBUG when using old ost_num_threads parameter
453 Details : Accept the old ost_num_threads parameter but warn that it
454 is deprecated, and fix an off-by-one error that caused an LBUG.
459 Description: Transient SCSI error results in persistent IO issue
460 Details : iobuf->dr_error is not reinitialized to 0 between two
464 Frequency : sometimes when underlying device returns I/O errors
466 Description: OSTs not going read-only during write failures
467 Details : OSTs are not remounted read-only when the journal commit threads
468 get I/O errors because fsfilt_ext3 calls journal_start/stop()
469 instead of the ext3 wrappers.
473 Description: poor connect scaling with increasing client count
474 Details : Don't run filter_grant_sanity_check for more than 100 exports
475 to improve scaling for large numbers of clients.
478 Frequency : SLES10 only
480 Description: sanity-quota.sh quotacheck failed: rc = -22
481 Details : Quotas cannot be enabled on SLES10.
484 Frequency : liblustre clients only
486 Description: getdirentries does not give error when run on compute nodes
487 Details : getdirentries does not fail when the size specified as an argument
488 is too small to contain at least one entry
490 Severity : enhancement
492 Description: Add LNET router traceability for debug purposes
493 Details : If a checksum failure occurs with a router as part of the
494 IO path, the NID of the last router that forwarded the bulk data
495 is printed so it can be identified.
500 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
501 Details : Due to a race condition, liblustre clients were occasionally
504 Severity : enhancement
506 Description: lfs setstripe use optional parameters instead of postional
509 Severity : enhancement
511 Description: Nanosecond timestamp support for ldiskfs
512 Details : The on-disk ldiskfs filesystem has added support for nanosecond
513 resolution timestamps. There is not yet support for this at
514 the Lustre filesystem level.
517 Frequency : during server recovery
519 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
520 Details : request with rq_no_resend flag not awake l_wait_event if they get a
524 Frequency : nfs export on patchless client
526 Description: connectathon hang when test nfs export over patchless client
527 Details : Disconnected dentry cannot be found with lookup, so we do not need
528 to unhash it or make it invalid
531 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
532 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
533 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
538 Description: open req refcounting wrong on reconnect
539 Details : If reconnect happened between getting open reply from server and
540 call to mdc_set_replay_data in ll_file_open, we will schedule
541 replay for unreferenced request that we are about to free.
542 Subsequent close will crash in variety of ways.
543 Check that request is still eligible for replay in
544 mdc_set_replay_data().
549 Description: disable writes to filesystem when reading health_check file
550 Details : the default for reading the health_check proc file has changed
551 to NOT do a journal transaction and write to disk, because this
552 can cause reads of the /proc file to hang and block HA state
553 checking on a healthy but otherwise heavily loaded system. It
554 is possible to return to the previous behaviour during configure
555 with --enable-health-write.
557 Severity : enhancement
559 Description: 64-bit inode version
560 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
561 to the inode. This will be required for version-based recovery.
566 Description: MDS fails to start if a duplicate client export is detected
567 Details : in some rare cases it was possible for a client to connect to
568 an MDS multiple times. Upon recovery the MDS would detect this
569 and fail during startup. Handle this more gracefully.
571 Severity : enhancement
573 Description: Add -o localflock option to simulate old noflock
575 Details : This will achieve local-only flock/fcntl locks
581 Description: log_commit_thread vs filter_destroy race leads to crash
582 Details : Take import reference before releasing llog record semaphore
587 Description: Wrong request locking in request set processing
588 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
589 imp_delayed_list, in this place should be used imp_lock.
592 Frequency : when reconnection
594 Description: Grant Leak when osc reconnect to OST
595 Details : When osc reconnect ost, OST(filter) should check whether it
596 should grant more space to client by comparing fed_grant and
597 cl_avail_grant, and return the granted space to client instead
598 of "new granted" space, because client will call osc_init_grant
599 to update the client grant space info.
602 Frequency : when client reconnect to OST
604 Description: Grant Leak when osc do resend and replay bulk write
605 Details : When osc reconnect to OST, OST(filter)should clear grant info of
606 bulk write request, because the grant info while be sync between
607 OSC and OST when reconnect, and we should ignore the grant info
608 these of resend/replay write req.
613 Description: Grant space more than avaiable left space sometimes.
614 Details : When then OST is about to be full, if two bulk writing from
615 different clients came to OST. Accord the avaliable space of the
616 OST, the first req should be permitted, and the second one
617 should be denied by ENOSPC. But if the seconde arrived before
618 the first one is commited. The OST might wrongly permit second
619 writing, which will cause grant space > avaiable space.
622 Frequency : when client is evicted
624 Description: Grant might be wrongly erased when osc is evicted by OST
625 Details : when the import is evicted by server, it will fork another
626 thread ptlrpc_invalidate_import_thread to invalidate the
627 import, where the grant will be set to 0. While the original
628 thread will update the grant it got when connecting. So if
629 the former happened latter, the grant will be wrongly errased
630 because of this race.
635 Description: Checking Stale with correct fid
636 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
637 to check whether it is stale, because sometimes, we want the
638 enqueue happened anyway, and op_data.fid2 will not be initialized.
640 Severity : enhancement
642 Description: update patchless client
643 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
646 Frequency : only with 2.4 kernel
648 Description: random memory corruption
649 Details : size of struct ll_inode_info is to big for union inode.u and this
650 can be cause of random memory corruption.
655 Description: Memory leak in recovery
656 Details : Lov_mds_md was not free in an error handler in mds_create_object.
657 It should also check obd_fail before fsfilt_start, otherwise if
658 fsfilt_start return -EROFS,(failover mds during mds recovery).
659 then the req will return with repmsg->transno = 0 and rc = EROFS.
660 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
661 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
662 be freed no matter whether fsfilt_commit success or not.
665 Frequency : only with huge count clients
667 Description: Prevents from taking the superblock lock in llap_from_page for
669 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
670 for llap_from_page call in ll_removepage prevents from taking the
671 superblock lock for a soon died page.
676 Description: Not check open intent error before release open handle
677 Details : in some rare cases, the open intent error is not checked before
678 release open handle, which may cause
679 ASSERTION(open_req->rq_transno != 0), because it tries to release
680 the failed open handle.
685 Description: Set cat log bitmap only after create log success.
686 Details : in some rare cases, the cat log bitmap is set too early. and it
687 should be set only after create log success.
691 Description: Accessing a block bevice can re-enable I/O when Lustre is
692 tearing down a device.
693 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
698 Description: service threads may hog cpus when there are a lot of requests
700 Details : Insert cond_resched to give other threads a chance to use some of
706 Description: the cat log was not initialized in recovery
707 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
708 unlink log on mds will not be initialized until mds post
709 recovery. And also in mds post recovery, the unlink log will
710 initialization will be done asynchronausly, so there will be race
711 between add unlink log and unlink log initialization.
715 Description: brw_stats were being printed incorrectly
716 Details : brw_stats were being printed as log2 but all of them were not
717 recorded as log2. Also remove some code duplication arising from
718 filter_tally_{read,write}.
722 Frequency : rare, only in recovery.
723 Description: ASSERTION(req->rq_type != LI_POISON) failed
724 Details : imp_lock should be held while iterating over imp_sending_list for
725 prevent destroy request after get timeout in ptlrpc_queue_wait.
729 Description: replay-single.sh test 52 fails
730 Details : A lock's skiplist need to be cleanup when it being unlinked
731 from its resource list.
735 Description: Short directio read returns full requested size rather than
737 Details : Direct I/O operations should return actual amount of bytes
738 transferred rather than requested size.
740 Severity : enhancement
742 Description: metadata RPC reduction (e.g. for rm performance)
743 Details : decrease the amount of synchronous RPC between clients and servers
744 by canceling conflicing lock before the operation on the client side
745 and packing thier handles into the main operation RPC to server.
747 Severity : enhancement
749 Description: Async OSC create to avoid the blocking unnecessarily.
750 Details : If a OST has no remain object, system will block on the creating
751 when need to create a new object on this OST. Now, ways use
752 pre-created objects when available, instead of blocking on an
753 empty osc while others are not empty. If we must block, we block
754 for the shortest possible period of time.
758 Description: improve handling recoverable errors
759 Details : if request processig with error which can be recoverable on server
760 request should be resend, otherwise page released from cache and
763 Severity : enhancement
765 Description: refine locking for avoid write wrong info into lov_objid file
766 Details : fix possible races with add new target and write/update data in
769 --------------------------------------------------------------------------------
771 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
776 Frequency : on some architectures
778 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
779 Details : Uninitialized flags sometimes cause configuration commands to
783 Frequency : patchless clients only
785 Description: missing __iget() symbol export
786 Details : The __iget() symbol export is missing. To avoid the need for
787 this on patchless clients the deathrow inode reaper is turned
788 off, and we depend on the VM to clean up old inodes. This
789 dependency was during via the fix for bug 12181.
791 --------------------------------------------------------------------------------
793 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
795 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
796 INTEROPERATE with older versions automatically. In many cases a
797 special upgrade step is needed. Please read the
798 user documentation before upgrading any part of a live system.
799 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
800 version will not interoperate with 1.6 betas before beta5 (1.5.95).
801 * WARNING: Lustre configuration and startup changes are required with
802 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
808 Severity : enhancement
810 Description: MountConf
811 Details : Lustre configuration is now managed via mkfs and mount
812 commands instead of lmc and lconf. New obd types (MGS, MGC)
813 are added for dynamic configuration management. See
814 https://mail.clusterfs.com/wikis/lustre/MountConf for
817 Severity : enhancement
819 Description: dynamic OST addition
820 Details : OSTs can now be added to a live filesystem
822 Severity : enhancement
824 Description: startup order invariance
825 Details : MDTs and OSTs can be started in any order. Clients only
826 require the MDT to complete startup.
828 Severity : enhancement
830 Description: parallel, asynchronous orphan cleanup
831 Details : orphan cleanup is now performed in separate threads for each
832 OST, allowing parallel non-blocking operation.
834 Severity : enhancement
836 Description: optimized stripe assignment
837 Details : stripe assignments are now made based on ost space available,
838 ost previous usage, and OSS previous usage, in order to try
839 to optimize storage space and networking resources.
841 Severity : enhancement
843 Description: Permanently set tunables
844 Details : All writable /proc/fs/lustre tunables can now be permanently
845 set on a per-server basis, at mkfs time or on a live
848 Severity : enhancement
850 Description: Lustre message v2
851 Details : Add lustre message format v2.
853 Severity : enhancement
855 Description: client OST exclusion list
856 Details : Clients can be started with a list of OSTs that should be
857 declared "inactive" for known non-responsive OSTs.
861 Description: SPEC SFS validation failure on NFS v2 over lustre.
862 Details : Changes the blocksize for regular files to be 2x RPC size,
863 and not depend on stripe size.
865 Severity : enhancement
867 Description: Multiple MD RPCs in flight.
868 Details : Further unserialise some read-only MDS RPCs - learn about intents.
869 To avoid overly-overloading MDS, introduce a limit on number of
870 MDS RPCs in flight for a single client and add /proc controls
871 to adjust this limit.
873 Severity : enhancement
875 Description: client read/write statistics
876 Details : Add client read/write call usage stats for performance
877 analysis of user processes.
878 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
879 file access. extents_stats shows chunk size distribution.
880 extents_stats_per_process show chunk size distribution per
883 Severity : enhancement
885 Description: mds statistics
886 Details : Add detailed mds operations statistics in
887 /proc/fs/lustre/mds/*/stats.
891 Description: Failure of copying files with lustre special EAs.
892 Details : Client side always return success for setxattr call for lustre
893 special xattr (currently only "trusted.lov").
898 Description: Refcount LNET uuids
899 Details : The global LNET uuid list grew linearly with every startup;
900 refcount repeated list entries instead of always adding to
903 Severity : enhancement
905 Description: Dynamic service threads
906 Details : Within a small range, start extra service threads
907 automatically when the request queue builds up.
910 Frequency : mixed-endian client/server environments
912 Description: mixed-endian crashes
913 Details : The new msg_v2 system had some failures in mixed-endian
916 Severity : enhancement
918 Description: Easy OST removal
919 Details : OSTs can be permanently deactivated with e.g. 'lctl
920 conf_param lustre-OST0001.osc.active=0'
922 Severity : enhancement
924 Description: MGS proc entries
925 Details : Added basic proc entries for the MGS showing what filesystems
928 Severity : enhancement
930 Description: provide MGS failover
931 Details : Added config lock reacquisition after MGS server failover.
933 Severity : enhancement
935 Description: add Linux 2.4 support
936 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
940 Description: a large application tries to do I/O to the same resource and dies
942 Details : Check the req->rq_arrival time after the call to
943 ost_brw_lock_get(), but before we do anything about
944 processing it & sending the BULK transfer request. This
945 should help move old stale pending locks off the queue as
946 quickly as obd_timeout.
949 Frequency : when an incorrect nid is specified during startup
951 Description: ptlrpc connect to non-existant node causes kernel crash
952 Details : LNET can't be re-entered from an event callback, which
953 happened when we expire a message after the export has been
954 cleaned up. Instead, hand the zombie cleanup off to another
957 Severity : enhancement
959 Description: plain/inodebits lock performance improvement
960 Details : Grouping plain/inodebits in granted list by their request modes
961 and bits policy, thus improving the performance of search through
965 Frequency : only if OST filesystem is corrupted
967 Description: client incorrectly hits assertion in ptlrpc_replay_req()
968 Details : for a short time RPCs with bulk IO are in the replay list,
969 but replay of bulk IOs is unimplemented. If the OST filesystem
970 is corrupted due to disk cache incoherency and then replay is
971 started it is possible to trip an assertion. Avoid putting
972 committed RPCs into the replay list at all to avoid this issue.
975 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
977 Description: System hang on startup
978 Details : This bug allowed the liblustre (e.g. catamount) client to
979 return to the app before handling all startup RPCs. This
980 could leave the node unresponsive to lustre network traffic
981 and manifested as a server ptllnd timeout.
983 Severity : enhancement
985 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
986 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
987 boolean to control whether to print peer debug info when a
988 client's RPC times out.
991 Frequency : only for kernels with patches from Lustre below 1.4.3
993 Description: Remove old rdonly API
994 Details : Remove old rdonly API which unused from at least lustre 1.4.3
997 Frequency : only for devices with external journals
999 Description: Set external device read-only also
1000 Details : During a commanded failover stop, we set the disk device
1001 read-only while the server shuts down. We now also set any
1002 external journal device read-only at the same time.
1005 Frequency : when upgrading from 1.4 while trying to change parameters
1007 Description: The wrong (new) MDC name was used when setting parameters for
1008 upgraded MDT's. Also allows changing of OSC (and MDC)
1009 parameters if --writeconf is specified at tunefs upgrade time.
1012 Frequency : when setting specific ost indicies
1014 Description: QOS code breaks on skipped indicies
1015 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1016 created with --index need not be sequential.
1018 Severity : enhancement
1020 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1021 Details : The uninit_groups feature works in conjunction with the kernel
1022 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1023 up the pass1 processing of e2fsck. This is a read-only feature
1024 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1025 mount filesystems that have had this feature enabled.
1027 Severity : enhancement
1029 Description: Improve multi-block allocation algorithm to avoid fragmentation
1030 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1031 allocation locality and avoid filesystem fragmentation.
1033 ------------------------------------------------------------------------------
1035 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1037 * Support for kernels:
1038 2.4.21-47.0.1.EL (RHEL 3)
1039 2.6.5-7.283 (SLES 9)
1040 2.6.9-42.0.10.EL (RHEL 4)
1041 2.6.12.6 vanilla (kernel.org)
1042 2.6.16.27-0.9 (SLES 10)
1043 * Recommended e2fsprogs version: 1.39.cfs5
1045 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1049 Frequency : occasional, depends on client load and configuration
1050 Bugzilla : 12181, 12203
1051 Description: data loss for recently-modified files
1053 Details : In some cases it is possible that recently written or created
1054 files may not be written to disk in a timely manner (this should
1055 normally be within 30s unless client IO load is very high).
1056 The problem appears as zero-length files or files that are a
1057 multiple of 1MB in size after a client crash or client eviction
1058 that are missing data at the end of the file.
1060 This problem is more likely to be hit on clients where files are
1061 repeatedly created and unlinked in the same directory, clients
1062 have a large amount of RAM, have many CPUs, the filesystem has
1063 many OSTs, the clients are rebooted frequently, and/or the files
1064 are not accessed by other nodes after being written.
1066 The presence of the problem can be detected by looking at
1067 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1068 smaller than the second (nr_unused) then dirty files will not
1069 be flushed automatically to disk. "sync; sleep 10" should be
1070 run several times on the node before unmounting it to update
1071 Lustre (this is also safe to run on nodes without this problem).
1073 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1074 that can cause this same problem, so customers using that kernel
1075 also need to update the kernel in addition to Lustre. In order
1076 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1078 It is normal that files written just before a client crash (less
1079 than 30s) may not yet have been flushed to disk, even for local
1083 Frequency : frequent on thin XT3 nodes
1085 Description: UUID collision on thin XT3 Linux nodes
1086 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1087 randomly, since we relied on an insufficiently-seeded PRNG.
1092 Description: OSS hangs after "All ost request buffers busy"
1093 Details : A deadlock between quota and journal operations caused OSS
1094 hangs after printing "All ost request buffers busy."
1097 Frequency : always on liblustre builds
1099 Description: Cleanup compiler warnings on liblustre
1102 Frequency : always on liblustre builds on XT3
1104 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1108 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1109 > 32000 subdirectories
1110 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1111 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1112 subdirectory count crosses 32000. This will aid e2fsck to
1113 correctly handle more than 32000 subdirectories.
1116 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1118 Description: System hang on startup
1119 Details : This bug allowed the liblustre (e.g. catamount) client to
1120 return to the app before handling all startup RPCs. This
1121 could leave the node unresponsive to lustre network traffic
1122 and manifested as a server ptllnd timeout.
1124 Severity : enhancement
1126 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1127 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1128 boolean to control whether to print peer debug info when a
1129 client's RPC times out.
1134 Description: make O_SYNC working on 2.6 kernels
1135 Details : 2.6 kernels use different method for mark pages for write,
1136 so need add a code to lustre for O_SYNC work.
1141 Description: Failure to close file and release space on NFS
1142 Details : Put inode details into lock acquired in ll_intent_file_open.
1143 Use mdc_intent_lock in ll_intent_open to properly
1144 detect all kind of errors unhandled by mdc_enqueue.
1149 Description: proc file read during shutdown sometimes raced obd removal,
1151 Details : Add lock to prevent obd access after proc file removal.
1154 Frequency : Only for files larger than 4GB on 32-bit clients.
1156 Description: improperly doing page alignment of locks
1157 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1158 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1163 Description: RPCs being resent when they shouldn't be
1164 Details : Some RPCs that should not be resent are being resent. This
1165 can cause inconsistencies in the RPC state machine. Do not
1166 resend such requests.
1169 Frequency : rare, only with NFS export
1171 Description: Crash on NFS re-export node
1172 Details : under very unusual load conditions an assertion is hit in
1173 ll_intent_file_open()
1176 Frequency : only if OST filesystem is corrupted
1178 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1179 Details : for a short time RPCs with bulk IO are in the replay list,
1180 but replay of bulk IOs is unimplemented. If the OST filesystem
1181 is corrupted due to disk cache incoherency and then replay is
1182 started it is possible to trip an assertion. Avoid putting
1183 committed RPCs into the replay list at all to avoid this issue.
1188 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1189 Details : Large single O_DIRECT read and write calls can fail to allocate
1190 a sufficiently large buffer to process the request. In case of
1191 allocation failure the allocation is retried with a smaller
1192 buffer and broken into smaller requests.
1194 Severity : enhancement
1196 Description: Add -o localflock option to simulate old noflock behaviour.
1197 Details : This will achieve local-only flock/fcntl locks coherentness.
1202 Description: versioning check is incomplete
1203 Details : Checking the version difference of client vs. server, report
1204 error if the gap is too big.
1209 Description: add support PG_writeback bit
1210 Details : add support for PG_writeback bit for Lustre, for more carefull
1211 work with page cache in 2.6 kernel. This also fix some deadlocks
1212 and remove hack for work O_SYNC with 2.6 kernel.
1214 Severity : enhancement
1216 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1217 Details : The uninit_groups feature works in conjunction with the kernel
1218 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1219 up the pass1 processing of e2fsck. This is a read-only feature
1220 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1221 mount filesystems that have had this feature enabled.
1223 Severity : enhancement
1225 Description: Improve multi-block allocation algorithm to avoid fragmentation
1226 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1227 allocation locality and avoid filesystem fragmentation.
1229 ------------------------------------------------------------------------------
1231 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1233 * Support for kernels:
1234 2.6.9-42.0.3.EL (RHEL 4)
1235 2.6.5-7.276 (SLES 9)
1236 2.4.21-47.0.1.EL (RHEL 3)
1237 2.6.12.6 vanilla (kernel.org)
1238 2.6.16.21-0.8 (SLES10)
1239 * Recommended e2fsprogs version: 1.39.cfs2-0
1241 * The backwards-compatible /proc/sys/portals symlink has been removed
1242 in this release. Before upgrading, please ensure that you change
1243 any configuration scripts or /etc/sysctl.conf files that access
1244 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1245 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1246 in advance of the upgrade on any system running Lustre 1.4.6 or
1247 newer, since /proc/sys/lnet was added in that version.
1248 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1252 Frequency : only when quota is used
1254 Description: avoid scanning export list for quota master
1255 Details : Change the algorithms to avoid scanning export list in order
1256 to improve the efficiency.
1259 Frequency : MDS failover only, very rarely
1261 Description: "went back in time" messages on mds failover
1262 Details : The greatest transno may be lost when the current operation
1263 finishes with an error (transno==0) and the client's last_rcvd
1264 record is over-written. Save the greatest transno in the
1265 mds_last_transno for this case.
1268 Frequency : always for specific kernels and striping counts
1270 Description: client may get "Matching packet too big" without ACL support
1271 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1272 when trying to access files in certain configurations. The
1273 clients should in fact be denied when mounting because they do
1274 not understand ACLs.
1277 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1279 Description: many clients connecting with IO in progress causes connect timeouts
1280 Details : Avoid synchronous journal commits to avoid delays caused by many
1281 clients connecting/disconnecting when bulk IO is in progress.
1282 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1283 OST_IO_PORTAL to avoid delays behind potentially many pending
1287 Frequency : occasionally with multiple writers to a single file
1289 Description: shared writes to file may result in wrong size reported by stat()
1290 Details : Allow growing of kms when extent lock is cancelled
1293 Frequency : always with random mmap IO to multi-striped file
1295 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1296 Details : Only if the hole is at the end of OST object so that kms is too
1297 small. Fix is to increase kms accordingly in ll_nopage.
1300 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1302 Description: writes to a missing object would leak memory on the OST
1303 Details : If there is an inconsistency between the MDS and OST filesystems,
1304 such that the MDS references an object that doesn't exist, writes
1305 to that object will leak memory due to incorrect cleanup in the
1306 error handling path, eventually running out of memory on the OST.
1311 Description: Creating too long symlink causes lustre errors
1312 Details : Check symlink and name lengths before sending requests to MDS.
1315 Frequency : only if flock is enabled (not on by default)
1317 Description: posix locks not released on fd closure on 2.6.9+
1318 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1319 kernels, this caused such locks not to be released on fd close and
1320 then assertions on fs unmount about still used locks.
1323 Frequency : MDS failover only, very rarely
1325 Description: clients may get ASSERTION(granted_lock != NULL)
1326 Details : When request was taking a long time, and a client was resending
1327 a getattr by name lock request. The were multiple lock requests
1328 with the same client lock handle and
1329 mds_getattr_name->fixup_handle_for_resent_request found one of the
1330 lock handles but later failed with ASSERTION(granted_lock != NULL).
1335 Description: handle->h_buffer_credits > 0, assertion failure
1336 Details : h_buffer_credits is zero after truncate, causing assertion
1337 failure. This patch extends the transaction or creates a new
1341 Frequency : NFS re-export or patchless client
1342 Bugzilla : 11179, 10796
1343 Description: Crash on NFS re-export node (__d_move)
1344 Details : We do not want to hash the dentry if we don't have a lock.
1345 But if this dentry is later used in d_move, we'd hit uninitialised
1346 list head d_hash, so we just do this to init d_hash field but
1347 leave dentry unhashed.
1350 Frequency : NFS re-export or patchless client
1352 Description: NFS exports has problem with symbolic link
1353 Details : lustre client didn't properly install dentry when re-exported
1354 to NFS or running patchless client.
1357 Frequency : NFS re-export or patchless client
1359 Description: Various nfs/patchless fixes.
1360 Details : fixes reuse disconected alias for lookup process - this fixes
1361 warning "find_exported_dentry: npd != pd",
1362 fix permission error with open files at nfs.
1363 fix apply umask when do revalidate.
1366 Frequency : occasional
1368 Description: Crash on NFS re-export node
1369 Details : calling clear_page() on the wrong pointer triggered oops in
1370 generic_mapping_read().
1373 Frequency : rarely, using O_DIRECT IO
1375 Description: unaligned directio crashes client with LASSERT
1376 Details : check for unaligned buffers before trying any requests.
1379 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1381 Description: stale data returned from RAID cache
1382 Details : If only a small amount of IO is done to the RAID device before
1383 reading it again it is possible to get stale data from the RAID
1384 cache instead of reading it from disk.
1387 Frequency : always for sles10 kernel
1389 Description: sles10 support
1390 Details : ll_follow_link: compile fixes and using of nd_set_link
1391 under newer kernels.
1394 Frequency : depends on arch, kernel and compiler version, always on sles10
1397 Description: recursive or deep enough symlinks cause stack overflow
1398 Details : getting rid of large stack-allocated variable in
1402 Frequency : depends on hardware
1404 Description: lustre write performance loss in the SLES10 kernel
1405 Details : the performance loss is caused by using of write barriers in the
1406 ext3 code. The SLES10 kernel turns barrier support on by
1407 default. The fix is to undo that change for ldiskfs.
1409 ------------------------------------------------------------------------------
1411 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1413 * Support for kernels:
1414 2.6.9-42.0.3EL (RHEL 4)
1415 2.6.5-7.276 (SLES 9)
1416 2.4.21-47.0.1.EL (RHEL 3)
1417 2.6.12.6 vanilla (kernel.org)
1421 Frequency : quota enabled and large files being deleted
1423 Description: releasing more than 4GB of quota at once hangs OST
1424 Details : If a user deletes more than 4GB of files on a single OST it
1425 will cause the OST to spin in an infinite loop. Release
1426 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1431 Description: statfs data retrieved from /proc may be stale or zero
1432 Details : When reading per-device statfs data from /proc, in the
1433 {kbytes,files}_{total,free,avail} files, it may appear
1434 as zero or be out of date.
1437 Frequency : systems with MD RAID1 external journal devices
1439 Description: lconf's call to blkid is confused by RAID1 journal devices
1440 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1441 of returning all block devices that match the journal UUID.
1444 Frequency : always, for aggregate stripe size over 4GB
1446 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1447 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1448 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1451 Frequency : always if "lfs find" used on a local file/directory
1453 Description: "lfs find" segfaults if used on a local file/directory
1454 Details : The case where a directory component was not specified wasn't
1455 handled correctly. Handle this properly.
1458 Frequency : always on ppc64
1460 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1461 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1462 perform endianness conversion.
1465 Frequency : rarely (truncate to non-zero file size after write under load)
1466 Bugzilla : 10730, 10687
1467 Description: Files padded with zeros to next 4K multiple
1468 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1469 it is possible that files that are truncated to a non-zero size
1470 immediately after being written are filled with zero bytes beyond
1471 the truncated size. No file data is lost.
1473 Severity : enhancement
1475 Description: Allow recovery/failover for liblustre clients.
1476 Details : liblustre clients were unaware of failover configurations until
1479 Severity : enhancement
1481 Description: user file locks should fail when not mounting with flock option
1482 Details : Set up an error-returning stub in ll_file_operations.lock field
1483 to prevent incorrect behaviour when client is mounted without
1484 flock option. Also, set up properly f_op->flock field for
1488 Frequency : always on ia64
1490 Description: "lfs df" loops on printing out MDS statfs information
1491 Details : The obd_ioctl_data was not initialized and in some systems
1492 this caused a failure during the ioctl that did not return
1493 an error. Initialize the struct and return an error on failure.
1496 Frequency : SLES 9 only
1498 Description: Error of copying files with lustre special EAs as root
1499 Details : Client side always return success for setxattr call for lustre
1500 special xattr (currently only "trusted.lov").
1503 Frequency : rarely on clusters with both ia64+i386 clients
1505 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1506 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1507 clients are doing concurrent writes to the same file it is
1508 possible that the ia64 clients may LASSERT because the OST
1509 extent locks are not PAGE_SIZE aligned. Ensure that grown
1510 locks are always aligned on the request boundary.
1513 Frequency : specific use, occasional
1515 Description: Overwriting in use executable truncates on-disk binary image
1516 Details : If one node attempts to overwrite an executable in use by
1517 another node, we now correctly return ETXTBSY instead of
1518 truncating the file.
1520 Severity : enhancement
1522 Description: Async OSC create to avoid the blocking unnecessarily.
1523 Details : If a OST has no remain object, system will block on the creating
1524 when need to create a new object on this OST. Now, ways use
1525 pre-created objects when available, instead of blocking on an
1526 empty osc while others are not empty. If we must block, we block
1527 for the shortest possible period of time.
1532 Description: chmod on Lustre root is propagated to other clients
1533 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1534 invalid by the follow_mount time.
1537 Frequency : liblustre clients only
1539 Description: Race in 'instant cancel' lock handling could lead to such locks
1540 never to be granted in case of SMP MDS
1541 Details : Do not destroy not yet granted but cbpending locks in
1545 Frequency : replay/resend of open
1547 Description: non null lock assetion failure in mds_intent_policy
1548 Details : Trying to replay/resend lockless open requests resulted in
1549 mds_open() returning 0 with no lock. Now it sets a flag if
1550 a lock is going to be returned.
1552 Severity : enhancement
1554 Description: Checksum enhancements
1555 Details : New checksum enhancements allow for resending RPCs that failed
1558 Severity : enhancement
1560 Description: Tunables on number of dirty pages in cacche
1561 Details : Allow to set limit on number of dirty pages cached.
1566 Description: client crash on unmount - lock still has references
1567 Details : In some error handling cases it was possible to leak a lock
1568 reference on a client while accessing a file. This was not
1569 harmful to the client during operation, but would cause the
1570 client to crash when the filesystem is unmounted.
1573 Frequency : specific case, rare
1575 Description: ETXTBSY on mds though file not in use by client
1576 Details : ETXTBSY is no longer incorrectly returned when attempting to
1577 chmod or chown a directory that the user previously tried to
1578 execute or a currently-executing binary.
1581 Frequency : extremely rare except on liblustre-based clients
1583 Description: Lustre space not freed when files are deleted
1584 Details : Clean up open-unlinked files after client eviction. Previously
1585 the unlink was skipped and the files remained as orphans.
1590 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1591 Details : In some cases it was possible to send a blocking callback to
1592 a client doing a glimpse, even though that client didn't get
1593 a lock granted. When the glimpse lock is cancelled on the OST
1594 the freed lock is left on the waiting list and corrupted the list.
1597 Frequency : all core dumps
1599 Description: Broke core dumps to lustre
1600 Details : Negative dentry may be unhashed if parent does not have UPDATE
1601 lock, but some callers, e.g. do_coredump, expect dentry to be
1602 hashed after successful create, hash it in ll_create_it.
1604 ------------------------------------------------------------------------------
1606 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1608 * Support for kernels:
1609 2.6.9-42.0.2.EL (RHEL 4)
1610 2.6.5-7.276 (SLES 9)
1611 2.4.21-40.EL (RHEL 3)
1612 2.6.12.6 vanilla (kernel.org)
1616 Frequency : always on RHEL 3
1618 Description: Number of open files grows over time
1619 Details : The number of open files grows over time, whether or not
1620 Lustre is started. This was due to a filp leak introduced
1621 by one of our kernel patches.
1623 ------------------------------------------------------------------------------
1625 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1627 * Support for kernels:
1628 2.6.9-42.EL (RHEL 4)
1629 2.6.5-7.276 (SLES 9)
1630 2.4.21-40.EL (RHEL 3)
1631 2.6.12.6 vanilla (kernel.org)
1636 Bugzilla : 5719, 9635, 9792, 9684
1637 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1638 Details : If a server is under heavy load and cannot reply to new
1639 connection requests before the client resends the (re)connect,
1640 the connection handling code can behave badly if two service
1641 threads are concurrently handing separate (re)connections from
1642 the same client. Add better locking to the connection handling
1643 code, and ensure that only a single connection will be processed
1644 for a given client UUID, even if the lock is dropped.
1646 Severity : enhancement
1648 Description: add TCP zero-copy support to kernel
1649 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1650 sends if the hardware supports scatter-gather and checksumming.
1651 This allows socklnd to do client-write and server-read more
1652 efficiently and reduce CPU utilization from skbuf copying.
1655 Frequency : only if NFS exporting from client
1657 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1658 Details : If setting attributes on a file created under NFS that had
1659 never been opened it would be possible to oops the client
1660 if the file had no objects.
1663 Frequency : always for liblustre
1665 Description: liblustre client does MDS+OSTs setattr RPC for each write
1666 Details : When doing a write from a liblustre client, the client
1667 incorrectly issued an RPC to the MDS and each OST the file was
1668 striped over in order to update the timestamps. When writing
1669 with small chunks and many clients this could overwhelm the MDS
1670 with RPCs. In all cases it would slow down the write because
1671 these RPCs are unnecessary.
1673 Severity : enhancement
1675 Description: allow number of MDS service threads to be changed at module load
1676 Details : It is now possible to change the number of MDS service threads
1677 running. Adding "options mds mds_num_threads={N}" to the MDS's
1678 /etc/modprobe.conf will set the number of threads for the next
1679 time Lustre is restarted (assuming the "mds" module is also
1680 reloaded at that time). The default number of threads will
1681 stay the same, 32 for most systems.
1686 Description: OST crash if filesystem is unformatted or corrupt
1687 Details : If an OST is started on a device that has never been formatted
1688 or if the filesystem is corrupt and cannot even mount then the
1689 error handling cleanup routines would dereference a NULL pointer.
1694 Description: NULL pointer deref in llap_from_page.
1695 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1696 incorrect) mapping if the page was truncated/reclaimed while it was
1697 searched for. Check for this condition and skip such pages when
1698 doing readahead. Introduce extra check to llap_from_page() to
1699 verify page->mapping->host is non-NULL (so page is not anonymous).
1702 Frequency : Sometimes when using sys_sendfile
1704 Description: "page not covered by a lock" warnings from ll_readpage
1705 Details : sendfile called ll_readpage without right page locks present.
1706 Now we introduced ll_file_sendfile that does necessary locking
1707 around call to generic_file_sendfile() much like we do in
1711 Frequency : with certain MDS communication failures at client mount time
1713 Description: NULL pointer deref after failed client mount
1714 Details : a client connection request may delayed by the network layer
1715 and not be sent until after the PTLRPC layer has timed out the
1716 request. If the client fails the mount immediately it will try
1717 to clean up before the network times out the request. Add a
1718 reference from the request import to the obd device and delay
1719 the cleanup until the network drops the request.
1722 Frequency : occasionally during client (re)connect
1724 Description: assertion failure during client (re)connect
1725 Details : processing a client connection request may be delayed by the
1726 client or server longer than the client connect timeout. This
1727 causes the client to resend the connection request. If the
1728 original connection request is replied in this interval, the
1729 client may trip an assertion failure in ptlrpc_connect_interpret()
1730 which thought it would be the only running connect process.
1733 Frequency : only with obd_echo servers and clients that are rebooted
1735 Description: kernel BUG accessing uninitialized data structure
1736 Details : When running an obd_echo server it did not start the ping_evictor
1737 thread, and when a client was evicted an uninitialized data
1738 structure was accessed. Start the ping_evictor in the RPC
1739 service startup instead of the OBD startup.
1741 Severity : enhancement
1742 Bugzilla : 10193 (patchless)
1743 Description: Remove dependency on various unexported kernel interfaces.
1744 Details : No longer need reparent_to_init, exit_mm, exit_files,
1745 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1748 Frequency : rare (only users of deprecated and unsupported LDAP config)
1750 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1751 Details : LDAP apparently contains 'lustreName' attributes instead of
1752 'name'. A simple remapping of the name is sufficient.
1755 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1757 Description: waiting_locks_callback trips kernel BUG if client is evicted
1758 Details : Running with the dump_on_timeout debug flag turned on makes
1759 it possible that the waiting_locks_callback() can try to dump
1760 the Lustre kernel debug logs from an interrupt handler. Defer
1761 this log dumping to the expired_lock_main() thread.
1763 Severity : enhancement
1765 Description: Support NFS exporting on 2.6 kernels.
1766 Details : Implement non-rawops metadata methods for NFS server to use without
1767 changing NFS server code.
1770 Frequency : very rare (synthetic metadata workload only)
1772 Description: two racing renames might cause an MDS thread to deadlock
1773 Details : Running the "racer" program may cause one MDS thread to rename
1774 a file from being the source of a rename to being the target of
1775 a rename at exactly the same time that another thread is doing
1776 so, and the second thread has already enqueued these locks after
1777 doing a lookup of the target and is trying to relock them in
1778 order. Ensure that we don't try to re-lock the same resource.
1781 Frequency : only very large systems with liblustre clients
1783 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1784 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1785 all OSTs in parallel. This allows the eviction of stale liblustre
1786 clients to proceed much faster than if they were done in series,
1787 and also offers similar improvements for other set_info RPCs.
1792 Description: excessive CPU usage during initial read phase on client
1793 Details : During the initial read phase on a client, it would agressively
1794 retry readahead on the file, consuming too much CPU and impacting
1795 performance (since 1.4.5.8). Improve the readahead algorithm
1796 to avoid this, and also improve some other common cases (read
1797 of small files in particular, where "small" is files smaller than
1798 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1803 Description: MDS crash when receiving packet with unknown intent.
1804 Details : Do not LBUG in unknown intent case, just return -EFAULT
1806 Severity : enhancement
1807 Bugzilla : 9293, 9385
1808 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1809 Details : Do not serialize getattr (non-intent version) and statfs.
1812 Frequency : occasional, when OST network is overloaded/intermittent
1814 Description: client evicted by OST after bulk IO timeout
1815 Details : If a client sends a bulk IO request (read or write) the OST
1816 may evict the client if it is unresposive to its data GET/PUT
1817 request. This is incorrect if the network is overloaded (takes
1818 too long to transfer the RPC data) or dropped the OST GET/PUT
1819 request. There is no need to evict the client at all, since
1820 the pinger and/or lock callbacks will handle this, and the
1821 client can restart the bulk request.
1824 Frequency : Always when mmapping file with no objects
1826 Description: client crashes when mmapping file with no objects
1827 Details : Check that we actually have objects in a file before doing any
1828 operations on objects in ll_vm_open, ll_vm_close and
1834 Description: Request leak when working with deleted CWD
1835 Details : Introduce advanced request refcount tracking for requests
1836 referenced from lustre intent.
1838 Severity : Enhancement
1840 Description: Cache open file handles on client.
1841 Details : MDS now will return special lock along with openhandle, if
1842 requested and client is allowed to hold openhandle, even if unused,
1843 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1844 closing files for every read/write openration.
1846 Severity : Enhancement
1848 Description: Cache open negative dentries on client when possible.
1849 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1850 negative dentries on lock revocation.
1855 Description: Remounting a client read-only wasn't possible with a zconf mount
1856 Details : It wasn't possible to remount a client read-only with llmount.
1858 Severity : enhancement
1859 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1860 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1861 MPICH is included to add Lustre-specific ADIO interfaces.
1862 This is based closely on the UFS ADIO layer and only differs
1863 in file creation, in order to allow the OST striping to be set.
1864 This is user-contributed code and not supported by CFS.
1869 Description: extended inode attributes (immutable, append-only) work improperly
1870 when 2.4 and 2.6 kernels are used on client/server or vice versa
1871 Details : Introduce kernel-independent values for these flags.
1873 Severity : enhancement
1876 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1877 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1878 granularity. Now, Fractional MB granularity is be supported,
1879 this is very useful for low memory system.
1881 Severity : enhancement
1883 Description: Getattr by fid
1884 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1885 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1886 have locks and avoids getattr rpc requests for it on every stat.
1889 Frequency : Always, for filesystems larger than 2TB
1891 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1892 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1893 filesystems larger than 2TB to be mounted with mballoc enabled.
1896 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1898 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1899 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1900 filesystems larger than 2TB is mounted then the kernel will
1901 silently corrupt the start of the filesystem. CONFIG_LBD is
1902 enabled for all CFS-supported kernels, but the possibility of
1903 this happening with a modified kernel config exists.
1905 Severity : enhancement
1907 Description: add client O_DIRECT support for 2.6 kernels
1908 Details : It is now possible to do O_DIRECT reads and writes to files
1909 in the Lustre client mountpoint on 2.6 kernel clients.
1911 Severity : enhancement
1913 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1914 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1915 parallel, not waiting for response from every OST before sending
1916 a rpc to the next OST.
1921 Description: setattr vs write race when updating file timestamps
1922 Details : Client processes that update a file timestamp into the past
1923 right after writing to the file (e.g. tar) it is possible that
1924 the updated file modification time can be reset to the current
1925 time due to a race between processing the setattr and write RPC.
1927 Severity : enhancement
1929 Description: Bring 'lfs find' closer in line with regular Linux find.
1930 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1931 -print0 options and obtains all the needed info through the lustre
1934 Severity : enhancement
1936 Description: support up to 1024 configured devices on one node
1937 Details : change obd_dev array from statically allocated to dynamically
1938 allocated structs as they are first used to reduce memory usage
1943 Description: Flush dirty partially truncated pages during truncate
1944 Details : Immediatelly flush partially truncated pages in filter_setattr,
1945 this way we completely avoid having any pages in page cache on OST
1946 and can retire ugly workarounds during writes to flush such pages.
1951 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1952 Details : protect inode from truncation within vfs_unlink() context
1953 just take a reference before calling vfs_unlink() and release it
1954 when parent's i_sem is free.
1959 Description: last_id value checked outside lock on OST caused LASSERT failure
1960 Details : If there were multiple MDS->OST object precreate requests in
1961 flight, it was possible that the OST's last object id was checked
1962 outside a lock and incorrectly tripped an assertion. Move checks
1963 inside locks, and discard old precreate requests.
1966 Frequency : always, if extents are used on OSTs
1968 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1969 Details : OSTs using the extents format would not zero the high 16 bits of
1970 the index physical block number. This is not a problem for any
1971 OST filesystems smaller than 16TB, and no kernels support ext3
1972 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1973 new/modified files) and can be fixed for existing filesystems
1974 with e2fsprogs-1.39-cfs1.
1979 Description: import connection selection may be incorrect if timer wraps
1980 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1981 import connections to be ignored if the 32-bit jiffies counter
1982 wraps. Use a 64-bit jiffies counter.
1985 Frequency : very large clusters immediately after boot
1987 Description: LNET request buffers exhausted under heavy short-term load
1988 Details : If a large number of client requests are generated on a service
1989 that has previously never seen so many requests it is possible
1990 that the request buffer growth cannot keep up with the spike in
1991 demand. Instead of dropping incoming requests, they are held in
1992 the LND until the RPC service can accept more requests.
1995 Frequency : Sometimes during replay
1997 Description: Assertion failure in ll_local_open after replay.
1998 Details : If replay happened on an open request reply before we were able
1999 to set replay handler, reply will become not swabbed tripping the
2000 assertion in ll_local_open. Now we set the handler right after
2001 recognising of open request
2004 Frequency : very rare
2006 Description: kernel reports "badness in vsnprintf"
2007 Details : Reading from the "recovery_status" /proc file in small chunks
2008 may cause a negative length in lprocfs_obd_rd_recovery_status()
2009 call to vsnprintf() (which is otherwise harmless). Exit early
2010 if there is no more space in the output buffer.
2012 Severity : enhancement
2014 Description: clear OBD RPC statistics by writing to them
2015 Details : It is now possible to clear the OBD RPC statistics by writing
2016 to the "stats" file.
2021 Description: Client mtime is not the same on different clients after utimes
2022 Details : In some cases, the client was using the utimes() syscall on
2023 a file cached on another node. The clients now validate the
2024 ctime from the MDS + OSTs to determine which one is right.
2029 Description: Inability to activate failout mode
2030 Details : lconf script incorrectly assumed that in pythong string's numeric
2031 value is used in comparisons.
2034 Frequency : always with multiple stripes per file
2036 Description: Inefficient object allocation for mutli-stripe files
2037 Details : When selecting which OSTs to stripe files over, for files with
2038 a stripe count that divides evenly into the number of OSTs,
2039 the MDS is always picking the same starting OST for each file.
2040 Return the OST selection heuristic to the original design.
2045 Description: mount failures may take full timeout to return an error
2046 Details : Under some heavy load conditions it is possible that a
2047 failed mount can wait for the full obd_timeout interval,
2048 possibly several minutes, before reporting an error.
2049 Instead return an error as soon as the status is known.
2051 Frequency : quota enabled and large files being deleted
2053 Description: releasing more than 4GB of quota at once hangs OST
2054 Details : If a user deletes more than 4GB of files on a single OST it
2055 will cause the OST to spin in an infinite loop. Release
2056 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2061 Description: statfs data retrieved from /proc may be stale or zero
2062 Details : When reading per-device statfs data from /proc, in the
2063 {kbytes,files}_{total,free,avail} files, it may appear
2064 as zero or be out of date.
2067 Frequency : systems with MD RAID1 external journal devices
2069 Description: lconf's call to blkid is confused by RAID1 journal devices
2070 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2071 of returning all block devices that match the journal UUID.
2074 Frequency : always, for aggregate stripe size over 4GB
2076 Description: assertion fails when trying to use 4GB stripe size
2077 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2078 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2081 Frequency : always on ppc64
2083 Description: the first write on an ext3 filesystem with mballoc got stuck
2084 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2085 perform endianness conversion.
2087 ------------------------------------------------------------------------------
2089 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2091 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2092 INTEROPERATE with older versions automatically. Please read the
2093 user documentation before upgrading any part of a live system.
2094 * WARNING: Lustre networking configuration changes are required with
2095 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2098 * Support for kernels:
2099 2.6.9-22.0.2.EL (RHEL 4)
2100 2.6.5-7.244 (SLES 9)
2101 2.6.12.6 vanilla (kernel.org)
2104 Severity : enhancement
2105 Bugzilla : 7981/8208
2106 Description: Introduced Lustre Networking (LNET)
2107 Details : LNET is new networking infrastructure for Lustre, it includes
2108 a reorganized network configuration mode (see the user
2109 documentation for full details) as well as support for routing
2110 between different network fabrics. Lustre Networking Devices
2111 (LNDs) for the supported network fabrics have also been
2112 created for this new infrastructure.
2114 Severity : enhancement
2115 Description: Introduced Access control lists
2116 Details : clients can set ACLs on files and directories in order to have
2117 more fine-grained permissions than the standard Unix UGO+RWX.
2118 The MDS must be started with the "-o acl" mount option.
2120 Severity : enhancement
2121 Description: Introduced filesystem quotas
2122 Details : Administrators may now establish per-user quotas on the
2125 Severity : enhancement
2127 Description: Configuration change for the XT3
2128 The PTLLND is now used to run Lustre over Portals on the XT3
2129 The configure option(s) --with-cray-portals are no longer used.
2130 Rather --with-portals=<path-to-portals-includes> is used to
2131 enable building on the XT3. In addition to enable XT3 specific
2132 features the option --enable-cray-xt3 must be used.
2137 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2138 Details : A race between allocating a new llcd and re-getting the llcd_lock
2139 allowed another thread to grab newly-allocated llcd.
2141 Severity : enhancement
2143 Description: 2.6 OST async journal commit and locking fix to improve performance
2144 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2145 2.6 kernels are now async as they already were in 2.4 kernels so
2146 that they can commit concurrently with the network bulk transfer.
2147 For block-allocated files the filter allocation semaphore is held
2148 to avoid filesystem fragmentation during allocation. BKL lock
2149 removed for 2.6 xattr operations where it is no longer needed.
2154 Description: lconf incorrectly determined whether two IP networks could talk
2155 Details : In some more complicated routing and multiple-network
2156 configurations, lconf will avoid trying to make a network
2157 connection to a disjoint part of the IP space. It was doing the
2158 math incorrectly for one set of cases.
2163 Description: Fix for potential infinite loop processing records in an llog.
2164 Details : If an llog record is corrupted/zeroed, it is possible to loop
2165 forever in llog_process(). Validate the llog record length
2166 and skip the remainder of the block on error.
2169 Frequency : occasional (liblustre only)
2171 Description: liblustre could not open files whose last component is a symlink
2172 Details : sysio_path_walk() would incorrectly pass the open intent to
2173 intermediate path components.
2176 Frequency : rare (liblustre only with non-standard tuning)
2177 Bugzilla : 7201 (7350)
2178 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2179 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2180 after it has been released, only do this for referenced locks
2182 Severity : enhancement
2184 Description: specifying an (invalid) directory default stripe_size of -1
2185 would reset the directory default striping
2186 Details : stripe_size -1 was used internally to signal directory stripe
2187 removal, now use "all default" to signal dir stripe removal
2188 as a directory striping of "all default" is not useful
2191 Frequency : common for large clusters running liblustre clients
2193 Description: doing an ls when liblustre clients are running is slow
2194 Details : sending a glimpse AST to a liblustre client waits for every AST
2195 to time out, as liblustre clients will not respond. Since they
2196 cannot cache data we refresh the OST lock LVB from disk instead.
2198 Severity : enhancement
2200 Description: doing an ls at the same time as file IO can be slow
2201 Details : enqueue and other "small" requests can be blocked behind many
2202 large IO requests. Create a new OST IO portal for non-IO
2203 requests so they can be processed faster.
2206 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2208 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2209 created file with mode 000
2210 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2211 allow an ftruncate() as a non-root user to a file with mode 000.
2212 Lustre now respects this flag to disable mode checks when
2213 truncating a file owned by the user
2216 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2219 Description: Revoking locks from clients that went dead or catatonic might take
2221 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2222 cancellation of such locks instant on servers without waiting for
2223 any reply from clients. Clients drops these locks when cancel
2224 notification from server is received without replying.
2227 Frequency : liblustre-only, when liblustre client dies or becomes busy
2229 Description: Doing ls on Linux clients can take a long time with active
2231 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2232 granting such locks to it in the first place if possible. Locks
2233 are taken by proxy on the OST during the read or write and
2234 dropped immediately afterward. Add connect flags handling, do
2235 not grant locks to liblustre clients for glimpse ASTs.
2237 Severity : enhancement
2239 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2240 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2241 sequential reads and behaves badly with applications doing random
2242 reads. Improve it by reading ahead at least read region, and
2243 avoiding excessive large RPC for small reads.
2245 Severity : enhancement
2247 Description: Creating more than 1000 files for a single job may cause a load
2248 imbalance on the OSTs if there are also a large number of OSTs.
2249 Details : qos_prep_create() uses an OST index reseed value that is an
2250 even multiple of the number of available OSTs so that if the
2251 reseed happens in the middle of the object allocation it will
2252 still utilize the OSTs as uniformly as possible.
2257 Description: OST or MDS may oops in ping_evictor_main()
2258 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2259 but doesn't restart at beginning of obd_exports_timed list
2262 Severity : enhancement
2264 Description: improve by-nid export eviction on the MDS and OST
2265 Details : allow multiple exports with the same NID to be evicted at one
2266 time without re-searching the exports list.
2269 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2271 Description: MDS may oops in groups_free()
2272 Details : in rare race conditions a newly allocated group_info struct is
2273 freed again, and this can be NULL. The 2.4 compatibility code
2274 for groups_free() checked for a NULL pointer, but 2.6 did not.
2277 Frequency : common for liblustre clients doing little filesystem IO
2278 Bugzilla : 9352, 7313
2279 Description: server may evict liblustre clients accessing contended locks
2280 Details : if a client is granted a lock or receives a completion AST
2281 with a blocking AST already set it would not reply to the AST
2282 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2285 Frequency : lfs setstripe, only systems with more than 160 OSTs
2287 Description: unable to set striping with a starting offset beyond OST 160
2288 Details : llapi_create_file() incorrectly limited the starting stripe
2289 index to the maximum single-file stripe count.
2292 Frequency : LDAP users only
2294 Description: lconf did not handle in-kernel recovery with LDAP properly
2295 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2297 Severity : enhancement
2299 Description: bind OST threads to NUMA nodes to improve performance
2300 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2301 node and do their allocations there to localize memory access
2303 Severity : enhancement
2305 Description: llmount can determine client NID directly from Myrinet (GM)
2306 Details : the client NID code from gmnalnid was moved directly into
2307 llmount, removing the need to use this or specifying the
2308 client NID explicitly when mounting GM clients with zeroconf
2311 Frequency : if client is started with down MDS
2313 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2314 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2315 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2316 ptlrpc_import_delayed_req() because we didn't check if the
2317 request was interrupted, and we also didn't break out of the
2318 event loop if there was no timeout
2323 Description: data loss during non-page-aligned writes to a single file from
2324 both multiple nodes and multiple threads on one node at same time
2325 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2326 inconsistency led to false short-reads, that were cached and later
2327 used by ->prepare_write() to fill in partially written page,
2328 leading to data loss.
2331 Frequency : always, if lconf --abort_recovery used
2333 Description: lconf --abort_recovery fails with 'Operation not supported'
2334 Details : lconf was attempting to abort recovery on the MDT device and not
2337 Severity : enhancement
2339 Description: remove cleanup logs
2340 Details : replace lconf-generated cleanup logs with lustre internal
2341 cleanup routines. Eliminates the need for client-cleanup and
2344 Severity : enhancement
2346 Description: add support for EAs (user and system) on lustre filesystems
2347 Details : it is now possible to store extended attributes in the Lustre
2348 client filesystem, and with the user_xattr mount option it
2349 is possible to allow users to store EAs on their files also
2351 Severity : enhancement
2353 Description: Add possibility (config option) to show minimal available OST free
2355 Details : When compiled with --enable-mindf configure option, statfs(2)
2356 (and so, df) will return least minimal free space available from
2357 all OSTs as amount of free space on FS, instead of summary of
2358 free spaces of all OSTs.
2360 Severity : enhancement
2362 Description: do not expand extent locks acquired on OST-side
2363 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2364 by server: they are not cached anyway.
2367 Frequency : when mmap is used/binaries executed from Lustre
2369 Description: Unmmap pages before throwing them away from read cache.
2370 Details : llap_shrink cache now attempts to unmap pages before discarding
2371 them (if unmapping failed - do not discard). SLES9 kernel has
2372 extra checks that trigger if this unmapping is not done first.
2377 Description: lconf didn't resolve symlinks before checking to see whether a
2378 given mountpoint was already in use
2381 Frequency : when migrating failover services
2382 Bugzilla : 6395, 9514
2383 Description: When migrating a subset of services from a node (e.g. failback
2384 from a failover service node) the remaining services would
2385 time out and evict clients.
2386 Details : lconf --force (implied by --failover) sets the global obd_timeout
2387 to 5 seconds in order to quickly disconnect, but this caused
2388 other RPCs to time out too quickly. Do not change the global
2389 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2391 Severity : enhancement
2392 Frequency : if MDS is started with down OST
2393 Bugzilla : 9439,5706
2394 Description: Allow startup/shutdown of an MDS without depending on the
2395 availability of the OSTs.
2396 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2397 Add appropriate locking and lov-osc refcounts for safe
2398 cleaning. Add osc abort_inflight calls in case the
2399 synchronize never started.
2402 Frequency : occasional (Cray XT3 only)
2404 Description: root not authorized to access files in CRAY_PORTALS environment
2405 Details : The client process capabilities were not honoured on the MDS in
2406 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2407 been accessed by an authorized user then root was able to access
2408 the file on the local client also. The root user capabilities
2409 are now allowed on the MDS, as this environment has secure UID.
2412 Frequency : occasional
2414 Description: ldiskfs "too long searching" message happens too often
2415 Details : A debugging message (otherwise harmless) prints too often on
2416 the OST console. This has been reduced to only happen when
2417 there are fragmentation problems on the filesystem.
2422 Description: Division by zero in statfs when all OSCs are inactive
2423 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2424 lov_statfs divides by value returned by lov_get_stripecnt().
2428 Bugzilla : 9489, 3273
2429 Description: First write from each client to each OST was only 4kB in size,
2430 to initialize client writeback cache, which caused sub-optimal
2431 RPCs and poor layout on disk for the first writen file.
2432 Details : Clients now request an initial cache grant at (re)connect time
2433 and so that they can start streaming writes to the cache right
2434 away and always do full-sized RPCs if there is enough data.
2435 If the OST is rebooted the client also re-establishes its grant
2436 so that client cached writes will be honoured under the grant.
2441 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2442 Details : Now I/O RPCs go to different portal number and (presumably) fast
2443 lock requests (and glimses) and other RPCs get their own service
2444 threads pool that should be able to service those RPCs
2447 Severity : enhancement
2449 Description: Ability to exchange lustre version between client and servers and
2450 issue warnings at client side if client is too old. Also for
2451 liblustre clients there is ability to refuse connection of too old
2453 Details : New 'version' field is added to connect data structure that is
2454 filled with version info. That info is later checked by server and
2458 Frequency : rare, liblustre only.
2459 Bugzilla : 9296, 9581
2460 Description: Two simultaneous writes from liblustre at offset within same page
2461 might proceed at the same time overwriting eachother with stale
2463 Details : I/O lock withing llu_file_prwv was released too early, before data
2464 actually was hitting the wire. Extended lock-holding time until
2465 server acknowledges receiving data.
2468 Frequency : extremely rare. Never observed in practice.
2470 Description: avoid generating lustre_handle cookie of 0.
2471 Details : class_handle_hash() generates handle cookies by incrementing
2472 global counter, and can hit 0 occasionaly (this is unlikely, but
2473 not impossible, because initial value of cookie counter is
2474 selected randonly). Value of 0 is used as a sentinel meaning
2475 "unassigned handle" --- avoid it. Also coalesce two critical
2476 sections in this function into one.
2478 Severity : enhancement
2480 Description: allow liblustre clients to delegate truncate locking to OST
2481 Details : To avoid overhead of locking, liblustre client instructs OST to
2482 take extent lock in ost_punch() on client's behalf. New connection
2483 flag is added to handle backward compatibility.
2485 Severity : enhancement
2486 Bugzilla : 4928, 7341, 9758
2487 Description: allow number of OST service threads to be specified
2488 Details : a module parameter allows the number of OST service threads
2489 to be specified via "options ost ost_num_threads={N}" in the
2490 OSS's /etc/modules.conf or /etc/modprobe.conf.
2494 Bugzilla : 6146, 9635, 9895
2495 Description: servers crash with bad pointer in target_handle_connect()
2496 Details : In rare cases when a client is reconnecting it was possible that
2497 the connection request was the last reference for that export.
2498 We would temporarily drop the export reference and get a new
2499 one, but this may have been the last reference and the export
2500 was just destroyed. Get new reference before dropping old one.
2502 Severity : enhancement
2503 Frequency : if client is started with failover MDS
2505 Description: Allow multiple MDS hostnames in the mount command
2506 Details : Try to read the configuration from all specified MDS
2507 hostnames during a client mount in case the "primary"
2510 Severity : enhancement
2512 Description: Stop sending data to evicted clients as soon as possible.
2513 Details : Check if the client we are about to send or are sending data to
2514 was evicted already. (Check is done every second of waiting,
2515 for which l_wait_event interface was extended to allow checking
2516 of exit condition at specified intervals).
2519 Frequency : rare, normally only when NFS exporting is done from client
2521 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2523 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2526 Frequency : always, if srand() is called before liblustre initialization
2528 Description: Liblustre uses system PRNG disturbing its usage by user application
2529 Details : Introduce internal to lustre fast and high-quality PRNG for
2530 lustre usage and make liblustre and some other places in generic
2531 lustre code to use it.
2533 Severity : enhancement
2534 Bugzilla : 9477, 9557, 9870
2535 Description: Verify that the MDS configuration logs are updated when xml is
2536 Details : Check if the .xml configuration logs are newer than the config
2537 logs stored on the MDS and report an error if this is the case.
2538 Request --write-conf, or allow starting with --old_conf.
2540 Severity : enhancement
2542 Description: Handle symlinks in the path when checking if Lustre is mounted.
2543 Details : Resolve intermediate symlinks when checking if a client has
2544 mounted a filesystem to avoid duplicate client mounts.
2549 Description: lconf can hit an error exception but still return success.
2550 Details : The lconf command catches the Command error exception at the top
2551 level script context and will exit with the associated exit
2552 status, but doesn't ensure that this exit status is non-zero.
2557 Description: failure of ptlrpc thread startup can cause oops
2558 Details : Starting a ptlrpc service thread can fail if there are a large
2559 number of threads or the server memory is very fragmented.
2560 Handle this without oopsing.
2563 Frequency : always, only if liblustre and non-default acceptor port was used
2565 Description: liblustre cannot connect to servers with non-default acceptor port
2566 Details : tcpnal_set_default_params() was not called and was therefore
2567 ignoring the environment varaible TCPNAL_PORT, as well as other
2568 TCPNAL_ environment variables
2573 Description: two objects could be created on the same OST for a single file
2574 Details : If an OST is down, in some cases it was possible to create two
2575 objects on a single OST for a single file. No problems other
2576 than potential performance impact and spurious error messages.
2580 Bugzilla : 5681, 9562
2581 Description: Client may oops in ll_unhash_aliases
2582 Details : Client dcache may become inconsistent in race condition.
2583 In some cases "getcwd" can fail if the current directory is
2589 Description: Inode refcounting problems in NFS export code
2590 Details : link_raw functions used to call d_instantiate without obtaining
2591 extra inode reference first.
2595 Bugzilla : 9942, 9903
2596 Description: Referencing freed requests leading to crash, memleaks with NFS.
2597 Details : We used to require that call to ll_revalidate_it was always
2598 followed by ll_lookup_it. Also with revalidate_special() it is
2599 possible to call ll_revalidate_it() twice for the same dentry
2600 even if first occurence returned success. This fix changes semantic
2601 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2602 reference on a request referred from the intent.
2603 ll_intent_release() then releases such a request.
2606 Frequency : rare, normally benchmark loads only
2608 Description: unlinked inodes were kept in memory on the client
2609 Details : If a client is repeatedly creating and unlinking files it
2610 can accumulate a lot of stale inodes in the inode slab cache.
2611 If there is no other client load running this can cause the
2612 client node to run out of memory. Instead flush old inodes
2613 from client cache that have the same inode number as a new inode.
2616 Frequency : SLES9 2.6.5 kernel and long filenames only
2617 Bugzilla : 9969, 10379
2618 Description: utime reports stale NFS file handle
2619 Details : SLES9 uses out-of-dentry names in some cases, which confused
2620 the lustre dentry revalidation. Change it to always use the
2624 Frequency : rare, unless heavy write-truncate concurrency is continuous
2625 Bugzilla : 4180, 6984, 7171, 9963, 9331
2626 Description: OST becomes very slow and/or deadlocked during object unlink
2627 Details : filter_destroy() was holding onto the parent directory lock
2628 while truncating+unlinking objects. For very large objects this
2629 may block other threads for a long time and slow overall OST
2630 responsiveness. It may also be possible to get a lock ordering
2631 deadlock in this case, or run out of journal credits because of
2632 the combined truncate+unlink. Solution is to do object truncate
2633 first in one transaction without parent lock, and then do the
2634 final unlink in a new transaction with the parent lock. This
2635 reduces the lock hold time dramatically.
2638 Frequency : rare, 2.4 kernels only
2640 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2641 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2642 before unmounting their filesystem in order to avoid deadlock.
2643 The kernel_locked() function in 2.4 kernels only checks whether
2644 the kernel lock is held, not whether it is this process that is
2645 holding it as 2.6 kernels do.
2650 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2651 Details : The client ptlrpc code may be trying to reconnect to a down
2652 server before a previous connection attempt has timed out.
2653 Increase the reconnect interval to be longer than the connection
2654 timeout interval to avoid sending duplicate connections to
2658 Frequency : echo_client brw_test command
2660 Description: fix echo_client to work with OST preallocated code
2661 Details : OST preallocation code (5137) didn't take echo_client IO path
2662 into account: echo_client calls filter methods outside of any
2663 OST thread and, hence, there is no per-thread preallocated
2664 pages and buffers to use. Solution: hijack pga pages for IO. As
2665 a byproduct, this avoids unnecessary data copying.
2669 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2670 Description: Client can oops in mdc_commit_close() after open replay
2671 Details : It was possible for the MDS to return an open request with no
2672 transaction number in mds_finish_transno() if the client was
2673 evicted, but without actually returning an error. Clients
2674 would later try to replay that open and may trip an assertion
2675 Simplify the client close codepath, and always return an error
2676 from the MDS in case the open is not successful.
2679 Frequency : rare, 2.6 OSTs only
2681 Description: OST may deadlock under high load on fragmented files
2682 Details : If there was a heavy load and highly-fragmented OST filesystems
2683 it was possible to have all the OST threads deadlock waiting on
2684 allocation of biovecs, because the biovecs were not released
2685 until the entire RPC IO was completed. Instead, release biovecs
2686 as soon as they are complete to ensure forward IO progress.
2688 Severity : enhancement
2690 Description: Support for specifying external journal device at mount
2691 Details : If an OST or MDS device is formatted with an external journal
2692 device, this device major/minor is stored in the ext3 superblock
2693 and may not be valid for failover. Allow detecting and
2694 specifying the external journal at mount time.
2699 Description: Mounting an MDS with pending unlinked files may cause oops
2700 Details : target_finish_recovery() calls mds_postrecov() which returned
2701 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2702 considers this an error and immediately begins cleaning up the
2703 lov, just after starting the mds_lov process
2705 Severity : enhancement
2707 Description: Implement 'lfs df' to report actual free space on per-OST basis
2708 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2709 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2710 usage of objects; '-h' to report in human readable format.
2712 ------------------------------------------------------------------------------
2714 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2721 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2722 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2723 at startup during orphan unlinking. Instead, skip buddy update
2724 before setup, as it will be regenerated after recovery is complete.
2729 Description: If an OST is inactive, its locks might reference stale inodes.
2730 Details : lov_change_cbdata() must iterate over all namespaces, even if
2731 they are inactive to clear inode references from the lock.
2733 Severity : enhancement
2734 Frequency : occasional, if non-standard max_dirty_mb used
2736 Description: Client will block write RPCs if not enough grant
2737 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2738 then the client will block writes while waiting for another RPC
2739 to complete instead of consuming its dirty limit. With change
2740 we get improved performance when max_dirty_mb is small.
2742 Severity : enhancement
2743 Bugzilla : 3389, 6253
2744 Description: Add support for supplementary groups on the MDS.
2745 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2746 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2747 lookups for user supplementary groups into a cache.
2751 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2752 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2753 for liblustre O_EXCL is enforced by MDS.
2757 Description: readdir never returns NULL in liblustre.
2758 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2759 dentry in struct dirent.
2763 Frequency : liblustre only, depends on application IO pattern
2764 Description: liblustre clients evicted if not contacting servers
2765 Details : Don't put liblustre clients into the ping_evictor list, so
2766 they will not be evicted by the pinger ever.
2768 Severity : enhancement
2770 Description: Add ability to evict clients by NID from MDS.
2771 Details : By echoing "nid:$NID" string into
2772 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2773 $NID would be instantly evicted from this MDS and from all active
2774 OSTs connected to it.
2778 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2779 Details : lookup_it_finish() used to query file size from OSTs that was not
2784 Description: service threads change working directory to that of init
2785 Details : Starting lustre service threads may pin the working directory
2786 of the parent thread, making that filesystem busy. Threads
2787 now change to the working directory of init to avoid this.
2791 Frequency : during shutdown only
2792 Description: shutdown with a failed MDS or OST can cause unmount to hang
2793 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2798 Frequency : 2.6 only
2799 Description: chmod/chown may include an extra supplementary group
2800 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2801 supplementary group and if none is specified this is used.
2804 Bugzilla : 5479 (6816)
2806 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2807 Details : If lookup is in progress on a file that is unlinked we might try
2808 to revalidate the inode and fail in revalidate after lookup is
2809 complete and ll_file_open() enqueues the open again but
2810 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2813 Frequency : always, if lconf --abort_recovery used
2815 Description: lconf --abort_recovery fails with 'Operation not supported'
2816 Details : lconf was attempting to abort recovery on the MDT device and not
2819 ------------------------------------------------------------------------------
2821 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2826 Frequency : rare (only unsupported configurations with a node running as an
2828 Bugzilla : 6514, 5137
2829 Description: Mounting a Lustre file system on a node running as an OST could
2831 Details : OSTs now preallocates memory needed to write out data at
2832 startup, instead of when needed, to avoid having to
2833 allocate memory in possibly low memory situations.
2834 Specifically, if the file system is mounted on on OST,
2835 memory pressure could force it to try to write out data,
2836 which it needed to allocate memory to do. Due to the low
2837 memory, it would be unable to do so and the node would
2838 become unresponsive.
2840 Severity : enhancement
2842 Description: Addition of lconf --service command line option
2843 Details : lconf now accepts a '--service <arg>' option, which is
2844 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2846 Severity : enhancement
2848 Description: Failover mode is now the default for OSTs.
2849 Details : By default, OSTs will now run in failover mode. To return to
2850 the old behaviour, add '--failout' to the lmc line for OSTs.
2852 Severity : enhancement
2854 Description: Health checks are now provided for MDS and OSTs
2855 Details : Additional detailed health check information on MSD and OSTs
2856 is now provided through the procfs health_check value.
2859 Frequency : occasional, depends on IO load
2861 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2862 after numerous create/delete cycles
2863 Details : The ext3 inode allocation policy would not allocate new inodes
2864 very well on the OSTs because there are no new directories
2865 being created. Instead we look for groups with free space if
2866 the parent directories are nearly full.
2871 Description: Network or server problems during mount may cause partially
2872 mounted clients instead of returning an error.
2873 Details : The config llog parsing code may overwrite the error return
2874 code during mount error handling, returning success instead
2880 Description: MDS can fail to allocate large reply buffers
2881 Details : After long uptimes the MDS can fail to allocate large reply
2882 buffers (e.g. zconf client mount config records) due to memory
2883 fragmentation or consumption by the buffer cache. Preallocate
2884 some large reply buffers so that these replies can be sent even
2885 under memory pressure.
2889 Frequency : rare (liblustre)
2890 Description: fsx running with liblustre complained that using truncate() to
2891 extend the file doesn't work. This patch corrects that issue.
2892 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2893 ATTR_SIZE and lsm use in llu_setattr_raw.
2897 Frequency : rare, only 2.6 kernels
2898 Description: Unusual file access patterns on the MDS may result in inode
2899 data being lost in very rare circumstances.
2900 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2901 code in the 2.6 kernel caused the inode and/or EA data not to
2902 be read from disk, causing single-file corruption.
2906 Frequency : rare, only 2.6 filesystems using extents
2907 Description: Heavy concurrent write and delete load may cause data corruption.
2908 Details : It was possible under high-load situations to have an extent
2909 metadata block in the block device cache from a just-unlinked
2910 file overwrite a newly-allocated data block. We now unmap any
2911 metadata buffers that alias just-allocated data blocks.
2915 Frequency : filesystems with default stripe_count larger than 77
2916 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2917 Details : lconf specifies an inode size of 4096 bytes when the default
2918 stripe_count is larger than 77. This conflicts with the default
2919 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2920 this case to avoid pinning too much memory for large EAs.
2922 ------------------------------------------------------------------------------
2924 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2929 Frequency : rare (extremely heavy IO load with hundreds of clients)
2931 Description: Client is evicted, gets IO error writing to file
2932 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2933 caused clients to be evicted for AST timeouts. The fixes in
2934 bug 5192 mean we no longer need to have such short AST timeouts
2935 so ldlm_timeout has been increased.
2938 Frequency : occasional during --force or --failover shutdown under load
2939 Bugzilla : 5949, 4834
2940 Description: Server oops/LBUG if stopped with --force or --failover under load
2941 Details : a collection of import/export refcount and cleanup ordering
2942 issues fixed for safer force cleanup
2945 Frequency : only filesystems larger than 120 OSTs
2946 Bugzilla : 5990, 6223
2947 Description: lfs getstripe would oops on a very large filesystem
2948 Details : lov_getconfig used kfree on vmalloc'd memory
2951 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2952 Bugzilla : 6242, 6243
2953 Description: reading from files that had been truncated to a non-zero size
2954 but never opened returned no data
2955 Details : ll_file_read() reads zeros from no-object files to EOF
2960 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2961 Details : There's little harm in aborting MDS/OSS recovery and letting it
2962 try again, so I removed the LASSERT and return an error instead.
2964 Severity : enhancement
2966 Description: New debugging infrastructure for tracking down data corruption
2967 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2968 (b) cover more of the client-side code path, and (c) try to narrow
2969 down where problems occurred
2973 Bugzilla : 3819, 4364, 4397, 6313
2974 Description: Racing close and eviction MDS could cause assertion in mds_close
2975 Details : It was possible to get multiple mfd references during close and
2976 client eviction, leading to one thread referencing a freed mfd.
2978 Severity: : enhancement
2979 Bugzilla : 3262, 6359
2980 Description: Attempts to reconnect to servers are now more aggressive.
2981 Details : This builds on the enhanced upcall-less recovery that was added
2982 in 1.4.2. When trying to reconnect to servers, clients will
2983 now try each server in the failover group every 10 seconds. By
2984 default, clients would previously try one server every 25 seconds.
2989 Description: After recovery, certain operations trigger a failed
2990 assertion on a client.
2991 Details : Failing over an mds, using lconf -d --failover, while a
2992 client was doing a readdir() call would cause the client to
2993 LBUG after recovery completed and the readdir() was resent.
2995 Severity : enhancement
2997 Description: Default groups are now added by lconf
2998 Details : You can now run lconf --group <servicename> without having to
2999 manually add groups with lmc.
3002 Frequency : occasional
3004 Description: Nodes with an elan id of 0 trigger a failed assertion
3007 Frequency : always when accessing e.g. tty/console device nodes
3009 Description: tty and some other devices nodes cannot be used on lustre
3010 Details : file's private_data field is used by device data and lustre
3011 values in there got lost. New field was added to struct file to
3012 store fs-specific private data.
3015 Frequency : when exporting Lustre via NFS
3017 Description: NFSD failed occasionally when looking up a path component
3018 Details : NFSD is looking up ".." which was broken in ext3 directories
3019 that had grown large enough to become hashed.
3022 Frequency : Clusters with multiple interfaces not on the same subnet
3024 Description: Nodes will repeatedly try to reconnect to an interface which it
3025 cannot reach and report an error to the log.
3026 Details : Extra peer list entries will be created by lconf with some peers
3027 unreachable. lconf now validates the peer before adding it.
3030 Frequency : Only if a default stripe is set on the filesystem root.
3032 Description: Setting a default stripe on the filesystem root prevented the
3033 filesystem from being remounted.
3034 Details : The client was sending extra request flags in the root getattr
3035 request and did not allocate a reply buffer for the dir EA.
3038 Frequency : occasional, higher if lots of files are accessed by one client
3039 Bugzilla : 6159, 6097
3040 Description: Client trips assertion regarding lsm mismatch/magic
3041 Details : While revalidating inodes the VFS looks up inodes with ifind()
3042 and in rare cases can find an inode that is being freed.
3043 The ll_test_inode() code will free the lsm during ifind()
3044 when it finds an existing inode and then the VFS later attaches
3045 this free lsm to a new inode.
3049 Bugzilla : 6422, 7030
3050 Description: MDS deadlock between mkdir and client eviction
3051 Details : Creating a new file via mkdir or mknod (starting a transaction
3052 and getting the ns lock) can deadlock with client eviction
3053 (gets ns lock and trying to finish a synchronous transaction).
3056 Frequency : occasional
3057 Description: While starting a server, the fsfilt_ext3 module could not be
3059 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3060 kernels. Previously, lconf would still use the ext3 name
3061 when trying to load modules. Now, it will correctly use
3062 ext3 on 2.4 and ldiskfs on 2.6.
3064 Severity : enhancement
3065 Description: The default stripe count has been changed to 1
3066 Details : The interpretation of the default stripe count (0, to lfs
3067 or lmc) has been changed to mean striping across a single
3068 OST, rather than all available. For general usage we have
3069 found a stripe count of 1 or 2 works best.
3071 Severity : enhancement
3072 Description: Add support for compiling against Cray portals.
3073 Details : Conditional compiling for some areas that are different
3077 Frequency : occasional
3078 Bugzilla : 6409, 6834
3079 Description: Creating files with an explicit stripe count may lead to
3080 a failed assertion on the MDS
3081 Details : If some OSTs are full or unavailable, creating files may
3082 trigger a failed assertion on the MDS. Now, Lustre will
3083 try to use other servers or return an error to the
3087 Frequency : occasional
3089 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3090 caused lock timeout during readahead (since 1.4.2).
3091 Details : Processes doing readahead might match a lock that hasn't been
3092 granted yet if there are overlapping and conflicting lock
3093 requests. The readahead process waits on ungranted lock
3094 (original lock is CBPENDING), while OST waits for that process
3095 to cancel CBPENDING read lock and eventually evicts client.
3097 Severity : enhancement
3099 Description: Initial enabling of flock support for clients
3100 Details : Implements fcntl advisory locking and file status functions.
3101 This feature is provided as an optional mount flag (default
3102 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3103 locking are implemented yet, and those that are are not guaranteed
3104 to be completely correct in production environments.
3105 mount -t lustre -o [flock|noflock] ...
3108 Frequency : occasional
3110 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3111 assertion in the ext3 JBD (journaling) layer.
3112 Details : The b_committed_data struct is protected by the big kernel lock
3113 in 2.4 kernels, serializing journal_commit_transaction() and
3114 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3115 access to this struct. In 2.6 kernels there is finer grained
3116 locking to improve SMP performance of the JBD layer.
3120 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3125 Description: merge and cleanup kernel patches.
3128 -----------------------------------------------------------------------------
3130 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3132 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3133 versions of Lustre. Please update all servers and clients to
3134 version 1.4.2 or later at the same time. You must also run
3135 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3136 to update the configuration logs.
3138 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3139 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3140 - swab llog records of type '0' so we get proper header size/idx (5861)
3141 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3142 - fix rename of one directory over another leaking an inode (5953)
3143 - avoid SetPageDirty on 2.6 (5981)
3144 - don't re-add just-being-destroyed locks to the waiting list (5653)
3145 - when creating new directories, inherit the parent's custom
3146 striping settings if present parent (3048)
3147 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3148 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3149 - don't hold client locks on temporary worklist from l_lru (5666)
3150 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3151 - automatically evict dead clients (5921)
3152 - Update file size properly in create+truncate+fstat case (6196)
3153 - Do not unhash mountpoint dentries, do not allow removal of
3155 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3156 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3157 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3158 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3159 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3160 - Correctly handle reads of files with no objects (6243)
3161 - lctl recover will also mark a device active if deactivate used (5933)
3163 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3164 - allow --write-conf on an MDS with different nettype than client (5619)
3165 - don't write config llogs to MDS for mounts not from that MDS (5617)
3166 - lconf should create multiple TCP connections from a client (5201)
3167 - init scripts are now turned off by default; run chkconfig --on
3168 lustre and chkconfig --on lustrefs to use them
3169 - upcalls are no longer needed for clients to recover to failover
3171 - add --abort-recovery option to lconf to abort recovery on device
3173 - add support for an arbitrary number of OSTs (3026)
3174 - Quota support protocol changes.
3175 - forward compatibility changes to wire structs (6007)
3176 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3177 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3178 - improved "lustre status" script
3179 - initialize blocksize for non-regular files (6062)
3180 - added --disable-server and --disable-client configure options (5782)
3181 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3182 - Vanilla 2.4.29 support
3183 - increase maximum number of obd devices to 520 (6242)
3184 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3185 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3186 SLES 9 2.6.5 kernels
3187 - limit stripes per file to 160 (the maximum EA size) (6093)
3189 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3192 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3193 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3194 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3195 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3196 - small ext3 extents cleanups and fixes (5733)
3197 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3198 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3199 - avoid needless client->OST connect, fix handle mismatch (5317)
3200 - fix DLM error path that led to out-of-sync client, long delays (5779)
3201 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3202 - fix several locking issues related to i_size (5492,5624,5654,5672)
3203 - don't move pending lock onto export if it is already evicted (5683)
3204 - fix kernel oops when creating .foo in unlinked directory (5548)
3205 - fix deadlock in obdfilter statistics vs. object create (5811)
3206 - use time_{before,after} to avoid timer jiffies wrap (5882)
3207 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3208 - Do not leak request if server was not able to process it (5154)
3209 - If mds_open unable to find parent dir, make that negative lookup(5154)
3210 - don't create new directories with extent-mapping (5909, 5936)
3212 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3213 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3214 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3215 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3216 - support for RHEL4 kernel on i686 (5773)
3217 - provide error messages when incompatible logs are encountered (5898)
3219 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3220 * version 1.4.0.10 (1.4.1 release candidate 1)
3222 - don't keep a lock reference when lock is not granted (4238)
3223 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3224 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3225 - drop import inflight refcount on signal_completed_replay error (5255)
3226 - unlock page after async write error during send (3677)
3227 - handle missing objects in filter_preprw_read properly (5265)
3228 - no transno return for symlink open, don't save no-trasno open (3440)
3229 - don't try to complete elan receive that already failed (4012)
3230 - free RPC server reply state on error (5406)
3231 - clean up thread from ptlrpc_start_thread() on error (5160)
3232 - readahead could read extra page into cache that wasn't ejected (5388)
3233 - prevent races in class_attach/setup/cleanup/detach (5260)
3234 - don't dereference de->d_inode after l_dput of de (5458)
3235 - use "int" for stripe value returned from lock_to_stripe (5544)
3236 - mballoc allocation and error-checking fixes in 2.6 (5504)
3237 - block device patches to fix I/O request sizes in 2.6 (5482)
3238 - look up hostnames for IB nals (5602)
3239 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3240 - don't start multiple acceptors for the same port (5277)
3241 - fix incorrect LASSERT in mds_getattr_name (5635)
3242 - export a proc file for general "ping" checking (5628)
3243 - fix "lfs check" to not block when the MDS is down (5628)
3245 - service request history (4965)
3246 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3247 - create an "evict_client" /proc entry on OSTs, like the MDS has
3248 - fix mount usage message, return errors per mount(8) (5168)
3249 - change grep [] to grep "[]" in tests so they work in more UMLs
3250 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3251 - remove ancient LOV_MAGIC_V0 EA support (5047)
3252 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3253 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3254 - allow readahead allocations to fail when low on memory (5383)
3255 - mmap locking landed again, after considerable improvement (2828)
3256 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3258 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3261 - send OST transaction number in read/write reply to free req (4966)
3262 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3263 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3264 - the watchdog thread now runs as interruptible (5246)
3265 - flock/lockf fixes (but it's still disabled, pending 5135)
3266 - don't use EXT3 constants in llite code (5094)
3267 - memory shortage at startup could cause assertion (5176)
3269 - reorganization of lov code
3270 - single portals codebase
3272 - add extents/mballoc support (5025)
3273 - direct I/O reads in the obdfilter (4048)
3274 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3276 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3279 - send OST transaction number in read/write reply to free req (4966)
3280 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3281 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3282 - the watchdog thread now runs as interruptible (5246)
3283 - handle missing objects in filter_preprw_read properly (5265)
3284 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3285 - drop import inflight refcount on signal_completed_replay error (5255)
3286 - unlock page after async write error during send (3677)
3287 - return original error code on reconstructed replies (3761)
3288 - no transno return for symlink open, don't save no-trasno open (3440)
3290 - add pid to ldlm debugging output (4922)
3291 - bump the watchdog timeouts -- we can't handle 30sec yet
3292 - extra debugging for orphan dentry/inode bug (5259)
3294 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3297 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3298 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3299 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3300 - fix NULL dereference in /proc/sys/portals/routes (4827)
3301 - allow failed mdc_close() operations to be interrupted (4561)
3302 - stop precreate on OST before MDS would time out on it (4778)
3303 - don't send partial-page writes before EOF from client (4410)
3304 - discard client grant for sub-page writes on large-page clients (4520)
3305 - don't free dentries not owned by NFS code, check generation (4806)
3306 - fix lsm leak if mds_create_objects() fails (4801)
3307 - limit debug_daemon file size, always print CERROR messages (4789)
3308 - use transno after validating reply (3892)
3309 - process timed out requests if import state changes (3754)
3310 - update mtime on OST during writes, return in glimpse (4829)
3311 - add mkfsoptions to LDAP (4679)
3312 - use ->max_readahead method instead of zapping global ra (5039)
3313 - don't interrupt __l_wait_event() during strace
3315 - add software watchdogs to catch hung threads quickly (4941)
3316 - make lustrefs init script start after nfs is mounted
3317 - fix CWARN/ERROR duplication (4930)
3318 - return async write errors to application if possible (2248)
3319 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3320 - print NAL number in %x format (4645)
3321 - update barely-supported suse-2.4.21-171 series (4842)
3322 - support for sles 9 %post scripts
3323 - support for building 2.6 kernel-source packages
3324 - support for sles km_* packages
3326 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3329 - ignore -ENOENT errors in osc_destroy (3639)
3330 - notify osc create thread that OSC is being cleaned up (4600)
3331 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3332 - reconstruct ost_handle() like mds_handle() (4657)
3333 - create a new thread to do import eviction to avoid deadlock (3969)
3334 - let lconf resolve symlinked-to devices (4629)
3335 - don't unlink "objects" from directory with default EA (4554)
3336 - hold socknal file ref over connect in case target is down (4394)
3337 - allow more than 32000 subdirectories in a single directory (3244)
3338 - fix blocks count for O_DIRECT writes (3751)
3339 - OST returns ENOSPC from object create when no space left (4539)
3340 - don't send truncate RPC if file size isn't changing (4410)
3341 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3342 - bind to privileged port in socknal and tcpnal (3689)
3344 - rate limit CERROR/CWARN console message to avoid overload (4519)
3345 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3346 - basic mmap support (3918)
3347 - kernel patch series update from b1_4 (4711)
3349 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3352 - avoid crash during MDS cleanup with OST shut down (2775)
3353 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3354 - don't use bad inodes on the MDS (3744)
3355 - dynamic object preallocation to improve recovery speed (4236)
3356 - don't hold spinlock over lock dumping or change debug flags (4401)
3357 - don't zero obd_dev when it is force cleaned (3651)
3358 - print grants to console if they go negative (4431)
3359 - "lctl deactivate" will stop automatic recovery attempts (3406)
3360 - look for existing locks in ldlm_handle_enqueue() (3764)
3361 - don't resolve lock handle twice in recovery avoiding race (4401)
3362 - revalidate should check working dir is a directory (4134)
3364 - don't always mark "slow" obdfilter messages as errors (4418)
3366 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3369 - don't close LustreDB during write_conf until it is done (3860)
3370 - fix typo in lconf for_each_profile (3821)
3371 - allow dumping logs from multiple threads at one time (3820)
3372 - don't allow multiple threads in OSC recovery (3812)
3373 - fix debug_size parameters (3864)
3374 - fix mds_postrecov to initialize import for llog ctxt (3121)
3375 - replace config semaphore with spinlock (3306)
3376 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3377 - don't allow enqueue to complete on a destroyed export (3822)
3378 - down write_lock before checking llog header bitmap (3825)
3379 - recover from lock replay timeout (3764)
3380 - up llog sem before sending rpc (3652)
3381 - reduce ns lock hold times when setting kms (3267)
3382 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3383 - fix NULL deref and obd_dev leak on setup error (3312)
3384 - replace some LBUG about llog ops with error handling (3841)
3385 - don't match INVALID dentries from d_lookup and spin (3784)
3386 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3387 - fix invalid assertion in ptlrpc_set_wait (3880)
3389 - add libwrap support for the TCP acceptor (3996)
3390 - add /proc/sys/portals/routes for non-root route listing (3994)
3391 - allow setting MDS UUID in .xml (2580)
3392 - print the stack of a process that LBUGs (4228)
3394 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3397 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3398 - make sure to unset replay flag from failed open requests (3440)
3399 - if default stripe count is 0, use OST count for inode size (3636)
3400 - update parent mtime/ctime on client for create/unlink (2611)
3401 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3402 - free recovery state on server during a forced cleanup (3571)
3403 - unregister_reply for resent reqs (3063)
3404 - loop back devices mounting and status check on 2.6 (3563)
3405 - fix resource-creation race that can provoke i_size == 0 (3513)
3406 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3407 - more debugging for page-accounting assertion (3746)
3408 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3409 - avoid deadlock after precreation failure (3758)
3410 - fix race and lock order deadlock in orphan handling (3450, 3750)
3411 - add validity checks when grabbing inodes from l_ast_data (3599)
3413 - add /proc/.../recovery_status to obdfilter (3428)
3414 - lightweight CDEBUG infrastructure, debug daemon (3668)
3415 - change default OSC RPC parameters to be better on small clusters
3416 - turn off OST read cache for files smaller than 32MB
3417 - install man pages and include them in rpms (3100)
3418 - add new init script for (un)mounting lustre filesystems (2593)
3419 - run chkconfig in %post for init scripts (3701)
3420 - drop scimac NAL (unmaintained)
3422 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3425 - clean kiobufs before and after use (3485)
3426 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3427 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3428 - change init script to fail more clearly if not run as root (1528)
3429 - allow clients to reconnect during replay (1742)
3430 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3431 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3432 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3433 - fix cond_resched (3554)
3434 - only evict unfinished clients after recovery (3515)
3435 - allow bulk resend, prevent data loss (3570)
3436 - dynamic ptlrpc request buffer allocation (2102)
3437 - don't allow unlinking open directory if it isn't empty (2904)
3438 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3439 - remove extraneous obd dereference causing LASSERT failure (3334)
3440 - don't use get_cycles() when creating temp. files on the mds (3156)
3441 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3442 - handle EEXIST for set-stripe, set proper directory name (3336)
3444 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3445 - fix ksocknal_fmb_callback() error messages (2918)
3447 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3450 - don't copy lvb into (possibly NULL) reply on error (2983)
3451 - don't deref dentry after dput, don't free lvb on error (2922)
3452 - use the kms to determine writeback rpc length (2947)
3453 - increment oti_logcookies when osc is inactive (2948)
3454 - update client's i_blocks count via lvb messages (2543)
3455 - handle intent open/close of special files properly (1557)
3456 - mount MDS with errors=remount-ro, like obdfilter (2009)
3457 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3458 - don't use cancelling-locks' kms values (2947)
3459 - use highest lock extent for kms, not last one (2925)
3460 - don't dereference ERR_PTR() dentry in error handling path (3107)
3461 - fix thread race in portals_debug_dumplog() (3122)
3462 - create lprocfs device entries at setup instead of at attach (1519)
3463 - common AST error handler, don't evict client on completion race (3145)
3464 - zero nameidata in detach_mnt in 2.6 (3118)
3465 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3466 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3467 - initialize RPC timeout timer earlier for 2.6 (3219)
3468 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3469 - print nal/nid for unknown nid (3258)
3470 - additional checks for oscc recovery before doing precreate (3284)
3471 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3472 - don't crash in mdc_close for bad permissions on open (3285)
3473 - zero i_rdev for non-device files (3147)
3474 - clear page->private before handing to FS, better assertion (3119)
3475 - tune the read pipeline (3236)
3476 - fix incorrect decref of invalidated dentry (2350)
3477 - provide read-ahead stats and refine rpc in flight stats (3328)
3478 - don't hold journal transaction open across create RPC (3313)
3479 - update atime on MDS at close time (3265)
3480 - close LDAP connection when recovering to avoid server load (3315)
3481 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3482 - don't leak open file on MDS after open resend (3325)
3483 - serialize filter_precreate and filter_destroy_precreated (3329)
3484 - loop device shouldn't call sync_dev() for nul device (3092)
3485 - clear page cache after eviction (2766)
3486 - resynchronize MDS->OST in background (2824)
3487 - refuse to mount the same filesystem twice on same mountpoint (3394)
3488 - allow llmount to create routes for mounting behind routers (3320)
3489 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3490 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3491 - fix rare problem with rename on htree directories (3417)
3493 - allow default OST striping configuration per directory (1414)
3494 - fix compilation for qswnal for 2.6 kernels (3125)
3495 - increase maximum number of MDS request buffers for large systems
3496 - change liblustreapi to be useful for external progs like lfsck (3098)
3497 - increase local configuration timeout for slow disks (3353)
3498 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3500 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3503 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3504 - don't overwrite extent policy data in reply if lock was blocked (2901)
3505 - drop filter export grants atomically with removal from device (2663)
3506 - del obd_self_export from work_list in class_disconnect_exports (2908)
3507 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3508 - swab reply message in mdc_close, other PPC fixes (2464)
3509 - fix destroying of named logs (2325)
3510 - overwrite old logs when running lconf --write_conf (2264)
3511 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3512 - fix race in target_handle_connect (2898)
3513 - mds_reint_create() should take same inode create lock (2926)
3514 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3515 - don't close files for self_export to avoid uninitialized obd (2936)
3516 - allow MDS with the same name as client node (2939)
3517 - hold dentry reference for closed log files for unlink (2325)
3518 - reserve space for all logs during transactions (2059)
3519 - don't evict page beyond end of stripe extent (2925)
3520 - don't oops on a deleted current working directory (2399)
3521 - handle hard links to targets without a parent properly (2517)
3522 - don't dereference NULL lock when racing during eviction (2867)
3523 - don't grow lock extents when lots of conflicting locks (2919)
3525 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3528 - account for cache space usage on clients to avoid data loss (974)
3529 - lfsck support in lustre kernel code (2349)
3530 - reduce journal credits needed for BRW writes (2370)
3531 - orphan handling to avoid losing space on client/server crashes
3532 - ptlrpcd can be blocked, stopping ALL progress (2477)
3533 - use lock value blocks to assist in proper KMS, faster stat (1021)
3534 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3535 - recovery for initial connections (2355)
3536 - fixes for mds_cleanup_orphans (1934)
3537 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3538 - block all file creations until orphan recovery completes (1901)
3539 - client remove rq_connection from request struct (2423)
3540 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3541 - recovery timer race (2670)
3542 - mdc_close recovey bug (2532)
3543 - ptlrpc cleanup bug (2710)
3544 - mds timeout on local locks (2588)
3545 - namespace lock held during RPCs (2431)
3546 - handle interrupted sync write properly (2503)
3547 - don't try to handle a message that hasn't been replied to (2699)
3548 - client assert failure during cleanup after abort recovery (2701)
3549 - leak mdc device after failed mount (2712)
3550 - ptlrpc_check_set allows timedout requests to complete (2714)
3551 - wait for inflight reqs when ptlrpcd finishes (2710)
3552 - make sure unregistered services are removed from the srv_list
3553 - reset bulk XID's when resending them (caught by 1138 test)
3554 - unregister_bulk after timeout
3555 - fix lconf error (2694)
3556 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3557 - readahead locks pages, leaves pending causing memory pressure (2673)
3558 - increase OST request buffers to 4096 on large machines (2729)
3559 - fix up permission of existing directories in simple_mkdir (2661)
3560 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3561 - don't assign transno to errored transactions (2742)
3562 - don't delete objects on OST if given a bogus objid from MDS (2751)
3563 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3564 - if rq_no_resend, then timeout request after recovery (2432)
3565 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3566 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3567 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3568 - MDS recovery shouldn't delete objects that it has given out (2730)
3569 - if enqueue arrives after completion, don't clobber LVB (2819)
3570 - don't unlock pages twice when trigger_group_io returns error (2814)
3571 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3572 - don't write pages to disk if there was an error (1450)
3573 - don't ping imports that have recovery disabled (2676)
3574 - take buffered bytes into account when balancing socknal conn (2817)
3575 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3576 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3577 - remove little-endian swabbing of llog records (1987)
3578 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3579 - retry reposting request buffers if they fail (1191)
3580 - grow extent at grant time to avoid granting a revoked lock (2809)
3581 - lock revoke doesn't evict page if covered by a second lock (2765)
3582 - disable VM readahead to avoid reading outside lock extents (2805)
3584 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3585 - updated kernel patches for hp-2.4.20 kernel (2681)
3587 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3590 - fix truncated write corruption (2366)
3591 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3593 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3594 - ensure len(uuid) < 37 in lmc (1171)
3595 - fix ia64 OOPS in llog_test (2255)
3596 - zero end of page at obdfilter for partial page writes (2648)
3597 - don't leave stale dentries around after renames (bug 2428)
3598 - fix timeouts when evicting a client with a single lock held (2642)
3599 - set deadline for the initial HELLO message to drain (2634)
3600 - print out dotted-quad IP addresses in the socknal (2302)
3602 - additional debugging for MDS client eviction problem (2443)
3603 - fix mkfsoptions support for osts (2603, 2604)
3605 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3608 - add series for the vanilla 2.6.0 kernel
3609 - add series for the vanilla 2.4.24 kernel
3610 - add series for a cray x86/64 UL kernel drop
3611 - fix xattr patches for the vanilla 2.4.19 series
3613 - generate true UUIDs in lmc (1171)
3614 - have portals stack dumping break in UML (2466)
3615 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3616 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3617 - wake callback waiting threads on client eviction (2460)
3618 - Add --ptldebug and --subsystem to lmc (1719)
3619 - update assertion to allow safe interrupt allocation
3620 - set rq_no_resend for cancel requests (2432)
3621 - recalculate ptlrpcd timeout after resend (2494)
3622 - call vfs_rmdir when removing pending directories (2368)
3623 - fix renaming a file to itself (2429)
3624 - lmc creates a default one-stripe lov (2454)
3625 - expand procfs space to handle large clusters (2326)
3626 - increase UML stack to avoid overflow
3627 - update lconf's list of debug and subsystem masks
3628 - fix lfs find --obd (2510)
3629 - /proc tunable for disabling filter read caching (2591)
3630 - stop rpm packages from altering slapd.conf (2301)
3631 - disable nagle in the socknal under 0conf (2578)
3632 - choose mds inode size based on stripe count (2572)
3633 - fix kernel-source rpm problems (2516)
3635 - add --disable-doc to avoid pdf generation (2421)
3636 - update documentation, tests, type-os, comments
3637 - avoid format warnings on ia64
3638 - remove the TOE NAL
3639 - tiny code cleanups by removing unused fields
3641 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3644 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3645 - remove the most blatant lies from BUILDING file (2371)
3646 - change default debug level to reasonable production setting
3647 - reduce client side cache size to reduce cache flush time
3648 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3649 - make TCP zerocopy and pinger support enabled by default (2476)
3650 - sync writes completed after process exits caused crashes (2319)
3651 - maintain correct mount count on the MDS (2356)
3652 - backout 1557, because 2316 wasn't really fixed
3653 - better file I/O statistics gathering in /proc
3654 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3655 - another kernel patch to fix zero-copy TCP function export
3656 - don't take duplicate lock when processing re-sent getattr (2420)
3657 - lctl uses obd_self_export instead of creating new conn (2353)
3658 - MDS/OST recovery case which requires object creation asserted (2425)
3659 - move lfs from /usr/sbin to /usr/bin in packages
3660 - fix race between mds_client_add and mds_client_free (2417)
3661 - use kmalloc instead of slabs in portals (2430)
3662 - don't create duplicate records when a failover MDS is present (2442)
3663 - remove unnecessary mount age check (2332)
3664 - don't remove directory inodes from locks prematurely (2451)
3665 - don't break if MDS service name is the same as hostname (2103)
3666 - fix races in client write RPC generation when cache full (2482)
3668 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3671 - remove now-unused request->rq_obd (278)
3672 - if an allocation fails, print out how much memory we've used (1933)
3673 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3674 - add the "configurable stack size" patch to most series files (1256)
3675 - ability to write large log records, for 100+ OST configs (2306)
3676 - fix NULL deref when filter_prep fails (2314)
3677 - fix operator precedence error in filter_sync
3678 - dynamic allocation of socknal TX descriptors (2315)
3679 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3680 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3681 - ensure that CWARN messages in Portals always get to the syslog
3682 - __init/__exit are not for prototype decls (ldlm_init/exit)
3683 - x86-64 compile warning fixes
3684 - fix gateway LMC keyword conflict (2318)
3685 - fix MDS lock inversions in getattr/reint paths (1844)
3686 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3687 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3688 - DLM race condition prevented some lock evictions (2328)
3689 - ENOMEM detection and retry on socknal sends (2230)
3690 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3691 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3692 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3693 - better error messages when a client is rejected during recovery (1505)
3694 - avoid cancelling locks which were never granted, after failure (2330)
3695 - fix i_sem/journal inversion in mds_client_add (2333)
3696 - fix truncate/getattr lock cycle deadlock (2334)
3697 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3698 - fix two rare exit paths which could leak an l_lock() ref (2321)
3699 - fencepost error in MDS/OST orphan recovery (2226)
3700 - make log record alignment 8 bytes (1988)
3701 - lstripe now fails when requested offset > ost_count (2237)
3702 - ensure that all kernel series have a complete list.h (1607)
3703 - fix crashes in special-file operations (2316)
3704 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3706 - allow configurable automake binary, for testing new versions
3707 - small update to the lfs documentation
3709 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3711 * fix negative export reference count in fsfilt_sync (2312)
3713 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3714 * release candidate 0.9.1
3716 - orphans are moved into the PENDING directory for possible recovery
3717 - replayed opens now open by fid for orphan/rename safety (1042)
3718 - last close of an orphan inode generates a transno (683)
3719 - chdir() and mount() now pin the directory entry (1020)
3720 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3721 - discard very old requests without processing them (1502)
3722 - remove some common, well-understood CERRORs (1505)
3723 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3724 - clear "grant" flags in OST replies until OST grant code lands (1644)
3725 - fix read performance by not clobbering i_blksize on client (1598)
3726 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3727 - make LRU size a /proc tunable, clears locks when reduced (707)
3728 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3729 - server locks take references on exports now (1558)
3730 - build fixes for 2.4.20-rh trees (1663)
3731 - return an error from lov_create if all OSCs are inactive (1751)
3732 - fix import levels when a reconnect happens without a timeout (1597)
3733 - exit early from mds_open if we get a lookup error (1749)
3734 - partial page read at EOF wouldn't wait for disk before sending (1642)
3735 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3736 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3737 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3738 - write/truncate lock inversion (1639)
3739 - Don't auto-load obdclass, portals modules during cleanup (1495)
3740 - fix timestamps from jumping to "now" (1763)
3741 - extra journal assertions (1648)
3742 - add an extra multiunlink test (1771)
3743 - fix read_record/write_record API (1776)
3744 - fix leak of offset_extent, possible incorrect i_size later (1772)
3745 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3746 - Debugging for the kqswnal_get_idle_tx problems (1820)
3747 - Allow recovery to be attempted multiple times (1536)
3748 - Write out MDS last_rcvd file after it is first created (1600)
3749 - Fix tx_descriptor leak in failed transmit situations (1827)
3750 - ext3 journaling fixes for assertion failure after IO error (1871)
3751 - class_export_put() on freed export after completion AST error (1896)
3752 - Fix revalidate looping in VFS (1322)
3753 - Don't access a freed export during MDS_REINT timeout (1521)
3754 - Add open-unlink recovery support on the MDS (1673,1764)
3755 - Return an error if no MDS data was read from last_rcvd (1946)
3756 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3757 - Don't setup a disk device that doesn't match exported UUID (317)
3758 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3759 - avoid committing NULL handle in force close
3760 - local.sh is now a one-stripe LOV configuration
3761 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3762 - fix ext3 htree duplicate directory entry corruption (1516)
3763 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3764 - update to new LOV EA format (2097)
3765 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3767 2003-06-15 Phil Schwan <phil@clusterfs.com>
3770 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3771 - per-import/export recovery handling (958, 931, 959)
3772 - multiple last-rcvd slots, for serving multiple FSes (949)
3773 - connections are again shared between multiple imp/exports (963, 964)
3774 - "umount -f" would hang if any requests needed to be sent (393, 978)
3775 - avoid pinning large req buffer by copying for queued messages (989)
3776 - add "uuid" to "lctl device" command to help upcalls (991)
3777 - "open" RPCs with transnos would confuse recovery counters (1037)
3778 - do proper endian conversion of all wire messages (288, 340, 891)
3779 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3780 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3781 - fix ptlrpc_connection leak in target_handle_connect (1174)
3782 - fix import refcounting bug in OST and MDS cleanup (1134)
3783 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3784 - fix very unlikely obd_types race condition (501)
3785 - remove osc_open hack for echo_client (1187)
3786 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3787 - a failure in read_inode2 leads to deadlock (1139)
3788 - cancel ack-locks as soon as transaction is committed (1072)
3789 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3790 - make sure to commitrw after any preprw to avoid deadlock (1162)
3791 - failing to execute a file in a lustre FS would lock inode (1203)
3792 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3793 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3794 - obd_uuid2tgt would walk off the end of the list (1255)
3795 - on IA64 the osc would give portals incorrect bulk size (1258)
3796 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3797 - fix lock inversion caused by new llite matching code (1282)
3798 - limit the number of dirty pages on a client to 10MB (1286)
3799 - timed out locks were not being corrected cancelled (1289)
3800 - fix O_DIRECT above 4GB on IA-32 (1292)
3801 * major user-visible changes
3802 - fail out/fail over policy now controlled by the upcall (993)
3804 - add OBD_PING to check server availability and failure (954)
3805 - lustre messages are now sent in sending host order (288, 340, 891)
3806 - add eadatalen to MDS getattr reply (340)
3807 - OST read replies may contain second buffer, with per-page status (593)
3809 2003-03-11 Phil Schwan <phil@clusterfs.com>
3812 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3813 - failed open()s could cause deadlock; fixed (867, 869)
3814 - stop cancelling OST locks when files are closed (481)
3815 - overlapping XID spaces caused network corruption (851, 853)
3816 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3817 - setattr_raw added, to do single-RPC, server-side setattrs
3818 - lmc/lconf syntax change for OST UUIDs
3819 - fix crashy race condition between ptlrpc_free_req and osc_close
3820 - don't use request in mdc_enqueue if we hit a timeout (889)
3821 - don't set the inode i_size for regular files from the MDS (896)
3822 - handle out of order completion AST (842)
3823 - don't LBUG if a lock request times out after receiving AST (913)
3824 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3825 - if a bad lock AST arrives, send an error instead of dropping entirely
3826 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3827 - fix leak in bulk IO when only partially completed (899, 900, 926)
3828 - fix O_DIRECT for ia64 (55)
3829 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3830 - C-z after timeout could hang a process forever; fixed (977)
3832 - client-side I/O cache (678, 924, 929, 941, 970)
3834 - READPAGE and SETATTRs which don't take server-side locks get
3837 2003-02-11 Phil Schwan <phil@clusterfs.com>
3840 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3841 - Fix fsfilt_extN_readpage() to read a full page of directory
3842 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3843 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3844 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3845 - Most intent operations were removed, in favour of a new RPC mode
3846 that does a single RPC to the server and bypasses most of the VFS
3847 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3848 - Aggressively cancel local locks on DLM servers
3849 - mds_reint_unlink sends EA to the client if it's the last nlink.
3850 client uses that EA to unlink OST objects.
3851 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3852 - recursive symlinks were fixed (439)
3853 - fixed NULL deref in DEBUG_REQ
3854 - filter_update_lastobjid no longer calls sync, which annoyed extN
3855 - fixed multi-client small-writes to a single file problem (445)
3856 - fixed mtime updates during file writes (607)
3857 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3858 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3859 - fixed Linux OST size reporting problem (444, 656)
3860 - OST now updates object mtime with writes or setattr (607, 619)
3861 - client verifies file size before zeroing page past EOF (445)
3862 - OST now writes last allocated objid to disk with allocation (108)
3863 - LOV on echo now works (409)
3865 - mds_reint_unlink sends a new buffer, with the EA included. this
3866 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3867 set if a regular file was being unlinked, and it was the last link
3868 - use PtlGet from the target for bulk writes (315)
3869 - OST now updates object mtime with writes or setattr (607, 619)
3870 - LDLM now has a grant-time callback to revalidate locked items, if
3872 - Many MDS operations were reorganized to combat race conditions
3874 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3875 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3877 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3880 - Fully reactivate OST imports after reconnection (512, others)
3881 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3882 - More graceful error handling for truncating on dead OST (515)
3883 - Don't error out unless we're actually accessing dead stripes (474)
3884 - Fix garbage sizes when stripes are missing (410)
3885 - LRU counters were broken, causing constant lock purge (433, 432)
3886 - garbage on read from stripes with failed OSTs (441)
3887 - mark OSCs as active before reconnecting during recovery (438)
3888 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3889 - lfind did not preserve OST order in output (443)
3890 - symlinks cause hung clients, incorrect data (439)
3891 - stop dereferencing request after dropping refcount (457)
3892 - don't LASSERT(spin_is_locked) on non-SMP (455)
3893 - fixes for many rename() bugs
3894 - fstat didn't correctly synchronize attributes (399)
3895 - server must handle lock cancellation during blocking AST prep (487)
3896 - bulk descriptors were free()d too soon (511)
3897 - fix paths in lconf, which would load incorrect modules (451, 507)
3898 - fix confusing lconf 'host not found' error message (386)
3899 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3900 - fix race condition in mdc_blocking_ast() for inode access (526)
3901 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3902 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3903 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3904 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3905 - fix lstripe in several ways (kernel side) (527)
3906 - fix request leak in ldlm_cli_enqueue (262)
3907 - incorrect OSC was marked inactive after OST failure
3908 - call mds_fs_cleanup before unmounting filesystem (524)
3909 - fix races between taking ns_lock and ldlm_lock_change_resource
3910 - fix races updating LOV export open file list
3911 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3912 - fix recovery NULL deref in ldlm_cli_cancel_unused
3913 - fix some DLM races by using new hash table for lock handles (419)
3914 - permit the client to specify desired inodes, at replay
3915 - duplicate requests when we queue them for replay reintegration
3916 - fix last_rcvd offset calculation
3917 - sync after each recovered transaction, so we always make progress
3918 - never, not always, ERESTART requests without transnos
3919 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3921 - skip replay if the MDS says that the client is already connected
3922 - don't check for a recovery-enabled export to match lctl's UUID
3923 - don't INC_USE_COUNT for phantom exports
3924 - don't crash when cleaning up phantom exports (567)
3925 - don't double-finish or set replay data for errored mdc_open requests
3926 - abort requests when they time out, so we don't get old replies
3927 - send/receive replies for AST messages again
3928 - if the client says that it doesn't have the lock, cancel it on the
3930 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3931 mark it as destroyed, it will all work out when decref is called
3932 - fix module use counts (22, 581)
3934 - ASTs now expect a reply (server cancels lock on error reply)
3936 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3939 - fix many simultaneous client startup (392)
3940 - fix dentry->d_it clobbering
3941 - credentials weren't being shipped for readdir/getattr operations
3942 - remove invalid assertions triggered during some concurrent MD
3944 - proper Lustre versions added (336, 389)
3945 - fix memory leak for create error case (398)
3946 - fix LOV locking bug that would get cli/srv out of sync
3947 - fix echo client over LOV (409)
3948 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3949 - fix double-O_EXCL intent crash (424)
3950 - avoid sending multiple lock CANCELs (352)
3952 - MDS can do multi-client recovery (modulo bugs in new code)
3954 - many updates, edits, cleanups
3956 2002-11-18 Phil Schwan <phil@clusterfs.com>
3959 - fix null d_it dereference (346)
3960 - fix full OST/dbench hang (333)
3961 - fix permission problem with file removal (286)
3962 - fix removal of OSCs from LOV when they fail
3963 - fix NULL deref during bulk timeout (214)
3964 - fix problems related to multiple filesystems on one MDS (241)
3965 - fixed serious subtle metadata locking bugs
3966 - free locks on clients when inodes are removed due to memory
3968 - fix inode pointer in lock data (285)
3969 - partial support for multiple MDS on a single host (241)
3970 - data locks weren't cancelled at clear_inode time (290, 311)
3971 - intent locks could lead to unbounded lock growth (205)
3972 - added a maximum lock count, an LRU list, and a flusher
3973 - fix multiple rename (365)
3974 - properly abstracted the echo client
3975 - OSC locked 1 byte too many; fixed
3976 - rewrote brw callback code:
3977 - fixed recovery bugs related to LOVs (306)
3978 - fixed too-many-pages-in-one-write crash (191)
3979 - fixed (again) crash in sync_io_timeout (214)
3980 - probably fixed callback-related race (385)
3982 - Add capability to MDS protocol
3983 - LDLM cancellations and callbacks on different portals
3985 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3988 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3989 - defer open object destruction to close time (601981, 138)
3990 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3991 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3992 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3993 - don't call obd_setattr in ll_file_release for destroyed objects
3994 * protocol change to lustre_msg: move |version| and add |flags|
3995 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3996 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3997 * added replay of create, unlink, link and rename operations during
3998 MDS failover; recovery should be much more robust now
3999 * remove failed OSCs from LOVs (only lov_create uses this so far)
4000 * the lustre-HOWTO was brought (more) up to date (582544)
4002 2002-10-23 Phil Schwan <phil@clusterfs.com>
4005 - in-use dentries weren't being reused properly (617851)
4006 - prevent multiple LDLM setup (599178)
4007 - fix LOV size calculations for truncate (617853)
4008 - fix client handling of MDS intent errors (POSIX)
4009 - fix permission bug in lovstripe.c test (624321)
4010 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4011 - truncate past end of file could corrupt data
4012 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4013 - a race in recovery could return ETIMEDOUT to apps (623947)
4014 - building outside the source directory was fixed
4015 * the lustre-HOWTO was brought (more) up to date (582544)
4016 * major progress was made on recovery functionality
4018 2002-10-10 Phil Schwan <phil@clusterfs.com>
4021 - recovery deadlock fix
4022 - rm -rf causes LBUG fix (617817)
4023 - file open by multiple tasks fix (618962)
4024 - directory permissions bugs (602707 and 620007)
4025 - journal_stop fixed with locking (611313)
4026 - O_APPEND failures resolved (618273, perhaps 614459)
4027 - lconf PATH fix (619770)
4028 - IA64 build fix (621450)
4029 - RPC buffer sizes scale with amount of memory
4031 2002-10-01 Phil Schwan <phil@clusterfs.com>
4034 - locks would be cancelled without throwing away data pages,
4035 resulting in inconsistent data (605627)
4036 - inode attributes were not always being refreshed (605627, 612449)
4037 - lconf now continues to cleanup after lctl reports an error
4038 - MDS now enforces user permissions (602707)
4039 - lprocfs cleanup fixed, but not yet enabled (614157)
4040 - fixed infinite server hang, should a client not respond to an AST
4041 - avoid going into recovery if user calls readlink() with a buffer
4042 that's too small (613941)
4043 - AST RPCs no longer require replies (614867) -- this may be changed
4044 - don't crash server if client sends an IOV that's too big (611336)
4045 - fixed lock conversion deadlock (611892)
4046 - fixed the following of symlinks (614622)
4047 * recovery: the server can remove locks from a client that dies, other
4048 clients can make progress
4049 * more extN patch fixes
4050 * compile-time configurable ptlrpc buffer allocations
4052 - collaborative read cache document
4053 - Lustre Lite Performance CDR document-in-progress
4055 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4058 - fix typo in patch-2.4.18
4060 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4063 - clear ptlrpc request each time in handle_incoming_request()
4064 - unlink of files now destroys the object on the OST
4066 2002-09-19 Peter Braam <braam@clusterfs.com>
4068 * add hard link support
4069 * change obdfile creation method
4070 * kernel patch changed
4072 2002-09-19 Peter Braam <braam@clusterfs.com>
4075 - stack overflow bug in extN fixed
4077 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4079 * documentation updates
4080 - add man pages for config tools
4081 - update tests/README to describe testing with new config tools
4082 - finish metadata API descriptions
4083 * bug fixes and cleanups
4084 - statfs workaround for 16TB limit
4085 - LOV stripe allocation improved, can stripe on subset of OSTs
4086 - LOV file size/IO offset was wrong for files > 4GB in size
4087 - object EA data was being dropped, caused files to be unreadable
4088 - memory overflow with non-LOV OST caused memory corruption
4089 - fixed regression tests to work with new config tools, obdfilter
4090 - fixed bug when directory size became larger than 1 block
4091 - fixed bug (for single client case) when PWD was deleted
4092 - invalidate local directory pages when doing intent-based ops
4093 - avoid LDLM oops when lock callback contained bad data
4095 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4097 * documentation updates
4098 * bug fixes and cleanups
4099 - configuration tools
4102 - 64-bit compile warnings
4103 - 64-bit internal statfs data
4105 * test_brw on persistent OST devices
4107 * lprocfs (disabled)
4109 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4111 * documentation updates
4112 * bug fixes and cleanups
4113 * configuration tools
4115 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4118 * many small fixes to 0_5_4
4119 * io/network handling
4120 * thinkos in MDS operations
4122 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4125 * crucial basic fixes to 0.5.3
4126 * IOR, Iozone work over Elan
4129 2002-08-07 Phil Schwan <phil@clusterfs.com>
4130 * version 0_5_3, our first alpha
4131 * we use the new Portals iovs
4132 * documentation updates
4133 * bug fixes and cleanups
4134 * small changes in the DLM wire protocol
4136 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4137 * version 0_5_1 with some initial stability,
4138 * locking on MD and file I/O.
4139 * documentation updates
4140 * several bug fixes since 0.5.0
4141 * small changes in wire protocol
4143 2002-07-18 Phil Schwan <phil@clusterfs.com>
4145 * delivered as Lustre Light Alpha
4146 * fixed a crash after handling invalid MDS requests
4147 * fixed directory pages for architectures with non-4k pages sizes
4149 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4150 * release version v0_4_4
4151 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4153 * Fixes a number of interruption problems with OST operations.
4154 * Update documentation for portals header changes
4155 * Move all wire protocol structs/defines to lustre_idl.h
4156 * Fixes symlink length bug.
4157 * Add tcpdump to repository.
4159 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4160 * release version v0_4_3
4161 * Fixes statfs for inodes on extN.
4162 * Fixes bug in runtests which would delete /etc/hosts.
4163 * Use 64-bit object IDs wherever possible (not into VFS though)
4164 Remove ost_get_info, which is unused by lustre, and out of date.
4166 2002-07-03 Peter Braam <braam@clusterfs.com>
4167 * release version v0_4_2 Fixes a lookup error (type not passed)
4168 * move forward to head of Portals
4169 * move forward to latest Lustre kernel
4171 2002-06-25 Peter Braam <braam@clusterfs.com>
4172 * release version v0_4_1. Hopefully stable on single node use.