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 Severity : enhancement
412 Description: adapt the lustre_config script to support the upgrade case
413 Details : Add "-u" option for lustre_config script to support upgrading 1.4
414 server targets to 1.6 in parallel.
416 --------------------------------------------------------------------------------
418 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
420 * Support for kernels:
421 2.6.5-7.283 (SLES 9),
422 2.6.9-55.EL (RHEL 4),
423 2.6.16.46-0.14 (SLES 10),
424 2.6.18.8 vanilla (kernel.org)
425 * Client support for unpatched kernels:
426 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
427 2.6.16 - 2.6.22 vanilla (kernel.org)
428 * Due to recently discovered recovery problems, we do not recommend
429 using patchless RHEL 4 clients with this or any earlier release.
430 * Recommended e2fsprogs version: 1.39.cfs8
431 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
432 * Starting with this release, the ldiskfs backing filesystem required
433 by Lustre is now in its own package, lustre-ldiskfs. This package
434 should be installed. It is versioned separately from Lustre and
435 may be released separately in future.
439 Description: block reactivating mgc import until all deactivates complete
440 Details : Fix race when failing back MDT/MGS to itself (testing)
442 Severity : enhancement
444 Description: add optional extra BUILD_VERSION info
445 Details : add a new environment variable (namely LUSTRE_VERS) which allows
446 to override the lustre version.
449 Frequency : 2.6.18 servers only
451 Description: ll_kern_mount() doesn't release the module reference
452 Details : The ldiskfs module reference count never drops down to 0
453 because ll_kern_mount() doesn't release the module reference.
458 Description: server LBUG when using old ost_num_threads parameter
459 Details : Accept the old ost_num_threads parameter but warn that it
460 is deprecated, and fix an off-by-one error that caused an LBUG.
465 Description: Transient SCSI error results in persistent IO issue
466 Details : iobuf->dr_error is not reinitialized to 0 between two
470 Frequency : sometimes when underlying device returns I/O errors
472 Description: OSTs not going read-only during write failures
473 Details : OSTs are not remounted read-only when the journal commit threads
474 get I/O errors because fsfilt_ext3 calls journal_start/stop()
475 instead of the ext3 wrappers.
479 Description: poor connect scaling with increasing client count
480 Details : Don't run filter_grant_sanity_check for more than 100 exports
481 to improve scaling for large numbers of clients.
484 Frequency : SLES10 only
486 Description: sanity-quota.sh quotacheck failed: rc = -22
487 Details : Quotas cannot be enabled on SLES10.
490 Frequency : liblustre clients only
492 Description: getdirentries does not give error when run on compute nodes
493 Details : getdirentries does not fail when the size specified as an argument
494 is too small to contain at least one entry
496 Severity : enhancement
498 Description: Add LNET router traceability for debug purposes
499 Details : If a checksum failure occurs with a router as part of the
500 IO path, the NID of the last router that forwarded the bulk data
501 is printed so it can be identified.
506 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
507 Details : Due to a race condition, liblustre clients were occasionally
510 Severity : enhancement
512 Description: lfs setstripe use optional parameters instead of postional
515 Severity : enhancement
517 Description: Nanosecond timestamp support for ldiskfs
518 Details : The on-disk ldiskfs filesystem has added support for nanosecond
519 resolution timestamps. There is not yet support for this at
520 the Lustre filesystem level.
523 Frequency : during server recovery
525 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
526 Details : request with rq_no_resend flag not awake l_wait_event if they get a
530 Frequency : nfs export on patchless client
532 Description: connectathon hang when test nfs export over patchless client
533 Details : Disconnected dentry cannot be found with lookup, so we do not need
534 to unhash it or make it invalid
537 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
538 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
539 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
544 Description: open req refcounting wrong on reconnect
545 Details : If reconnect happened between getting open reply from server and
546 call to mdc_set_replay_data in ll_file_open, we will schedule
547 replay for unreferenced request that we are about to free.
548 Subsequent close will crash in variety of ways.
549 Check that request is still eligible for replay in
550 mdc_set_replay_data().
555 Description: disable writes to filesystem when reading health_check file
556 Details : the default for reading the health_check proc file has changed
557 to NOT do a journal transaction and write to disk, because this
558 can cause reads of the /proc file to hang and block HA state
559 checking on a healthy but otherwise heavily loaded system. It
560 is possible to return to the previous behaviour during configure
561 with --enable-health-write.
563 Severity : enhancement
565 Description: 64-bit inode version
566 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
567 to the inode. This will be required for version-based recovery.
572 Description: MDS fails to start if a duplicate client export is detected
573 Details : in some rare cases it was possible for a client to connect to
574 an MDS multiple times. Upon recovery the MDS would detect this
575 and fail during startup. Handle this more gracefully.
577 Severity : enhancement
579 Description: Add -o localflock option to simulate old noflock
581 Details : This will achieve local-only flock/fcntl locks
587 Description: log_commit_thread vs filter_destroy race leads to crash
588 Details : Take import reference before releasing llog record semaphore
593 Description: Wrong request locking in request set processing
594 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
595 imp_delayed_list, in this place should be used imp_lock.
598 Frequency : when reconnection
600 Description: Grant Leak when osc reconnect to OST
601 Details : When osc reconnect ost, OST(filter) should check whether it
602 should grant more space to client by comparing fed_grant and
603 cl_avail_grant, and return the granted space to client instead
604 of "new granted" space, because client will call osc_init_grant
605 to update the client grant space info.
608 Frequency : when client reconnect to OST
610 Description: Grant Leak when osc do resend and replay bulk write
611 Details : When osc reconnect to OST, OST(filter)should clear grant info of
612 bulk write request, because the grant info while be sync between
613 OSC and OST when reconnect, and we should ignore the grant info
614 these of resend/replay write req.
619 Description: Grant space more than avaiable left space sometimes.
620 Details : When then OST is about to be full, if two bulk writing from
621 different clients came to OST. Accord the avaliable space of the
622 OST, the first req should be permitted, and the second one
623 should be denied by ENOSPC. But if the seconde arrived before
624 the first one is commited. The OST might wrongly permit second
625 writing, which will cause grant space > avaiable space.
628 Frequency : when client is evicted
630 Description: Grant might be wrongly erased when osc is evicted by OST
631 Details : when the import is evicted by server, it will fork another
632 thread ptlrpc_invalidate_import_thread to invalidate the
633 import, where the grant will be set to 0. While the original
634 thread will update the grant it got when connecting. So if
635 the former happened latter, the grant will be wrongly errased
636 because of this race.
641 Description: Checking Stale with correct fid
642 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
643 to check whether it is stale, because sometimes, we want the
644 enqueue happened anyway, and op_data.fid2 will not be initialized.
646 Severity : enhancement
648 Description: update patchless client
649 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
652 Frequency : only with 2.4 kernel
654 Description: random memory corruption
655 Details : size of struct ll_inode_info is to big for union inode.u and this
656 can be cause of random memory corruption.
661 Description: Memory leak in recovery
662 Details : Lov_mds_md was not free in an error handler in mds_create_object.
663 It should also check obd_fail before fsfilt_start, otherwise if
664 fsfilt_start return -EROFS,(failover mds during mds recovery).
665 then the req will return with repmsg->transno = 0 and rc = EROFS.
666 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
667 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
668 be freed no matter whether fsfilt_commit success or not.
671 Frequency : only with huge count clients
673 Description: Prevents from taking the superblock lock in llap_from_page for
675 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
676 for llap_from_page call in ll_removepage prevents from taking the
677 superblock lock for a soon died page.
682 Description: Not check open intent error before release open handle
683 Details : in some rare cases, the open intent error is not checked before
684 release open handle, which may cause
685 ASSERTION(open_req->rq_transno != 0), because it tries to release
686 the failed open handle.
691 Description: Set cat log bitmap only after create log success.
692 Details : in some rare cases, the cat log bitmap is set too early. and it
693 should be set only after create log success.
697 Description: Accessing a block bevice can re-enable I/O when Lustre is
698 tearing down a device.
699 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
704 Description: service threads may hog cpus when there are a lot of requests
706 Details : Insert cond_resched to give other threads a chance to use some of
712 Description: the cat log was not initialized in recovery
713 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
714 unlink log on mds will not be initialized until mds post
715 recovery. And also in mds post recovery, the unlink log will
716 initialization will be done asynchronausly, so there will be race
717 between add unlink log and unlink log initialization.
721 Description: brw_stats were being printed incorrectly
722 Details : brw_stats were being printed as log2 but all of them were not
723 recorded as log2. Also remove some code duplication arising from
724 filter_tally_{read,write}.
728 Frequency : rare, only in recovery.
729 Description: ASSERTION(req->rq_type != LI_POISON) failed
730 Details : imp_lock should be held while iterating over imp_sending_list for
731 prevent destroy request after get timeout in ptlrpc_queue_wait.
735 Description: replay-single.sh test 52 fails
736 Details : A lock's skiplist need to be cleanup when it being unlinked
737 from its resource list.
741 Description: Short directio read returns full requested size rather than
743 Details : Direct I/O operations should return actual amount of bytes
744 transferred rather than requested size.
746 Severity : enhancement
748 Description: metadata RPC reduction (e.g. for rm performance)
749 Details : decrease the amount of synchronous RPC between clients and servers
750 by canceling conflicing lock before the operation on the client side
751 and packing thier handles into the main operation RPC to server.
753 Severity : enhancement
755 Description: Async OSC create to avoid the blocking unnecessarily.
756 Details : If a OST has no remain object, system will block on the creating
757 when need to create a new object on this OST. Now, ways use
758 pre-created objects when available, instead of blocking on an
759 empty osc while others are not empty. If we must block, we block
760 for the shortest possible period of time.
764 Description: improve handling recoverable errors
765 Details : if request processig with error which can be recoverable on server
766 request should be resend, otherwise page released from cache and
769 Severity : enhancement
771 Description: refine locking for avoid write wrong info into lov_objid file
772 Details : fix possible races with add new target and write/update data in
775 --------------------------------------------------------------------------------
777 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
782 Frequency : on some architectures
784 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
785 Details : Uninitialized flags sometimes cause configuration commands to
789 Frequency : patchless clients only
791 Description: missing __iget() symbol export
792 Details : The __iget() symbol export is missing. To avoid the need for
793 this on patchless clients the deathrow inode reaper is turned
794 off, and we depend on the VM to clean up old inodes. This
795 dependency was during via the fix for bug 12181.
797 --------------------------------------------------------------------------------
799 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
801 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
802 INTEROPERATE with older versions automatically. In many cases a
803 special upgrade step is needed. Please read the
804 user documentation before upgrading any part of a live system.
805 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
806 version will not interoperate with 1.6 betas before beta5 (1.5.95).
807 * WARNING: Lustre configuration and startup changes are required with
808 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
814 Severity : enhancement
816 Description: MountConf
817 Details : Lustre configuration is now managed via mkfs and mount
818 commands instead of lmc and lconf. New obd types (MGS, MGC)
819 are added for dynamic configuration management. See
820 https://mail.clusterfs.com/wikis/lustre/MountConf for
823 Severity : enhancement
825 Description: dynamic OST addition
826 Details : OSTs can now be added to a live filesystem
828 Severity : enhancement
830 Description: startup order invariance
831 Details : MDTs and OSTs can be started in any order. Clients only
832 require the MDT to complete startup.
834 Severity : enhancement
836 Description: parallel, asynchronous orphan cleanup
837 Details : orphan cleanup is now performed in separate threads for each
838 OST, allowing parallel non-blocking operation.
840 Severity : enhancement
842 Description: optimized stripe assignment
843 Details : stripe assignments are now made based on ost space available,
844 ost previous usage, and OSS previous usage, in order to try
845 to optimize storage space and networking resources.
847 Severity : enhancement
849 Description: Permanently set tunables
850 Details : All writable /proc/fs/lustre tunables can now be permanently
851 set on a per-server basis, at mkfs time or on a live
854 Severity : enhancement
856 Description: Lustre message v2
857 Details : Add lustre message format v2.
859 Severity : enhancement
861 Description: client OST exclusion list
862 Details : Clients can be started with a list of OSTs that should be
863 declared "inactive" for known non-responsive OSTs.
867 Description: SPEC SFS validation failure on NFS v2 over lustre.
868 Details : Changes the blocksize for regular files to be 2x RPC size,
869 and not depend on stripe size.
871 Severity : enhancement
873 Description: Multiple MD RPCs in flight.
874 Details : Further unserialise some read-only MDS RPCs - learn about intents.
875 To avoid overly-overloading MDS, introduce a limit on number of
876 MDS RPCs in flight for a single client and add /proc controls
877 to adjust this limit.
879 Severity : enhancement
881 Description: client read/write statistics
882 Details : Add client read/write call usage stats for performance
883 analysis of user processes.
884 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
885 file access. extents_stats shows chunk size distribution.
886 extents_stats_per_process show chunk size distribution per
889 Severity : enhancement
891 Description: mds statistics
892 Details : Add detailed mds operations statistics in
893 /proc/fs/lustre/mds/*/stats.
897 Description: Failure of copying files with lustre special EAs.
898 Details : Client side always return success for setxattr call for lustre
899 special xattr (currently only "trusted.lov").
904 Description: Refcount LNET uuids
905 Details : The global LNET uuid list grew linearly with every startup;
906 refcount repeated list entries instead of always adding to
909 Severity : enhancement
911 Description: Dynamic service threads
912 Details : Within a small range, start extra service threads
913 automatically when the request queue builds up.
916 Frequency : mixed-endian client/server environments
918 Description: mixed-endian crashes
919 Details : The new msg_v2 system had some failures in mixed-endian
922 Severity : enhancement
924 Description: Easy OST removal
925 Details : OSTs can be permanently deactivated with e.g. 'lctl
926 conf_param lustre-OST0001.osc.active=0'
928 Severity : enhancement
930 Description: MGS proc entries
931 Details : Added basic proc entries for the MGS showing what filesystems
934 Severity : enhancement
936 Description: provide MGS failover
937 Details : Added config lock reacquisition after MGS server failover.
939 Severity : enhancement
941 Description: add Linux 2.4 support
942 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
946 Description: a large application tries to do I/O to the same resource and dies
948 Details : Check the req->rq_arrival time after the call to
949 ost_brw_lock_get(), but before we do anything about
950 processing it & sending the BULK transfer request. This
951 should help move old stale pending locks off the queue as
952 quickly as obd_timeout.
955 Frequency : when an incorrect nid is specified during startup
957 Description: ptlrpc connect to non-existant node causes kernel crash
958 Details : LNET can't be re-entered from an event callback, which
959 happened when we expire a message after the export has been
960 cleaned up. Instead, hand the zombie cleanup off to another
963 Severity : enhancement
965 Description: plain/inodebits lock performance improvement
966 Details : Grouping plain/inodebits in granted list by their request modes
967 and bits policy, thus improving the performance of search through
971 Frequency : only if OST filesystem is corrupted
973 Description: client incorrectly hits assertion in ptlrpc_replay_req()
974 Details : for a short time RPCs with bulk IO are in the replay list,
975 but replay of bulk IOs is unimplemented. If the OST filesystem
976 is corrupted due to disk cache incoherency and then replay is
977 started it is possible to trip an assertion. Avoid putting
978 committed RPCs into the replay list at all to avoid this issue.
981 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
983 Description: System hang on startup
984 Details : This bug allowed the liblustre (e.g. catamount) client to
985 return to the app before handling all startup RPCs. This
986 could leave the node unresponsive to lustre network traffic
987 and manifested as a server ptllnd timeout.
989 Severity : enhancement
991 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
992 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
993 boolean to control whether to print peer debug info when a
994 client's RPC times out.
997 Frequency : only for kernels with patches from Lustre below 1.4.3
999 Description: Remove old rdonly API
1000 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1003 Frequency : only for devices with external journals
1005 Description: Set external device read-only also
1006 Details : During a commanded failover stop, we set the disk device
1007 read-only while the server shuts down. We now also set any
1008 external journal device read-only at the same time.
1011 Frequency : when upgrading from 1.4 while trying to change parameters
1013 Description: The wrong (new) MDC name was used when setting parameters for
1014 upgraded MDT's. Also allows changing of OSC (and MDC)
1015 parameters if --writeconf is specified at tunefs upgrade time.
1018 Frequency : when setting specific ost indicies
1020 Description: QOS code breaks on skipped indicies
1021 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1022 created with --index need not be sequential.
1024 Severity : enhancement
1026 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1027 Details : The uninit_groups feature works in conjunction with the kernel
1028 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1029 up the pass1 processing of e2fsck. This is a read-only feature
1030 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1031 mount filesystems that have had this feature enabled.
1033 Severity : enhancement
1035 Description: Improve multi-block allocation algorithm to avoid fragmentation
1036 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1037 allocation locality and avoid filesystem fragmentation.
1039 ------------------------------------------------------------------------------
1041 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1043 * Support for kernels:
1044 2.4.21-47.0.1.EL (RHEL 3)
1045 2.6.5-7.283 (SLES 9)
1046 2.6.9-42.0.10.EL (RHEL 4)
1047 2.6.12.6 vanilla (kernel.org)
1048 2.6.16.27-0.9 (SLES 10)
1049 * Recommended e2fsprogs version: 1.39.cfs5
1051 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1055 Frequency : occasional, depends on client load and configuration
1056 Bugzilla : 12181, 12203
1057 Description: data loss for recently-modified files
1059 Details : In some cases it is possible that recently written or created
1060 files may not be written to disk in a timely manner (this should
1061 normally be within 30s unless client IO load is very high).
1062 The problem appears as zero-length files or files that are a
1063 multiple of 1MB in size after a client crash or client eviction
1064 that are missing data at the end of the file.
1066 This problem is more likely to be hit on clients where files are
1067 repeatedly created and unlinked in the same directory, clients
1068 have a large amount of RAM, have many CPUs, the filesystem has
1069 many OSTs, the clients are rebooted frequently, and/or the files
1070 are not accessed by other nodes after being written.
1072 The presence of the problem can be detected by looking at
1073 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1074 smaller than the second (nr_unused) then dirty files will not
1075 be flushed automatically to disk. "sync; sleep 10" should be
1076 run several times on the node before unmounting it to update
1077 Lustre (this is also safe to run on nodes without this problem).
1079 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1080 that can cause this same problem, so customers using that kernel
1081 also need to update the kernel in addition to Lustre. In order
1082 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1084 It is normal that files written just before a client crash (less
1085 than 30s) may not yet have been flushed to disk, even for local
1089 Frequency : frequent on thin XT3 nodes
1091 Description: UUID collision on thin XT3 Linux nodes
1092 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1093 randomly, since we relied on an insufficiently-seeded PRNG.
1098 Description: OSS hangs after "All ost request buffers busy"
1099 Details : A deadlock between quota and journal operations caused OSS
1100 hangs after printing "All ost request buffers busy."
1103 Frequency : always on liblustre builds
1105 Description: Cleanup compiler warnings on liblustre
1108 Frequency : always on liblustre builds on XT3
1110 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1114 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1115 > 32000 subdirectories
1116 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1117 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1118 subdirectory count crosses 32000. This will aid e2fsck to
1119 correctly handle more than 32000 subdirectories.
1122 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1124 Description: System hang on startup
1125 Details : This bug allowed the liblustre (e.g. catamount) client to
1126 return to the app before handling all startup RPCs. This
1127 could leave the node unresponsive to lustre network traffic
1128 and manifested as a server ptllnd timeout.
1130 Severity : enhancement
1132 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1133 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1134 boolean to control whether to print peer debug info when a
1135 client's RPC times out.
1140 Description: make O_SYNC working on 2.6 kernels
1141 Details : 2.6 kernels use different method for mark pages for write,
1142 so need add a code to lustre for O_SYNC work.
1147 Description: Failure to close file and release space on NFS
1148 Details : Put inode details into lock acquired in ll_intent_file_open.
1149 Use mdc_intent_lock in ll_intent_open to properly
1150 detect all kind of errors unhandled by mdc_enqueue.
1155 Description: proc file read during shutdown sometimes raced obd removal,
1157 Details : Add lock to prevent obd access after proc file removal.
1160 Frequency : Only for files larger than 4GB on 32-bit clients.
1162 Description: improperly doing page alignment of locks
1163 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1164 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1169 Description: RPCs being resent when they shouldn't be
1170 Details : Some RPCs that should not be resent are being resent. This
1171 can cause inconsistencies in the RPC state machine. Do not
1172 resend such requests.
1175 Frequency : rare, only with NFS export
1177 Description: Crash on NFS re-export node
1178 Details : under very unusual load conditions an assertion is hit in
1179 ll_intent_file_open()
1182 Frequency : only if OST filesystem is corrupted
1184 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1185 Details : for a short time RPCs with bulk IO are in the replay list,
1186 but replay of bulk IOs is unimplemented. If the OST filesystem
1187 is corrupted due to disk cache incoherency and then replay is
1188 started it is possible to trip an assertion. Avoid putting
1189 committed RPCs into the replay list at all to avoid this issue.
1194 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1195 Details : Large single O_DIRECT read and write calls can fail to allocate
1196 a sufficiently large buffer to process the request. In case of
1197 allocation failure the allocation is retried with a smaller
1198 buffer and broken into smaller requests.
1200 Severity : enhancement
1202 Description: Add -o localflock option to simulate old noflock behaviour.
1203 Details : This will achieve local-only flock/fcntl locks coherentness.
1208 Description: versioning check is incomplete
1209 Details : Checking the version difference of client vs. server, report
1210 error if the gap is too big.
1215 Description: add support PG_writeback bit
1216 Details : add support for PG_writeback bit for Lustre, for more carefull
1217 work with page cache in 2.6 kernel. This also fix some deadlocks
1218 and remove hack for work O_SYNC with 2.6 kernel.
1220 Severity : enhancement
1222 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1223 Details : The uninit_groups feature works in conjunction with the kernel
1224 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1225 up the pass1 processing of e2fsck. This is a read-only feature
1226 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1227 mount filesystems that have had this feature enabled.
1229 Severity : enhancement
1231 Description: Improve multi-block allocation algorithm to avoid fragmentation
1232 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1233 allocation locality and avoid filesystem fragmentation.
1235 ------------------------------------------------------------------------------
1237 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1239 * Support for kernels:
1240 2.6.9-42.0.3.EL (RHEL 4)
1241 2.6.5-7.276 (SLES 9)
1242 2.4.21-47.0.1.EL (RHEL 3)
1243 2.6.12.6 vanilla (kernel.org)
1244 2.6.16.21-0.8 (SLES10)
1245 * Recommended e2fsprogs version: 1.39.cfs2-0
1247 * The backwards-compatible /proc/sys/portals symlink has been removed
1248 in this release. Before upgrading, please ensure that you change
1249 any configuration scripts or /etc/sysctl.conf files that access
1250 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1251 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1252 in advance of the upgrade on any system running Lustre 1.4.6 or
1253 newer, since /proc/sys/lnet was added in that version.
1254 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1258 Frequency : only when quota is used
1260 Description: avoid scanning export list for quota master
1261 Details : Change the algorithms to avoid scanning export list in order
1262 to improve the efficiency.
1265 Frequency : MDS failover only, very rarely
1267 Description: "went back in time" messages on mds failover
1268 Details : The greatest transno may be lost when the current operation
1269 finishes with an error (transno==0) and the client's last_rcvd
1270 record is over-written. Save the greatest transno in the
1271 mds_last_transno for this case.
1274 Frequency : always for specific kernels and striping counts
1276 Description: client may get "Matching packet too big" without ACL support
1277 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1278 when trying to access files in certain configurations. The
1279 clients should in fact be denied when mounting because they do
1280 not understand ACLs.
1283 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1285 Description: many clients connecting with IO in progress causes connect timeouts
1286 Details : Avoid synchronous journal commits to avoid delays caused by many
1287 clients connecting/disconnecting when bulk IO is in progress.
1288 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1289 OST_IO_PORTAL to avoid delays behind potentially many pending
1293 Frequency : occasionally with multiple writers to a single file
1295 Description: shared writes to file may result in wrong size reported by stat()
1296 Details : Allow growing of kms when extent lock is cancelled
1299 Frequency : always with random mmap IO to multi-striped file
1301 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1302 Details : Only if the hole is at the end of OST object so that kms is too
1303 small. Fix is to increase kms accordingly in ll_nopage.
1306 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1308 Description: writes to a missing object would leak memory on the OST
1309 Details : If there is an inconsistency between the MDS and OST filesystems,
1310 such that the MDS references an object that doesn't exist, writes
1311 to that object will leak memory due to incorrect cleanup in the
1312 error handling path, eventually running out of memory on the OST.
1317 Description: Creating too long symlink causes lustre errors
1318 Details : Check symlink and name lengths before sending requests to MDS.
1321 Frequency : only if flock is enabled (not on by default)
1323 Description: posix locks not released on fd closure on 2.6.9+
1324 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1325 kernels, this caused such locks not to be released on fd close and
1326 then assertions on fs unmount about still used locks.
1329 Frequency : MDS failover only, very rarely
1331 Description: clients may get ASSERTION(granted_lock != NULL)
1332 Details : When request was taking a long time, and a client was resending
1333 a getattr by name lock request. The were multiple lock requests
1334 with the same client lock handle and
1335 mds_getattr_name->fixup_handle_for_resent_request found one of the
1336 lock handles but later failed with ASSERTION(granted_lock != NULL).
1341 Description: handle->h_buffer_credits > 0, assertion failure
1342 Details : h_buffer_credits is zero after truncate, causing assertion
1343 failure. This patch extends the transaction or creates a new
1347 Frequency : NFS re-export or patchless client
1348 Bugzilla : 11179, 10796
1349 Description: Crash on NFS re-export node (__d_move)
1350 Details : We do not want to hash the dentry if we don't have a lock.
1351 But if this dentry is later used in d_move, we'd hit uninitialised
1352 list head d_hash, so we just do this to init d_hash field but
1353 leave dentry unhashed.
1356 Frequency : NFS re-export or patchless client
1358 Description: NFS exports has problem with symbolic link
1359 Details : lustre client didn't properly install dentry when re-exported
1360 to NFS or running patchless client.
1363 Frequency : NFS re-export or patchless client
1365 Description: Various nfs/patchless fixes.
1366 Details : fixes reuse disconected alias for lookup process - this fixes
1367 warning "find_exported_dentry: npd != pd",
1368 fix permission error with open files at nfs.
1369 fix apply umask when do revalidate.
1372 Frequency : occasional
1374 Description: Crash on NFS re-export node
1375 Details : calling clear_page() on the wrong pointer triggered oops in
1376 generic_mapping_read().
1379 Frequency : rarely, using O_DIRECT IO
1381 Description: unaligned directio crashes client with LASSERT
1382 Details : check for unaligned buffers before trying any requests.
1385 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1387 Description: stale data returned from RAID cache
1388 Details : If only a small amount of IO is done to the RAID device before
1389 reading it again it is possible to get stale data from the RAID
1390 cache instead of reading it from disk.
1393 Frequency : always for sles10 kernel
1395 Description: sles10 support
1396 Details : ll_follow_link: compile fixes and using of nd_set_link
1397 under newer kernels.
1400 Frequency : depends on arch, kernel and compiler version, always on sles10
1403 Description: recursive or deep enough symlinks cause stack overflow
1404 Details : getting rid of large stack-allocated variable in
1408 Frequency : depends on hardware
1410 Description: lustre write performance loss in the SLES10 kernel
1411 Details : the performance loss is caused by using of write barriers in the
1412 ext3 code. The SLES10 kernel turns barrier support on by
1413 default. The fix is to undo that change for ldiskfs.
1415 ------------------------------------------------------------------------------
1417 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1419 * Support for kernels:
1420 2.6.9-42.0.3EL (RHEL 4)
1421 2.6.5-7.276 (SLES 9)
1422 2.4.21-47.0.1.EL (RHEL 3)
1423 2.6.12.6 vanilla (kernel.org)
1427 Frequency : quota enabled and large files being deleted
1429 Description: releasing more than 4GB of quota at once hangs OST
1430 Details : If a user deletes more than 4GB of files on a single OST it
1431 will cause the OST to spin in an infinite loop. Release
1432 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1437 Description: statfs data retrieved from /proc may be stale or zero
1438 Details : When reading per-device statfs data from /proc, in the
1439 {kbytes,files}_{total,free,avail} files, it may appear
1440 as zero or be out of date.
1443 Frequency : systems with MD RAID1 external journal devices
1445 Description: lconf's call to blkid is confused by RAID1 journal devices
1446 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1447 of returning all block devices that match the journal UUID.
1450 Frequency : always, for aggregate stripe size over 4GB
1452 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1453 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1454 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1457 Frequency : always if "lfs find" used on a local file/directory
1459 Description: "lfs find" segfaults if used on a local file/directory
1460 Details : The case where a directory component was not specified wasn't
1461 handled correctly. Handle this properly.
1464 Frequency : always on ppc64
1466 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1467 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1468 perform endianness conversion.
1471 Frequency : rarely (truncate to non-zero file size after write under load)
1472 Bugzilla : 10730, 10687
1473 Description: Files padded with zeros to next 4K multiple
1474 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1475 it is possible that files that are truncated to a non-zero size
1476 immediately after being written are filled with zero bytes beyond
1477 the truncated size. No file data is lost.
1479 Severity : enhancement
1481 Description: Allow recovery/failover for liblustre clients.
1482 Details : liblustre clients were unaware of failover configurations until
1485 Severity : enhancement
1487 Description: user file locks should fail when not mounting with flock option
1488 Details : Set up an error-returning stub in ll_file_operations.lock field
1489 to prevent incorrect behaviour when client is mounted without
1490 flock option. Also, set up properly f_op->flock field for
1494 Frequency : always on ia64
1496 Description: "lfs df" loops on printing out MDS statfs information
1497 Details : The obd_ioctl_data was not initialized and in some systems
1498 this caused a failure during the ioctl that did not return
1499 an error. Initialize the struct and return an error on failure.
1502 Frequency : SLES 9 only
1504 Description: Error of copying files with lustre special EAs as root
1505 Details : Client side always return success for setxattr call for lustre
1506 special xattr (currently only "trusted.lov").
1509 Frequency : rarely on clusters with both ia64+i386 clients
1511 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1512 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1513 clients are doing concurrent writes to the same file it is
1514 possible that the ia64 clients may LASSERT because the OST
1515 extent locks are not PAGE_SIZE aligned. Ensure that grown
1516 locks are always aligned on the request boundary.
1519 Frequency : specific use, occasional
1521 Description: Overwriting in use executable truncates on-disk binary image
1522 Details : If one node attempts to overwrite an executable in use by
1523 another node, we now correctly return ETXTBSY instead of
1524 truncating the file.
1526 Severity : enhancement
1528 Description: Async OSC create to avoid the blocking unnecessarily.
1529 Details : If a OST has no remain object, system will block on the creating
1530 when need to create a new object on this OST. Now, ways use
1531 pre-created objects when available, instead of blocking on an
1532 empty osc while others are not empty. If we must block, we block
1533 for the shortest possible period of time.
1538 Description: chmod on Lustre root is propagated to other clients
1539 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1540 invalid by the follow_mount time.
1543 Frequency : liblustre clients only
1545 Description: Race in 'instant cancel' lock handling could lead to such locks
1546 never to be granted in case of SMP MDS
1547 Details : Do not destroy not yet granted but cbpending locks in
1551 Frequency : replay/resend of open
1553 Description: non null lock assetion failure in mds_intent_policy
1554 Details : Trying to replay/resend lockless open requests resulted in
1555 mds_open() returning 0 with no lock. Now it sets a flag if
1556 a lock is going to be returned.
1558 Severity : enhancement
1560 Description: Checksum enhancements
1561 Details : New checksum enhancements allow for resending RPCs that failed
1564 Severity : enhancement
1566 Description: Tunables on number of dirty pages in cacche
1567 Details : Allow to set limit on number of dirty pages cached.
1572 Description: client crash on unmount - lock still has references
1573 Details : In some error handling cases it was possible to leak a lock
1574 reference on a client while accessing a file. This was not
1575 harmful to the client during operation, but would cause the
1576 client to crash when the filesystem is unmounted.
1579 Frequency : specific case, rare
1581 Description: ETXTBSY on mds though file not in use by client
1582 Details : ETXTBSY is no longer incorrectly returned when attempting to
1583 chmod or chown a directory that the user previously tried to
1584 execute or a currently-executing binary.
1587 Frequency : extremely rare except on liblustre-based clients
1589 Description: Lustre space not freed when files are deleted
1590 Details : Clean up open-unlinked files after client eviction. Previously
1591 the unlink was skipped and the files remained as orphans.
1596 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1597 Details : In some cases it was possible to send a blocking callback to
1598 a client doing a glimpse, even though that client didn't get
1599 a lock granted. When the glimpse lock is cancelled on the OST
1600 the freed lock is left on the waiting list and corrupted the list.
1603 Frequency : all core dumps
1605 Description: Broke core dumps to lustre
1606 Details : Negative dentry may be unhashed if parent does not have UPDATE
1607 lock, but some callers, e.g. do_coredump, expect dentry to be
1608 hashed after successful create, hash it in ll_create_it.
1610 ------------------------------------------------------------------------------
1612 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1614 * Support for kernels:
1615 2.6.9-42.0.2.EL (RHEL 4)
1616 2.6.5-7.276 (SLES 9)
1617 2.4.21-40.EL (RHEL 3)
1618 2.6.12.6 vanilla (kernel.org)
1622 Frequency : always on RHEL 3
1624 Description: Number of open files grows over time
1625 Details : The number of open files grows over time, whether or not
1626 Lustre is started. This was due to a filp leak introduced
1627 by one of our kernel patches.
1629 ------------------------------------------------------------------------------
1631 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1633 * Support for kernels:
1634 2.6.9-42.EL (RHEL 4)
1635 2.6.5-7.276 (SLES 9)
1636 2.4.21-40.EL (RHEL 3)
1637 2.6.12.6 vanilla (kernel.org)
1642 Bugzilla : 5719, 9635, 9792, 9684
1643 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1644 Details : If a server is under heavy load and cannot reply to new
1645 connection requests before the client resends the (re)connect,
1646 the connection handling code can behave badly if two service
1647 threads are concurrently handing separate (re)connections from
1648 the same client. Add better locking to the connection handling
1649 code, and ensure that only a single connection will be processed
1650 for a given client UUID, even if the lock is dropped.
1652 Severity : enhancement
1654 Description: add TCP zero-copy support to kernel
1655 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1656 sends if the hardware supports scatter-gather and checksumming.
1657 This allows socklnd to do client-write and server-read more
1658 efficiently and reduce CPU utilization from skbuf copying.
1661 Frequency : only if NFS exporting from client
1663 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1664 Details : If setting attributes on a file created under NFS that had
1665 never been opened it would be possible to oops the client
1666 if the file had no objects.
1669 Frequency : always for liblustre
1671 Description: liblustre client does MDS+OSTs setattr RPC for each write
1672 Details : When doing a write from a liblustre client, the client
1673 incorrectly issued an RPC to the MDS and each OST the file was
1674 striped over in order to update the timestamps. When writing
1675 with small chunks and many clients this could overwhelm the MDS
1676 with RPCs. In all cases it would slow down the write because
1677 these RPCs are unnecessary.
1679 Severity : enhancement
1681 Description: allow number of MDS service threads to be changed at module load
1682 Details : It is now possible to change the number of MDS service threads
1683 running. Adding "options mds mds_num_threads={N}" to the MDS's
1684 /etc/modprobe.conf will set the number of threads for the next
1685 time Lustre is restarted (assuming the "mds" module is also
1686 reloaded at that time). The default number of threads will
1687 stay the same, 32 for most systems.
1692 Description: OST crash if filesystem is unformatted or corrupt
1693 Details : If an OST is started on a device that has never been formatted
1694 or if the filesystem is corrupt and cannot even mount then the
1695 error handling cleanup routines would dereference a NULL pointer.
1700 Description: NULL pointer deref in llap_from_page.
1701 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1702 incorrect) mapping if the page was truncated/reclaimed while it was
1703 searched for. Check for this condition and skip such pages when
1704 doing readahead. Introduce extra check to llap_from_page() to
1705 verify page->mapping->host is non-NULL (so page is not anonymous).
1708 Frequency : Sometimes when using sys_sendfile
1710 Description: "page not covered by a lock" warnings from ll_readpage
1711 Details : sendfile called ll_readpage without right page locks present.
1712 Now we introduced ll_file_sendfile that does necessary locking
1713 around call to generic_file_sendfile() much like we do in
1717 Frequency : with certain MDS communication failures at client mount time
1719 Description: NULL pointer deref after failed client mount
1720 Details : a client connection request may delayed by the network layer
1721 and not be sent until after the PTLRPC layer has timed out the
1722 request. If the client fails the mount immediately it will try
1723 to clean up before the network times out the request. Add a
1724 reference from the request import to the obd device and delay
1725 the cleanup until the network drops the request.
1728 Frequency : occasionally during client (re)connect
1730 Description: assertion failure during client (re)connect
1731 Details : processing a client connection request may be delayed by the
1732 client or server longer than the client connect timeout. This
1733 causes the client to resend the connection request. If the
1734 original connection request is replied in this interval, the
1735 client may trip an assertion failure in ptlrpc_connect_interpret()
1736 which thought it would be the only running connect process.
1739 Frequency : only with obd_echo servers and clients that are rebooted
1741 Description: kernel BUG accessing uninitialized data structure
1742 Details : When running an obd_echo server it did not start the ping_evictor
1743 thread, and when a client was evicted an uninitialized data
1744 structure was accessed. Start the ping_evictor in the RPC
1745 service startup instead of the OBD startup.
1747 Severity : enhancement
1748 Bugzilla : 10193 (patchless)
1749 Description: Remove dependency on various unexported kernel interfaces.
1750 Details : No longer need reparent_to_init, exit_mm, exit_files,
1751 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1754 Frequency : rare (only users of deprecated and unsupported LDAP config)
1756 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1757 Details : LDAP apparently contains 'lustreName' attributes instead of
1758 'name'. A simple remapping of the name is sufficient.
1761 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1763 Description: waiting_locks_callback trips kernel BUG if client is evicted
1764 Details : Running with the dump_on_timeout debug flag turned on makes
1765 it possible that the waiting_locks_callback() can try to dump
1766 the Lustre kernel debug logs from an interrupt handler. Defer
1767 this log dumping to the expired_lock_main() thread.
1769 Severity : enhancement
1771 Description: Support NFS exporting on 2.6 kernels.
1772 Details : Implement non-rawops metadata methods for NFS server to use without
1773 changing NFS server code.
1776 Frequency : very rare (synthetic metadata workload only)
1778 Description: two racing renames might cause an MDS thread to deadlock
1779 Details : Running the "racer" program may cause one MDS thread to rename
1780 a file from being the source of a rename to being the target of
1781 a rename at exactly the same time that another thread is doing
1782 so, and the second thread has already enqueued these locks after
1783 doing a lookup of the target and is trying to relock them in
1784 order. Ensure that we don't try to re-lock the same resource.
1787 Frequency : only very large systems with liblustre clients
1789 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1790 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1791 all OSTs in parallel. This allows the eviction of stale liblustre
1792 clients to proceed much faster than if they were done in series,
1793 and also offers similar improvements for other set_info RPCs.
1798 Description: excessive CPU usage during initial read phase on client
1799 Details : During the initial read phase on a client, it would agressively
1800 retry readahead on the file, consuming too much CPU and impacting
1801 performance (since 1.4.5.8). Improve the readahead algorithm
1802 to avoid this, and also improve some other common cases (read
1803 of small files in particular, where "small" is files smaller than
1804 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1809 Description: MDS crash when receiving packet with unknown intent.
1810 Details : Do not LBUG in unknown intent case, just return -EFAULT
1812 Severity : enhancement
1813 Bugzilla : 9293, 9385
1814 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1815 Details : Do not serialize getattr (non-intent version) and statfs.
1818 Frequency : occasional, when OST network is overloaded/intermittent
1820 Description: client evicted by OST after bulk IO timeout
1821 Details : If a client sends a bulk IO request (read or write) the OST
1822 may evict the client if it is unresposive to its data GET/PUT
1823 request. This is incorrect if the network is overloaded (takes
1824 too long to transfer the RPC data) or dropped the OST GET/PUT
1825 request. There is no need to evict the client at all, since
1826 the pinger and/or lock callbacks will handle this, and the
1827 client can restart the bulk request.
1830 Frequency : Always when mmapping file with no objects
1832 Description: client crashes when mmapping file with no objects
1833 Details : Check that we actually have objects in a file before doing any
1834 operations on objects in ll_vm_open, ll_vm_close and
1840 Description: Request leak when working with deleted CWD
1841 Details : Introduce advanced request refcount tracking for requests
1842 referenced from lustre intent.
1844 Severity : Enhancement
1846 Description: Cache open file handles on client.
1847 Details : MDS now will return special lock along with openhandle, if
1848 requested and client is allowed to hold openhandle, even if unused,
1849 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1850 closing files for every read/write openration.
1852 Severity : Enhancement
1854 Description: Cache open negative dentries on client when possible.
1855 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1856 negative dentries on lock revocation.
1861 Description: Remounting a client read-only wasn't possible with a zconf mount
1862 Details : It wasn't possible to remount a client read-only with llmount.
1864 Severity : enhancement
1865 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1866 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1867 MPICH is included to add Lustre-specific ADIO interfaces.
1868 This is based closely on the UFS ADIO layer and only differs
1869 in file creation, in order to allow the OST striping to be set.
1870 This is user-contributed code and not supported by CFS.
1875 Description: extended inode attributes (immutable, append-only) work improperly
1876 when 2.4 and 2.6 kernels are used on client/server or vice versa
1877 Details : Introduce kernel-independent values for these flags.
1879 Severity : enhancement
1882 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1883 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1884 granularity. Now, Fractional MB granularity is be supported,
1885 this is very useful for low memory system.
1887 Severity : enhancement
1889 Description: Getattr by fid
1890 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1891 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1892 have locks and avoids getattr rpc requests for it on every stat.
1895 Frequency : Always, for filesystems larger than 2TB
1897 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1898 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1899 filesystems larger than 2TB to be mounted with mballoc enabled.
1902 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1904 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1905 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1906 filesystems larger than 2TB is mounted then the kernel will
1907 silently corrupt the start of the filesystem. CONFIG_LBD is
1908 enabled for all CFS-supported kernels, but the possibility of
1909 this happening with a modified kernel config exists.
1911 Severity : enhancement
1913 Description: add client O_DIRECT support for 2.6 kernels
1914 Details : It is now possible to do O_DIRECT reads and writes to files
1915 in the Lustre client mountpoint on 2.6 kernel clients.
1917 Severity : enhancement
1919 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1920 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1921 parallel, not waiting for response from every OST before sending
1922 a rpc to the next OST.
1927 Description: setattr vs write race when updating file timestamps
1928 Details : Client processes that update a file timestamp into the past
1929 right after writing to the file (e.g. tar) it is possible that
1930 the updated file modification time can be reset to the current
1931 time due to a race between processing the setattr and write RPC.
1933 Severity : enhancement
1935 Description: Bring 'lfs find' closer in line with regular Linux find.
1936 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1937 -print0 options and obtains all the needed info through the lustre
1940 Severity : enhancement
1942 Description: support up to 1024 configured devices on one node
1943 Details : change obd_dev array from statically allocated to dynamically
1944 allocated structs as they are first used to reduce memory usage
1949 Description: Flush dirty partially truncated pages during truncate
1950 Details : Immediatelly flush partially truncated pages in filter_setattr,
1951 this way we completely avoid having any pages in page cache on OST
1952 and can retire ugly workarounds during writes to flush such pages.
1957 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1958 Details : protect inode from truncation within vfs_unlink() context
1959 just take a reference before calling vfs_unlink() and release it
1960 when parent's i_sem is free.
1965 Description: last_id value checked outside lock on OST caused LASSERT failure
1966 Details : If there were multiple MDS->OST object precreate requests in
1967 flight, it was possible that the OST's last object id was checked
1968 outside a lock and incorrectly tripped an assertion. Move checks
1969 inside locks, and discard old precreate requests.
1972 Frequency : always, if extents are used on OSTs
1974 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1975 Details : OSTs using the extents format would not zero the high 16 bits of
1976 the index physical block number. This is not a problem for any
1977 OST filesystems smaller than 16TB, and no kernels support ext3
1978 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1979 new/modified files) and can be fixed for existing filesystems
1980 with e2fsprogs-1.39-cfs1.
1985 Description: import connection selection may be incorrect if timer wraps
1986 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1987 import connections to be ignored if the 32-bit jiffies counter
1988 wraps. Use a 64-bit jiffies counter.
1991 Frequency : very large clusters immediately after boot
1993 Description: LNET request buffers exhausted under heavy short-term load
1994 Details : If a large number of client requests are generated on a service
1995 that has previously never seen so many requests it is possible
1996 that the request buffer growth cannot keep up with the spike in
1997 demand. Instead of dropping incoming requests, they are held in
1998 the LND until the RPC service can accept more requests.
2001 Frequency : Sometimes during replay
2003 Description: Assertion failure in ll_local_open after replay.
2004 Details : If replay happened on an open request reply before we were able
2005 to set replay handler, reply will become not swabbed tripping the
2006 assertion in ll_local_open. Now we set the handler right after
2007 recognising of open request
2010 Frequency : very rare
2012 Description: kernel reports "badness in vsnprintf"
2013 Details : Reading from the "recovery_status" /proc file in small chunks
2014 may cause a negative length in lprocfs_obd_rd_recovery_status()
2015 call to vsnprintf() (which is otherwise harmless). Exit early
2016 if there is no more space in the output buffer.
2018 Severity : enhancement
2020 Description: clear OBD RPC statistics by writing to them
2021 Details : It is now possible to clear the OBD RPC statistics by writing
2022 to the "stats" file.
2027 Description: Client mtime is not the same on different clients after utimes
2028 Details : In some cases, the client was using the utimes() syscall on
2029 a file cached on another node. The clients now validate the
2030 ctime from the MDS + OSTs to determine which one is right.
2035 Description: Inability to activate failout mode
2036 Details : lconf script incorrectly assumed that in pythong string's numeric
2037 value is used in comparisons.
2040 Frequency : always with multiple stripes per file
2042 Description: Inefficient object allocation for mutli-stripe files
2043 Details : When selecting which OSTs to stripe files over, for files with
2044 a stripe count that divides evenly into the number of OSTs,
2045 the MDS is always picking the same starting OST for each file.
2046 Return the OST selection heuristic to the original design.
2051 Description: mount failures may take full timeout to return an error
2052 Details : Under some heavy load conditions it is possible that a
2053 failed mount can wait for the full obd_timeout interval,
2054 possibly several minutes, before reporting an error.
2055 Instead return an error as soon as the status is known.
2057 Frequency : quota enabled and large files being deleted
2059 Description: releasing more than 4GB of quota at once hangs OST
2060 Details : If a user deletes more than 4GB of files on a single OST it
2061 will cause the OST to spin in an infinite loop. Release
2062 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2067 Description: statfs data retrieved from /proc may be stale or zero
2068 Details : When reading per-device statfs data from /proc, in the
2069 {kbytes,files}_{total,free,avail} files, it may appear
2070 as zero or be out of date.
2073 Frequency : systems with MD RAID1 external journal devices
2075 Description: lconf's call to blkid is confused by RAID1 journal devices
2076 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2077 of returning all block devices that match the journal UUID.
2080 Frequency : always, for aggregate stripe size over 4GB
2082 Description: assertion fails when trying to use 4GB stripe size
2083 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2084 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2087 Frequency : always on ppc64
2089 Description: the first write on an ext3 filesystem with mballoc got stuck
2090 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2091 perform endianness conversion.
2093 ------------------------------------------------------------------------------
2095 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2097 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2098 INTEROPERATE with older versions automatically. Please read the
2099 user documentation before upgrading any part of a live system.
2100 * WARNING: Lustre networking configuration changes are required with
2101 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2104 * Support for kernels:
2105 2.6.9-22.0.2.EL (RHEL 4)
2106 2.6.5-7.244 (SLES 9)
2107 2.6.12.6 vanilla (kernel.org)
2110 Severity : enhancement
2111 Bugzilla : 7981/8208
2112 Description: Introduced Lustre Networking (LNET)
2113 Details : LNET is new networking infrastructure for Lustre, it includes
2114 a reorganized network configuration mode (see the user
2115 documentation for full details) as well as support for routing
2116 between different network fabrics. Lustre Networking Devices
2117 (LNDs) for the supported network fabrics have also been
2118 created for this new infrastructure.
2120 Severity : enhancement
2121 Description: Introduced Access control lists
2122 Details : clients can set ACLs on files and directories in order to have
2123 more fine-grained permissions than the standard Unix UGO+RWX.
2124 The MDS must be started with the "-o acl" mount option.
2126 Severity : enhancement
2127 Description: Introduced filesystem quotas
2128 Details : Administrators may now establish per-user quotas on the
2131 Severity : enhancement
2133 Description: Configuration change for the XT3
2134 The PTLLND is now used to run Lustre over Portals on the XT3
2135 The configure option(s) --with-cray-portals are no longer used.
2136 Rather --with-portals=<path-to-portals-includes> is used to
2137 enable building on the XT3. In addition to enable XT3 specific
2138 features the option --enable-cray-xt3 must be used.
2143 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2144 Details : A race between allocating a new llcd and re-getting the llcd_lock
2145 allowed another thread to grab newly-allocated llcd.
2147 Severity : enhancement
2149 Description: 2.6 OST async journal commit and locking fix to improve performance
2150 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2151 2.6 kernels are now async as they already were in 2.4 kernels so
2152 that they can commit concurrently with the network bulk transfer.
2153 For block-allocated files the filter allocation semaphore is held
2154 to avoid filesystem fragmentation during allocation. BKL lock
2155 removed for 2.6 xattr operations where it is no longer needed.
2160 Description: lconf incorrectly determined whether two IP networks could talk
2161 Details : In some more complicated routing and multiple-network
2162 configurations, lconf will avoid trying to make a network
2163 connection to a disjoint part of the IP space. It was doing the
2164 math incorrectly for one set of cases.
2169 Description: Fix for potential infinite loop processing records in an llog.
2170 Details : If an llog record is corrupted/zeroed, it is possible to loop
2171 forever in llog_process(). Validate the llog record length
2172 and skip the remainder of the block on error.
2175 Frequency : occasional (liblustre only)
2177 Description: liblustre could not open files whose last component is a symlink
2178 Details : sysio_path_walk() would incorrectly pass the open intent to
2179 intermediate path components.
2182 Frequency : rare (liblustre only with non-standard tuning)
2183 Bugzilla : 7201 (7350)
2184 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2185 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2186 after it has been released, only do this for referenced locks
2188 Severity : enhancement
2190 Description: specifying an (invalid) directory default stripe_size of -1
2191 would reset the directory default striping
2192 Details : stripe_size -1 was used internally to signal directory stripe
2193 removal, now use "all default" to signal dir stripe removal
2194 as a directory striping of "all default" is not useful
2197 Frequency : common for large clusters running liblustre clients
2199 Description: doing an ls when liblustre clients are running is slow
2200 Details : sending a glimpse AST to a liblustre client waits for every AST
2201 to time out, as liblustre clients will not respond. Since they
2202 cannot cache data we refresh the OST lock LVB from disk instead.
2204 Severity : enhancement
2206 Description: doing an ls at the same time as file IO can be slow
2207 Details : enqueue and other "small" requests can be blocked behind many
2208 large IO requests. Create a new OST IO portal for non-IO
2209 requests so they can be processed faster.
2212 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2214 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2215 created file with mode 000
2216 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2217 allow an ftruncate() as a non-root user to a file with mode 000.
2218 Lustre now respects this flag to disable mode checks when
2219 truncating a file owned by the user
2222 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2225 Description: Revoking locks from clients that went dead or catatonic might take
2227 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2228 cancellation of such locks instant on servers without waiting for
2229 any reply from clients. Clients drops these locks when cancel
2230 notification from server is received without replying.
2233 Frequency : liblustre-only, when liblustre client dies or becomes busy
2235 Description: Doing ls on Linux clients can take a long time with active
2237 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2238 granting such locks to it in the first place if possible. Locks
2239 are taken by proxy on the OST during the read or write and
2240 dropped immediately afterward. Add connect flags handling, do
2241 not grant locks to liblustre clients for glimpse ASTs.
2243 Severity : enhancement
2245 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2246 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2247 sequential reads and behaves badly with applications doing random
2248 reads. Improve it by reading ahead at least read region, and
2249 avoiding excessive large RPC for small reads.
2251 Severity : enhancement
2253 Description: Creating more than 1000 files for a single job may cause a load
2254 imbalance on the OSTs if there are also a large number of OSTs.
2255 Details : qos_prep_create() uses an OST index reseed value that is an
2256 even multiple of the number of available OSTs so that if the
2257 reseed happens in the middle of the object allocation it will
2258 still utilize the OSTs as uniformly as possible.
2263 Description: OST or MDS may oops in ping_evictor_main()
2264 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2265 but doesn't restart at beginning of obd_exports_timed list
2268 Severity : enhancement
2270 Description: improve by-nid export eviction on the MDS and OST
2271 Details : allow multiple exports with the same NID to be evicted at one
2272 time without re-searching the exports list.
2275 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2277 Description: MDS may oops in groups_free()
2278 Details : in rare race conditions a newly allocated group_info struct is
2279 freed again, and this can be NULL. The 2.4 compatibility code
2280 for groups_free() checked for a NULL pointer, but 2.6 did not.
2283 Frequency : common for liblustre clients doing little filesystem IO
2284 Bugzilla : 9352, 7313
2285 Description: server may evict liblustre clients accessing contended locks
2286 Details : if a client is granted a lock or receives a completion AST
2287 with a blocking AST already set it would not reply to the AST
2288 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2291 Frequency : lfs setstripe, only systems with more than 160 OSTs
2293 Description: unable to set striping with a starting offset beyond OST 160
2294 Details : llapi_create_file() incorrectly limited the starting stripe
2295 index to the maximum single-file stripe count.
2298 Frequency : LDAP users only
2300 Description: lconf did not handle in-kernel recovery with LDAP properly
2301 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2303 Severity : enhancement
2305 Description: bind OST threads to NUMA nodes to improve performance
2306 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2307 node and do their allocations there to localize memory access
2309 Severity : enhancement
2311 Description: llmount can determine client NID directly from Myrinet (GM)
2312 Details : the client NID code from gmnalnid was moved directly into
2313 llmount, removing the need to use this or specifying the
2314 client NID explicitly when mounting GM clients with zeroconf
2317 Frequency : if client is started with down MDS
2319 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2320 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2321 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2322 ptlrpc_import_delayed_req() because we didn't check if the
2323 request was interrupted, and we also didn't break out of the
2324 event loop if there was no timeout
2329 Description: data loss during non-page-aligned writes to a single file from
2330 both multiple nodes and multiple threads on one node at same time
2331 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2332 inconsistency led to false short-reads, that were cached and later
2333 used by ->prepare_write() to fill in partially written page,
2334 leading to data loss.
2337 Frequency : always, if lconf --abort_recovery used
2339 Description: lconf --abort_recovery fails with 'Operation not supported'
2340 Details : lconf was attempting to abort recovery on the MDT device and not
2343 Severity : enhancement
2345 Description: remove cleanup logs
2346 Details : replace lconf-generated cleanup logs with lustre internal
2347 cleanup routines. Eliminates the need for client-cleanup and
2350 Severity : enhancement
2352 Description: add support for EAs (user and system) on lustre filesystems
2353 Details : it is now possible to store extended attributes in the Lustre
2354 client filesystem, and with the user_xattr mount option it
2355 is possible to allow users to store EAs on their files also
2357 Severity : enhancement
2359 Description: Add possibility (config option) to show minimal available OST free
2361 Details : When compiled with --enable-mindf configure option, statfs(2)
2362 (and so, df) will return least minimal free space available from
2363 all OSTs as amount of free space on FS, instead of summary of
2364 free spaces of all OSTs.
2366 Severity : enhancement
2368 Description: do not expand extent locks acquired on OST-side
2369 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2370 by server: they are not cached anyway.
2373 Frequency : when mmap is used/binaries executed from Lustre
2375 Description: Unmmap pages before throwing them away from read cache.
2376 Details : llap_shrink cache now attempts to unmap pages before discarding
2377 them (if unmapping failed - do not discard). SLES9 kernel has
2378 extra checks that trigger if this unmapping is not done first.
2383 Description: lconf didn't resolve symlinks before checking to see whether a
2384 given mountpoint was already in use
2387 Frequency : when migrating failover services
2388 Bugzilla : 6395, 9514
2389 Description: When migrating a subset of services from a node (e.g. failback
2390 from a failover service node) the remaining services would
2391 time out and evict clients.
2392 Details : lconf --force (implied by --failover) sets the global obd_timeout
2393 to 5 seconds in order to quickly disconnect, but this caused
2394 other RPCs to time out too quickly. Do not change the global
2395 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2397 Severity : enhancement
2398 Frequency : if MDS is started with down OST
2399 Bugzilla : 9439,5706
2400 Description: Allow startup/shutdown of an MDS without depending on the
2401 availability of the OSTs.
2402 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2403 Add appropriate locking and lov-osc refcounts for safe
2404 cleaning. Add osc abort_inflight calls in case the
2405 synchronize never started.
2408 Frequency : occasional (Cray XT3 only)
2410 Description: root not authorized to access files in CRAY_PORTALS environment
2411 Details : The client process capabilities were not honoured on the MDS in
2412 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2413 been accessed by an authorized user then root was able to access
2414 the file on the local client also. The root user capabilities
2415 are now allowed on the MDS, as this environment has secure UID.
2418 Frequency : occasional
2420 Description: ldiskfs "too long searching" message happens too often
2421 Details : A debugging message (otherwise harmless) prints too often on
2422 the OST console. This has been reduced to only happen when
2423 there are fragmentation problems on the filesystem.
2428 Description: Division by zero in statfs when all OSCs are inactive
2429 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2430 lov_statfs divides by value returned by lov_get_stripecnt().
2434 Bugzilla : 9489, 3273
2435 Description: First write from each client to each OST was only 4kB in size,
2436 to initialize client writeback cache, which caused sub-optimal
2437 RPCs and poor layout on disk for the first writen file.
2438 Details : Clients now request an initial cache grant at (re)connect time
2439 and so that they can start streaming writes to the cache right
2440 away and always do full-sized RPCs if there is enough data.
2441 If the OST is rebooted the client also re-establishes its grant
2442 so that client cached writes will be honoured under the grant.
2447 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2448 Details : Now I/O RPCs go to different portal number and (presumably) fast
2449 lock requests (and glimses) and other RPCs get their own service
2450 threads pool that should be able to service those RPCs
2453 Severity : enhancement
2455 Description: Ability to exchange lustre version between client and servers and
2456 issue warnings at client side if client is too old. Also for
2457 liblustre clients there is ability to refuse connection of too old
2459 Details : New 'version' field is added to connect data structure that is
2460 filled with version info. That info is later checked by server and
2464 Frequency : rare, liblustre only.
2465 Bugzilla : 9296, 9581
2466 Description: Two simultaneous writes from liblustre at offset within same page
2467 might proceed at the same time overwriting eachother with stale
2469 Details : I/O lock withing llu_file_prwv was released too early, before data
2470 actually was hitting the wire. Extended lock-holding time until
2471 server acknowledges receiving data.
2474 Frequency : extremely rare. Never observed in practice.
2476 Description: avoid generating lustre_handle cookie of 0.
2477 Details : class_handle_hash() generates handle cookies by incrementing
2478 global counter, and can hit 0 occasionaly (this is unlikely, but
2479 not impossible, because initial value of cookie counter is
2480 selected randonly). Value of 0 is used as a sentinel meaning
2481 "unassigned handle" --- avoid it. Also coalesce two critical
2482 sections in this function into one.
2484 Severity : enhancement
2486 Description: allow liblustre clients to delegate truncate locking to OST
2487 Details : To avoid overhead of locking, liblustre client instructs OST to
2488 take extent lock in ost_punch() on client's behalf. New connection
2489 flag is added to handle backward compatibility.
2491 Severity : enhancement
2492 Bugzilla : 4928, 7341, 9758
2493 Description: allow number of OST service threads to be specified
2494 Details : a module parameter allows the number of OST service threads
2495 to be specified via "options ost ost_num_threads={N}" in the
2496 OSS's /etc/modules.conf or /etc/modprobe.conf.
2500 Bugzilla : 6146, 9635, 9895
2501 Description: servers crash with bad pointer in target_handle_connect()
2502 Details : In rare cases when a client is reconnecting it was possible that
2503 the connection request was the last reference for that export.
2504 We would temporarily drop the export reference and get a new
2505 one, but this may have been the last reference and the export
2506 was just destroyed. Get new reference before dropping old one.
2508 Severity : enhancement
2509 Frequency : if client is started with failover MDS
2511 Description: Allow multiple MDS hostnames in the mount command
2512 Details : Try to read the configuration from all specified MDS
2513 hostnames during a client mount in case the "primary"
2516 Severity : enhancement
2518 Description: Stop sending data to evicted clients as soon as possible.
2519 Details : Check if the client we are about to send or are sending data to
2520 was evicted already. (Check is done every second of waiting,
2521 for which l_wait_event interface was extended to allow checking
2522 of exit condition at specified intervals).
2525 Frequency : rare, normally only when NFS exporting is done from client
2527 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2529 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2532 Frequency : always, if srand() is called before liblustre initialization
2534 Description: Liblustre uses system PRNG disturbing its usage by user application
2535 Details : Introduce internal to lustre fast and high-quality PRNG for
2536 lustre usage and make liblustre and some other places in generic
2537 lustre code to use it.
2539 Severity : enhancement
2540 Bugzilla : 9477, 9557, 9870
2541 Description: Verify that the MDS configuration logs are updated when xml is
2542 Details : Check if the .xml configuration logs are newer than the config
2543 logs stored on the MDS and report an error if this is the case.
2544 Request --write-conf, or allow starting with --old_conf.
2546 Severity : enhancement
2548 Description: Handle symlinks in the path when checking if Lustre is mounted.
2549 Details : Resolve intermediate symlinks when checking if a client has
2550 mounted a filesystem to avoid duplicate client mounts.
2555 Description: lconf can hit an error exception but still return success.
2556 Details : The lconf command catches the Command error exception at the top
2557 level script context and will exit with the associated exit
2558 status, but doesn't ensure that this exit status is non-zero.
2563 Description: failure of ptlrpc thread startup can cause oops
2564 Details : Starting a ptlrpc service thread can fail if there are a large
2565 number of threads or the server memory is very fragmented.
2566 Handle this without oopsing.
2569 Frequency : always, only if liblustre and non-default acceptor port was used
2571 Description: liblustre cannot connect to servers with non-default acceptor port
2572 Details : tcpnal_set_default_params() was not called and was therefore
2573 ignoring the environment varaible TCPNAL_PORT, as well as other
2574 TCPNAL_ environment variables
2579 Description: two objects could be created on the same OST for a single file
2580 Details : If an OST is down, in some cases it was possible to create two
2581 objects on a single OST for a single file. No problems other
2582 than potential performance impact and spurious error messages.
2586 Bugzilla : 5681, 9562
2587 Description: Client may oops in ll_unhash_aliases
2588 Details : Client dcache may become inconsistent in race condition.
2589 In some cases "getcwd" can fail if the current directory is
2595 Description: Inode refcounting problems in NFS export code
2596 Details : link_raw functions used to call d_instantiate without obtaining
2597 extra inode reference first.
2601 Bugzilla : 9942, 9903
2602 Description: Referencing freed requests leading to crash, memleaks with NFS.
2603 Details : We used to require that call to ll_revalidate_it was always
2604 followed by ll_lookup_it. Also with revalidate_special() it is
2605 possible to call ll_revalidate_it() twice for the same dentry
2606 even if first occurence returned success. This fix changes semantic
2607 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2608 reference on a request referred from the intent.
2609 ll_intent_release() then releases such a request.
2612 Frequency : rare, normally benchmark loads only
2614 Description: unlinked inodes were kept in memory on the client
2615 Details : If a client is repeatedly creating and unlinking files it
2616 can accumulate a lot of stale inodes in the inode slab cache.
2617 If there is no other client load running this can cause the
2618 client node to run out of memory. Instead flush old inodes
2619 from client cache that have the same inode number as a new inode.
2622 Frequency : SLES9 2.6.5 kernel and long filenames only
2623 Bugzilla : 9969, 10379
2624 Description: utime reports stale NFS file handle
2625 Details : SLES9 uses out-of-dentry names in some cases, which confused
2626 the lustre dentry revalidation. Change it to always use the
2630 Frequency : rare, unless heavy write-truncate concurrency is continuous
2631 Bugzilla : 4180, 6984, 7171, 9963, 9331
2632 Description: OST becomes very slow and/or deadlocked during object unlink
2633 Details : filter_destroy() was holding onto the parent directory lock
2634 while truncating+unlinking objects. For very large objects this
2635 may block other threads for a long time and slow overall OST
2636 responsiveness. It may also be possible to get a lock ordering
2637 deadlock in this case, or run out of journal credits because of
2638 the combined truncate+unlink. Solution is to do object truncate
2639 first in one transaction without parent lock, and then do the
2640 final unlink in a new transaction with the parent lock. This
2641 reduces the lock hold time dramatically.
2644 Frequency : rare, 2.4 kernels only
2646 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2647 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2648 before unmounting their filesystem in order to avoid deadlock.
2649 The kernel_locked() function in 2.4 kernels only checks whether
2650 the kernel lock is held, not whether it is this process that is
2651 holding it as 2.6 kernels do.
2656 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2657 Details : The client ptlrpc code may be trying to reconnect to a down
2658 server before a previous connection attempt has timed out.
2659 Increase the reconnect interval to be longer than the connection
2660 timeout interval to avoid sending duplicate connections to
2664 Frequency : echo_client brw_test command
2666 Description: fix echo_client to work with OST preallocated code
2667 Details : OST preallocation code (5137) didn't take echo_client IO path
2668 into account: echo_client calls filter methods outside of any
2669 OST thread and, hence, there is no per-thread preallocated
2670 pages and buffers to use. Solution: hijack pga pages for IO. As
2671 a byproduct, this avoids unnecessary data copying.
2675 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2676 Description: Client can oops in mdc_commit_close() after open replay
2677 Details : It was possible for the MDS to return an open request with no
2678 transaction number in mds_finish_transno() if the client was
2679 evicted, but without actually returning an error. Clients
2680 would later try to replay that open and may trip an assertion
2681 Simplify the client close codepath, and always return an error
2682 from the MDS in case the open is not successful.
2685 Frequency : rare, 2.6 OSTs only
2687 Description: OST may deadlock under high load on fragmented files
2688 Details : If there was a heavy load and highly-fragmented OST filesystems
2689 it was possible to have all the OST threads deadlock waiting on
2690 allocation of biovecs, because the biovecs were not released
2691 until the entire RPC IO was completed. Instead, release biovecs
2692 as soon as they are complete to ensure forward IO progress.
2694 Severity : enhancement
2696 Description: Support for specifying external journal device at mount
2697 Details : If an OST or MDS device is formatted with an external journal
2698 device, this device major/minor is stored in the ext3 superblock
2699 and may not be valid for failover. Allow detecting and
2700 specifying the external journal at mount time.
2705 Description: Mounting an MDS with pending unlinked files may cause oops
2706 Details : target_finish_recovery() calls mds_postrecov() which returned
2707 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2708 considers this an error and immediately begins cleaning up the
2709 lov, just after starting the mds_lov process
2711 Severity : enhancement
2713 Description: Implement 'lfs df' to report actual free space on per-OST basis
2714 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2715 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2716 usage of objects; '-h' to report in human readable format.
2718 ------------------------------------------------------------------------------
2720 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2727 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2728 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2729 at startup during orphan unlinking. Instead, skip buddy update
2730 before setup, as it will be regenerated after recovery is complete.
2735 Description: If an OST is inactive, its locks might reference stale inodes.
2736 Details : lov_change_cbdata() must iterate over all namespaces, even if
2737 they are inactive to clear inode references from the lock.
2739 Severity : enhancement
2740 Frequency : occasional, if non-standard max_dirty_mb used
2742 Description: Client will block write RPCs if not enough grant
2743 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2744 then the client will block writes while waiting for another RPC
2745 to complete instead of consuming its dirty limit. With change
2746 we get improved performance when max_dirty_mb is small.
2748 Severity : enhancement
2749 Bugzilla : 3389, 6253
2750 Description: Add support for supplementary groups on the MDS.
2751 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2752 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2753 lookups for user supplementary groups into a cache.
2757 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2758 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2759 for liblustre O_EXCL is enforced by MDS.
2763 Description: readdir never returns NULL in liblustre.
2764 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2765 dentry in struct dirent.
2769 Frequency : liblustre only, depends on application IO pattern
2770 Description: liblustre clients evicted if not contacting servers
2771 Details : Don't put liblustre clients into the ping_evictor list, so
2772 they will not be evicted by the pinger ever.
2774 Severity : enhancement
2776 Description: Add ability to evict clients by NID from MDS.
2777 Details : By echoing "nid:$NID" string into
2778 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2779 $NID would be instantly evicted from this MDS and from all active
2780 OSTs connected to it.
2784 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2785 Details : lookup_it_finish() used to query file size from OSTs that was not
2790 Description: service threads change working directory to that of init
2791 Details : Starting lustre service threads may pin the working directory
2792 of the parent thread, making that filesystem busy. Threads
2793 now change to the working directory of init to avoid this.
2797 Frequency : during shutdown only
2798 Description: shutdown with a failed MDS or OST can cause unmount to hang
2799 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2804 Frequency : 2.6 only
2805 Description: chmod/chown may include an extra supplementary group
2806 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2807 supplementary group and if none is specified this is used.
2810 Bugzilla : 5479 (6816)
2812 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2813 Details : If lookup is in progress on a file that is unlinked we might try
2814 to revalidate the inode and fail in revalidate after lookup is
2815 complete and ll_file_open() enqueues the open again but
2816 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2819 Frequency : always, if lconf --abort_recovery used
2821 Description: lconf --abort_recovery fails with 'Operation not supported'
2822 Details : lconf was attempting to abort recovery on the MDT device and not
2825 ------------------------------------------------------------------------------
2827 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2832 Frequency : rare (only unsupported configurations with a node running as an
2834 Bugzilla : 6514, 5137
2835 Description: Mounting a Lustre file system on a node running as an OST could
2837 Details : OSTs now preallocates memory needed to write out data at
2838 startup, instead of when needed, to avoid having to
2839 allocate memory in possibly low memory situations.
2840 Specifically, if the file system is mounted on on OST,
2841 memory pressure could force it to try to write out data,
2842 which it needed to allocate memory to do. Due to the low
2843 memory, it would be unable to do so and the node would
2844 become unresponsive.
2846 Severity : enhancement
2848 Description: Addition of lconf --service command line option
2849 Details : lconf now accepts a '--service <arg>' option, which is
2850 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2852 Severity : enhancement
2854 Description: Failover mode is now the default for OSTs.
2855 Details : By default, OSTs will now run in failover mode. To return to
2856 the old behaviour, add '--failout' to the lmc line for OSTs.
2858 Severity : enhancement
2860 Description: Health checks are now provided for MDS and OSTs
2861 Details : Additional detailed health check information on MSD and OSTs
2862 is now provided through the procfs health_check value.
2865 Frequency : occasional, depends on IO load
2867 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2868 after numerous create/delete cycles
2869 Details : The ext3 inode allocation policy would not allocate new inodes
2870 very well on the OSTs because there are no new directories
2871 being created. Instead we look for groups with free space if
2872 the parent directories are nearly full.
2877 Description: Network or server problems during mount may cause partially
2878 mounted clients instead of returning an error.
2879 Details : The config llog parsing code may overwrite the error return
2880 code during mount error handling, returning success instead
2886 Description: MDS can fail to allocate large reply buffers
2887 Details : After long uptimes the MDS can fail to allocate large reply
2888 buffers (e.g. zconf client mount config records) due to memory
2889 fragmentation or consumption by the buffer cache. Preallocate
2890 some large reply buffers so that these replies can be sent even
2891 under memory pressure.
2895 Frequency : rare (liblustre)
2896 Description: fsx running with liblustre complained that using truncate() to
2897 extend the file doesn't work. This patch corrects that issue.
2898 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2899 ATTR_SIZE and lsm use in llu_setattr_raw.
2903 Frequency : rare, only 2.6 kernels
2904 Description: Unusual file access patterns on the MDS may result in inode
2905 data being lost in very rare circumstances.
2906 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2907 code in the 2.6 kernel caused the inode and/or EA data not to
2908 be read from disk, causing single-file corruption.
2912 Frequency : rare, only 2.6 filesystems using extents
2913 Description: Heavy concurrent write and delete load may cause data corruption.
2914 Details : It was possible under high-load situations to have an extent
2915 metadata block in the block device cache from a just-unlinked
2916 file overwrite a newly-allocated data block. We now unmap any
2917 metadata buffers that alias just-allocated data blocks.
2921 Frequency : filesystems with default stripe_count larger than 77
2922 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2923 Details : lconf specifies an inode size of 4096 bytes when the default
2924 stripe_count is larger than 77. This conflicts with the default
2925 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2926 this case to avoid pinning too much memory for large EAs.
2928 ------------------------------------------------------------------------------
2930 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2935 Frequency : rare (extremely heavy IO load with hundreds of clients)
2937 Description: Client is evicted, gets IO error writing to file
2938 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2939 caused clients to be evicted for AST timeouts. The fixes in
2940 bug 5192 mean we no longer need to have such short AST timeouts
2941 so ldlm_timeout has been increased.
2944 Frequency : occasional during --force or --failover shutdown under load
2945 Bugzilla : 5949, 4834
2946 Description: Server oops/LBUG if stopped with --force or --failover under load
2947 Details : a collection of import/export refcount and cleanup ordering
2948 issues fixed for safer force cleanup
2951 Frequency : only filesystems larger than 120 OSTs
2952 Bugzilla : 5990, 6223
2953 Description: lfs getstripe would oops on a very large filesystem
2954 Details : lov_getconfig used kfree on vmalloc'd memory
2957 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2958 Bugzilla : 6242, 6243
2959 Description: reading from files that had been truncated to a non-zero size
2960 but never opened returned no data
2961 Details : ll_file_read() reads zeros from no-object files to EOF
2966 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2967 Details : There's little harm in aborting MDS/OSS recovery and letting it
2968 try again, so I removed the LASSERT and return an error instead.
2970 Severity : enhancement
2972 Description: New debugging infrastructure for tracking down data corruption
2973 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2974 (b) cover more of the client-side code path, and (c) try to narrow
2975 down where problems occurred
2979 Bugzilla : 3819, 4364, 4397, 6313
2980 Description: Racing close and eviction MDS could cause assertion in mds_close
2981 Details : It was possible to get multiple mfd references during close and
2982 client eviction, leading to one thread referencing a freed mfd.
2984 Severity: : enhancement
2985 Bugzilla : 3262, 6359
2986 Description: Attempts to reconnect to servers are now more aggressive.
2987 Details : This builds on the enhanced upcall-less recovery that was added
2988 in 1.4.2. When trying to reconnect to servers, clients will
2989 now try each server in the failover group every 10 seconds. By
2990 default, clients would previously try one server every 25 seconds.
2995 Description: After recovery, certain operations trigger a failed
2996 assertion on a client.
2997 Details : Failing over an mds, using lconf -d --failover, while a
2998 client was doing a readdir() call would cause the client to
2999 LBUG after recovery completed and the readdir() was resent.
3001 Severity : enhancement
3003 Description: Default groups are now added by lconf
3004 Details : You can now run lconf --group <servicename> without having to
3005 manually add groups with lmc.
3008 Frequency : occasional
3010 Description: Nodes with an elan id of 0 trigger a failed assertion
3013 Frequency : always when accessing e.g. tty/console device nodes
3015 Description: tty and some other devices nodes cannot be used on lustre
3016 Details : file's private_data field is used by device data and lustre
3017 values in there got lost. New field was added to struct file to
3018 store fs-specific private data.
3021 Frequency : when exporting Lustre via NFS
3023 Description: NFSD failed occasionally when looking up a path component
3024 Details : NFSD is looking up ".." which was broken in ext3 directories
3025 that had grown large enough to become hashed.
3028 Frequency : Clusters with multiple interfaces not on the same subnet
3030 Description: Nodes will repeatedly try to reconnect to an interface which it
3031 cannot reach and report an error to the log.
3032 Details : Extra peer list entries will be created by lconf with some peers
3033 unreachable. lconf now validates the peer before adding it.
3036 Frequency : Only if a default stripe is set on the filesystem root.
3038 Description: Setting a default stripe on the filesystem root prevented the
3039 filesystem from being remounted.
3040 Details : The client was sending extra request flags in the root getattr
3041 request and did not allocate a reply buffer for the dir EA.
3044 Frequency : occasional, higher if lots of files are accessed by one client
3045 Bugzilla : 6159, 6097
3046 Description: Client trips assertion regarding lsm mismatch/magic
3047 Details : While revalidating inodes the VFS looks up inodes with ifind()
3048 and in rare cases can find an inode that is being freed.
3049 The ll_test_inode() code will free the lsm during ifind()
3050 when it finds an existing inode and then the VFS later attaches
3051 this free lsm to a new inode.
3055 Bugzilla : 6422, 7030
3056 Description: MDS deadlock between mkdir and client eviction
3057 Details : Creating a new file via mkdir or mknod (starting a transaction
3058 and getting the ns lock) can deadlock with client eviction
3059 (gets ns lock and trying to finish a synchronous transaction).
3062 Frequency : occasional
3063 Description: While starting a server, the fsfilt_ext3 module could not be
3065 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3066 kernels. Previously, lconf would still use the ext3 name
3067 when trying to load modules. Now, it will correctly use
3068 ext3 on 2.4 and ldiskfs on 2.6.
3070 Severity : enhancement
3071 Description: The default stripe count has been changed to 1
3072 Details : The interpretation of the default stripe count (0, to lfs
3073 or lmc) has been changed to mean striping across a single
3074 OST, rather than all available. For general usage we have
3075 found a stripe count of 1 or 2 works best.
3077 Severity : enhancement
3078 Description: Add support for compiling against Cray portals.
3079 Details : Conditional compiling for some areas that are different
3083 Frequency : occasional
3084 Bugzilla : 6409, 6834
3085 Description: Creating files with an explicit stripe count may lead to
3086 a failed assertion on the MDS
3087 Details : If some OSTs are full or unavailable, creating files may
3088 trigger a failed assertion on the MDS. Now, Lustre will
3089 try to use other servers or return an error to the
3093 Frequency : occasional
3095 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3096 caused lock timeout during readahead (since 1.4.2).
3097 Details : Processes doing readahead might match a lock that hasn't been
3098 granted yet if there are overlapping and conflicting lock
3099 requests. The readahead process waits on ungranted lock
3100 (original lock is CBPENDING), while OST waits for that process
3101 to cancel CBPENDING read lock and eventually evicts client.
3103 Severity : enhancement
3105 Description: Initial enabling of flock support for clients
3106 Details : Implements fcntl advisory locking and file status functions.
3107 This feature is provided as an optional mount flag (default
3108 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3109 locking are implemented yet, and those that are are not guaranteed
3110 to be completely correct in production environments.
3111 mount -t lustre -o [flock|noflock] ...
3114 Frequency : occasional
3116 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3117 assertion in the ext3 JBD (journaling) layer.
3118 Details : The b_committed_data struct is protected by the big kernel lock
3119 in 2.4 kernels, serializing journal_commit_transaction() and
3120 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3121 access to this struct. In 2.6 kernels there is finer grained
3122 locking to improve SMP performance of the JBD layer.
3126 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3131 Description: merge and cleanup kernel patches.
3134 -----------------------------------------------------------------------------
3136 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3138 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3139 versions of Lustre. Please update all servers and clients to
3140 version 1.4.2 or later at the same time. You must also run
3141 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3142 to update the configuration logs.
3144 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3145 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3146 - swab llog records of type '0' so we get proper header size/idx (5861)
3147 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3148 - fix rename of one directory over another leaking an inode (5953)
3149 - avoid SetPageDirty on 2.6 (5981)
3150 - don't re-add just-being-destroyed locks to the waiting list (5653)
3151 - when creating new directories, inherit the parent's custom
3152 striping settings if present parent (3048)
3153 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3154 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3155 - don't hold client locks on temporary worklist from l_lru (5666)
3156 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3157 - automatically evict dead clients (5921)
3158 - Update file size properly in create+truncate+fstat case (6196)
3159 - Do not unhash mountpoint dentries, do not allow removal of
3161 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3162 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3163 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3164 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3165 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3166 - Correctly handle reads of files with no objects (6243)
3167 - lctl recover will also mark a device active if deactivate used (5933)
3169 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3170 - allow --write-conf on an MDS with different nettype than client (5619)
3171 - don't write config llogs to MDS for mounts not from that MDS (5617)
3172 - lconf should create multiple TCP connections from a client (5201)
3173 - init scripts are now turned off by default; run chkconfig --on
3174 lustre and chkconfig --on lustrefs to use them
3175 - upcalls are no longer needed for clients to recover to failover
3177 - add --abort-recovery option to lconf to abort recovery on device
3179 - add support for an arbitrary number of OSTs (3026)
3180 - Quota support protocol changes.
3181 - forward compatibility changes to wire structs (6007)
3182 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3183 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3184 - improved "lustre status" script
3185 - initialize blocksize for non-regular files (6062)
3186 - added --disable-server and --disable-client configure options (5782)
3187 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3188 - Vanilla 2.4.29 support
3189 - increase maximum number of obd devices to 520 (6242)
3190 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3191 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3192 SLES 9 2.6.5 kernels
3193 - limit stripes per file to 160 (the maximum EA size) (6093)
3195 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3198 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3199 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3200 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3201 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3202 - small ext3 extents cleanups and fixes (5733)
3203 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3204 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3205 - avoid needless client->OST connect, fix handle mismatch (5317)
3206 - fix DLM error path that led to out-of-sync client, long delays (5779)
3207 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3208 - fix several locking issues related to i_size (5492,5624,5654,5672)
3209 - don't move pending lock onto export if it is already evicted (5683)
3210 - fix kernel oops when creating .foo in unlinked directory (5548)
3211 - fix deadlock in obdfilter statistics vs. object create (5811)
3212 - use time_{before,after} to avoid timer jiffies wrap (5882)
3213 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3214 - Do not leak request if server was not able to process it (5154)
3215 - If mds_open unable to find parent dir, make that negative lookup(5154)
3216 - don't create new directories with extent-mapping (5909, 5936)
3218 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3219 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3220 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3221 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3222 - support for RHEL4 kernel on i686 (5773)
3223 - provide error messages when incompatible logs are encountered (5898)
3225 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3226 * version 1.4.0.10 (1.4.1 release candidate 1)
3228 - don't keep a lock reference when lock is not granted (4238)
3229 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3230 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3231 - drop import inflight refcount on signal_completed_replay error (5255)
3232 - unlock page after async write error during send (3677)
3233 - handle missing objects in filter_preprw_read properly (5265)
3234 - no transno return for symlink open, don't save no-trasno open (3440)
3235 - don't try to complete elan receive that already failed (4012)
3236 - free RPC server reply state on error (5406)
3237 - clean up thread from ptlrpc_start_thread() on error (5160)
3238 - readahead could read extra page into cache that wasn't ejected (5388)
3239 - prevent races in class_attach/setup/cleanup/detach (5260)
3240 - don't dereference de->d_inode after l_dput of de (5458)
3241 - use "int" for stripe value returned from lock_to_stripe (5544)
3242 - mballoc allocation and error-checking fixes in 2.6 (5504)
3243 - block device patches to fix I/O request sizes in 2.6 (5482)
3244 - look up hostnames for IB nals (5602)
3245 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3246 - don't start multiple acceptors for the same port (5277)
3247 - fix incorrect LASSERT in mds_getattr_name (5635)
3248 - export a proc file for general "ping" checking (5628)
3249 - fix "lfs check" to not block when the MDS is down (5628)
3251 - service request history (4965)
3252 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3253 - create an "evict_client" /proc entry on OSTs, like the MDS has
3254 - fix mount usage message, return errors per mount(8) (5168)
3255 - change grep [] to grep "[]" in tests so they work in more UMLs
3256 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3257 - remove ancient LOV_MAGIC_V0 EA support (5047)
3258 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3259 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3260 - allow readahead allocations to fail when low on memory (5383)
3261 - mmap locking landed again, after considerable improvement (2828)
3262 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3264 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3267 - send OST transaction number in read/write reply to free req (4966)
3268 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3269 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3270 - the watchdog thread now runs as interruptible (5246)
3271 - flock/lockf fixes (but it's still disabled, pending 5135)
3272 - don't use EXT3 constants in llite code (5094)
3273 - memory shortage at startup could cause assertion (5176)
3275 - reorganization of lov code
3276 - single portals codebase
3278 - add extents/mballoc support (5025)
3279 - direct I/O reads in the obdfilter (4048)
3280 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3282 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3285 - send OST transaction number in read/write reply to free req (4966)
3286 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3287 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3288 - the watchdog thread now runs as interruptible (5246)
3289 - handle missing objects in filter_preprw_read properly (5265)
3290 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3291 - drop import inflight refcount on signal_completed_replay error (5255)
3292 - unlock page after async write error during send (3677)
3293 - return original error code on reconstructed replies (3761)
3294 - no transno return for symlink open, don't save no-trasno open (3440)
3296 - add pid to ldlm debugging output (4922)
3297 - bump the watchdog timeouts -- we can't handle 30sec yet
3298 - extra debugging for orphan dentry/inode bug (5259)
3300 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3303 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3304 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3305 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3306 - fix NULL dereference in /proc/sys/portals/routes (4827)
3307 - allow failed mdc_close() operations to be interrupted (4561)
3308 - stop precreate on OST before MDS would time out on it (4778)
3309 - don't send partial-page writes before EOF from client (4410)
3310 - discard client grant for sub-page writes on large-page clients (4520)
3311 - don't free dentries not owned by NFS code, check generation (4806)
3312 - fix lsm leak if mds_create_objects() fails (4801)
3313 - limit debug_daemon file size, always print CERROR messages (4789)
3314 - use transno after validating reply (3892)
3315 - process timed out requests if import state changes (3754)
3316 - update mtime on OST during writes, return in glimpse (4829)
3317 - add mkfsoptions to LDAP (4679)
3318 - use ->max_readahead method instead of zapping global ra (5039)
3319 - don't interrupt __l_wait_event() during strace
3321 - add software watchdogs to catch hung threads quickly (4941)
3322 - make lustrefs init script start after nfs is mounted
3323 - fix CWARN/ERROR duplication (4930)
3324 - return async write errors to application if possible (2248)
3325 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3326 - print NAL number in %x format (4645)
3327 - update barely-supported suse-2.4.21-171 series (4842)
3328 - support for sles 9 %post scripts
3329 - support for building 2.6 kernel-source packages
3330 - support for sles km_* packages
3332 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3335 - ignore -ENOENT errors in osc_destroy (3639)
3336 - notify osc create thread that OSC is being cleaned up (4600)
3337 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3338 - reconstruct ost_handle() like mds_handle() (4657)
3339 - create a new thread to do import eviction to avoid deadlock (3969)
3340 - let lconf resolve symlinked-to devices (4629)
3341 - don't unlink "objects" from directory with default EA (4554)
3342 - hold socknal file ref over connect in case target is down (4394)
3343 - allow more than 32000 subdirectories in a single directory (3244)
3344 - fix blocks count for O_DIRECT writes (3751)
3345 - OST returns ENOSPC from object create when no space left (4539)
3346 - don't send truncate RPC if file size isn't changing (4410)
3347 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3348 - bind to privileged port in socknal and tcpnal (3689)
3350 - rate limit CERROR/CWARN console message to avoid overload (4519)
3351 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3352 - basic mmap support (3918)
3353 - kernel patch series update from b1_4 (4711)
3355 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3358 - avoid crash during MDS cleanup with OST shut down (2775)
3359 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3360 - don't use bad inodes on the MDS (3744)
3361 - dynamic object preallocation to improve recovery speed (4236)
3362 - don't hold spinlock over lock dumping or change debug flags (4401)
3363 - don't zero obd_dev when it is force cleaned (3651)
3364 - print grants to console if they go negative (4431)
3365 - "lctl deactivate" will stop automatic recovery attempts (3406)
3366 - look for existing locks in ldlm_handle_enqueue() (3764)
3367 - don't resolve lock handle twice in recovery avoiding race (4401)
3368 - revalidate should check working dir is a directory (4134)
3370 - don't always mark "slow" obdfilter messages as errors (4418)
3372 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3375 - don't close LustreDB during write_conf until it is done (3860)
3376 - fix typo in lconf for_each_profile (3821)
3377 - allow dumping logs from multiple threads at one time (3820)
3378 - don't allow multiple threads in OSC recovery (3812)
3379 - fix debug_size parameters (3864)
3380 - fix mds_postrecov to initialize import for llog ctxt (3121)
3381 - replace config semaphore with spinlock (3306)
3382 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3383 - don't allow enqueue to complete on a destroyed export (3822)
3384 - down write_lock before checking llog header bitmap (3825)
3385 - recover from lock replay timeout (3764)
3386 - up llog sem before sending rpc (3652)
3387 - reduce ns lock hold times when setting kms (3267)
3388 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3389 - fix NULL deref and obd_dev leak on setup error (3312)
3390 - replace some LBUG about llog ops with error handling (3841)
3391 - don't match INVALID dentries from d_lookup and spin (3784)
3392 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3393 - fix invalid assertion in ptlrpc_set_wait (3880)
3395 - add libwrap support for the TCP acceptor (3996)
3396 - add /proc/sys/portals/routes for non-root route listing (3994)
3397 - allow setting MDS UUID in .xml (2580)
3398 - print the stack of a process that LBUGs (4228)
3400 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3403 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3404 - make sure to unset replay flag from failed open requests (3440)
3405 - if default stripe count is 0, use OST count for inode size (3636)
3406 - update parent mtime/ctime on client for create/unlink (2611)
3407 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3408 - free recovery state on server during a forced cleanup (3571)
3409 - unregister_reply for resent reqs (3063)
3410 - loop back devices mounting and status check on 2.6 (3563)
3411 - fix resource-creation race that can provoke i_size == 0 (3513)
3412 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3413 - more debugging for page-accounting assertion (3746)
3414 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3415 - avoid deadlock after precreation failure (3758)
3416 - fix race and lock order deadlock in orphan handling (3450, 3750)
3417 - add validity checks when grabbing inodes from l_ast_data (3599)
3419 - add /proc/.../recovery_status to obdfilter (3428)
3420 - lightweight CDEBUG infrastructure, debug daemon (3668)
3421 - change default OSC RPC parameters to be better on small clusters
3422 - turn off OST read cache for files smaller than 32MB
3423 - install man pages and include them in rpms (3100)
3424 - add new init script for (un)mounting lustre filesystems (2593)
3425 - run chkconfig in %post for init scripts (3701)
3426 - drop scimac NAL (unmaintained)
3428 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3431 - clean kiobufs before and after use (3485)
3432 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3433 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3434 - change init script to fail more clearly if not run as root (1528)
3435 - allow clients to reconnect during replay (1742)
3436 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3437 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3438 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3439 - fix cond_resched (3554)
3440 - only evict unfinished clients after recovery (3515)
3441 - allow bulk resend, prevent data loss (3570)
3442 - dynamic ptlrpc request buffer allocation (2102)
3443 - don't allow unlinking open directory if it isn't empty (2904)
3444 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3445 - remove extraneous obd dereference causing LASSERT failure (3334)
3446 - don't use get_cycles() when creating temp. files on the mds (3156)
3447 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3448 - handle EEXIST for set-stripe, set proper directory name (3336)
3450 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3451 - fix ksocknal_fmb_callback() error messages (2918)
3453 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3456 - don't copy lvb into (possibly NULL) reply on error (2983)
3457 - don't deref dentry after dput, don't free lvb on error (2922)
3458 - use the kms to determine writeback rpc length (2947)
3459 - increment oti_logcookies when osc is inactive (2948)
3460 - update client's i_blocks count via lvb messages (2543)
3461 - handle intent open/close of special files properly (1557)
3462 - mount MDS with errors=remount-ro, like obdfilter (2009)
3463 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3464 - don't use cancelling-locks' kms values (2947)
3465 - use highest lock extent for kms, not last one (2925)
3466 - don't dereference ERR_PTR() dentry in error handling path (3107)
3467 - fix thread race in portals_debug_dumplog() (3122)
3468 - create lprocfs device entries at setup instead of at attach (1519)
3469 - common AST error handler, don't evict client on completion race (3145)
3470 - zero nameidata in detach_mnt in 2.6 (3118)
3471 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3472 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3473 - initialize RPC timeout timer earlier for 2.6 (3219)
3474 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3475 - print nal/nid for unknown nid (3258)
3476 - additional checks for oscc recovery before doing precreate (3284)
3477 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3478 - don't crash in mdc_close for bad permissions on open (3285)
3479 - zero i_rdev for non-device files (3147)
3480 - clear page->private before handing to FS, better assertion (3119)
3481 - tune the read pipeline (3236)
3482 - fix incorrect decref of invalidated dentry (2350)
3483 - provide read-ahead stats and refine rpc in flight stats (3328)
3484 - don't hold journal transaction open across create RPC (3313)
3485 - update atime on MDS at close time (3265)
3486 - close LDAP connection when recovering to avoid server load (3315)
3487 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3488 - don't leak open file on MDS after open resend (3325)
3489 - serialize filter_precreate and filter_destroy_precreated (3329)
3490 - loop device shouldn't call sync_dev() for nul device (3092)
3491 - clear page cache after eviction (2766)
3492 - resynchronize MDS->OST in background (2824)
3493 - refuse to mount the same filesystem twice on same mountpoint (3394)
3494 - allow llmount to create routes for mounting behind routers (3320)
3495 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3496 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3497 - fix rare problem with rename on htree directories (3417)
3499 - allow default OST striping configuration per directory (1414)
3500 - fix compilation for qswnal for 2.6 kernels (3125)
3501 - increase maximum number of MDS request buffers for large systems
3502 - change liblustreapi to be useful for external progs like lfsck (3098)
3503 - increase local configuration timeout for slow disks (3353)
3504 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3506 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3509 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3510 - don't overwrite extent policy data in reply if lock was blocked (2901)
3511 - drop filter export grants atomically with removal from device (2663)
3512 - del obd_self_export from work_list in class_disconnect_exports (2908)
3513 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3514 - swab reply message in mdc_close, other PPC fixes (2464)
3515 - fix destroying of named logs (2325)
3516 - overwrite old logs when running lconf --write_conf (2264)
3517 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3518 - fix race in target_handle_connect (2898)
3519 - mds_reint_create() should take same inode create lock (2926)
3520 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3521 - don't close files for self_export to avoid uninitialized obd (2936)
3522 - allow MDS with the same name as client node (2939)
3523 - hold dentry reference for closed log files for unlink (2325)
3524 - reserve space for all logs during transactions (2059)
3525 - don't evict page beyond end of stripe extent (2925)
3526 - don't oops on a deleted current working directory (2399)
3527 - handle hard links to targets without a parent properly (2517)
3528 - don't dereference NULL lock when racing during eviction (2867)
3529 - don't grow lock extents when lots of conflicting locks (2919)
3531 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3534 - account for cache space usage on clients to avoid data loss (974)
3535 - lfsck support in lustre kernel code (2349)
3536 - reduce journal credits needed for BRW writes (2370)
3537 - orphan handling to avoid losing space on client/server crashes
3538 - ptlrpcd can be blocked, stopping ALL progress (2477)
3539 - use lock value blocks to assist in proper KMS, faster stat (1021)
3540 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3541 - recovery for initial connections (2355)
3542 - fixes for mds_cleanup_orphans (1934)
3543 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3544 - block all file creations until orphan recovery completes (1901)
3545 - client remove rq_connection from request struct (2423)
3546 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3547 - recovery timer race (2670)
3548 - mdc_close recovey bug (2532)
3549 - ptlrpc cleanup bug (2710)
3550 - mds timeout on local locks (2588)
3551 - namespace lock held during RPCs (2431)
3552 - handle interrupted sync write properly (2503)
3553 - don't try to handle a message that hasn't been replied to (2699)
3554 - client assert failure during cleanup after abort recovery (2701)
3555 - leak mdc device after failed mount (2712)
3556 - ptlrpc_check_set allows timedout requests to complete (2714)
3557 - wait for inflight reqs when ptlrpcd finishes (2710)
3558 - make sure unregistered services are removed from the srv_list
3559 - reset bulk XID's when resending them (caught by 1138 test)
3560 - unregister_bulk after timeout
3561 - fix lconf error (2694)
3562 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3563 - readahead locks pages, leaves pending causing memory pressure (2673)
3564 - increase OST request buffers to 4096 on large machines (2729)
3565 - fix up permission of existing directories in simple_mkdir (2661)
3566 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3567 - don't assign transno to errored transactions (2742)
3568 - don't delete objects on OST if given a bogus objid from MDS (2751)
3569 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3570 - if rq_no_resend, then timeout request after recovery (2432)
3571 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3572 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3573 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3574 - MDS recovery shouldn't delete objects that it has given out (2730)
3575 - if enqueue arrives after completion, don't clobber LVB (2819)
3576 - don't unlock pages twice when trigger_group_io returns error (2814)
3577 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3578 - don't write pages to disk if there was an error (1450)
3579 - don't ping imports that have recovery disabled (2676)
3580 - take buffered bytes into account when balancing socknal conn (2817)
3581 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3582 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3583 - remove little-endian swabbing of llog records (1987)
3584 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3585 - retry reposting request buffers if they fail (1191)
3586 - grow extent at grant time to avoid granting a revoked lock (2809)
3587 - lock revoke doesn't evict page if covered by a second lock (2765)
3588 - disable VM readahead to avoid reading outside lock extents (2805)
3590 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3591 - updated kernel patches for hp-2.4.20 kernel (2681)
3593 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3596 - fix truncated write corruption (2366)
3597 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3599 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3600 - ensure len(uuid) < 37 in lmc (1171)
3601 - fix ia64 OOPS in llog_test (2255)
3602 - zero end of page at obdfilter for partial page writes (2648)
3603 - don't leave stale dentries around after renames (bug 2428)
3604 - fix timeouts when evicting a client with a single lock held (2642)
3605 - set deadline for the initial HELLO message to drain (2634)
3606 - print out dotted-quad IP addresses in the socknal (2302)
3608 - additional debugging for MDS client eviction problem (2443)
3609 - fix mkfsoptions support for osts (2603, 2604)
3611 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3614 - add series for the vanilla 2.6.0 kernel
3615 - add series for the vanilla 2.4.24 kernel
3616 - add series for a cray x86/64 UL kernel drop
3617 - fix xattr patches for the vanilla 2.4.19 series
3619 - generate true UUIDs in lmc (1171)
3620 - have portals stack dumping break in UML (2466)
3621 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3622 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3623 - wake callback waiting threads on client eviction (2460)
3624 - Add --ptldebug and --subsystem to lmc (1719)
3625 - update assertion to allow safe interrupt allocation
3626 - set rq_no_resend for cancel requests (2432)
3627 - recalculate ptlrpcd timeout after resend (2494)
3628 - call vfs_rmdir when removing pending directories (2368)
3629 - fix renaming a file to itself (2429)
3630 - lmc creates a default one-stripe lov (2454)
3631 - expand procfs space to handle large clusters (2326)
3632 - increase UML stack to avoid overflow
3633 - update lconf's list of debug and subsystem masks
3634 - fix lfs find --obd (2510)
3635 - /proc tunable for disabling filter read caching (2591)
3636 - stop rpm packages from altering slapd.conf (2301)
3637 - disable nagle in the socknal under 0conf (2578)
3638 - choose mds inode size based on stripe count (2572)
3639 - fix kernel-source rpm problems (2516)
3641 - add --disable-doc to avoid pdf generation (2421)
3642 - update documentation, tests, type-os, comments
3643 - avoid format warnings on ia64
3644 - remove the TOE NAL
3645 - tiny code cleanups by removing unused fields
3647 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3650 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3651 - remove the most blatant lies from BUILDING file (2371)
3652 - change default debug level to reasonable production setting
3653 - reduce client side cache size to reduce cache flush time
3654 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3655 - make TCP zerocopy and pinger support enabled by default (2476)
3656 - sync writes completed after process exits caused crashes (2319)
3657 - maintain correct mount count on the MDS (2356)
3658 - backout 1557, because 2316 wasn't really fixed
3659 - better file I/O statistics gathering in /proc
3660 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3661 - another kernel patch to fix zero-copy TCP function export
3662 - don't take duplicate lock when processing re-sent getattr (2420)
3663 - lctl uses obd_self_export instead of creating new conn (2353)
3664 - MDS/OST recovery case which requires object creation asserted (2425)
3665 - move lfs from /usr/sbin to /usr/bin in packages
3666 - fix race between mds_client_add and mds_client_free (2417)
3667 - use kmalloc instead of slabs in portals (2430)
3668 - don't create duplicate records when a failover MDS is present (2442)
3669 - remove unnecessary mount age check (2332)
3670 - don't remove directory inodes from locks prematurely (2451)
3671 - don't break if MDS service name is the same as hostname (2103)
3672 - fix races in client write RPC generation when cache full (2482)
3674 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3677 - remove now-unused request->rq_obd (278)
3678 - if an allocation fails, print out how much memory we've used (1933)
3679 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3680 - add the "configurable stack size" patch to most series files (1256)
3681 - ability to write large log records, for 100+ OST configs (2306)
3682 - fix NULL deref when filter_prep fails (2314)
3683 - fix operator precedence error in filter_sync
3684 - dynamic allocation of socknal TX descriptors (2315)
3685 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3686 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3687 - ensure that CWARN messages in Portals always get to the syslog
3688 - __init/__exit are not for prototype decls (ldlm_init/exit)
3689 - x86-64 compile warning fixes
3690 - fix gateway LMC keyword conflict (2318)
3691 - fix MDS lock inversions in getattr/reint paths (1844)
3692 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3693 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3694 - DLM race condition prevented some lock evictions (2328)
3695 - ENOMEM detection and retry on socknal sends (2230)
3696 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3697 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3698 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3699 - better error messages when a client is rejected during recovery (1505)
3700 - avoid cancelling locks which were never granted, after failure (2330)
3701 - fix i_sem/journal inversion in mds_client_add (2333)
3702 - fix truncate/getattr lock cycle deadlock (2334)
3703 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3704 - fix two rare exit paths which could leak an l_lock() ref (2321)
3705 - fencepost error in MDS/OST orphan recovery (2226)
3706 - make log record alignment 8 bytes (1988)
3707 - lstripe now fails when requested offset > ost_count (2237)
3708 - ensure that all kernel series have a complete list.h (1607)
3709 - fix crashes in special-file operations (2316)
3710 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3712 - allow configurable automake binary, for testing new versions
3713 - small update to the lfs documentation
3715 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3717 * fix negative export reference count in fsfilt_sync (2312)
3719 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3720 * release candidate 0.9.1
3722 - orphans are moved into the PENDING directory for possible recovery
3723 - replayed opens now open by fid for orphan/rename safety (1042)
3724 - last close of an orphan inode generates a transno (683)
3725 - chdir() and mount() now pin the directory entry (1020)
3726 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3727 - discard very old requests without processing them (1502)
3728 - remove some common, well-understood CERRORs (1505)
3729 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3730 - clear "grant" flags in OST replies until OST grant code lands (1644)
3731 - fix read performance by not clobbering i_blksize on client (1598)
3732 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3733 - make LRU size a /proc tunable, clears locks when reduced (707)
3734 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3735 - server locks take references on exports now (1558)
3736 - build fixes for 2.4.20-rh trees (1663)
3737 - return an error from lov_create if all OSCs are inactive (1751)
3738 - fix import levels when a reconnect happens without a timeout (1597)
3739 - exit early from mds_open if we get a lookup error (1749)
3740 - partial page read at EOF wouldn't wait for disk before sending (1642)
3741 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3742 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3743 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3744 - write/truncate lock inversion (1639)
3745 - Don't auto-load obdclass, portals modules during cleanup (1495)
3746 - fix timestamps from jumping to "now" (1763)
3747 - extra journal assertions (1648)
3748 - add an extra multiunlink test (1771)
3749 - fix read_record/write_record API (1776)
3750 - fix leak of offset_extent, possible incorrect i_size later (1772)
3751 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3752 - Debugging for the kqswnal_get_idle_tx problems (1820)
3753 - Allow recovery to be attempted multiple times (1536)
3754 - Write out MDS last_rcvd file after it is first created (1600)
3755 - Fix tx_descriptor leak in failed transmit situations (1827)
3756 - ext3 journaling fixes for assertion failure after IO error (1871)
3757 - class_export_put() on freed export after completion AST error (1896)
3758 - Fix revalidate looping in VFS (1322)
3759 - Don't access a freed export during MDS_REINT timeout (1521)
3760 - Add open-unlink recovery support on the MDS (1673,1764)
3761 - Return an error if no MDS data was read from last_rcvd (1946)
3762 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3763 - Don't setup a disk device that doesn't match exported UUID (317)
3764 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3765 - avoid committing NULL handle in force close
3766 - local.sh is now a one-stripe LOV configuration
3767 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3768 - fix ext3 htree duplicate directory entry corruption (1516)
3769 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3770 - update to new LOV EA format (2097)
3771 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3773 2003-06-15 Phil Schwan <phil@clusterfs.com>
3776 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3777 - per-import/export recovery handling (958, 931, 959)
3778 - multiple last-rcvd slots, for serving multiple FSes (949)
3779 - connections are again shared between multiple imp/exports (963, 964)
3780 - "umount -f" would hang if any requests needed to be sent (393, 978)
3781 - avoid pinning large req buffer by copying for queued messages (989)
3782 - add "uuid" to "lctl device" command to help upcalls (991)
3783 - "open" RPCs with transnos would confuse recovery counters (1037)
3784 - do proper endian conversion of all wire messages (288, 340, 891)
3785 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3786 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3787 - fix ptlrpc_connection leak in target_handle_connect (1174)
3788 - fix import refcounting bug in OST and MDS cleanup (1134)
3789 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3790 - fix very unlikely obd_types race condition (501)
3791 - remove osc_open hack for echo_client (1187)
3792 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3793 - a failure in read_inode2 leads to deadlock (1139)
3794 - cancel ack-locks as soon as transaction is committed (1072)
3795 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3796 - make sure to commitrw after any preprw to avoid deadlock (1162)
3797 - failing to execute a file in a lustre FS would lock inode (1203)
3798 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3799 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3800 - obd_uuid2tgt would walk off the end of the list (1255)
3801 - on IA64 the osc would give portals incorrect bulk size (1258)
3802 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3803 - fix lock inversion caused by new llite matching code (1282)
3804 - limit the number of dirty pages on a client to 10MB (1286)
3805 - timed out locks were not being corrected cancelled (1289)
3806 - fix O_DIRECT above 4GB on IA-32 (1292)
3807 * major user-visible changes
3808 - fail out/fail over policy now controlled by the upcall (993)
3810 - add OBD_PING to check server availability and failure (954)
3811 - lustre messages are now sent in sending host order (288, 340, 891)
3812 - add eadatalen to MDS getattr reply (340)
3813 - OST read replies may contain second buffer, with per-page status (593)
3815 2003-03-11 Phil Schwan <phil@clusterfs.com>
3818 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3819 - failed open()s could cause deadlock; fixed (867, 869)
3820 - stop cancelling OST locks when files are closed (481)
3821 - overlapping XID spaces caused network corruption (851, 853)
3822 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3823 - setattr_raw added, to do single-RPC, server-side setattrs
3824 - lmc/lconf syntax change for OST UUIDs
3825 - fix crashy race condition between ptlrpc_free_req and osc_close
3826 - don't use request in mdc_enqueue if we hit a timeout (889)
3827 - don't set the inode i_size for regular files from the MDS (896)
3828 - handle out of order completion AST (842)
3829 - don't LBUG if a lock request times out after receiving AST (913)
3830 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3831 - if a bad lock AST arrives, send an error instead of dropping entirely
3832 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3833 - fix leak in bulk IO when only partially completed (899, 900, 926)
3834 - fix O_DIRECT for ia64 (55)
3835 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3836 - C-z after timeout could hang a process forever; fixed (977)
3838 - client-side I/O cache (678, 924, 929, 941, 970)
3840 - READPAGE and SETATTRs which don't take server-side locks get
3843 2003-02-11 Phil Schwan <phil@clusterfs.com>
3846 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3847 - Fix fsfilt_extN_readpage() to read a full page of directory
3848 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3849 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3850 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3851 - Most intent operations were removed, in favour of a new RPC mode
3852 that does a single RPC to the server and bypasses most of the VFS
3853 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3854 - Aggressively cancel local locks on DLM servers
3855 - mds_reint_unlink sends EA to the client if it's the last nlink.
3856 client uses that EA to unlink OST objects.
3857 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3858 - recursive symlinks were fixed (439)
3859 - fixed NULL deref in DEBUG_REQ
3860 - filter_update_lastobjid no longer calls sync, which annoyed extN
3861 - fixed multi-client small-writes to a single file problem (445)
3862 - fixed mtime updates during file writes (607)
3863 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3864 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3865 - fixed Linux OST size reporting problem (444, 656)
3866 - OST now updates object mtime with writes or setattr (607, 619)
3867 - client verifies file size before zeroing page past EOF (445)
3868 - OST now writes last allocated objid to disk with allocation (108)
3869 - LOV on echo now works (409)
3871 - mds_reint_unlink sends a new buffer, with the EA included. this
3872 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3873 set if a regular file was being unlinked, and it was the last link
3874 - use PtlGet from the target for bulk writes (315)
3875 - OST now updates object mtime with writes or setattr (607, 619)
3876 - LDLM now has a grant-time callback to revalidate locked items, if
3878 - Many MDS operations were reorganized to combat race conditions
3880 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3881 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3883 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3886 - Fully reactivate OST imports after reconnection (512, others)
3887 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3888 - More graceful error handling for truncating on dead OST (515)
3889 - Don't error out unless we're actually accessing dead stripes (474)
3890 - Fix garbage sizes when stripes are missing (410)
3891 - LRU counters were broken, causing constant lock purge (433, 432)
3892 - garbage on read from stripes with failed OSTs (441)
3893 - mark OSCs as active before reconnecting during recovery (438)
3894 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3895 - lfind did not preserve OST order in output (443)
3896 - symlinks cause hung clients, incorrect data (439)
3897 - stop dereferencing request after dropping refcount (457)
3898 - don't LASSERT(spin_is_locked) on non-SMP (455)
3899 - fixes for many rename() bugs
3900 - fstat didn't correctly synchronize attributes (399)
3901 - server must handle lock cancellation during blocking AST prep (487)
3902 - bulk descriptors were free()d too soon (511)
3903 - fix paths in lconf, which would load incorrect modules (451, 507)
3904 - fix confusing lconf 'host not found' error message (386)
3905 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3906 - fix race condition in mdc_blocking_ast() for inode access (526)
3907 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3908 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3909 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3910 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3911 - fix lstripe in several ways (kernel side) (527)
3912 - fix request leak in ldlm_cli_enqueue (262)
3913 - incorrect OSC was marked inactive after OST failure
3914 - call mds_fs_cleanup before unmounting filesystem (524)
3915 - fix races between taking ns_lock and ldlm_lock_change_resource
3916 - fix races updating LOV export open file list
3917 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3918 - fix recovery NULL deref in ldlm_cli_cancel_unused
3919 - fix some DLM races by using new hash table for lock handles (419)
3920 - permit the client to specify desired inodes, at replay
3921 - duplicate requests when we queue them for replay reintegration
3922 - fix last_rcvd offset calculation
3923 - sync after each recovered transaction, so we always make progress
3924 - never, not always, ERESTART requests without transnos
3925 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3927 - skip replay if the MDS says that the client is already connected
3928 - don't check for a recovery-enabled export to match lctl's UUID
3929 - don't INC_USE_COUNT for phantom exports
3930 - don't crash when cleaning up phantom exports (567)
3931 - don't double-finish or set replay data for errored mdc_open requests
3932 - abort requests when they time out, so we don't get old replies
3933 - send/receive replies for AST messages again
3934 - if the client says that it doesn't have the lock, cancel it on the
3936 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3937 mark it as destroyed, it will all work out when decref is called
3938 - fix module use counts (22, 581)
3940 - ASTs now expect a reply (server cancels lock on error reply)
3942 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3945 - fix many simultaneous client startup (392)
3946 - fix dentry->d_it clobbering
3947 - credentials weren't being shipped for readdir/getattr operations
3948 - remove invalid assertions triggered during some concurrent MD
3950 - proper Lustre versions added (336, 389)
3951 - fix memory leak for create error case (398)
3952 - fix LOV locking bug that would get cli/srv out of sync
3953 - fix echo client over LOV (409)
3954 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3955 - fix double-O_EXCL intent crash (424)
3956 - avoid sending multiple lock CANCELs (352)
3958 - MDS can do multi-client recovery (modulo bugs in new code)
3960 - many updates, edits, cleanups
3962 2002-11-18 Phil Schwan <phil@clusterfs.com>
3965 - fix null d_it dereference (346)
3966 - fix full OST/dbench hang (333)
3967 - fix permission problem with file removal (286)
3968 - fix removal of OSCs from LOV when they fail
3969 - fix NULL deref during bulk timeout (214)
3970 - fix problems related to multiple filesystems on one MDS (241)
3971 - fixed serious subtle metadata locking bugs
3972 - free locks on clients when inodes are removed due to memory
3974 - fix inode pointer in lock data (285)
3975 - partial support for multiple MDS on a single host (241)
3976 - data locks weren't cancelled at clear_inode time (290, 311)
3977 - intent locks could lead to unbounded lock growth (205)
3978 - added a maximum lock count, an LRU list, and a flusher
3979 - fix multiple rename (365)
3980 - properly abstracted the echo client
3981 - OSC locked 1 byte too many; fixed
3982 - rewrote brw callback code:
3983 - fixed recovery bugs related to LOVs (306)
3984 - fixed too-many-pages-in-one-write crash (191)
3985 - fixed (again) crash in sync_io_timeout (214)
3986 - probably fixed callback-related race (385)
3988 - Add capability to MDS protocol
3989 - LDLM cancellations and callbacks on different portals
3991 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3994 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3995 - defer open object destruction to close time (601981, 138)
3996 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3997 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3998 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3999 - don't call obd_setattr in ll_file_release for destroyed objects
4000 * protocol change to lustre_msg: move |version| and add |flags|
4001 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4002 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4003 * added replay of create, unlink, link and rename operations during
4004 MDS failover; recovery should be much more robust now
4005 * remove failed OSCs from LOVs (only lov_create uses this so far)
4006 * the lustre-HOWTO was brought (more) up to date (582544)
4008 2002-10-23 Phil Schwan <phil@clusterfs.com>
4011 - in-use dentries weren't being reused properly (617851)
4012 - prevent multiple LDLM setup (599178)
4013 - fix LOV size calculations for truncate (617853)
4014 - fix client handling of MDS intent errors (POSIX)
4015 - fix permission bug in lovstripe.c test (624321)
4016 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4017 - truncate past end of file could corrupt data
4018 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4019 - a race in recovery could return ETIMEDOUT to apps (623947)
4020 - building outside the source directory was fixed
4021 * the lustre-HOWTO was brought (more) up to date (582544)
4022 * major progress was made on recovery functionality
4024 2002-10-10 Phil Schwan <phil@clusterfs.com>
4027 - recovery deadlock fix
4028 - rm -rf causes LBUG fix (617817)
4029 - file open by multiple tasks fix (618962)
4030 - directory permissions bugs (602707 and 620007)
4031 - journal_stop fixed with locking (611313)
4032 - O_APPEND failures resolved (618273, perhaps 614459)
4033 - lconf PATH fix (619770)
4034 - IA64 build fix (621450)
4035 - RPC buffer sizes scale with amount of memory
4037 2002-10-01 Phil Schwan <phil@clusterfs.com>
4040 - locks would be cancelled without throwing away data pages,
4041 resulting in inconsistent data (605627)
4042 - inode attributes were not always being refreshed (605627, 612449)
4043 - lconf now continues to cleanup after lctl reports an error
4044 - MDS now enforces user permissions (602707)
4045 - lprocfs cleanup fixed, but not yet enabled (614157)
4046 - fixed infinite server hang, should a client not respond to an AST
4047 - avoid going into recovery if user calls readlink() with a buffer
4048 that's too small (613941)
4049 - AST RPCs no longer require replies (614867) -- this may be changed
4050 - don't crash server if client sends an IOV that's too big (611336)
4051 - fixed lock conversion deadlock (611892)
4052 - fixed the following of symlinks (614622)
4053 * recovery: the server can remove locks from a client that dies, other
4054 clients can make progress
4055 * more extN patch fixes
4056 * compile-time configurable ptlrpc buffer allocations
4058 - collaborative read cache document
4059 - Lustre Lite Performance CDR document-in-progress
4061 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4064 - fix typo in patch-2.4.18
4066 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4069 - clear ptlrpc request each time in handle_incoming_request()
4070 - unlink of files now destroys the object on the OST
4072 2002-09-19 Peter Braam <braam@clusterfs.com>
4074 * add hard link support
4075 * change obdfile creation method
4076 * kernel patch changed
4078 2002-09-19 Peter Braam <braam@clusterfs.com>
4081 - stack overflow bug in extN fixed
4083 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4085 * documentation updates
4086 - add man pages for config tools
4087 - update tests/README to describe testing with new config tools
4088 - finish metadata API descriptions
4089 * bug fixes and cleanups
4090 - statfs workaround for 16TB limit
4091 - LOV stripe allocation improved, can stripe on subset of OSTs
4092 - LOV file size/IO offset was wrong for files > 4GB in size
4093 - object EA data was being dropped, caused files to be unreadable
4094 - memory overflow with non-LOV OST caused memory corruption
4095 - fixed regression tests to work with new config tools, obdfilter
4096 - fixed bug when directory size became larger than 1 block
4097 - fixed bug (for single client case) when PWD was deleted
4098 - invalidate local directory pages when doing intent-based ops
4099 - avoid LDLM oops when lock callback contained bad data
4101 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4103 * documentation updates
4104 * bug fixes and cleanups
4105 - configuration tools
4108 - 64-bit compile warnings
4109 - 64-bit internal statfs data
4111 * test_brw on persistent OST devices
4113 * lprocfs (disabled)
4115 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4117 * documentation updates
4118 * bug fixes and cleanups
4119 * configuration tools
4121 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4124 * many small fixes to 0_5_4
4125 * io/network handling
4126 * thinkos in MDS operations
4128 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4131 * crucial basic fixes to 0.5.3
4132 * IOR, Iozone work over Elan
4135 2002-08-07 Phil Schwan <phil@clusterfs.com>
4136 * version 0_5_3, our first alpha
4137 * we use the new Portals iovs
4138 * documentation updates
4139 * bug fixes and cleanups
4140 * small changes in the DLM wire protocol
4142 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4143 * version 0_5_1 with some initial stability,
4144 * locking on MD and file I/O.
4145 * documentation updates
4146 * several bug fixes since 0.5.0
4147 * small changes in wire protocol
4149 2002-07-18 Phil Schwan <phil@clusterfs.com>
4151 * delivered as Lustre Light Alpha
4152 * fixed a crash after handling invalid MDS requests
4153 * fixed directory pages for architectures with non-4k pages sizes
4155 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4156 * release version v0_4_4
4157 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4159 * Fixes a number of interruption problems with OST operations.
4160 * Update documentation for portals header changes
4161 * Move all wire protocol structs/defines to lustre_idl.h
4162 * Fixes symlink length bug.
4163 * Add tcpdump to repository.
4165 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4166 * release version v0_4_3
4167 * Fixes statfs for inodes on extN.
4168 * Fixes bug in runtests which would delete /etc/hosts.
4169 * Use 64-bit object IDs wherever possible (not into VFS though)
4170 Remove ost_get_info, which is unused by lustre, and out of date.
4172 2002-07-03 Peter Braam <braam@clusterfs.com>
4173 * release version v0_4_2 Fixes a lookup error (type not passed)
4174 * move forward to head of Portals
4175 * move forward to latest Lustre kernel
4177 2002-06-25 Peter Braam <braam@clusterfs.com>
4178 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob