1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5 2.6.9-55.0.6.EL (RHEL 4),
6 2.6.16.46-0.14 (SLES 10),
7 2.6.18-8.1.14.el5 (RHEL 5),
8 2.6.18.8 vanilla (kernel.org)
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.16 - 2.6.22 vanilla (kernel.org)
12 * Due to recently discovered recovery problems, we do not recommend
13 using patchless RHEL 4 clients with this or any earlier release.
14 * Recommended e2fsprogs version: 1.40.2-cfs1
15 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Severity : enhancement
19 Description: eliminate client locks in face of contention
20 Details : file contention detection and lockless i/o implementation
25 Description: Remove client patches from SLES 10 kernel.
26 Details : This causes SLES 10 clients to behave as patchless clients
27 even on a Lustre-patched (server) kernel.
31 Description: use i_size_read and i_size_write in 2.6 port
32 Details : replace inode->i_size access with i_size_read/write()
35 Frequency : when removing large files
37 Description: scheduling issue during removal of large Lustre files
38 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
39 It causes scheduling issues when removing large files (17TB in the
44 Description: 1.4.11 Can't handle directories with stripe set and extended acls
45 Details : Impossible (EPROTO is returned) to access a directory that has a
46 non-default striping and ACLs.
49 Frequency : only on ppc
51 Description: /proc/fs/lustre/devices broken on ppc
52 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
53 We should make sure the type of 'index' is loff_t and then cast
54 explicitly as needed below. Do not assign an explicitly cast
58 Frequency : only for rhel5
60 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
61 Details : Modify the target file & which_kernel.
64 Frequency : if the uninit_groups feature is enabled on ldiskfs
66 Description: e2fsck reports "invalid unused inodes count"
67 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
68 feature and only a single inode is created in a group then the
69 "bg_unused_inodes" count is incorrectly updated. Creating a
70 second inode in that group would update it correctly.
74 Description: buffer overruns could theoretically occur
75 Details : llapi_semantic_traverse() modifies the "path" argument by
76 appending values to the end of the origin string, and a buffer
77 overrun may occur. Adding buffer overrun check in liblustreapi.
81 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
85 Description: handle "serious error: objid * already exists" more gracefully
86 Details : If LAST_ID value on disk is smaller than the objects existing in
87 the O/0/d* directories, it indicates disk corruption and causes an
88 LBUG(). If the object is 0-length, then we should use the existing
89 object. This will help to avoid a full fsck in most cases.
93 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
94 Details : Modify vm-tunables-rhel4.patch.
98 Description: Kernel config for 2.6.18-vanilla.
99 Details : Modify targets/2.6-vanilla.target.in.
100 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
101 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
102 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
103 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
107 Description: improve handling recoverable errors
108 Details : if request processig with error which can be recoverable on server
109 request should be resend, otherwise page released from cache and
114 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
115 Details : Modify target file & which_patch.
116 A flaw was found in the IA32 system call emulation provided
117 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
118 value could be stored in the %RAX register, which could trigger an
119 out-of-bounds system call table access. An untrusted local user
120 could exploit this flaw to run code in the kernel
121 (ie a root privilege escalation). (CVE-2007-4573).
123 --------------------------------------------------------------------------------
125 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
127 * Support for kernels:
128 2.6.5-7.286 (SLES 9),
129 2.6.9-55.0.2.EL (RHEL 4),
130 2.6.16.46-0.14 (SLES 10),
131 2.6.18-8.1.8.el5 (RHEL 5),
132 2.6.18.8 vanilla (kernel.org)
133 * Client support for unpatched kernels:
134 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
135 2.6.16 - 2.6.22 vanilla (kernel.org)
136 * Due to recently discovered recovery problems, we do not recommend
137 using patchless RHEL 4 clients with this or any earlier release.
138 * Recommended e2fsprogs version: 1.40.2-cfs1
139 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
143 Description: Fix errors in lfs documentation
144 Details : Fixes man pages
146 Severity : enhancement
148 Description: Adaptive timeouts
149 Details : RPC timeouts adapt to changing server load and network
150 conditions to reduce resend attempts and improve recovery time.
152 Severity : enhancement
154 Description: llapi_file_create() does not allow some changes
155 Details : add llapi_file_open() that allows specifying the file creation
156 mode and open flags, and also returns an open file handle.
158 Severity : enhancement
160 Description: df doesn't work properly if diskfs blocksize != 4K
161 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
164 Frequency : always on directories with default striping set
166 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
167 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
169 Severity : enhancement
171 Description: merge and cleanup kernel patches.
172 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
174 Severity : enhancement
176 Description: RedHat Update kernel for RHEL5
177 Details : Kernel config file for RHEL5.
179 Severity : enhancement
181 Description: OSS needs mutliple precreate threads
182 Details : Add ability to start more than one create thread per OSS.
185 Frequency : only with quota on the root user
187 Description: mds_obd_create error creating tmp object
188 Details : When the user sets quota on root, llog will be affected and can't
189 create files and write files.
192 Frequency : Always on ia64 patchless client, and possibly others.
194 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
195 Details : This allows the patchless client to be loaded on architectures
201 Description: disorder of journal start and llog_add cause deadlock.
202 Details : in llog_origin_connect, journal start should happen before
203 llog_add keep the same order as other functions to avoid
206 Severity : enhancement
208 Description: RedHat Update kernel for RHEL5
209 Details : Modify the kernel config file more closer RHEL5.
211 Severity : enhancement
213 Description: Build failure against Centos5 (RHEL5)
214 Details : Define PAGE_SIZE when it isn't present.
217 Frequency : occasionally when using NFS
219 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
220 Details : with NFS, the anon dentry's parent was set to itself in
221 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
222 corresponding dentry other than use rec->ur_name.
224 Severity : enhancement
226 Description: enable data checksumming by default
227 Details : enable checksum by default, allow --disable-checksum
228 configure option and "-o nochecksum" mount option. Checksums
229 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
232 Frequency : Occasionally with failover
234 Description: Client eviction due to failover config
235 Details : after a connection loss, the lustre client should attempt to
236 reconnect to the last active server first before trying the
237 other potential connections.
239 Severity : enhancement
241 Description: client-side metadata stat-ahead during readdir(directory readahead)
242 Details : perform client-side metadata stat-ahead when the client detects
243 readdir and sequential stat of dir entries therein
248 Description: when mds and osts use different quota unit(32bit and 64bit),
249 quota will be released repeatly.
250 Details : void sending multiple quota reqs to mds, which will keep the
251 status between the reqs.
254 Frequency : only with liblustre clients on XT3
256 Description: evictions taking too long
257 Details : allow llrd to evict clients directly on OSTs
261 Description: osts not allocated evenly to files
262 Details : change the condition to increase offset_idx
264 Severity : enhancement
266 Description: self-adjustable client's lru lists
267 Details : use adaptive algorithm for managing client cached locks lru
268 lists according to current server load, other client's work
269 pattern, memory activities, etc. Both, server and client
270 side namespaces provide number of proc tunables for controlling
275 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
276 Details : rewrite inherited code (uuid parsing code from ext2 utils and
277 readdir code from ext3) from scratch preserving functionality.
281 Description: Only those disconnect error should be returned by rq_status.
282 Details : In open/enqueue processs, Some errors, which will cause client
283 disconnected, should be returned by rq_status, while other
284 errors should still be returned by intent, then mdc or llite will
287 Severity : enhancement
289 Description: Tune the kernel for good SCSI performance.
290 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
291 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
295 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
296 Bugzilla : 13547, 13627
297 Description: Data corruption for OSTs that are formatted larger than 2TB
299 Details : When generating the bio request for lustre file writes the
300 sector number would overflow a temporary variable before being
301 used for the IO. The data reads correctly from Lustre (which
302 will overflow in a similar manner) but other file data or
303 filesystem metadata may be corrupted in some cases.
307 Description: "lfs find -obd UUID" prints directories
308 Details : "lfs find -obd UUID" will return all directory names instead
309 of just file names. It is incorrect because the directories
310 do not reside on the OSTs.
314 Description: MDS hang after unclean shutdown of lots of clients
315 Details : Never resend AST requests.
319 Frequency : Always, for kernels after 2.6.16
320 Description: Fix warning idr_remove called for id=.. which is not allocated.
321 Details : Last kernels save old s_dev before kill super and not allow
322 to restore from callback - restore it before call kill_anon_super.
326 Description: client sometimes tried revalidate root dentry.
327 Details : not revalidate dentry if it root dentry.
329 --------------------------------------------------------------------------------
331 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
333 * Support for kernels:
334 2.6.5-7.286 (SLES 9),
335 2.6.9-55.0.2.EL (RHEL 4),
336 2.6.16.46-0.14 (SLES 10),
337 2.6.18-8.1.8.el5 (RHEL 5),
338 2.6.18.8 vanilla (kernel.org)
339 * Client support for unpatched kernels:
340 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
341 2.6.16 - 2.6.22 vanilla (kernel.org)
342 * Due to recently discovered recovery problems, we do not recommend
343 using patchless RHEL 4 clients with this or any earlier release.
344 * Recommended e2fsprogs version: 1.39.cfs8
345 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
350 Description: block reactivating mgc import until all deactivates complete
351 Details : Fix race when failing back MDT/MGS to itself (testing)
353 Severity : enhancement
355 Description: lfs setstripe enhancement
356 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
359 Frequency : mds/oss recovery
361 Description: llog ctxt is refrenced after it has been freed.
362 Details : llog ctxt refcount was added to avoide the race between ctxt free
363 and llog recovery process. Each llog user must hold ctxt refcount
364 before it access the llog. And the llog ctxt can only be freed
365 when its refcount is zero.
367 Severity : enhancement
369 Description: randomly memory allocation failure util
370 Details : Make lustre randomly failed allocating memory for testing purpose.
373 Frequency : only for SLES10
375 Description: Update kernel patch for SLES10 SP1
376 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
378 Severity : enhancement
380 Description: omit set fsid for export NFS
381 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
382 as FS_REQUIRES_DEV for avoid problems with generate fsid.
385 Frequency : after network failures
387 Description: Add sync option to mount_lustre.c
388 Details : Client loses data written to lustre after a network interruption.
390 Severity : enhancement
392 Description: Error message improvement.
393 Details : Merging of two LCONSOLE_ERROR_MSG into one.
395 Severity : enhancement
397 Description: don't use GFP_* in generic Lustre code.
398 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
400 Severity : enhancement
402 Description: obdclass is limited by single OBD_ALLOC(idarray)
403 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
405 Severity : enhancement
407 Description: updated patchess for new RHEL4 kernel
408 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
410 Severity : enhancement
412 Description: warnings with build patchless client with vanila 2.6.19 and up
413 Details : change old ctl_table style and replace ctl_table/ctl_table_header
414 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
416 Severity : enhancement
418 Description: O_DIRECT bypasses client statistics.
419 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
420 read_ahead_stats were updated. Copied stats section from
421 osc_send_oap_rpc() into async_internal().
424 Frequency : only for Cray XT3
426 Description: peer credits not enough on many OST per OSS systems.
427 Details : Use new lnet way to add credits as we need those for pings and ASTs
430 Frequency : only with liblustre
432 Description: Liblustre is not releasing flock locks on file close.
433 Details : Release flock locks on file close.
436 Frequency : only for RHEL4
438 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
439 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
440 Update target file and kernel config.
444 Description: lustre support for RHEL5
445 Details : Add support for RHEL5.
450 Description: ASSERTION(export != NULL) failed in target_handle_connect
451 Details : Assetion hit is result of rare race between disconnect and connet
452 to same nid. target_handle_connect found old connect cockie and
453 tried to reconnect, but can't find export for this cockie.
458 Description: umount blocks forever on error
459 Details : In result of wrong using obd_no_recov and obd_force flags client
460 can hand if cancel or some other requests is lost.
463 Frequency : Only for SLES
465 Description: sanity_quota fail test_1
466 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
467 file, which makes TSTID[2] inunique.
469 Severity : enhancement
471 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
472 Details : Update target/ChangeLog/which_patch .
474 Severity : enhancement
476 Description: jbd statistics
477 Details : Port older jbd statistics patch for sles10
482 Description: lvbo_init failed for resource with missing objects.
483 Details : Fix returning error if we do stat for file with missing/corrupted
484 objects and i_size set to all sum of size of all avaible objects.
485 if we to truncate/write to missing object - it is recreated.
490 Description: Oops in read and write path when failing to allocate lock.
491 Details : Check if lock allocation failed and return error back.
494 Frequency : When flocks are used.
496 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
497 Details : Flock locks might destroy just granted lock if it could be merged
498 with another existing flock, this is done in completion handler,
499 so teach ldlm_cli_enquque_fini that this is a valid case for
505 Description: reply_lock_interpret crash due to race with it and lock cancel.
506 Details : Do not replay locks that are being cancelled. Do not reference
507 locks by their address during replay, just by their handle.
510 Frequency : only with deactivated OSTs
512 Description: lstripe command fails for valid OST index
513 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
514 instead of lov->desc.ld_active_tgt_count.
516 Severity : enhancement
518 Description: Build failure against Centos5 (RHEL5)
519 Details : Use getpagesize() instead of PAGE_SIZE.
521 --------------------------------------------------------------------------------
523 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
525 * Support for kernels:
526 2.6.5-7.283 (SLES 9),
527 2.6.9-55.EL (RHEL 4),
528 2.6.16.46-0.14 (SLES 10),
529 2.6.18.8 vanilla (kernel.org)
530 * Client support for unpatched kernels:
531 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
532 2.6.16 - 2.6.22 vanilla (kernel.org)
533 * Due to recently discovered recovery problems, we do not recommend
534 using patchless RHEL 4 clients with this or any earlier release.
535 * Recommended e2fsprogs version: 1.39.cfs8
536 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
537 * Starting with this release, the ldiskfs backing filesystem required
538 by Lustre is now in its own package, lustre-ldiskfs. This package
539 should be installed. It is versioned separately from Lustre and
540 may be released separately in future.
545 Description: server LBUG when shutting down
546 Details : Block umount forever until the mount refcount is zero rather
547 than giving up after an arbitrary timeout.
549 Severity : enhancement
551 Description: add optional extra BUILD_VERSION info
552 Details : add a new environment variable (namely LUSTRE_VERS) which allows
553 to override the lustre version.
556 Frequency : 2.6.18 servers only
558 Description: ll_kern_mount() doesn't release the module reference
559 Details : The ldiskfs module reference count never drops down to 0
560 because ll_kern_mount() doesn't release the module reference.
565 Description: server LBUG when using old ost_num_threads parameter
566 Details : Accept the old ost_num_threads parameter but warn that it
567 is deprecated, and fix an off-by-one error that caused an LBUG.
572 Description: Transient SCSI error results in persistent IO issue
573 Details : iobuf->dr_error is not reinitialized to 0 between two
577 Frequency : sometimes when underlying device returns I/O errors
579 Description: OSTs not going read-only during write failures
580 Details : OSTs are not remounted read-only when the journal commit threads
581 get I/O errors because fsfilt_ext3 calls journal_start/stop()
582 instead of the ext3 wrappers.
585 Frequency : only with 10000 clients or more
587 Description: poor connect scaling with increasing client count
588 Details : Don't run filter_grant_sanity_check for more than 100 exports
589 to improve scaling for large numbers of clients.
592 Frequency : SLES10 only
594 Description: sanity-quota.sh quotacheck failed: rc = -22
595 Details : Quotas cannot be enabled on SLES10.
598 Frequency : liblustre clients only
600 Description: getdirentries does not give error when run on compute nodes
601 Details : getdirentries does not fail when the size specified as an argument
602 is too small to contain at least one entry
604 Severity : enhancement
606 Description: Add LNET router traceability for debug purposes
607 Details : If a checksum failure occurs with a router as part of the
608 IO path, the NID of the last router that forwarded the bulk data
609 is printed so it can be identified.
614 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
615 Details : Due to a race condition, liblustre clients were occasionally
618 Severity : enhancement
620 Description: lfs setstripe use optional parameters instead of postional
623 Severity : enhancement
625 Description: Nanosecond timestamp support for ldiskfs
626 Details : The on-disk ldiskfs filesystem has added support for nanosecond
627 resolution timestamps. There is not yet support for this at
628 the Lustre filesystem level.
631 Frequency : during server recovery
633 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
634 Details : request with rq_no_resend flag not awake l_wait_event if they get a
638 Frequency : nfs export on patchless client
640 Description: connectathon hang when test nfs export over patchless client
641 Details : Disconnected dentry cannot be found with lookup, so we do not need
642 to unhash it or make it invalid
645 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
646 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
647 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
652 Description: open req refcounting wrong on reconnect
653 Details : If reconnect happened between getting open reply from server and
654 call to mdc_set_replay_data in ll_file_open, we will schedule
655 replay for unreferenced request that we are about to free.
656 Subsequent close will crash in variety of ways.
657 Check that request is still eligible for replay in
658 mdc_set_replay_data().
663 Description: disable writes to filesystem when reading health_check file
664 Details : the default for reading the health_check proc file has changed
665 to NOT do a journal transaction and write to disk, because this
666 can cause reads of the /proc file to hang and block HA state
667 checking on a healthy but otherwise heavily loaded system. It
668 is possible to return to the previous behaviour during configure
669 with --enable-health-write.
671 Severity : enhancement
673 Description: 64-bit inode version
674 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
675 to the inode. This will be required for version-based recovery.
680 Description: MDS fails to start if a duplicate client export is detected
681 Details : in some rare cases it was possible for a client to connect to
682 an MDS multiple times. Upon recovery the MDS would detect this
683 and fail during startup. Handle this more gracefully.
685 Severity : enhancement
687 Description: Add -o localflock option to simulate old noflock behaviour.
688 Details : This will achieve local-only flock/fcntl locks coherentness.
693 Description: log_commit_thread vs filter_destroy race leads to crash
694 Details : Take import reference before releasing llog record semaphore
699 Description: Wrong request locking in request set processing
700 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
701 imp_delayed_list, in this place should be used imp_lock.
704 Frequency : when reconnecting
706 Description: Grant leak when OSC reconnect to OST
707 Details : When osc reconnect ost, OST(filter) should check whether it
708 should grant more space to client by comparing fed_grant and
709 cl_avail_grant, and return the granted space to client instead
710 of "new granted" space, because client will call osc_init_grant
711 to update the client grant space info.
714 Frequency : when client reconnects to OST
716 Description: Grant leak when OSC does a resend and replays bulk write
717 Details : When osc reconnect to OST, OST(filter) should clear grant info of
718 bulk write request, because the grant info while be sync between
719 OSC and OST when reconnect, and we should ignore the grant info
720 these of resend/replay write req.
725 Description: Grant space more than avaiable space sometimes.
726 Details : When then OST is about to be full, if two bulk writing from
727 different clients came to OST. Accord the avaliable space of the
728 OST, the first req should be permitted, and the second one
729 should be denied by ENOSPC. But if the seconde arrived before
730 the first one is commited. The OST might wrongly permit second
731 writing, which will cause grant space > avaiable space.
734 Frequency : when client is evicted
736 Description: Grant might be wrongly erased when osc is evicted by OST
737 Details : when the import is evicted by server, it will fork another
738 thread ptlrpc_invalidate_import_thread to invalidate the
739 import, where the grant will be set to 0. While the original
740 thread will update the grant it got when connecting. So if
741 the former happened latter, the grant will be wrongly errased
742 because of this race.
747 Description: Checking Stale with correct fid
748 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
749 to check whether it is stale, because sometimes, we want the
750 enqueue happened anyway, and op_data.fid2 will not be initialized.
752 Severity : enhancement
754 Description: update patchless client
755 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
758 Frequency : only with 2.4 kernel
760 Description: random memory corruption
761 Details : size of struct ll_inode_info is to big for union inode.u and this
762 can be cause of random memory corruption.
767 Description: Memory leak in recovery
768 Details : Lov_mds_md was not free in an error handler in mds_create_object.
769 It should also check obd_fail before fsfilt_start, otherwise if
770 fsfilt_start return -EROFS,(failover mds during mds recovery).
771 then the req will return with repmsg->transno = 0 and rc = EROFS.
772 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
773 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
774 be freed no matter whether fsfilt_commit success or not.
777 Frequency : only with huge numbers of clients
779 Description: Prevents from taking the superblock lock in llap_from_page for
781 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
782 for llap_from_page call in ll_removepage() prevents from taking
783 the superblock lock for a soon died page.
788 Description: Not check open intent error before release open handle
789 Details : in some rare cases, the open intent error is not checked before
790 release open handle, which may cause
791 ASSERTION(open_req->rq_transno != 0), because it tries to release
792 the failed open handle.
797 Description: Set cat log bitmap only after create log success.
798 Details : in some rare cases, the cat log bitmap is set too early. and it
799 should be set only after create log success.
803 Description: Accessing a block bevice can re-enable I/O when Lustre is
804 tearing down a device.
805 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
810 Description: service threads may hog cpus when there are a lot of requests
811 Details : Insert cond_resched to give other threads a chance to use some CPU
816 Description: the cat log was not initialized in recovery
817 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
818 unlink log on mds will not be initialized until mds post
819 recovery. And also in mds post recovery, the unlink log will
820 initialization will be done asynchronausly, so there will be race
821 between add unlink log and unlink log initialization.
825 Description: brw_stats were being printed incorrectly
826 Details : brw_stats were being printed as log2 but all of them were not
827 recorded as log2. Also remove some code duplication arising from
828 filter_tally_{read,write}.
832 Frequency : rare, only in recovery.
833 Description: ASSERTION(req->rq_type != LI_POISON) failed
834 Details : imp_lock should be held while iterating over imp_sending_list for
835 prevent destroy request after get timeout in ptlrpc_queue_wait.
839 Description: replay-single.sh test 52 fails
840 Details : A lock's skiplist need to be cleanup when it being unlinked
841 from its resource list.
846 Description: Short directio read returns full requested size rather than
848 Details : Direct I/O operations should return actual amount of bytes
849 transferred rather than requested size.
851 Severity : enhancement
853 Description: metadata RPC reduction (e.g. for rm performance)
854 Details : decrease the amount of synchronous RPC between clients and servers
855 by canceling conflicing lock before the operation on the client
856 and packing thier handles into the main operation RPC to server.
858 Severity : enhancement
860 Description: add #ifdef HAVE_KERNEL_CONFIG_H
861 Details : kernels from 2.6.19 not need include linux/config.h, but add
862 include linux/autoconf.h in commpiler command line.
864 Severity : enhancement
866 Description: patchless client support for 2.6.22 kernel
867 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
868 removed. In this case we need drop using os depended interface to
869 kmem_cache and use cfs_mem_cache API.
874 Description: fix mal-formatted messages
875 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
879 Frequency : always in liblustre
880 Description: wrong IS_ERR implementation in liblustre.h
881 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
886 Description: Correct condition for output debug message.
887 Details : inode i_nlink equal zero is not enough for output message about
888 disk corruption, i_ctime and i_mode should be also checked.
892 Frequency : always in patchless client
893 Description: add configure check for truncate_complete_page
894 Details : improve checks for exported symbols. This allow run check without
895 sources, but with Module.symvers shipped with kernel distribution.
896 add check for truncate_complete_page used by patchless client.
900 Description: sanity.sh test_77h fails with "test_77h file compare failed"
901 Details : test_77h uses a file which was messed by other test case.
905 Description: Not Check whether lov_tgts is NULL in some lov functions
906 Details : Checking whether lov_tgts is NULL in some functions.
910 Description: replace obdo_alloc() with OBDO_ALLOC macro
911 Details : nothing special is done in obdo_alloc() function, and for
912 debugging purpose, it needs to be replaced with macros.
916 Description: bad return value and errno from fcntl call
917 Details : In liblustre API, errno should be a negative value if error
922 Description: ptlrpc_check_set() LBUG
923 Details : In case of positive reply from server and failed client bulk
924 callback after bulk transfer shouldn't LBUG, but process this
925 request as erroneous.
927 Severity : enhancement
929 Description: VFS operations stats tool.
930 Details : Tool which collects stats by tracking value written in pid,
931 ppid, gid and uses llstat to generate output to plot graph using
933 Updated lustre/utils/Makefile.am
934 Added lustre/utils/ltrack_stats.c
936 Severity : enhancement
938 Description: 2.6.18 server support (lustre 1.6.1)
939 Details : Support for 2.6.18 kernels on the server side.
944 Description: ASSERTION(imp->imp_conn_current) failed
945 Details : an assertion failure is hit if a client node boots and attempts to
946 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
949 Frequency : only for i686
951 Description: 1.4.11 RC1 build fails for RHEL 4, i686
952 Details : Fixed config variable for build.
957 Description: Updated patchess for new RHEL4 kernel
958 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
959 Updated patch jbd-stats-2.6.9.patch
960 Updated patch qsnet-rhel4-2.6.patch
961 Updated patch quota-deadlock-on-pagelock-core.patch
962 Updated patch vfs_intent-2.6-rhel4.patch
963 Updated patch vfs_races-2.6-rhel4.patch
964 Updated series file 2.6-rhel4-titech.series
965 Updated series file 2.6-rhel4.series
966 Updated kernel config files
971 Description: lquota slave complains LBUG when reconnecting with mds
973 Details : quota slave depends qctxt->lqc_import to send its quota request.
974 This pointer will be invalid if mds did failover or broke its
975 connect to osts, which leads to LBUG.
978 Frequency : when qunit size is too small(less than 20M)
980 Description: write is stopped by improper -EDQUOT
981 Details : If the master is busy and qunit size is small enough(let's say 1M),
982 the slave can not get quota from master on time, which will lead
983 slave to trigger a -EQUOTA to client.
988 Description: Deadlock during metadata tests
989 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
990 called with the per-dentry lock held.
993 Frequency : SLES9 only
995 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
996 Details : set CONFIG_NR_CPUS to 128 instead of 8.
998 Severity : enhancement
1000 Description: remove fs_prep_san_write operation and related patches
1001 Details : remove the ext3-san-jdike patches which are no longer useful.
1006 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
1007 Details : a disk corruption can cause the mballoc code to assert on a
1008 double free or other extent corruptions. Handle these with
1009 ext3_error() instead of assertions.
1012 Frequency : only with mballoc3 code and deep extent trees
1014 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
1015 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
1016 can cause the filesystem to be remounted read-only.
1021 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
1022 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
1023 few external users of this ioctl number is lfsck's call to
1024 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
1025 Add the old ioctl number to the handler so both old and new
1030 Description: parallel lock callbacks
1031 Details : Instead of sending blocking and completion callbacks as separated
1032 requests, adding them to a set and sending in parallel.
1036 Description: Disable most debugging by default
1037 Details : To improve performance, disable most logging (for debug purposes)
1038 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
1039 default, and HA includes fewer messages.
1043 Frequency : only run on patchless client.
1044 Description: use do_facet on sanity.sh for test handling recoverables errors
1045 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
1049 Description: hash tables for lists of nids, connections and uuids
1050 Details : Hash tables noticeably help when a lot of clients connect to a
1051 server, to faster identify duplicate connections or reconnects,
1052 also to faster find export to evict in manual eviction case.
1056 Description: Sometimes, when the server evict a client, and the client will
1057 not be evicted as soon as possible.
1058 Details : In enqueue req, the error was returned by intent, instead of
1059 rq_status which make ptlrpc layer not detect this error, and
1060 does not evict the client. So enqueue error should be returned
1064 Frequency : only at startup
1066 Description: Delay client connections to MDT until fist MDT->OST connect
1067 Details : If a client tried to create a new file before the MDT had
1068 connected to any OSTs, the create would return EIO. Now
1069 the client will simply block until the MDT connects to the
1070 first OST and the create can succeed.
1073 Frequency : only for SLES9
1075 Description: Routinely utilize latest Quadrics drivers in CFS releases
1076 Details : Update patch qsnet-suse-2.6.patch.
1079 Frequency : only for sles10
1081 Description: Update patches for SLES 10 SP1 kernel.
1082 Details : Update the patch vfs_intent-2.6-sles10.patch.
1087 Description: Routinely utilize latest Quadrics drivers in CFS releases
1088 Details : Update patch qsnet-rhel4-2.6.patch.
1091 Frequency : at statup only
1093 Description: mds_lov_synchronize race leads to various problems
1094 Details : simultaneous MDT->OST connections at startup can cause the
1095 sync to abort, leaving the OSC in a bad state.
1100 Description: Using wrong pointer in osc_brw_prep_request
1101 Details : Access to array[-1] can produce panic if kernel compiled with
1102 CONFIG_PAGE_ALLOC enabled
1104 Severity : enhancement
1106 Description: Async OSC create to avoid the blocking unnecessarily.
1107 Details : If a OST has no remain object, system will block on the creating
1108 when need to create a new object on this OST. Now, ways use
1109 pre-created objects when available, instead of blocking on an
1110 empty osc while others are not empty. If we must block, we block
1111 for the shortest possible period of time.
1115 Frequency : only in recovery
1116 Description: Mark OST as early accessible if his start SYNC.
1117 Details : osc_precreate return flag early accessible if oscc marked as
1118 OSCC_FLAG_SYNC_IN_PROGRESS.
1123 Description: Sometimes precreate code can triger create object on wrong ost
1124 Details : Wrong protected or not not restored variables aftre precreate loop
1125 can produce creation object on wrong ost.
1128 Frequency : oss recovery
1130 Description: llog_commit_thread cleanup should sync with llog_commit_thread
1132 Details : llog_commit_thread_count should be synced between llog_commit
1133 start and cleanup, so new llog_commit thread should not be started
1134 when llog_commit threads being stopped to avoid accessing some
1137 Severity : enhancement
1139 Description: Add printing inode info into message about error in writepage.
1141 --------------------------------------------------------------------------------
1143 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1148 Frequency : on some architectures
1150 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1151 Details : Uninitialized flags sometimes cause configuration commands to
1155 Frequency : patchless clients only
1157 Description: missing __iget() symbol export
1158 Details : The __iget() symbol export is missing. To avoid the need for
1159 this on patchless clients the deathrow inode reaper is turned
1160 off, and we depend on the VM to clean up old inodes. This
1161 dependency was during via the fix for bug 12181.
1166 Description: sanity.sh fail: test_52b
1167 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
1168 incorrect inode's flags to client.
1170 --------------------------------------------------------------------------------
1172 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1174 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1175 INTEROPERATE with older versions automatically. In many cases a
1176 special upgrade step is needed. Please read the
1177 user documentation before upgrading any part of a 1.4.x system.
1178 * WARNING: Lustre configuration and startup changes are required with
1179 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1181 * Support for kernels:
1182 2.4.21-47.0.1.EL (RHEL 3),
1183 2.6.5-7.283 (SLES 9),
1184 2.6.9-42.0.10.EL (RHEL 4),
1185 2.6.12.6 vanilla (kernel.org),
1186 2.6.16.27-0.9 (SLES10)
1187 * Client support for unpatched kernels:
1188 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
1189 2.6.16 - 2.6.19 vanilla (kernel.org),
1190 2.6.9-42.0.8EL (RHEL 4)
1191 * Recommended e2fsprogs version: 1.39.cfs6
1192 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1195 Severity : enhancement
1197 Description: Async OSC create to avoid the blocking unnecessarily.
1198 Details : If a OST has no remain object, system will block on the creating
1199 when need to create a new object on this OST. Now, ways use
1200 pre-created objects when available, instead of blocking on an
1201 empty osc while others are not empty. If we must block, we block
1202 for the shortest possible period of time.
1204 Severity : enhancement
1206 Description: MountConf
1207 Details : Lustre configuration is now managed via mkfs and mount
1208 commands instead of lmc and lconf. New obd types (MGS, MGC)
1209 are added for dynamic configuration management. See
1210 https://mail.clusterfs.com/wikis/lustre/MountConf for
1213 Severity : enhancement
1215 Description: dynamic OST addition
1216 Details : OSTs can now be added to a live filesystem
1218 Severity : enhancement
1220 Description: startup order invariance
1221 Details : MDTs and OSTs can be started in any order. Clients only
1222 require the MDT to complete startup.
1224 Severity : enhancement
1226 Description: parallel, asynchronous orphan cleanup
1227 Details : orphan cleanup is now performed in separate threads for each
1228 OST, allowing parallel non-blocking operation.
1230 Severity : enhancement
1232 Description: optimized stripe assignment
1233 Details : stripe assignments are now made based on ost space available,
1234 ost previous usage, and OSS previous usage, in order to try
1235 to optimize storage space and networking resources.
1237 Severity : enhancement
1239 Description: Permanently set tunables
1240 Details : All writable /proc/fs/lustre tunables can now be permanently
1241 set on a per-server basis, at mkfs time or on a live
1244 Severity : enhancement
1246 Description: Lustre message v2
1247 Details : Add lustre message format v2.
1249 Severity : enhancement
1251 Description: client OST exclusion list
1252 Details : Clients can be started with a list of OSTs that should be
1253 declared "inactive" for known non-responsive OSTs.
1257 Description: ENOENT returned for valid filehandle during dbench.
1258 Details : Check if a directory has children when invalidating dentries
1259 associated with an inode during lock cancellation. This fixes
1260 an incorrect ENOENT sometimes seen for valid filehandles during
1261 testing with dbench.
1264 Frequency : SFS test only (otherwise harmless)
1266 Description: SPEC SFS validation failure on NFS v2 over lustre.
1267 Details : Changes the blocksize for regular files to be 2x RPC size,
1268 and not depend on stripe size.
1270 Severity : enhancement
1272 Description: fine-grained SMP locking inside DLM
1273 Details : Improve DLM performance on SMP systems by removing the single
1274 per-namespace lock and replace it with per-resource locks.
1276 Severity : enhancement
1278 Description: don't hold multiple extent locks at one time
1279 Details : To avoid client eviction during large writes, locks are not
1280 held on multiple stripes at one time or for very large writes.
1281 Otherwise, clients can block waiting for a lock on a failed OST
1282 while holding locks on other OSTs and be evicted.
1284 Severity : enhancement
1286 Description: Multiple MD RPCs in flight.
1287 Details : Further unserialise some read-only MDT RPCs - learn about intents.
1288 To avoid overly-overloading MDT, introduce a limit on number of
1289 MDT RPCs in flight for a single client and add /proc controls
1290 to adjust this limit.
1292 Severity : enhancement
1294 Description: client read/write statistics
1295 Details : Add client read/write call usage stats for performance
1296 analysis of user processes.
1297 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1298 file access. extents_stats shows chunk size distribution.
1299 extents_stats_per_process show chunk size distribution per
1302 Severity : enhancement
1304 Description: per-client statistics on server
1305 Details : Add ldlm and operations statistics for each client in
1306 /proc/fs/lustre/mds|obdfilter/*/exports/
1308 Severity : enhancement
1310 Description: improved MDT statistics
1311 Details : Add detailed MDT operations statistics in
1312 /proc/fs/lustre/mds/*/stats
1314 Severity : enhancement
1316 Description: VFS operations stats
1317 Details : Add client VFS call stats, trackable by pid, ppid, or gid
1318 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
1323 Description: Fix client-side osc byte counters
1324 Details : The osc read/write byte counters in
1325 /proc/fs/lustre/osc/*/stats are now working
1328 Frequency : always as root on SLES
1330 Description: Failure of copying files with lustre special EAs.
1331 Details : Client side always return success for setxattr call for lustre
1332 special xattr (currently only "trusted.lov").
1337 Description: Refcount LNET uuids
1338 Details : The global LNET uuid list grew linearly with every startup;
1339 refcount repeated list entries instead of always adding to
1342 Severity : enhancement
1344 Description: Dynamic service threads
1345 Details : Within a small range, start extra service threads
1346 automatically when the request queue builds up.
1349 Frequency : mixed-endian client/server environments
1351 Description: mixed-endian crashes
1352 Details : The new msg_v2 system had some failures in mixed-endian
1355 Severity : enhancement
1357 Description: Easy OST removal
1358 Details : OSTs can be permanently deactivated with e.g. 'lctl
1359 conf_param lustre-OST0001.osc.active=0'
1361 Severity : enhancement
1363 Description: MGS proc entries
1364 Details : Added basic proc entries for the MGS showing what filesystems
1367 Severity : enhancement
1369 Description: provide MGS failover
1370 Details : Added config lock reacquisition after MGS server failover.
1372 Severity : enhancement
1374 Description: add Linux 2.4 support
1375 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1379 Description: a large application tries to do I/O to the same resource and dies
1380 in the middle of it.
1381 Details : Check the req->rq_arrival time after the call to
1382 ost_brw_lock_get(), but before we do anything about
1383 processing it & sending the BULK transfer request. This
1384 should help move old stale pending locks off the queue as
1385 quickly as obd_timeout.
1388 Frequency : when an incorrect nid is specified during startup
1390 Description: ptlrpc connect to non-existant node causes kernel crash
1391 Details : LNET can't be re-entered from an event callback, which
1392 happened when we expire a message after the export has been
1393 cleaned up. Instead, hand the zombie cleanup off to another
1396 Severity : enhancement
1398 Description: plain/inodebits lock performance improvement
1399 Details : Grouping plain/inodebits in granted list by their request modes
1400 and bits policy, thus improving the performance of search through
1404 Frequency : only if OST filesystem is corrupted
1406 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1407 Details : for a short time RPCs with bulk IO are in the replay list,
1408 but replay of bulk IOs is unimplemented. If the OST filesystem
1409 is corrupted due to disk cache incoherency and then replay is
1410 started it is possible to trip an assertion. Avoid putting
1411 committed RPCs into the replay list at all to avoid this issue.
1414 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1416 Description: System hang on startup
1417 Details : This bug allowed the liblustre (e.g. catamount) client to
1418 return to the app before handling all startup RPCs. This
1419 could leave the node unresponsive to lustre network traffic
1420 and manifested as a server ptllnd timeout.
1422 Severity : enhancement
1424 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1425 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1426 boolean to control whether to print peer debug info when a
1427 client's RPC times out.
1430 Frequency : only for kernels with patches from Lustre below 1.4.3
1432 Description: Remove old rdonly API
1433 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1436 Frequency : only for devices with external journals
1438 Description: Set external device read-only also
1439 Details : During a commanded failover stop, we set the disk device
1440 read-only while the server shuts down. We now also set any
1441 external journal device read-only at the same time.
1444 Frequency : when upgrading from 1.4 while trying to change parameters
1446 Description: The wrong (new) MDC name was used when setting parameters for
1447 upgraded MDT's. Also allows changing of OSC (and MDC)
1448 parameters if --writeconf is specified at tunefs upgrade time.
1451 Frequency : when setting specific ost indicies
1453 Description: QOS code breaks on skipped indicies
1454 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1455 created with --index need not be sequential.
1457 Severity : enhancement
1459 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1460 Details : The uninit_groups feature works in conjunction with the kernel
1461 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1462 up the pass1 processing of e2fsck. This is a read-only feature
1463 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1464 mount filesystems that have had this feature enabled.
1466 Severity : enhancement
1468 Description: Improve multi-block allocation algorithm to avoid fragmentation
1469 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1470 allocation locality and avoid filesystem fragmentation.
1472 ------------------------------------------------------------------------------
1474 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1476 * Support for kernels:
1477 2.4.21-47.0.1.EL (RHEL 3)
1478 2.6.5-7.283 (SLES 9)
1479 2.6.9-42.0.10.EL (RHEL 4)
1480 2.6.12.6 vanilla (kernel.org)
1481 2.6.16.27-0.9 (SLES 10)
1482 * Recommended e2fsprogs version: 1.39.cfs5
1484 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1488 Frequency : occasional, depends on client load and configuration
1489 Bugzilla : 12181, 12203
1490 Description: data loss for recently-modified files
1492 Details : In some cases it is possible that recently written or created
1493 files may not be written to disk in a timely manner (this should
1494 normally be within 30s unless client IO load is very high).
1495 The problem appears as zero-length files or files that are a
1496 multiple of 1MB in size after a client crash or client eviction
1497 that are missing data at the end of the file.
1499 This problem is more likely to be hit on clients where files are
1500 repeatedly created and unlinked in the same directory, clients
1501 have a large amount of RAM, have many CPUs, the filesystem has
1502 many OSTs, the clients are rebooted frequently, and/or the files
1503 are not accessed by other nodes after being written.
1505 The presence of the problem can be detected by looking at
1506 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1507 smaller than the second (nr_unused) then dirty files will not
1508 be flushed automatically to disk. "sync; sleep 10" should be
1509 run several times on the node before unmounting it to update
1510 Lustre (this is also safe to run on nodes without this problem).
1512 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1513 that can cause this same problem, so customers using that kernel
1514 also need to update the kernel in addition to Lustre. In order
1515 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1517 It is normal that files written just before a client crash (less
1518 than 30s) may not yet have been flushed to disk, even for local
1522 Frequency : frequent on thin XT3 nodes
1524 Description: UUID collision on thin XT3 Linux nodes
1525 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1526 randomly, since we relied on an insufficiently-seeded PRNG.
1531 Description: OSS hangs after "All ost request buffers busy"
1532 Details : A deadlock between quota and journal operations caused OSS
1533 hangs after printing "All ost request buffers busy."
1536 Frequency : always on liblustre builds
1538 Description: Cleanup compiler warnings on liblustre
1541 Frequency : always on liblustre builds on XT3
1543 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1547 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1548 > 32000 subdirectories
1549 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1550 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1551 subdirectory count crosses 32000. This will aid e2fsck to
1552 correctly handle more than 32000 subdirectories.
1555 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1557 Description: System hang on startup
1558 Details : This bug allowed the liblustre (e.g. catamount) client to
1559 return to the app before handling all startup RPCs. This
1560 could leave the node unresponsive to lustre network traffic
1561 and manifested as a server ptllnd timeout.
1563 Severity : enhancement
1565 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1566 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1567 boolean to control whether to print peer debug info when a
1568 client's RPC times out.
1573 Description: make O_SYNC working on 2.6 kernels
1574 Details : 2.6 kernels use different method for mark pages for write,
1575 so need add a code to lustre for O_SYNC work.
1580 Description: Failure to close file and release space on NFS
1581 Details : Put inode details into lock acquired in ll_intent_file_open.
1582 Use mdc_intent_lock in ll_intent_open to properly
1583 detect all kind of errors unhandled by mdc_enqueue.
1588 Description: proc file read during shutdown sometimes raced obd removal,
1590 Details : Add lock to prevent obd access after proc file removal.
1593 Frequency : Only for files larger than 4GB on 32-bit clients.
1595 Description: improperly doing page alignment of locks
1596 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1597 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1602 Description: RPCs being resent when they shouldn't be
1603 Details : Some RPCs that should not be resent are being resent. This
1604 can cause inconsistencies in the RPC state machine. Do not
1605 resend such requests.
1608 Frequency : rare, only with NFS export
1610 Description: Crash on NFS re-export node
1611 Details : under very unusual load conditions an assertion is hit in
1612 ll_intent_file_open()
1615 Frequency : only if OST filesystem is corrupted
1617 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1618 Details : for a short time RPCs with bulk IO are in the replay list,
1619 but replay of bulk IOs is unimplemented. If the OST filesystem
1620 is corrupted due to disk cache incoherency and then replay is
1621 started it is possible to trip an assertion. Avoid putting
1622 committed RPCs into the replay list at all to avoid this issue.
1627 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1628 Details : Large single O_DIRECT read and write calls can fail to allocate
1629 a sufficiently large buffer to process the request. In case of
1630 allocation failure the allocation is retried with a smaller
1631 buffer and broken into smaller requests.
1633 Severity : enhancement
1635 Description: Add -o localflock option to simulate old noflock behaviour.
1636 Details : This will achieve local-only flock/fcntl locks coherentness.
1641 Description: versioning check is incomplete
1642 Details : Checking the version difference of client vs. server, report
1643 error if the gap is too big.
1648 Description: add support PG_writeback bit
1649 Details : add support for PG_writeback bit for Lustre, for more carefull
1650 work with page cache in 2.6 kernel. This also fix some deadlocks
1651 and remove hack for work O_SYNC with 2.6 kernel.
1653 Severity : enhancement
1655 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1656 Details : The uninit_groups feature works in conjunction with the kernel
1657 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1658 up the pass1 processing of e2fsck. This is a read-only feature
1659 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1660 mount filesystems that have had this feature enabled.
1662 Severity : enhancement
1664 Description: Improve multi-block allocation algorithm to avoid fragmentation
1665 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1666 allocation locality and avoid filesystem fragmentation.
1668 ------------------------------------------------------------------------------
1670 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1672 * Support for kernels:
1673 2.6.9-42.0.3.EL (RHEL 4)
1674 2.6.5-7.276 (SLES 9)
1675 2.4.21-47.0.1.EL (RHEL 3)
1676 2.6.12.6 vanilla (kernel.org)
1677 2.6.16.21-0.8 (SLES10)
1678 * Recommended e2fsprogs version: 1.39.cfs2-0
1680 * The backwards-compatible /proc/sys/portals symlink has been removed
1681 in this release. Before upgrading, please ensure that you change
1682 any configuration scripts or /etc/sysctl.conf files that access
1683 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1684 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1685 in advance of the upgrade on any system running Lustre 1.4.6 or
1686 newer, since /proc/sys/lnet was added in that version.
1687 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1691 Frequency : only when quota is used
1693 Description: avoid scanning export list for quota master
1694 Details : Change the algorithms to avoid scanning export list in order
1695 to improve the efficiency.
1698 Frequency : MDS failover only, very rarely
1700 Description: "went back in time" messages on mds failover
1701 Details : The greatest transno may be lost when the current operation
1702 finishes with an error (transno==0) and the client's last_rcvd
1703 record is over-written. Save the greatest transno in the
1704 mds_last_transno for this case.
1707 Frequency : always for specific kernels and striping counts
1709 Description: client may get "Matching packet too big" without ACL support
1710 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1711 when trying to access files in certain configurations. The
1712 clients should in fact be denied when mounting because they do
1713 not understand ACLs.
1716 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1718 Description: many clients connecting with IO in progress causes connect timeouts
1719 Details : Avoid synchronous journal commits to avoid delays caused by many
1720 clients connecting/disconnecting when bulk IO is in progress.
1721 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1722 OST_IO_PORTAL to avoid delays behind potentially many pending
1726 Frequency : occasionally with multiple writers to a single file
1728 Description: shared writes to file may result in wrong size reported by stat()
1729 Details : Allow growing of kms when extent lock is cancelled
1732 Frequency : always with random mmap IO to multi-striped file
1734 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1735 Details : Only if the hole is at the end of OST object so that kms is too
1736 small. Fix is to increase kms accordingly in ll_nopage.
1739 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1741 Description: writes to a missing object would leak memory on the OST
1742 Details : If there is an inconsistency between the MDS and OST filesystems,
1743 such that the MDS references an object that doesn't exist, writes
1744 to that object will leak memory due to incorrect cleanup in the
1745 error handling path, eventually running out of memory on the OST.
1750 Description: Creating too long symlink causes lustre errors
1751 Details : Check symlink and name lengths before sending requests to MDS.
1754 Frequency : only if flock is enabled (not on by default)
1756 Description: posix locks not released on fd closure on 2.6.9+
1757 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1758 kernels, this caused such locks not to be released on fd close and
1759 then assertions on fs unmount about still used locks.
1762 Frequency : MDS failover only, very rarely
1764 Description: clients may get ASSERTION(granted_lock != NULL)
1765 Details : When request was taking a long time, and a client was resending
1766 a getattr by name lock request. The were multiple lock requests
1767 with the same client lock handle and
1768 mds_getattr_name->fixup_handle_for_resent_request found one of the
1769 lock handles but later failed with ASSERTION(granted_lock != NULL).
1774 Description: handle->h_buffer_credits > 0, assertion failure
1775 Details : h_buffer_credits is zero after truncate, causing assertion
1776 failure. This patch extends the transaction or creates a new
1780 Frequency : NFS re-export or patchless client
1781 Bugzilla : 11179, 10796
1782 Description: Crash on NFS re-export node (__d_move)
1783 Details : We do not want to hash the dentry if we don't have a lock.
1784 But if this dentry is later used in d_move, we'd hit uninitialised
1785 list head d_hash, so we just do this to init d_hash field but
1786 leave dentry unhashed.
1789 Frequency : NFS re-export or patchless client
1791 Description: NFS exports has problem with symbolic link
1792 Details : lustre client didn't properly install dentry when re-exported
1793 to NFS or running patchless client.
1796 Frequency : NFS re-export or patchless client
1798 Description: Various nfs/patchless fixes.
1799 Details : fixes reuse disconected alias for lookup process - this fixes
1800 warning "find_exported_dentry: npd != pd",
1801 fix permission error with open files at nfs.
1802 fix apply umask when do revalidate.
1805 Frequency : occasional
1807 Description: Crash on NFS re-export node
1808 Details : calling clear_page() on the wrong pointer triggered oops in
1809 generic_mapping_read().
1812 Frequency : rarely, using O_DIRECT IO
1814 Description: unaligned directio crashes client with LASSERT
1815 Details : check for unaligned buffers before trying any requests.
1818 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1820 Description: stale data returned from RAID cache
1821 Details : If only a small amount of IO is done to the RAID device before
1822 reading it again it is possible to get stale data from the RAID
1823 cache instead of reading it from disk.
1826 Frequency : always for sles10 kernel
1828 Description: sles10 support
1829 Details : ll_follow_link: compile fixes and using of nd_set_link
1830 under newer kernels.
1833 Frequency : depends on arch, kernel and compiler version, always on sles10
1836 Description: recursive or deep enough symlinks cause stack overflow
1837 Details : getting rid of large stack-allocated variable in
1841 Frequency : depends on hardware
1843 Description: lustre write performance loss in the SLES10 kernel
1844 Details : the performance loss is caused by using of write barriers in the
1845 ext3 code. The SLES10 kernel turns barrier support on by
1846 default. The fix is to undo that change for ldiskfs.
1848 ------------------------------------------------------------------------------
1850 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1852 * Support for kernels:
1853 2.6.9-42.0.3EL (RHEL 4)
1854 2.6.5-7.276 (SLES 9)
1855 2.4.21-47.0.1.EL (RHEL 3)
1856 2.6.12.6 vanilla (kernel.org)
1860 Frequency : quota enabled and large files being deleted
1862 Description: releasing more than 4GB of quota at once hangs OST
1863 Details : If a user deletes more than 4GB of files on a single OST it
1864 will cause the OST to spin in an infinite loop. Release
1865 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1870 Description: statfs data retrieved from /proc may be stale or zero
1871 Details : When reading per-device statfs data from /proc, in the
1872 {kbytes,files}_{total,free,avail} files, it may appear
1873 as zero or be out of date.
1876 Frequency : systems with MD RAID1 external journal devices
1878 Description: lconf's call to blkid is confused by RAID1 journal devices
1879 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1880 of returning all block devices that match the journal UUID.
1883 Frequency : always, for aggregate stripe size over 4GB
1885 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1886 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1887 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1890 Frequency : always if "lfs find" used on a local file/directory
1892 Description: "lfs find" segfaults if used on a local file/directory
1893 Details : The case where a directory component was not specified wasn't
1894 handled correctly. Handle this properly.
1897 Frequency : always on ppc64
1899 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1900 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1901 perform endianness conversion.
1904 Frequency : rarely (truncate to non-zero file size after write under load)
1905 Bugzilla : 10730, 10687
1906 Description: Files padded with zeros to next 4K multiple
1907 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1908 it is possible that files that are truncated to a non-zero size
1909 immediately after being written are filled with zero bytes beyond
1910 the truncated size. No file data is lost.
1912 Severity : enhancement
1914 Description: Allow recovery/failover for liblustre clients.
1915 Details : liblustre clients were unaware of failover configurations until
1918 Severity : enhancement
1920 Description: user file locks should fail when not mounting with flock option
1921 Details : Set up an error-returning stub in ll_file_operations.lock field
1922 to prevent incorrect behaviour when client is mounted without
1923 flock option. Also, set up properly f_op->flock field for
1927 Frequency : always on ia64
1929 Description: "lfs df" loops on printing out MDS statfs information
1930 Details : The obd_ioctl_data was not initialized and in some systems
1931 this caused a failure during the ioctl that did not return
1932 an error. Initialize the struct and return an error on failure.
1935 Frequency : SLES 9 only
1937 Description: Error of copying files with lustre special EAs as root
1938 Details : Client side always return success for setxattr call for lustre
1939 special xattr (currently only "trusted.lov").
1942 Frequency : rarely on clusters with both ia64+i386 clients
1944 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1945 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1946 clients are doing concurrent writes to the same file it is
1947 possible that the ia64 clients may LASSERT because the OST
1948 extent locks are not PAGE_SIZE aligned. Ensure that grown
1949 locks are always aligned on the request boundary.
1952 Frequency : specific use, occasional
1954 Description: Overwriting in use executable truncates on-disk binary image
1955 Details : If one node attempts to overwrite an executable in use by
1956 another node, we now correctly return ETXTBSY instead of
1957 truncating the file.
1959 Severity : enhancement
1961 Description: Async OSC create to avoid the blocking unnecessarily.
1962 Details : If a OST has no remain object, system will block on the creating
1963 when need to create a new object on this OST. Now, ways use
1964 pre-created objects when available, instead of blocking on an
1965 empty osc while others are not empty. If we must block, we block
1966 for the shortest possible period of time.
1971 Description: chmod on Lustre root is propagated to other clients
1972 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1973 invalid by the follow_mount time.
1976 Frequency : liblustre clients only
1978 Description: Race in 'instant cancel' lock handling could lead to such locks
1979 never to be granted in case of SMP MDS
1980 Details : Do not destroy not yet granted but cbpending locks in
1984 Frequency : replay/resend of open
1986 Description: non null lock assetion failure in mds_intent_policy
1987 Details : Trying to replay/resend lockless open requests resulted in
1988 mds_open() returning 0 with no lock. Now it sets a flag if
1989 a lock is going to be returned.
1991 Severity : enhancement
1993 Description: Checksum enhancements
1994 Details : New checksum enhancements allow for resending RPCs that failed
1997 Severity : enhancement
1999 Description: Tunables on number of dirty pages in cacche
2000 Details : Allow to set limit on number of dirty pages cached.
2005 Description: client crash on unmount - lock still has references
2006 Details : In some error handling cases it was possible to leak a lock
2007 reference on a client while accessing a file. This was not
2008 harmful to the client during operation, but would cause the
2009 client to crash when the filesystem is unmounted.
2012 Frequency : specific case, rare
2014 Description: ETXTBSY on mds though file not in use by client
2015 Details : ETXTBSY is no longer incorrectly returned when attempting to
2016 chmod or chown a directory that the user previously tried to
2017 execute or a currently-executing binary.
2020 Frequency : extremely rare except on liblustre-based clients
2022 Description: Lustre space not freed when files are deleted
2023 Details : Clean up open-unlinked files after client eviction. Previously
2024 the unlink was skipped and the files remained as orphans.
2029 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2030 Details : In some cases it was possible to send a blocking callback to
2031 a client doing a glimpse, even though that client didn't get
2032 a lock granted. When the glimpse lock is cancelled on the OST
2033 the freed lock is left on the waiting list and corrupted the list.
2036 Frequency : all core dumps
2038 Description: Broke core dumps to lustre
2039 Details : Negative dentry may be unhashed if parent does not have UPDATE
2040 lock, but some callers, e.g. do_coredump, expect dentry to be
2041 hashed after successful create, hash it in ll_create_it.
2043 ------------------------------------------------------------------------------
2045 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2047 * Support for kernels:
2048 2.6.9-42.0.2.EL (RHEL 4)
2049 2.6.5-7.276 (SLES 9)
2050 2.4.21-40.EL (RHEL 3)
2051 2.6.12.6 vanilla (kernel.org)
2055 Frequency : always on RHEL 3
2057 Description: Number of open files grows over time
2058 Details : The number of open files grows over time, whether or not
2059 Lustre is started. This was due to a filp leak introduced
2060 by one of our kernel patches.
2062 ------------------------------------------------------------------------------
2064 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
2066 * Support for kernels:
2067 2.6.9-42.EL (RHEL 4)
2068 2.6.5-7.267 (SLES 9)
2069 2.4.21-40.EL (RHEL 3)
2070 2.6.12.6 vanilla (kernel.org)
2075 Bugzilla : 5719, 9635, 9792, 9684
2076 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2077 Details : If a server is under heavy load and cannot reply to new
2078 connection requests before the client resends the (re)connect,
2079 the connection handling code can behave badly if two service
2080 threads are concurrently handing separate (re)connections from
2081 the same client. Add better locking to the connection handling
2082 code, and ensure that only a single connection will be processed
2083 for a given client UUID, even if the lock is dropped.
2085 Severity : enhancement
2087 Description: add TCP zero-copy support to kernel
2088 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2089 sends if the hardware supports scatter-gather and checksumming.
2090 This allows socklnd to do client-write and server-read more
2091 efficiently and reduce CPU utilization from skbuf copying.
2094 Frequency : only if NFS exporting from client
2096 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2097 Details : If setting attributes on a file created under NFS that had
2098 never been opened it would be possible to oops the client
2099 if the file had no objects.
2103 Bugzilla : 9326, 10402, 10897
2104 Description: client crash in ptlrpcd_wake() thread when sending async RPC
2105 Details : It is possible that ptlrpcd_wake() dereferences a freed async
2106 RPC. In rare cases the ptlrpcd thread alread processed the RPC
2107 before ptlrpcd_wake() was called and the request was freed.
2110 Frequency : always for liblustre
2112 Description: liblustre client does MDS+OSTs setattr RPC for each write
2113 Details : When doing a write from a liblustre client, the client
2114 incorrectly issued an RPC to the MDS and each OST the file was
2115 striped over in order to update the timestamps. When writing
2116 with small chunks and many clients this could overwhelm the MDS
2117 with RPCs. In all cases it would slow down the write because
2118 these RPCs are unnecessary.
2120 Severity : enhancement
2122 Description: allow number of MDS service threads to be changed at module load
2123 Details : It is now possible to change the number of MDS service threads
2124 running. Adding "options mds mds_num_threads={N}" to the MDS's
2125 /etc/modprobe.conf will set the number of threads for the next
2126 time Lustre is restarted (assuming the "mds" module is also
2127 reloaded at that time). The default number of threads will
2128 stay the same, 32 for most systems.
2133 Description: OST crash if filesystem is unformatted or corrupt
2134 Details : If an OST is started on a device that has never been formatted
2135 or if the filesystem is corrupt and cannot even mount then the
2136 error handling cleanup routines would dereference a NULL pointer.
2141 Description: NULL pointer deref in llap_from_page.
2142 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2143 incorrect) mapping if the page was truncated/reclaimed while it was
2144 searched for. Check for this condition and skip such pages when
2145 doing readahead. Introduce extra check to llap_from_page() to
2146 verify page->mapping->host is non-NULL (so page is not anonymous).
2149 Frequency : Sometimes when using sys_sendfile
2151 Description: "page not covered by a lock" warnings from ll_readpage
2152 Details : sendfile called ll_readpage without right page locks present.
2153 Now we introduced ll_file_sendfile that does necessary locking
2154 around call to generic_file_sendfile() much like we do in
2158 Frequency : with certain MDS communication failures at client mount time
2160 Description: NULL pointer deref after failed client mount
2161 Details : a client connection request may delayed by the network layer
2162 and not be sent until after the PTLRPC layer has timed out the
2163 request. If the client fails the mount immediately it will try
2164 to clean up before the network times out the request. Add a
2165 reference from the request import to the obd device and delay
2166 the cleanup until the network drops the request.
2169 Frequency : occasionally during client (re)connect
2171 Description: assertion failure during client (re)connect
2172 Details : processing a client connection request may be delayed by the
2173 client or server longer than the client connect timeout. This
2174 causes the client to resend the connection request. If the
2175 original connection request is replied in this interval, the
2176 client may trip an assertion failure in ptlrpc_connect_interpret()
2177 which thought it would be the only running connect process.
2180 Frequency : only with obd_echo servers and clients that are rebooted
2182 Description: kernel BUG accessing uninitialized data structure
2183 Details : When running an obd_echo server it did not start the ping_evictor
2184 thread, and when a client was evicted an uninitialized data
2185 structure was accessed. Start the ping_evictor in the RPC
2186 service startup instead of the OBD startup.
2188 Severity : enhancement
2189 Bugzilla : 10193 (patchless)
2190 Description: Remove dependency on various unexported kernel interfaces.
2191 Details : No longer need reparent_to_init, exit_mm, exit_files,
2192 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2195 Frequency : rare (only users of deprecated and unsupported LDAP config)
2197 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2198 Details : LDAP apparently contains 'lustreName' attributes instead of
2199 'name'. A simple remapping of the name is sufficient.
2202 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2204 Description: waiting_locks_callback trips kernel BUG if client is evicted
2205 Details : Running with the dump_on_timeout debug flag turned on makes
2206 it possible that the waiting_locks_callback() can try to dump
2207 the Lustre kernel debug logs from an interrupt handler. Defer
2208 this log dumping to the expired_lock_main() thread.
2210 Severity : enhancement
2212 Description: Support NFS exporting on 2.6 kernels.
2213 Details : Implement non-rawops metadata methods for NFS server to use without
2214 changing NFS server code.
2217 Frequency : very rare (synthetic metadata workload only)
2219 Description: two racing renames might cause an MDS thread to deadlock
2220 Details : Running the "racer" program may cause one MDS thread to rename
2221 a file from being the source of a rename to being the target of
2222 a rename at exactly the same time that another thread is doing
2223 so, and the second thread has already enqueued these locks after
2224 doing a lookup of the target and is trying to relock them in
2225 order. Ensure that we don't try to re-lock the same resource.
2228 Frequency : only very large systems with liblustre clients
2230 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2231 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2232 all OSTs in parallel. This allows the eviction of stale liblustre
2233 clients to proceed much faster than if they were done in series,
2234 and also offers similar improvements for other set_info RPCs.
2239 Description: excessive CPU usage during initial read phase on client
2240 Details : During the initial read phase on a client, it would agressively
2241 retry readahead on the file, consuming too much CPU and impacting
2242 performance (since 1.4.5.8). Improve the readahead algorithm
2243 to avoid this, and also improve some other common cases (read
2244 of small files in particular, where "small" is files smaller than
2245 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2250 Description: MDS crash when receiving packet with unknown intent.
2251 Details : Do not LBUG in unknown intent case, just return -EFAULT
2253 Severity : enhancement
2254 Bugzilla : 9293, 9385
2255 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2256 Details : Do not serialize getattr (non-intent version) and statfs.
2259 Frequency : occasional, when OST network is overloaded/intermittent
2261 Description: client evicted by OST after bulk IO timeout
2262 Details : If a client sends a bulk IO request (read or write) the OST
2263 may evict the client if it is unresposive to its data GET/PUT
2264 request. This is incorrect if the network is overloaded (takes
2265 too long to transfer the RPC data) or dropped the OST GET/PUT
2266 request. There is no need to evict the client at all, since
2267 the pinger and/or lock callbacks will handle this, and the
2268 client can restart the bulk request.
2271 Frequency : Always when mmapping file with no objects
2273 Description: client crashes when mmapping file with no objects
2274 Details : Check that we actually have objects in a file before doing any
2275 operations on objects in ll_vm_open, ll_vm_close and
2281 Description: Request leak when working with deleted CWD
2282 Details : Introduce advanced request refcount tracking for requests
2283 referenced from lustre intent.
2285 Severity : Enhancement
2287 Description: Cache open file handles on client.
2288 Details : MDS now will return special lock along with openhandle, if
2289 requested and client is allowed to hold openhandle, even if unused,
2290 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2291 closing files for every read/write openration.
2293 Severity : Enhancement
2295 Description: Cache open negative dentries on client when possible.
2296 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2297 negative dentries on lock revocation.
2302 Description: Remounting a client read-only wasn't possible with a zconf mount
2303 Details : It wasn't possible to remount a client read-only with llmount.
2305 Severity : enhancement
2306 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2307 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2308 MPICH is included to add Lustre-specific ADIO interfaces.
2309 This is based closely on the UFS ADIO layer and only differs
2310 in file creation, in order to allow the OST striping to be set.
2311 This is user-contributed code and not supported by CFS.
2316 Description: extended inode attributes (immutable, append-only) work improperly
2317 when 2.4 and 2.6 kernels are used on client/server or vice versa
2318 Details : Introduce kernel-independent values for these flags.
2320 Severity : enhancement
2323 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2324 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2325 granularity. Now, Fractional MB granularity is be supported,
2326 this is very useful for low memory system.
2328 Severity : enhancement
2330 Description: Getattr by fid
2331 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2332 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2333 have locks and avoids getattr rpc requests for it on every stat.
2336 Frequency : Always, for filesystems larger than 2TB
2338 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2339 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2340 filesystems larger than 2TB to be mounted with mballoc enabled.
2343 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2345 Description: filesystem corruption for non-standard kernels and very large OSTs
2346 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2347 filesystems larger than 2TB is mounted then the kernel will
2348 silently corrupt the start of the filesystem. CONFIG_LBD is
2349 enabled for all CFS-supported kernels, but the possibility of
2350 this happening with a modified kernel config exists.
2352 Severity : enhancement
2354 Description: add client O_DIRECT support for 2.6 kernels
2355 Details : It is now possible to do O_DIRECT reads and writes to files
2356 in the Lustre client mountpoint on 2.6 kernel clients.
2358 Severity : enhancement
2360 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2361 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2362 parallel, not waiting for response from every OST before sending
2363 a rpc to the next OST.
2368 Description: setattr vs write race when updating file timestamps
2369 Details : Client processes that update a file timestamp into the past
2370 right after writing to the file (e.g. tar) it is possible that
2371 the updated file modification time can be reset to the current
2372 time due to a race between processing the setattr and write RPC.
2374 Severity : enhancement
2376 Description: Bring 'lfs find' closer in line with regular Linux find.
2377 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2378 -print0 options and obtains all the needed info through the lustre
2381 Severity : enhancement
2383 Description: support up to 1024 configured devices on one node
2384 Details : change obd_dev array from statically allocated to dynamically
2385 allocated structs as they are first used to reduce memory usage
2390 Description: Flush dirty partially truncated pages during truncate
2391 Details : Immediatelly flush partially truncated pages in filter_setattr,
2392 this way we completely avoid having any pages in page cache on OST
2393 and can retire ugly workarounds during writes to flush such pages.
2398 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2399 Details : protect inode from truncation within vfs_unlink() context
2400 just take a reference before calling vfs_unlink() and release it
2401 when parent's i_sem is free.
2404 Frequency : always, if extents are used on OSTs
2406 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2407 Details : OSTs using the extents format would not zero the high 16 bits of
2408 the index physical block number. This is not a problem for any
2409 OST filesystems smaller than 16TB, and no kernels support ext3
2410 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2411 new/modified files) and can be fixed for existing filesystems
2412 with e2fsprogs-1.39-cfs1.
2417 Description: import connection selection may be incorrect if timer wraps
2418 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2419 import connections to be ignored if the 32-bit jiffies counter
2420 wraps. Use a 64-bit jiffies counter.
2423 Frequency : during server recovery
2425 Description: crash after server is denying duplicate export
2426 Details : If clients are resending connect requests to the server, the
2427 server refuses to allow a client to connect multiple times.
2428 Fixed a bug in the handling of this case.
2431 Frequency : very large clusters immediately after boot
2433 Description: LNET request buffers exhausted under heavy short-term load
2434 Details : If a large number of client requests are generated on a service
2435 that has previously never seen so many requests it is possible
2436 that the request buffer growth cannot keep up with the spike in
2437 demand. Instead of dropping incoming requests, they are held in
2438 the LND until the RPC service can accept more requests.
2441 Frequency : Sometimes during replay
2443 Description: Assertion failure in ll_local_open after replay.
2444 Details : If replay happened on an open request reply before we were able
2445 to set replay handler, reply will become not swabbed tripping the
2446 assertion in ll_local_open. Now we set the handler right after
2447 recognising of open request
2450 Frequency : very rare
2452 Description: kernel reports "badness in vsnprintf"
2453 Details : Reading from the "recovery_status" /proc file in small chunks
2454 may cause a negative length in lprocfs_obd_rd_recovery_status()
2455 call to vsnprintf() (which is otherwise harmless). Exit early
2456 if there is no more space in the output buffer.
2458 Severity : enhancement
2460 Description: clear OBD RPC statistics by writing to them
2461 Details : It is now possible to clear the OBD RPC statistics by writing
2462 to the "stats" file.
2467 Description: Client mtime is not the same on different clients after utimes
2468 Details : In some cases, the client was using the utimes() syscall on
2469 a file cached on another node. The clients now validate the
2470 ctime from the MDS + OSTs to determine which one is right.
2475 Description: Inability to activate failout mode
2476 Details : lconf script incorrectly assumed that in python string's numeric
2477 value is used in comparisons.
2480 Frequency : always with multiple stripes per file
2482 Description: Inefficient object allocation for mutli-stripe files
2483 Details : When selecting which OSTs to stripe files over, for files with
2484 a stripe count that divides evenly into the number of OSTs,
2485 the MDS is always picking the same starting OST for each file.
2486 Return the OST selection heuristic to the original design.
2491 Description: mount failures may take full timeout to return an error
2492 Details : Under some heavy load conditions it is possible that a
2493 failed mount can wait for the full obd_timeout interval,
2494 possibly several minutes, before reporting an error.
2495 Instead return an error as soon as the status is known.
2497 ------------------------------------------------------------------------------
2499 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
2501 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2502 INTEROPERATE with older versions automatically. Please read the
2503 user documentation before upgrading any part of a live system.
2504 * WARNING: Lustre networking configuration changes are required with
2505 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2508 * Support for kernels:
2509 2.6.9-22.0.2.EL (RHEL 4)
2510 2.6.5-7.244 (SLES 9)
2511 2.6.12.6 vanilla (kernel.org)
2514 Severity : enhancement
2515 Bugzilla : 7981/8208
2516 Description: Introduced Lustre Networking (LNET)
2517 Details : LNET is new networking infrastructure for Lustre, it includes
2518 a reorganized network configuration mode (see the user
2519 documentation for full details) as well as support for routing
2520 between different network fabrics. Lustre Networking Devices
2521 (LNDs) for the supported network fabrics have also been
2522 created for this new infrastructure.
2524 Severity : enhancement
2525 Description: Introduced Access control lists
2526 Details : clients can set ACLs on files and directories in order to have
2527 more fine-grained permissions than the standard Unix UGO+RWX.
2528 The MDS must be started with the "-o acl" mount option.
2530 Severity : enhancement
2531 Description: Introduced filesystem quotas
2532 Details : Administrators may now establish per-user quotas on the
2535 Severity : enhancement
2537 Description: Configuration change for the XT3
2538 The PTLLND is now used to run Lustre over Portals on the XT3
2539 The configure option(s) --with-cray-portals are no longer used.
2540 Rather --with-portals=<path-to-portals-includes> is used to
2541 enable building on the XT3. In addition to enable XT3 specific
2542 features the option --enable-cray-xt3 must be used.
2547 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2548 Details : A race between allocating a new llcd and re-getting the llcd_lock
2549 allowed another thread to grab newly-allocated llcd.
2551 Severity : enhancement
2553 Description: 2.6 OST async journal commit and locking fix to improve performance
2554 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2555 2.6 kernels are now async as they already were in 2.4 kernels so
2556 that they can commit concurrently with the network bulk transfer.
2557 For block-allocated files the filter allocation semaphore is held
2558 to avoid filesystem fragmentation during allocation. BKL lock
2559 removed for 2.6 xattr operations where it is no longer needed.
2564 Description: lconf incorrectly determined whether two IP networks could talk
2565 Details : In some more complicated routing and multiple-network
2566 configurations, lconf will avoid trying to make a network
2567 connection to a disjoint part of the IP space. It was doing the
2568 math incorrectly for one set of cases.
2573 Description: Fix for potential infinite loop processing records in an llog.
2574 Details : If an llog record is corrupted/zeroed, it is possible to loop
2575 forever in llog_process(). Validate the llog record length
2576 and skip the remainder of the block on error.
2579 Frequency : occasional (liblustre only)
2581 Description: liblustre could not open files whose last component is a symlink
2582 Details : sysio_path_walk() would incorrectly pass the open intent to
2583 intermediate path components.
2586 Frequency : rare (liblustre only with non-standard tuning)
2587 Bugzilla : 7201 (7350)
2588 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2589 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2590 after it has been released, only do this for referenced locks
2592 Severity : enhancement
2594 Description: specifying an (invalid) directory default stripe_size of -1
2595 would reset the directory default striping
2596 Details : stripe_size -1 was used internally to signal directory stripe
2597 removal, now use "all default" to signal dir stripe removal
2598 as a directory striping of "all default" is not useful
2601 Frequency : common for large clusters running liblustre clients
2603 Description: doing an ls when liblustre clients are running is slow
2604 Details : sending a glimpse AST to a liblustre client waits for every AST
2605 to time out, as liblustre clients will not respond. Since they
2606 cannot cache data we refresh the OST lock LVB from disk instead.
2608 Severity : enhancement
2610 Description: doing an ls at the same time as file IO can be slow
2611 Details : enqueue and other "small" requests can be blocked behind many
2612 large IO requests. Create a new OST IO portal for non-IO
2613 requests so they can be processed faster.
2616 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2618 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2619 created file with mode 000
2620 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2621 allow an ftruncate() as a non-root user to a file with mode 000.
2622 Lustre now respects this flag to disable mode checks when
2623 truncating a file owned by the user
2626 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2629 Description: Revoking locks from clients that went dead or catatonic might take
2631 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2632 cancellation of such locks instant on servers without waiting for
2633 any reply from clients. Clients drops these locks when cancel
2634 notification from server is received without replying.
2637 Frequency : liblustre-only, when liblustre client dies or becomes busy
2639 Description: Doing ls on Linux clients can take a long time with active
2641 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2642 granting such locks to it in the first place if possible. Locks
2643 are taken by proxy on the OST during the read or write and
2644 dropped immediately afterward. Add connect flags handling, do
2645 not grant locks to liblustre clients for glimpse ASTs.
2647 Severity : enhancement
2649 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2650 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2651 sequential reads and behaves badly with applications doing random
2652 reads. Improve it by reading ahead at least read region, and
2653 avoiding excessive large RPC for small reads.
2655 Severity : enhancement
2657 Description: Creating more than 1000 files for a single job may cause a load
2658 imbalance on the OSTs if there are also a large number of OSTs.
2659 Details : qos_prep_create() uses an OST index reseed value that is an
2660 even multiple of the number of available OSTs so that if the
2661 reseed happens in the middle of the object allocation it will
2662 still utilize the OSTs as uniformly as possible.
2667 Description: OST or MDS may oops in ping_evictor_main()
2668 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2669 but doesn't restart at beginning of obd_exports_timed list
2672 Severity : enhancement
2674 Description: improve by-nid export eviction on the MDS and OST
2675 Details : allow multiple exports with the same NID to be evicted at one
2676 time without re-searching the exports list.
2679 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2681 Description: MDS may oops in groups_free()
2682 Details : in rare race conditions a newly allocated group_info struct is
2683 freed again, and this can be NULL. The 2.4 compatibility code
2684 for groups_free() checked for a NULL pointer, but 2.6 did not.
2687 Frequency : common for liblustre clients doing little filesystem IO
2688 Bugzilla : 9352, 7313
2689 Description: server may evict liblustre clients accessing contended locks
2690 Details : if a client is granted a lock or receives a completion AST
2691 with a blocking AST already set it would not reply to the AST
2692 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2695 Frequency : lfs setstripe, only systems with more than 160 OSTs
2697 Description: unable to set striping with a starting offset beyond OST 160
2698 Details : llapi_create_file() incorrectly limited the starting stripe
2699 index to the maximum single-file stripe count.
2702 Frequency : LDAP users only
2704 Description: lconf did not handle in-kernel recovery with LDAP properly
2705 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2707 Severity : enhancement
2709 Description: bind OST threads to NUMA nodes to improve performance
2710 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2711 node and do their allocations there to localize memory access
2713 Severity : enhancement
2715 Description: llmount can determine client NID directly from Myrinet (GM)
2716 Details : the client NID code from gmnalnid was moved directly into
2717 llmount, removing the need to use this or specifying the
2718 client NID explicitly when mounting GM clients with zeroconf
2721 Frequency : if client is started with down MDS
2723 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2724 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2725 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2726 ptlrpc_import_delayed_req() because we didn't check if the
2727 request was interrupted, and we also didn't break out of the
2728 event loop if there was no timeout
2733 Description: data loss during non-page-aligned writes to a single file from
2734 both multiple nodes and multiple threads on one node at same time
2735 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2736 inconsistency led to false short-reads, that were cached and later
2737 used by ->prepare_write() to fill in partially written page,
2738 leading to data loss.
2741 Frequency : always, if lconf --abort_recovery used
2743 Description: lconf --abort_recovery fails with 'Operation not supported'
2744 Details : lconf was attempting to abort recovery on the MDT device and not
2747 Severity : enhancement
2749 Description: remove cleanup logs
2750 Details : replace lconf-generated cleanup logs with lustre internal
2751 cleanup routines. Eliminates the need for client-cleanup and
2754 Severity : enhancement
2756 Description: add support for EAs (user and system) on lustre filesystems
2757 Details : it is now possible to store extended attributes in the Lustre
2758 client filesystem, and with the user_xattr mount option it
2759 is possible to allow users to store EAs on their files also
2761 Severity : enhancement
2763 Description: Add possibility (config option) to show minimal available OST free
2765 Details : When compiled with --enable-mindf configure option, statfs(2)
2766 (and so, df) will return least minimal free space available from
2767 all OSTs as amount of free space on FS, instead of summary of
2768 free spaces of all OSTs.
2770 Severity : enhancement
2772 Description: do not expand extent locks acquired on OST-side
2773 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2774 by server: they are not cached anyway.
2777 Frequency : when mmap is used/binaries executed from Lustre
2779 Description: Unmmap pages before throwing them away from read cache.
2780 Details : llap_shrink cache now attempts to unmap pages before discarding
2781 them (if unmapping failed - do not discard). SLES9 kernel has
2782 extra checks that trigger if this unmapping is not done first.
2787 Description: lconf didn't resolve symlinks before checking to see whether a
2788 given mountpoint was already in use
2791 Frequency : when migrating failover services
2792 Bugzilla : 6395, 9514
2793 Description: When migrating a subset of services from a node (e.g. failback
2794 from a failover service node) the remaining services would
2795 time out and evict clients.
2796 Details : lconf --force (implied by --failover) sets the global obd_timeout
2797 to 5 seconds in order to quickly disconnect, but this caused
2798 other RPCs to time out too quickly. Do not change the global
2799 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2801 Severity : enhancement
2802 Frequency : if MDS is started with down OST
2803 Bugzilla : 9439,5706
2804 Description: Allow startup/shutdown of an MDS without depending on the
2805 availability of the OSTs.
2806 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2807 Add appropriate locking and lov-osc refcounts for safe
2808 cleaning. Add osc abort_inflight calls in case the
2809 synchronize never started.
2812 Frequency : occasional (Cray XT3 only)
2814 Description: root not authorized to access files in CRAY_PORTALS environment
2815 Details : The client process capabilities were not honoured on the MDS in
2816 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2817 been accessed by an authorized user then root was able to access
2818 the file on the local client also. The root user capabilities
2819 are now allowed on the MDS, as this environment has secure UID.
2822 Frequency : occasional
2824 Description: ldiskfs "too long searching" message happens too often
2825 Details : A debugging message (otherwise harmless) prints too often on
2826 the OST console. This has been reduced to only happen when
2827 there are fragmentation problems on the filesystem.
2832 Description: Division by zero in statfs when all OSCs are inactive
2833 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2834 lov_statfs divides by value returned by lov_get_stripecnt().
2838 Bugzilla : 9489, 3273
2839 Description: First write from each client to each OST was only 4kB in size,
2840 to initialize client writeback cache, which caused sub-optimal
2841 RPCs and poor layout on disk for the first writen file.
2842 Details : Clients now request an initial cache grant at (re)connect time
2843 and so that they can start streaming writes to the cache right
2844 away and always do full-sized RPCs if there is enough data.
2845 If the OST is rebooted the client also re-establishes its grant
2846 so that client cached writes will be honoured under the grant.
2851 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2852 Details : Now I/O RPCs go to different portal number and (presumably) fast
2853 lock requests (and glimses) and other RPCs get their own service
2854 threads pool that should be able to service those RPCs
2857 Severity : enhancement
2859 Description: Ability to exchange lustre version between client and servers and
2860 issue warnings at client side if client is too old. Also for
2861 liblustre clients there is ability to refuse connection of too old
2863 Details : New 'version' field is added to connect data structure that is
2864 filled with version info. That info is later checked by server and
2868 Frequency : rare, liblustre only.
2869 Bugzilla : 9296, 9581
2870 Description: Two simultaneous writes from liblustre at offset within same page
2871 might proceed at the same time overwriting eachother with stale
2873 Details : I/O lock withing llu_file_prwv was released too early, before data
2874 actually was hitting the wire. Extended lock-holding time until
2875 server acknowledges receiving data.
2878 Frequency : extremely rare. Never observed in practice.
2880 Description: avoid generating lustre_handle cookie of 0.
2881 Details : class_handle_hash() generates handle cookies by incrementing
2882 global counter, and can hit 0 occasionaly (this is unlikely, but
2883 not impossible, because initial value of cookie counter is
2884 selected randonly). Value of 0 is used as a sentinel meaning
2885 "unassigned handle" --- avoid it. Also coalesce two critical
2886 sections in this function into one.
2888 Severity : enhancement
2890 Description: allow liblustre clients to delegate truncate locking to OST
2891 Details : To avoid overhead of locking, liblustre client instructs OST to
2892 take extent lock in ost_punch() on client's behalf. New connection
2893 flag is added to handle backward compatibility.
2895 Severity : enhancement
2896 Bugzilla : 4928, 7341, 9758
2897 Description: allow number of OST service threads to be specified
2898 Details : a module parameter allows the number of OST service threads
2899 to be specified via "options ost ost_num_threads={N}" in the
2900 OSS's /etc/modules.conf or /etc/modprobe.conf.
2904 Bugzilla : 6146, 9635, 9895
2905 Description: servers crash with bad pointer in target_handle_connect()
2906 Details : In rare cases when a client is reconnecting it was possible that
2907 the connection request was the last reference for that export.
2908 We would temporarily drop the export reference and get a new
2909 one, but this may have been the last reference and the export
2910 was just destroyed. Get new reference before dropping old one.
2912 Severity : enhancement
2913 Frequency : if client is started with failover MDS
2915 Description: Allow multiple MDS hostnames in the mount command
2916 Details : Try to read the configuration from all specified MDS
2917 hostnames during a client mount in case the "primary"
2920 Severity : enhancement
2922 Description: Stop sending data to evicted clients as soon as possible.
2923 Details : Check if the client we are about to send or are sending data to
2924 was evicted already. (Check is done every second of waiting,
2925 for which l_wait_event interface was extended to allow checking
2926 of exit condition at specified intervals).
2929 Frequency : rare, normally only when NFS exporting is done from client
2931 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2933 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2936 Frequency : always, if srand() is called before liblustre initialization
2938 Description: Liblustre uses system PRNG disturbing its usage by user application
2939 Details : Introduce internal to lustre fast and high-quality PRNG for
2940 lustre usage and make liblustre and some other places in generic
2941 lustre code to use it.
2943 Severity : enhancement
2944 Bugzilla : 9477, 9557, 9870
2945 Description: Verify that the MDS configuration logs are updated when xml is
2946 Details : Check if the .xml configuration logs are newer than the config
2947 logs stored on the MDS and report an error if this is the case.
2948 Request --write-conf, or allow starting with --old_conf.
2950 Severity : enhancement
2952 Description: Handle symlinks in the path when checking if Lustre is mounted.
2953 Details : Resolve intermediate symlinks when checking if a client has
2954 mounted a filesystem to avoid duplicate client mounts.
2959 Description: lconf can hit an error exception but still return success.
2960 Details : The lconf command catches the Command error exception at the top
2961 level script context and will exit with the associated exit
2962 status, but doesn't ensure that this exit status is non-zero.
2967 Description: failure of ptlrpc thread startup can cause oops
2968 Details : Starting a ptlrpc service thread can fail if there are a large
2969 number of threads or the server memory is very fragmented.
2970 Handle this without oopsing.
2973 Frequency : always, only if liblustre and non-default acceptor port was used
2975 Description: liblustre cannot connect to servers with non-default acceptor port
2976 Details : tcpnal_set_default_params() was not called and was therefore
2977 ignoring the environment varaible TCPNAL_PORT, as well as other
2978 TCPNAL_ environment variables
2983 Description: two objects could be created on the same OST for a single file
2984 Details : If an OST is down, in some cases it was possible to create two
2985 objects on a single OST for a single file. No problems other
2986 than potential performance impact and spurious error messages.
2990 Bugzilla : 5681, 9562
2991 Description: Client may oops in ll_unhash_aliases
2992 Details : Client dcache may become inconsistent in race condition.
2993 In some cases "getcwd" can fail if the current directory is
2999 Description: Inode refcounting problems in NFS export code
3000 Details : link_raw functions used to call d_instantiate without obtaining
3001 extra inode reference first.
3005 Bugzilla : 9942, 9903
3006 Description: Referencing freed requests leading to crash, memleaks with NFS.
3007 Details : We used to require that call to ll_revalidate_it was always
3008 followed by ll_lookup_it. Also with revalidate_special() it is
3009 possible to call ll_revalidate_it() twice for the same dentry
3010 even if first occurence returned success. This fix changes semantic
3011 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3012 reference on a request referred from the intent.
3013 ll_intent_release() then releases such a request.
3016 Frequency : rare, normally benchmark loads only
3018 Description: unlinked inodes were kept in memory on the client
3019 Details : If a client is repeatedly creating and unlinking files it
3020 can accumulate a lot of stale inodes in the inode slab cache.
3021 If there is no other client load running this can cause the
3022 client node to run out of memory. Instead flush old inodes
3023 from client cache that have the same inode number as a new inode.
3026 Frequency : SLES9 2.6.5 kernel and long filenames only
3027 Bugzilla : 9969, 10379
3028 Description: utime reports stale NFS file handle
3029 Details : SLES9 uses out-of-dentry names in some cases, which confused
3030 the lustre dentry revalidation. Change it to always use the
3034 Frequency : rare, unless heavy write-truncate concurrency is continuous
3035 Bugzilla : 4180, 6984, 7171, 9963, 9331
3036 Description: OST becomes very slow and/or deadlocked during object unlink
3037 Details : filter_destroy() was holding onto the parent directory lock
3038 while truncating+unlinking objects. For very large objects this
3039 may block other threads for a long time and slow overall OST
3040 responsiveness. It may also be possible to get a lock ordering
3041 deadlock in this case, or run out of journal credits because of
3042 the combined truncate+unlink. Solution is to do object truncate
3043 first in one transaction without parent lock, and then do the
3044 final unlink in a new transaction with the parent lock. This
3045 reduces the lock hold time dramatically.
3048 Frequency : rare, 2.4 kernels only
3050 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3051 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3052 before unmounting their filesystem in order to avoid deadlock.
3053 The kernel_locked() function in 2.4 kernels only checks whether
3054 the kernel lock is held, not whether it is this process that is
3055 holding it as 2.6 kernels do.
3060 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3061 Details : The client ptlrpc code may be trying to reconnect to a down
3062 server before a previous connection attempt has timed out.
3063 Increase the reconnect interval to be longer than the connection
3064 timeout interval to avoid sending duplicate connections to
3068 Frequency : echo_client brw_test command
3070 Description: fix echo_client to work with OST preallocated code
3071 Details : OST preallocation code (5137) didn't take echo_client IO path
3072 into account: echo_client calls filter methods outside of any
3073 OST thread and, hence, there is no per-thread preallocated
3074 pages and buffers to use. Solution: hijack pga pages for IO. As
3075 a byproduct, this avoids unnecessary data copying.
3079 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3080 Description: Client can oops in mdc_commit_close() after open replay
3081 Details : It was possible for the MDS to return an open request with no
3082 transaction number in mds_finish_transno() if the client was
3083 evicted, but without actually returning an error. Clients
3084 would later try to replay that open and may trip an assertion
3085 Simplify the client close codepath, and always return an error
3086 from the MDS in case the open is not successful.
3089 Frequency : rare, 2.6 OSTs only
3091 Description: OST may deadlock under high load on fragmented files
3092 Details : If there was a heavy load and highly-fragmented OST filesystems
3093 it was possible to have all the OST threads deadlock waiting on
3094 allocation of biovecs, because the biovecs were not released
3095 until the entire RPC IO was completed. Instead, release biovecs
3096 as soon as they are complete to ensure forward IO progress.
3098 Severity : enhancement
3100 Description: Support for specifying external journal device at mount
3101 Details : If an OST or MDS device is formatted with an external journal
3102 device, this device major/minor is stored in the ext3 superblock
3103 and may not be valid for failover. Allow detecting and
3104 specifying the external journal at mount time.
3109 Description: Mounting an MDS with pending unlinked files may cause oops
3110 Details : target_finish_recovery() calls mds_postrecov() which returned
3111 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3112 considers this an error and immediately begins cleaning up the
3113 lov, just after starting the mds_lov process
3115 Severity : enhancement
3117 Description: Implement 'lfs df' to report actual free space on per-OST basis
3118 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3119 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3120 usage of objects; '-h' to report in human readable format.
3122 ------------------------------------------------------------------------------
3124 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
3131 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3132 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3133 at startup during orphan unlinking. Instead, skip buddy update
3134 before setup, as it will be regenerated after recovery is complete.
3139 Description: If an OST is inactive, its locks might reference stale inodes.
3140 Details : lov_change_cbdata() must iterate over all namespaces, even if
3141 they are inactive to clear inode references from the lock.
3143 Severity : enhancement
3144 Frequency : occasional, if non-standard max_dirty_mb used
3146 Description: Client will block write RPCs if not enough grant
3147 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3148 then the client will block writes while waiting for another RPC
3149 to complete instead of consuming its dirty limit. With change
3150 we get improved performance when max_dirty_mb is small.
3152 Severity : enhancement
3153 Bugzilla : 3389, 6253
3154 Description: Add support for supplementary groups on the MDS.
3155 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3156 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3157 lookups for user supplementary groups into a cache.
3161 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3162 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3163 for liblustre O_EXCL is enforced by MDS.
3167 Description: readdir never returns NULL in liblustre.
3168 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3169 dentry in struct dirent.
3173 Frequency : liblustre only, depends on application IO pattern
3174 Description: liblustre clients evicted if not contacting servers
3175 Details : Don't put liblustre clients into the ping_evictor list, so
3176 they will not be evicted by the pinger ever.
3178 Severity : enhancement
3180 Description: Add ability to evict clients by NID from MDS.
3181 Details : By echoing "nid:$NID" string into
3182 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3183 $NID would be instantly evicted from this MDS and from all active
3184 OSTs connected to it.
3188 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3189 Details : lookup_it_finish() used to query file size from OSTs that was not
3194 Description: service threads change working directory to that of init
3195 Details : Starting lustre service threads may pin the working directory
3196 of the parent thread, making that filesystem busy. Threads
3197 now change to the working directory of init to avoid this.
3201 Frequency : during shutdown only
3202 Description: shutdown with a failed MDS or OST can cause unmount to hang
3203 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3208 Frequency : 2.6 only
3209 Description: chmod/chown may include an extra supplementary group
3210 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3211 supplementary group and if none is specified this is used.
3214 Bugzilla : 5479 (6816)
3216 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3217 Details : If lookup is in progress on a file that is unlinked we might try
3218 to revalidate the inode and fail in revalidate after lookup is
3219 complete and ll_file_open() enqueues the open again but
3220 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3223 Frequency : always, if lconf --abort_recovery used
3225 Description: lconf --abort_recovery fails with 'Operation not supported'
3226 Details : lconf was attempting to abort recovery on the MDT device and not
3229 ------------------------------------------------------------------------------
3231 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3236 Frequency : rare (only unsupported configurations with a node running as an
3238 Bugzilla : 6514, 5137
3239 Description: Mounting a Lustre file system on a node running as an OST could
3241 Details : OSTs now preallocates memory needed to write out data at
3242 startup, instead of when needed, to avoid having to
3243 allocate memory in possibly low memory situations.
3244 Specifically, if the file system is mounted on on OST,
3245 memory pressure could force it to try to write out data,
3246 which it needed to allocate memory to do. Due to the low
3247 memory, it would be unable to do so and the node would
3248 become unresponsive.
3250 Severity : enhancement
3252 Description: Addition of lconf --service command line option
3253 Details : lconf now accepts a '--service <arg>' option, which is
3254 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3256 Severity : enhancement
3258 Description: Failover mode is now the default for OSTs.
3259 Details : By default, OSTs will now run in failover mode. To return to
3260 the old behaviour, add '--failout' to the lmc line for OSTs.
3262 Severity : enhancement
3264 Description: Health checks are now provided for MDS and OSTs
3265 Details : Additional detailed health check information on MSD and OSTs
3266 is now provided through the procfs health_check value.
3269 Frequency : occasional, depends on IO load
3271 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3272 after numerous create/delete cycles
3273 Details : The ext3 inode allocation policy would not allocate new inodes
3274 very well on the OSTs because there are no new directories
3275 being created. Instead we look for groups with free space if
3276 the parent directories are nearly full.
3281 Description: Network or server problems during mount may cause partially
3282 mounted clients instead of returning an error.
3283 Details : The config llog parsing code may overwrite the error return
3284 code during mount error handling, returning success instead
3290 Description: MDS can fail to allocate large reply buffers
3291 Details : After long uptimes the MDS can fail to allocate large reply
3292 buffers (e.g. zconf client mount config records) due to memory
3293 fragmentation or consumption by the buffer cache. Preallocate
3294 some large reply buffers so that these replies can be sent even
3295 under memory pressure.
3299 Frequency : rare (liblustre)
3300 Description: fsx running with liblustre complained that using truncate() to
3301 extend the file doesn't work. This patch corrects that issue.
3302 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3303 ATTR_SIZE and lsm use in llu_setattr_raw.
3307 Frequency : rare, only 2.6 kernels
3308 Description: Unusual file access patterns on the MDS may result in inode
3309 data being lost in very rare circumstances.
3310 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3311 code in the 2.6 kernel caused the inode and/or EA data not to
3312 be read from disk, causing single-file corruption.
3316 Frequency : rare, only 2.6 filesystems using extents
3317 Description: Heavy concurrent write and delete load may cause data corruption.
3318 Details : It was possible under high-load situations to have an extent
3319 metadata block in the block device cache from a just-unlinked
3320 file overwrite a newly-allocated data block. We now unmap any
3321 metadata buffers that alias just-allocated data blocks.
3325 Frequency : filesystems with default stripe_count larger than 77
3326 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3327 Details : lconf specifies an inode size of 4096 bytes when the default
3328 stripe_count is larger than 77. This conflicts with the default
3329 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3330 this case to avoid pinning too much memory for large EAs.
3332 ------------------------------------------------------------------------------
3334 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3339 Frequency : rare (extremely heavy IO load with hundreds of clients)
3341 Description: Client is evicted, gets IO error writing to file
3342 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3343 caused clients to be evicted for AST timeouts. The fixes in
3344 bug 5192 mean we no longer need to have such short AST timeouts
3345 so ldlm_timeout has been increased.
3348 Frequency : occasional during --force or --failover shutdown under load
3349 Bugzilla : 5949, 4834
3350 Description: Server oops/LBUG if stopped with --force or --failover under load
3351 Details : a collection of import/export refcount and cleanup ordering
3352 issues fixed for safer force cleanup
3355 Frequency : only filesystems larger than 120 OSTs
3356 Bugzilla : 5990, 6223
3357 Description: lfs getstripe would oops on a very large filesystem
3358 Details : lov_getconfig used kfree on vmalloc'd memory
3361 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3362 Bugzilla : 6242, 6243
3363 Description: reading from files that had been truncated to a non-zero size
3364 but never opened returned no data
3365 Details : ll_file_read() reads zeros from no-object files to EOF
3370 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3371 Details : There's little harm in aborting MDS/OSS recovery and letting it
3372 try again, so I removed the LASSERT and return an error instead.
3374 Severity : enhancement
3376 Description: New debugging infrastructure for tracking down data corruption
3377 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3378 (b) cover more of the client-side code path, and (c) try to narrow
3379 down where problems occurred
3383 Bugzilla : 3819, 4364, 4397, 6313
3384 Description: Racing close and eviction MDS could cause assertion in mds_close
3385 Details : It was possible to get multiple mfd references during close and
3386 client eviction, leading to one thread referencing a freed mfd.
3388 Severity: : enhancement
3389 Bugzilla : 3262, 6359
3390 Description: Attempts to reconnect to servers are now more aggressive.
3391 Details : This builds on the enhanced upcall-less recovery that was added
3392 in 1.4.2. When trying to reconnect to servers, clients will
3393 now try each server in the failover group every 10 seconds. By
3394 default, clients would previously try one server every 25 seconds.
3399 Description: After recovery, certain operations trigger a failed
3400 assertion on a client.
3401 Details : Failing over an mds, using lconf -d --failover, while a
3402 client was doing a readdir() call would cause the client to
3403 LBUG after recovery completed and the readdir() was resent.
3405 Severity : enhancement
3407 Description: Default groups are now added by lconf
3408 Details : You can now run lconf --group <servicename> without having to
3409 manually add groups with lmc.
3412 Frequency : occasional
3414 Description: Nodes with an elan id of 0 trigger a failed assertion
3417 Frequency : always when accessing e.g. tty/console device nodes
3419 Description: tty and some other devices nodes cannot be used on lustre
3420 Details : file's private_data field is used by device data and lustre
3421 values in there got lost. New field was added to struct file to
3422 store fs-specific private data.
3425 Frequency : when exporting Lustre via NFS
3427 Description: NFSD failed occasionally when looking up a path component
3428 Details : NFSD is looking up ".." which was broken in ext3 directories
3429 that had grown large enough to become hashed.
3432 Frequency : Clusters with multiple interfaces not on the same subnet
3434 Description: Nodes will repeatedly try to reconnect to an interface which it
3435 cannot reach and report an error to the log.
3436 Details : Extra peer list entries will be created by lconf with some peers
3437 unreachable. lconf now validates the peer before adding it.
3440 Frequency : Only if a default stripe is set on the filesystem root.
3442 Description: Setting a default stripe on the filesystem root prevented the
3443 filesystem from being remounted.
3444 Details : The client was sending extra request flags in the root getattr
3445 request and did not allocate a reply buffer for the dir EA.
3448 Frequency : occasional, higher if lots of files are accessed by one client
3449 Bugzilla : 6159, 6097
3450 Description: Client trips assertion regarding lsm mismatch/magic
3451 Details : While revalidating inodes the VFS looks up inodes with ifind()
3452 and in rare cases can find an inode that is being freed.
3453 The ll_test_inode() code will free the lsm during ifind()
3454 when it finds an existing inode and then the VFS later attaches
3455 this free lsm to a new inode.
3459 Bugzilla : 6422, 7030
3460 Description: MDS deadlock between mkdir and client eviction
3461 Details : Creating a new file via mkdir or mknod (starting a transaction
3462 and getting the ns lock) can deadlock with client eviction
3463 (gets ns lock and trying to finish a synchronous transaction).
3466 Frequency : occasional
3467 Description: While starting a server, the fsfilt_ext3 module could not be
3469 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3470 kernels. Previously, lconf would still use the ext3 name
3471 when trying to load modules. Now, it will correctly use
3472 ext3 on 2.4 and ldiskfs on 2.6.
3474 Severity : enhancement
3475 Description: The default stripe count has been changed to 1
3476 Details : The interpretation of the default stripe count (0, to lfs
3477 or lmc) has been changed to mean striping across a single
3478 OST, rather than all available. For general usage we have
3479 found a stripe count of 1 or 2 works best.
3481 Severity : enhancement
3482 Description: Add support for compiling against Cray portals.
3483 Details : Conditional compiling for some areas that are different
3487 Frequency : occasional
3488 Bugzilla : 6409, 6834
3489 Description: Creating files with an explicit stripe count may lead to
3490 a failed assertion on the MDS
3491 Details : If some OSTs are full or unavailable, creating files may
3492 trigger a failed assertion on the MDS. Now, Lustre will
3493 try to use other servers or return an error to the
3497 Frequency : occasional
3499 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3500 caused lock timeout during readahead (since 1.4.2).
3501 Details : Processes doing readahead might match a lock that hasn't been
3502 granted yet if there are overlapping and conflicting lock
3503 requests. The readahead process waits on ungranted lock
3504 (original lock is CBPENDING), while OST waits for that process
3505 to cancel CBPENDING read lock and eventually evicts client.
3507 Severity : enhancement
3509 Description: Initial enabling of flock support for clients
3510 Details : Implements fcntl advisory locking and file status functions.
3511 This feature is provided as an optional mount flag (default
3512 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3513 locking are implemented yet, and those that are are not guaranteed
3514 to be completely correct in production environments.
3515 mount -t lustre -o [flock|noflock] ...
3518 Frequency : occasional
3520 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3521 assertion in the ext3 JBD (journaling) layer.
3522 Details : The b_committed_data struct is protected by the big kernel lock
3523 in 2.4 kernels, serializing journal_commit_transaction() and
3524 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3525 access to this struct. In 2.6 kernels there is finer grained
3526 locking to improve SMP performance of the JBD layer.
3530 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3532 -----------------------------------------------------------------------------
3534 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3536 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3537 versions of Lustre. Please update all servers and clients to
3538 version 1.4.2 or later at the same time. You must also run
3539 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3540 to update the configuration logs.
3542 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3543 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3544 - swab llog records of type '0' so we get proper header size/idx (5861)
3545 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3546 - fix rename of one directory over another leaking an inode (5953)
3547 - avoid SetPageDirty on 2.6 (5981)
3548 - don't re-add just-being-destroyed locks to the waiting list (5653)
3549 - when creating new directories, inherit the parent's custom
3550 striping settings if present parent (3048)
3551 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3552 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3553 - don't hold client locks on temporary worklist from l_lru (5666)
3554 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3555 - automatically evict dead clients (5921)
3556 - Update file size properly in create+truncate+fstat case (6196)
3557 - Do not unhash mountpoint dentries, do not allow removal of
3559 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3560 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3561 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3562 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3563 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3564 - Correctly handle reads of files with no objects (6243)
3565 - lctl recover will also mark a device active if deactivate used (5933)
3567 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3568 - allow --write-conf on an MDS with different nettype than client (5619)
3569 - don't write config llogs to MDS for mounts not from that MDS (5617)
3570 - lconf should create multiple TCP connections from a client (5201)
3571 - init scripts are now turned off by default; run chkconfig --on
3572 lustre and chkconfig --on lustrefs to use them
3573 - upcalls are no longer needed for clients to recover to failover
3575 - add --abort-recovery option to lconf to abort recovery on device
3577 - add support for an arbitrary number of OSTs (3026)
3578 - Quota support protocol changes.
3579 - forward compatibility changes to wire structs (6007)
3580 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3581 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3582 - improved "lustre status" script
3583 - initialize blocksize for non-regular files (6062)
3584 - added --disable-server and --disable-client configure options (5782)
3585 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3586 - Vanilla 2.4.29 support
3587 - increase maximum number of obd devices to 520 (6242)
3588 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3589 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3590 SLES 9 2.6.5 kernels
3591 - limit stripes per file to 160 (the maximum EA size) (6093)
3593 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3596 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3597 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3598 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3599 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3600 - small ext3 extents cleanups and fixes (5733)
3601 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3602 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3603 - avoid needless client->OST connect, fix handle mismatch (5317)
3604 - fix DLM error path that led to out-of-sync client, long delays (5779)
3605 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3606 - fix several locking issues related to i_size (5492,5624,5654,5672)
3607 - don't move pending lock onto export if it is already evicted (5683)
3608 - fix kernel oops when creating .foo in unlinked directory (5548)
3609 - fix deadlock in obdfilter statistics vs. object create (5811)
3610 - use time_{before,after} to avoid timer jiffies wrap (5882)
3611 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3612 - Do not leak request if server was not able to process it (5154)
3613 - If mds_open unable to find parent dir, make that negative lookup(5154)
3614 - don't create new directories with extent-mapping (5909, 5936)
3616 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3617 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3618 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3619 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3620 - support for RHEL4 kernel on i686 (5773)
3621 - provide error messages when incompatible logs are encountered (5898)
3623 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3624 * version 1.4.0.10 (1.4.1 release candidate 1)
3626 - don't keep a lock reference when lock is not granted (4238)
3627 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3628 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3629 - drop import inflight refcount on signal_completed_replay error (5255)
3630 - unlock page after async write error during send (3677)
3631 - handle missing objects in filter_preprw_read properly (5265)
3632 - no transno return for symlink open, don't save no-trasno open (3440)
3633 - don't try to complete elan receive that already failed (4012)
3634 - free RPC server reply state on error (5406)
3635 - clean up thread from ptlrpc_start_thread() on error (5160)
3636 - readahead could read extra page into cache that wasn't ejected (5388)
3637 - prevent races in class_attach/setup/cleanup/detach (5260)
3638 - don't dereference de->d_inode after l_dput of de (5458)
3639 - use "int" for stripe value returned from lock_to_stripe (5544)
3640 - mballoc allocation and error-checking fixes in 2.6 (5504)
3641 - block device patches to fix I/O request sizes in 2.6 (5482)
3642 - look up hostnames for IB nals (5602)
3643 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3644 - don't start multiple acceptors for the same port (5277)
3645 - fix incorrect LASSERT in mds_getattr_name (5635)
3646 - export a proc file for general "ping" checking (5628)
3647 - fix "lfs check" to not block when the MDS is down (5628)
3649 - service request history (4965)
3650 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3651 - create an "evict_client" /proc entry on OSTs, like the MDS has
3652 - fix mount usage message, return errors per mount(8) (5168)
3653 - change grep [] to grep "[]" in tests so they work in more UMLs
3654 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3655 - remove ancient LOV_MAGIC_V0 EA support (5047)
3656 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3657 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3658 - allow readahead allocations to fail when low on memory (5383)
3659 - mmap locking landed again, after considerable improvement (2828)
3660 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3662 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3665 - send OST transaction number in read/write reply to free req (4966)
3666 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3667 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3668 - the watchdog thread now runs as interruptible (5246)
3669 - flock/lockf fixes (but it's still disabled, pending 5135)
3670 - don't use EXT3 constants in llite code (5094)
3671 - memory shortage at startup could cause assertion (5176)
3673 - reorganization of lov code
3674 - single portals codebase
3676 - add extents/mballoc support (5025)
3677 - direct I/O reads in the obdfilter (4048)
3678 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3680 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3683 - send OST transaction number in read/write reply to free req (4966)
3684 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3685 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3686 - the watchdog thread now runs as interruptible (5246)
3687 - handle missing objects in filter_preprw_read properly (5265)
3688 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3689 - drop import inflight refcount on signal_completed_replay error (5255)
3690 - unlock page after async write error during send (3677)
3691 - return original error code on reconstructed replies (3761)
3692 - no transno return for symlink open, don't save no-trasno open (3440)
3694 - add pid to ldlm debugging output (4922)
3695 - bump the watchdog timeouts -- we can't handle 30sec yet
3696 - extra debugging for orphan dentry/inode bug (5259)
3698 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3701 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3702 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3703 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3704 - fix NULL dereference in /proc/sys/portals/routes (4827)
3705 - allow failed mdc_close() operations to be interrupted (4561)
3706 - stop precreate on OST before MDS would time out on it (4778)
3707 - don't send partial-page writes before EOF from client (4410)
3708 - discard client grant for sub-page writes on large-page clients (4520)
3709 - don't free dentries not owned by NFS code, check generation (4806)
3710 - fix lsm leak if mds_create_objects() fails (4801)
3711 - limit debug_daemon file size, always print CERROR messages (4789)
3712 - use transno after validating reply (3892)
3713 - process timed out requests if import state changes (3754)
3714 - update mtime on OST during writes, return in glimpse (4829)
3715 - add mkfsoptions to LDAP (4679)
3716 - use ->max_readahead method instead of zapping global ra (5039)
3717 - don't interrupt __l_wait_event() during strace
3719 - add software watchdogs to catch hung threads quickly (4941)
3720 - make lustrefs init script start after nfs is mounted
3721 - fix CWARN/ERROR duplication (4930)
3722 - return async write errors to application if possible (2248)
3723 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3724 - print NAL number in %x format (4645)
3725 - update barely-supported suse-2.4.21-171 series (4842)
3726 - support for sles 9 %post scripts
3727 - support for building 2.6 kernel-source packages
3728 - support for sles km_* packages
3730 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3733 - ignore -ENOENT errors in osc_destroy (3639)
3734 - notify osc create thread that OSC is being cleaned up (4600)
3735 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3736 - reconstruct ost_handle() like mds_handle() (4657)
3737 - create a new thread to do import eviction to avoid deadlock (3969)
3738 - let lconf resolve symlinked-to devices (4629)
3739 - don't unlink "objects" from directory with default EA (4554)
3740 - hold socknal file ref over connect in case target is down (4394)
3741 - allow more than 32000 subdirectories in a single directory (3244)
3742 - fix blocks count for O_DIRECT writes (3751)
3743 - OST returns ENOSPC from object create when no space left (4539)
3744 - don't send truncate RPC if file size isn't changing (4410)
3745 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3746 - bind to privileged port in socknal and tcpnal (3689)
3748 - rate limit CERROR/CWARN console message to avoid overload (4519)
3749 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3750 - basic mmap support (3918)
3751 - kernel patch series update from b1_4 (4711)
3753 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3756 - avoid crash during MDS cleanup with OST shut down (2775)
3757 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3758 - don't use bad inodes on the MDS (3744)
3759 - dynamic object preallocation to improve recovery speed (4236)
3760 - don't hold spinlock over lock dumping or change debug flags (4401)
3761 - don't zero obd_dev when it is force cleaned (3651)
3762 - print grants to console if they go negative (4431)
3763 - "lctl deactivate" will stop automatic recovery attempts (3406)
3764 - look for existing locks in ldlm_handle_enqueue() (3764)
3765 - don't resolve lock handle twice in recovery avoiding race (4401)
3766 - revalidate should check working dir is a directory (4134)
3768 - don't always mark "slow" obdfilter messages as errors (4418)
3770 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3773 - don't close LustreDB during write_conf until it is done (3860)
3774 - fix typo in lconf for_each_profile (3821)
3775 - allow dumping logs from multiple threads at one time (3820)
3776 - don't allow multiple threads in OSC recovery (3812)
3777 - fix debug_size parameters (3864)
3778 - fix mds_postrecov to initialize import for llog ctxt (3121)
3779 - replace config semaphore with spinlock (3306)
3780 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3781 - don't allow enqueue to complete on a destroyed export (3822)
3782 - down write_lock before checking llog header bitmap (3825)
3783 - recover from lock replay timeout (3764)
3784 - up llog sem before sending rpc (3652)
3785 - reduce ns lock hold times when setting kms (3267)
3786 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3787 - fix NULL deref and obd_dev leak on setup error (3312)
3788 - replace some LBUG about llog ops with error handling (3841)
3789 - don't match INVALID dentries from d_lookup and spin (3784)
3790 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3791 - fix invalid assertion in ptlrpc_set_wait (3880)
3793 - add libwrap support for the TCP acceptor (3996)
3794 - add /proc/sys/portals/routes for non-root route listing (3994)
3795 - allow setting MDS UUID in .xml (2580)
3796 - print the stack of a process that LBUGs (4228)
3798 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3801 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3802 - make sure to unset replay flag from failed open requests (3440)
3803 - if default stripe count is 0, use OST count for inode size (3636)
3804 - update parent mtime/ctime on client for create/unlink (2611)
3805 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3806 - free recovery state on server during a forced cleanup (3571)
3807 - unregister_reply for resent reqs (3063)
3808 - loop back devices mounting and status check on 2.6 (3563)
3809 - fix resource-creation race that can provoke i_size == 0 (3513)
3810 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3811 - more debugging for page-accounting assertion (3746)
3812 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3813 - avoid deadlock after precreation failure (3758)
3814 - fix race and lock order deadlock in orphan handling (3450, 3750)
3815 - add validity checks when grabbing inodes from l_ast_data (3599)
3817 - add /proc/.../recovery_status to obdfilter (3428)
3818 - lightweight CDEBUG infrastructure, debug daemon (3668)
3819 - change default OSC RPC parameters to be better on small clusters
3820 - turn off OST read cache for files smaller than 32MB
3821 - install man pages and include them in rpms (3100)
3822 - add new init script for (un)mounting lustre filesystems (2593)
3823 - run chkconfig in %post for init scripts (3701)
3824 - drop scimac NAL (unmaintained)
3826 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3829 - clean kiobufs before and after use (3485)
3830 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3831 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3832 - change init script to fail more clearly if not run as root (1528)
3833 - allow clients to reconnect during replay (1742)
3834 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3835 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3836 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3837 - fix cond_resched (3554)
3838 - only evict unfinished clients after recovery (3515)
3839 - allow bulk resend, prevent data loss (3570)
3840 - dynamic ptlrpc request buffer allocation (2102)
3841 - don't allow unlinking open directory if it isn't empty (2904)
3842 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3843 - remove extraneous obd dereference causing LASSERT failure (3334)
3844 - don't use get_cycles() when creating temp. files on the mds (3156)
3845 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3846 - handle EEXIST for set-stripe, set proper directory name (3336)
3848 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3849 - fix ksocknal_fmb_callback() error messages (2918)
3851 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3854 - don't copy lvb into (possibly NULL) reply on error (2983)
3855 - don't deref dentry after dput, don't free lvb on error (2922)
3856 - use the kms to determine writeback rpc length (2947)
3857 - increment oti_logcookies when osc is inactive (2948)
3858 - update client's i_blocks count via lvb messages (2543)
3859 - handle intent open/close of special files properly (1557)
3860 - mount MDS with errors=remount-ro, like obdfilter (2009)
3861 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3862 - don't use cancelling-locks' kms values (2947)
3863 - use highest lock extent for kms, not last one (2925)
3864 - don't dereference ERR_PTR() dentry in error handling path (3107)
3865 - fix thread race in portals_debug_dumplog() (3122)
3866 - create lprocfs device entries at setup instead of at attach (1519)
3867 - common AST error handler, don't evict client on completion race (3145)
3868 - zero nameidata in detach_mnt in 2.6 (3118)
3869 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3870 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3871 - initialize RPC timeout timer earlier for 2.6 (3219)
3872 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3873 - print nal/nid for unknown nid (3258)
3874 - additional checks for oscc recovery before doing precreate (3284)
3875 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3876 - don't crash in mdc_close for bad permissions on open (3285)
3877 - zero i_rdev for non-device files (3147)
3878 - clear page->private before handing to FS, better assertion (3119)
3879 - tune the read pipeline (3236)
3880 - fix incorrect decref of invalidated dentry (2350)
3881 - provide read-ahead stats and refine rpc in flight stats (3328)
3882 - don't hold journal transaction open across create RPC (3313)
3883 - update atime on MDS at close time (3265)
3884 - close LDAP connection when recovering to avoid server load (3315)
3885 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3886 - don't leak open file on MDS after open resend (3325)
3887 - serialize filter_precreate and filter_destroy_precreated (3329)
3888 - loop device shouldn't call sync_dev() for nul device (3092)
3889 - clear page cache after eviction (2766)
3890 - resynchronize MDS->OST in background (2824)
3891 - refuse to mount the same filesystem twice on same mountpoint (3394)
3892 - allow llmount to create routes for mounting behind routers (3320)
3893 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3894 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3895 - fix rare problem with rename on htree directories (3417)
3897 - allow default OST striping configuration per directory (1414)
3898 - fix compilation for qswnal for 2.6 kernels (3125)
3899 - increase maximum number of MDS request buffers for large systems
3900 - change liblustreapi to be useful for external progs like lfsck (3098)
3901 - increase local configuration timeout for slow disks (3353)
3902 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3904 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3907 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3908 - don't overwrite extent policy data in reply if lock was blocked (2901)
3909 - drop filter export grants atomically with removal from device (2663)
3910 - del obd_self_export from work_list in class_disconnect_exports (2908)
3911 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3912 - swab reply message in mdc_close, other PPC fixes (2464)
3913 - fix destroying of named logs (2325)
3914 - overwrite old logs when running lconf --write_conf (2264)
3915 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3916 - fix race in target_handle_connect (2898)
3917 - mds_reint_create() should take same inode create lock (2926)
3918 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3919 - don't close files for self_export to avoid uninitialized obd (2936)
3920 - allow MDS with the same name as client node (2939)
3921 - hold dentry reference for closed log files for unlink (2325)
3922 - reserve space for all logs during transactions (2059)
3923 - don't evict page beyond end of stripe extent (2925)
3924 - don't oops on a deleted current working directory (2399)
3925 - handle hard links to targets without a parent properly (2517)
3926 - don't dereference NULL lock when racing during eviction (2867)
3927 - don't grow lock extents when lots of conflicting locks (2919)
3929 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3932 - account for cache space usage on clients to avoid data loss (974)
3933 - lfsck support in lustre kernel code (2349)
3934 - reduce journal credits needed for BRW writes (2370)
3935 - orphan handling to avoid losing space on client/server crashes
3936 - ptlrpcd can be blocked, stopping ALL progress (2477)
3937 - use lock value blocks to assist in proper KMS, faster stat (1021)
3938 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3939 - recovery for initial connections (2355)
3940 - fixes for mds_cleanup_orphans (1934)
3941 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3942 - block all file creations until orphan recovery completes (1901)
3943 - client remove rq_connection from request struct (2423)
3944 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3945 - recovery timer race (2670)
3946 - mdc_close recovey bug (2532)
3947 - ptlrpc cleanup bug (2710)
3948 - mds timeout on local locks (2588)
3949 - namespace lock held during RPCs (2431)
3950 - handle interrupted sync write properly (2503)
3951 - don't try to handle a message that hasn't been replied to (2699)
3952 - client assert failure during cleanup after abort recovery (2701)
3953 - leak mdc device after failed mount (2712)
3954 - ptlrpc_check_set allows timedout requests to complete (2714)
3955 - wait for inflight reqs when ptlrpcd finishes (2710)
3956 - make sure unregistered services are removed from the srv_list
3957 - reset bulk XID's when resending them (caught by 1138 test)
3958 - unregister_bulk after timeout
3959 - fix lconf error (2694)
3960 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3961 - readahead locks pages, leaves pending causing memory pressure (2673)
3962 - increase OST request buffers to 4096 on large machines (2729)
3963 - fix up permission of existing directories in simple_mkdir (2661)
3964 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3965 - don't assign transno to errored transactions (2742)
3966 - don't delete objects on OST if given a bogus objid from MDS (2751)
3967 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3968 - if rq_no_resend, then timeout request after recovery (2432)
3969 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3970 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3971 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3972 - MDS recovery shouldn't delete objects that it has given out (2730)
3973 - if enqueue arrives after completion, don't clobber LVB (2819)
3974 - don't unlock pages twice when trigger_group_io returns error (2814)
3975 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3976 - don't write pages to disk if there was an error (1450)
3977 - don't ping imports that have recovery disabled (2676)
3978 - take buffered bytes into account when balancing socknal conn (2817)
3979 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3980 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3981 - remove little-endian swabbing of llog records (1987)
3982 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3983 - retry reposting request buffers if they fail (1191)
3984 - grow extent at grant time to avoid granting a revoked lock (2809)
3985 - lock revoke doesn't evict page if covered by a second lock (2765)
3986 - disable VM readahead to avoid reading outside lock extents (2805)
3988 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3989 - updated kernel patches for hp-2.4.20 kernel (2681)
3991 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3994 - fix truncated write corruption (2366)
3995 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3997 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3998 - ensure len(uuid) < 37 in lmc (1171)
3999 - fix ia64 OOPS in llog_test (2255)
4000 - zero end of page at obdfilter for partial page writes (2648)
4001 - don't leave stale dentries around after renames (bug 2428)
4002 - fix timeouts when evicting a client with a single lock held (2642)
4003 - set deadline for the initial HELLO message to drain (2634)
4004 - print out dotted-quad IP addresses in the socknal (2302)
4006 - additional debugging for MDS client eviction problem (2443)
4007 - fix mkfsoptions support for osts (2603, 2604)
4009 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4012 - add series for the vanilla 2.6.0 kernel
4013 - add series for the vanilla 2.4.24 kernel
4014 - add series for a cray x86/64 UL kernel drop
4015 - fix xattr patches for the vanilla 2.4.19 series
4017 - generate true UUIDs in lmc (1171)
4018 - have portals stack dumping break in UML (2466)
4019 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4020 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4021 - wake callback waiting threads on client eviction (2460)
4022 - Add --ptldebug and --subsystem to lmc (1719)
4023 - update assertion to allow safe interrupt allocation
4024 - set rq_no_resend for cancel requests (2432)
4025 - recalculate ptlrpcd timeout after resend (2494)
4026 - call vfs_rmdir when removing pending directories (2368)
4027 - fix renaming a file to itself (2429)
4028 - lmc creates a default one-stripe lov (2454)
4029 - expand procfs space to handle large clusters (2326)
4030 - increase UML stack to avoid overflow
4031 - update lconf's list of debug and subsystem masks
4032 - fix lfs find --obd (2510)
4033 - /proc tunable for disabling filter read caching (2591)
4034 - stop rpm packages from altering slapd.conf (2301)
4035 - disable nagle in the socknal under 0conf (2578)
4036 - choose mds inode size based on stripe count (2572)
4037 - fix kernel-source rpm problems (2516)
4039 - add --disable-doc to avoid pdf generation (2421)
4040 - update documentation, tests, type-os, comments
4041 - avoid format warnings on ia64
4042 - remove the TOE NAL
4043 - tiny code cleanups by removing unused fields
4045 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4048 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4049 - remove the most blatant lies from BUILDING file (2371)
4050 - change default debug level to reasonable production setting
4051 - reduce client side cache size to reduce cache flush time
4052 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4053 - make TCP zerocopy and pinger support enabled by default (2476)
4054 - sync writes completed after process exits caused crashes (2319)
4055 - maintain correct mount count on the MDS (2356)
4056 - backout 1557, because 2316 wasn't really fixed
4057 - better file I/O statistics gathering in /proc
4058 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4059 - another kernel patch to fix zero-copy TCP function export
4060 - don't take duplicate lock when processing re-sent getattr (2420)
4061 - lctl uses obd_self_export instead of creating new conn (2353)
4062 - MDS/OST recovery case which requires object creation asserted (2425)
4063 - move lfs from /usr/sbin to /usr/bin in packages
4064 - fix race between mds_client_add and mds_client_free (2417)
4065 - use kmalloc instead of slabs in portals (2430)
4066 - don't create duplicate records when a failover MDS is present (2442)
4067 - remove unnecessary mount age check (2332)
4068 - don't remove directory inodes from locks prematurely (2451)
4069 - don't break if MDS service name is the same as hostname (2103)
4070 - fix races in client write RPC generation when cache full (2482)
4072 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4075 - remove now-unused request->rq_obd (278)
4076 - if an allocation fails, print out how much memory we've used (1933)
4077 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4078 - add the "configurable stack size" patch to most series files (1256)
4079 - ability to write large log records, for 100+ OST configs (2306)
4080 - fix NULL deref when filter_prep fails (2314)
4081 - fix operator precedence error in filter_sync
4082 - dynamic allocation of socknal TX descriptors (2315)
4083 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4084 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4085 - ensure that CWARN messages in Portals always get to the syslog
4086 - __init/__exit are not for prototype decls (ldlm_init/exit)
4087 - x86-64 compile warning fixes
4088 - fix gateway LMC keyword conflict (2318)
4089 - fix MDS lock inversions in getattr/reint paths (1844)
4090 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4091 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4092 - DLM race condition prevented some lock evictions (2328)
4093 - ENOMEM detection and retry on socknal sends (2230)
4094 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4095 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4096 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4097 - better error messages when a client is rejected during recovery (1505)
4098 - avoid cancelling locks which were never granted, after failure (2330)
4099 - fix i_sem/journal inversion in mds_client_add (2333)
4100 - fix truncate/getattr lock cycle deadlock (2334)
4101 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4102 - fix two rare exit paths which could leak an l_lock() ref (2321)
4103 - fencepost error in MDS/OST orphan recovery (2226)
4104 - make log record alignment 8 bytes (1988)
4105 - lstripe now fails when requested offset > ost_count (2237)
4106 - ensure that all kernel series have a complete list.h (1607)
4107 - fix crashes in special-file operations (2316)
4108 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4110 - allow configurable automake binary, for testing new versions
4111 - small update to the lfs documentation
4113 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4115 * fix negative export reference count in fsfilt_sync (2312)
4117 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4118 * release candidate 0.9.1
4120 - orphans are moved into the PENDING directory for possible recovery
4121 - replayed opens now open by fid for orphan/rename safety (1042)
4122 - last close of an orphan inode generates a transno (683)
4123 - chdir() and mount() now pin the directory entry (1020)
4124 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4125 - discard very old requests without processing them (1502)
4126 - remove some common, well-understood CERRORs (1505)
4127 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4128 - clear "grant" flags in OST replies until OST grant code lands (1644)
4129 - fix read performance by not clobbering i_blksize on client (1598)
4130 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4131 - make LRU size a /proc tunable, clears locks when reduced (707)
4132 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4133 - server locks take references on exports now (1558)
4134 - build fixes for 2.4.20-rh trees (1663)
4135 - return an error from lov_create if all OSCs are inactive (1751)
4136 - fix import levels when a reconnect happens without a timeout (1597)
4137 - exit early from mds_open if we get a lookup error (1749)
4138 - partial page read at EOF wouldn't wait for disk before sending (1642)
4139 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4140 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4141 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4142 - write/truncate lock inversion (1639)
4143 - Don't auto-load obdclass, portals modules during cleanup (1495)
4144 - fix timestamps from jumping to "now" (1763)
4145 - extra journal assertions (1648)
4146 - add an extra multiunlink test (1771)
4147 - fix read_record/write_record API (1776)
4148 - fix leak of offset_extent, possible incorrect i_size later (1772)
4149 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4150 - Debugging for the kqswnal_get_idle_tx problems (1820)
4151 - Allow recovery to be attempted multiple times (1536)
4152 - Write out MDS last_rcvd file after it is first created (1600)
4153 - Fix tx_descriptor leak in failed transmit situations (1827)
4154 - ext3 journaling fixes for assertion failure after IO error (1871)
4155 - class_export_put() on freed export after completion AST error (1896)
4156 - Fix revalidate looping in VFS (1322)
4157 - Don't access a freed export during MDS_REINT timeout (1521)
4158 - Add open-unlink recovery support on the MDS (1673,1764)
4159 - Return an error if no MDS data was read from last_rcvd (1946)
4160 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4161 - Don't setup a disk device that doesn't match exported UUID (317)
4162 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4163 - avoid committing NULL handle in force close
4164 - local.sh is now a one-stripe LOV configuration
4165 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4166 - fix ext3 htree duplicate directory entry corruption (1516)
4167 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4168 - update to new LOV EA format (2097)
4169 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4171 2003-06-15 Phil Schwan <phil@clusterfs.com>
4174 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4175 - per-import/export recovery handling (958, 931, 959)
4176 - multiple last-rcvd slots, for serving multiple FSes (949)
4177 - connections are again shared between multiple imp/exports (963, 964)
4178 - "umount -f" would hang if any requests needed to be sent (393, 978)
4179 - avoid pinning large req buffer by copying for queued messages (989)
4180 - add "uuid" to "lctl device" command to help upcalls (991)
4181 - "open" RPCs with transnos would confuse recovery counters (1037)
4182 - do proper endian conversion of all wire messages (288, 340, 891)
4183 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4184 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4185 - fix ptlrpc_connection leak in target_handle_connect (1174)
4186 - fix import refcounting bug in OST and MDS cleanup (1134)
4187 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4188 - fix very unlikely obd_types race condition (501)
4189 - remove osc_open hack for echo_client (1187)
4190 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4191 - a failure in read_inode2 leads to deadlock (1139)
4192 - cancel ack-locks as soon as transaction is committed (1072)
4193 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4194 - make sure to commitrw after any preprw to avoid deadlock (1162)
4195 - failing to execute a file in a lustre FS would lock inode (1203)
4196 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4197 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4198 - obd_uuid2tgt would walk off the end of the list (1255)
4199 - on IA64 the osc would give portals incorrect bulk size (1258)
4200 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4201 - fix lock inversion caused by new llite matching code (1282)
4202 - limit the number of dirty pages on a client to 10MB (1286)
4203 - timed out locks were not being corrected cancelled (1289)
4204 - fix O_DIRECT above 4GB on IA-32 (1292)
4205 * major user-visible changes
4206 - fail out/fail over policy now controlled by the upcall (993)
4208 - add OBD_PING to check server availability and failure (954)
4209 - lustre messages are now sent in sending host order (288, 340, 891)
4210 - add eadatalen to MDS getattr reply (340)
4211 - OST read replies may contain second buffer, with per-page status (593)
4213 2003-03-11 Phil Schwan <phil@clusterfs.com>
4216 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4217 - failed open()s could cause deadlock; fixed (867, 869)
4218 - stop cancelling OST locks when files are closed (481)
4219 - overlapping XID spaces caused network corruption (851, 853)
4220 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4221 - setattr_raw added, to do single-RPC, server-side setattrs
4222 - lmc/lconf syntax change for OST UUIDs
4223 - fix crashy race condition between ptlrpc_free_req and osc_close
4224 - don't use request in mdc_enqueue if we hit a timeout (889)
4225 - don't set the inode i_size for regular files from the MDS (896)
4226 - handle out of order completion AST (842)
4227 - don't LBUG if a lock request times out after receiving AST (913)
4228 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4229 - if a bad lock AST arrives, send an error instead of dropping entirely
4230 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4231 - fix leak in bulk IO when only partially completed (899, 900, 926)
4232 - fix O_DIRECT for ia64 (55)
4233 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4234 - C-z after timeout could hang a process forever; fixed (977)
4236 - client-side I/O cache (678, 924, 929, 941, 970)
4238 - READPAGE and SETATTRs which don't take server-side locks get
4241 2003-02-11 Phil Schwan <phil@clusterfs.com>
4244 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4245 - Fix fsfilt_extN_readpage() to read a full page of directory
4246 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4247 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4248 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4249 - Most intent operations were removed, in favour of a new RPC mode
4250 that does a single RPC to the server and bypasses most of the VFS
4251 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4252 - Aggressively cancel local locks on DLM servers
4253 - mds_reint_unlink sends EA to the client if it's the last nlink.
4254 client uses that EA to unlink OST objects.
4255 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4256 - recursive symlinks were fixed (439)
4257 - fixed NULL deref in DEBUG_REQ
4258 - filter_update_lastobjid no longer calls sync, which annoyed extN
4259 - fixed multi-client small-writes to a single file problem (445)
4260 - fixed mtime updates during file writes (607)
4261 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4262 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4263 - fixed Linux OST size reporting problem (444, 656)
4264 - OST now updates object mtime with writes or setattr (607, 619)
4265 - client verifies file size before zeroing page past EOF (445)
4266 - OST now writes last allocated objid to disk with allocation (108)
4267 - LOV on echo now works (409)
4269 - mds_reint_unlink sends a new buffer, with the EA included. this
4270 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4271 set if a regular file was being unlinked, and it was the last link
4272 - use PtlGet from the target for bulk writes (315)
4273 - OST now updates object mtime with writes or setattr (607, 619)
4274 - LDLM now has a grant-time callback to revalidate locked items, if
4276 - Many MDS operations were reorganized to combat race conditions
4278 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4279 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4281 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4284 - Fully reactivate OST imports after reconnection (512, others)
4285 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4286 - More graceful error handling for truncating on dead OST (515)
4287 - Don't error out unless we're actually accessing dead stripes (474)
4288 - Fix garbage sizes when stripes are missing (410)
4289 - LRU counters were broken, causing constant lock purge (433, 432)
4290 - garbage on read from stripes with failed OSTs (441)
4291 - mark OSCs as active before reconnecting during recovery (438)
4292 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4293 - lfind did not preserve OST order in output (443)
4294 - symlinks cause hung clients, incorrect data (439)
4295 - stop dereferencing request after dropping refcount (457)
4296 - don't LASSERT(spin_is_locked) on non-SMP (455)
4297 - fixes for many rename() bugs
4298 - fstat didn't correctly synchronize attributes (399)
4299 - server must handle lock cancellation during blocking AST prep (487)
4300 - bulk descriptors were free()d too soon (511)
4301 - fix paths in lconf, which would load incorrect modules (451, 507)
4302 - fix confusing lconf 'host not found' error message (386)
4303 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4304 - fix race condition in mdc_blocking_ast() for inode access (526)
4305 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4306 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4307 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4308 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4309 - fix lstripe in several ways (kernel side) (527)
4310 - fix request leak in ldlm_cli_enqueue (262)
4311 - incorrect OSC was marked inactive after OST failure
4312 - call mds_fs_cleanup before unmounting filesystem (524)
4313 - fix races between taking ns_lock and ldlm_lock_change_resource
4314 - fix races updating LOV export open file list
4315 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4316 - fix recovery NULL deref in ldlm_cli_cancel_unused
4317 - fix some DLM races by using new hash table for lock handles (419)
4318 - permit the client to specify desired inodes, at replay
4319 - duplicate requests when we queue them for replay reintegration
4320 - fix last_rcvd offset calculation
4321 - sync after each recovered transaction, so we always make progress
4322 - never, not always, ERESTART requests without transnos
4323 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4325 - skip replay if the MDS says that the client is already connected
4326 - don't check for a recovery-enabled export to match lctl's UUID
4327 - don't INC_USE_COUNT for phantom exports
4328 - don't crash when cleaning up phantom exports (567)
4329 - don't double-finish or set replay data for errored mdc_open requests
4330 - abort requests when they time out, so we don't get old replies
4331 - send/receive replies for AST messages again
4332 - if the client says that it doesn't have the lock, cancel it on the
4334 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4335 mark it as destroyed, it will all work out when decref is called
4336 - fix module use counts (22, 581)
4338 - ASTs now expect a reply (server cancels lock on error reply)
4340 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4343 - fix many simultaneous client startup (392)
4344 - fix dentry->d_it clobbering
4345 - credentials weren't being shipped for readdir/getattr operations
4346 - remove invalid assertions triggered during some concurrent MD
4348 - proper Lustre versions added (336, 389)
4349 - fix memory leak for create error case (398)
4350 - fix LOV locking bug that would get cli/srv out of sync
4351 - fix echo client over LOV (409)
4352 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4353 - fix double-O_EXCL intent crash (424)
4354 - avoid sending multiple lock CANCELs (352)
4356 - MDS can do multi-client recovery (modulo bugs in new code)
4358 - many updates, edits, cleanups
4360 2002-11-18 Phil Schwan <phil@clusterfs.com>
4363 - fix null d_it dereference (346)
4364 - fix full OST/dbench hang (333)
4365 - fix permission problem with file removal (286)
4366 - fix removal of OSCs from LOV when they fail
4367 - fix NULL deref during bulk timeout (214)
4368 - fix problems related to multiple filesystems on one MDS (241)
4369 - fixed serious subtle metadata locking bugs
4370 - free locks on clients when inodes are removed due to memory
4372 - fix inode pointer in lock data (285)
4373 - partial support for multiple MDS on a single host (241)
4374 - data locks weren't cancelled at clear_inode time (290, 311)
4375 - intent locks could lead to unbounded lock growth (205)
4376 - added a maximum lock count, an LRU list, and a flusher
4377 - fix multiple rename (365)
4378 - properly abstracted the echo client
4379 - OSC locked 1 byte too many; fixed
4380 - rewrote brw callback code:
4381 - fixed recovery bugs related to LOVs (306)
4382 - fixed too-many-pages-in-one-write crash (191)
4383 - fixed (again) crash in sync_io_timeout (214)
4384 - probably fixed callback-related race (385)
4386 - Add capability to MDS protocol
4387 - LDLM cancellations and callbacks on different portals
4389 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4392 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4393 - defer open object destruction to close time (601981, 138)
4394 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4395 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4396 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4397 - don't call obd_setattr in ll_file_release for destroyed objects
4398 * protocol change to lustre_msg: move |version| and add |flags|
4399 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4400 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4401 * added replay of create, unlink, link and rename operations during
4402 MDS failover; recovery should be much more robust now
4403 * remove failed OSCs from LOVs (only lov_create uses this so far)
4404 * the lustre-HOWTO was brought (more) up to date (582544)
4406 2002-10-23 Phil Schwan <phil@clusterfs.com>
4409 - in-use dentries weren't being reused properly (617851)
4410 - prevent multiple LDLM setup (599178)
4411 - fix LOV size calculations for truncate (617853)
4412 - fix client handling of MDS intent errors (POSIX)
4413 - fix permission bug in lovstripe.c test (624321)
4414 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4415 - truncate past end of file could corrupt data
4416 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4417 - a race in recovery could return ETIMEDOUT to apps (623947)
4418 - building outside the source directory was fixed
4419 * the lustre-HOWTO was brought (more) up to date (582544)
4420 * major progress was made on recovery functionality
4422 2002-10-10 Phil Schwan <phil@clusterfs.com>
4425 - recovery deadlock fix
4426 - rm -rf causes LBUG fix (617817)
4427 - file open by multiple tasks fix (618962)
4428 - directory permissions bugs (602707 and 620007)
4429 - journal_stop fixed with locking (611313)
4430 - O_APPEND failures resolved (618273, perhaps 614459)
4431 - lconf PATH fix (619770)
4432 - IA64 build fix (621450)
4433 - RPC buffer sizes scale with amount of memory
4435 2002-10-01 Phil Schwan <phil@clusterfs.com>
4438 - locks would be cancelled without throwing away data pages,
4439 resulting in inconsistent data (605627)
4440 - inode attributes were not always being refreshed (605627, 612449)
4441 - lconf now continues to cleanup after lctl reports an error
4442 - MDS now enforces user permissions (602707)
4443 - lprocfs cleanup fixed, but not yet enabled (614157)
4444 - fixed infinite server hang, should a client not respond to an AST
4445 - avoid going into recovery if user calls readlink() with a buffer
4446 that's too small (613941)
4447 - AST RPCs no longer require replies (614867) -- this may be changed
4448 - don't crash server if client sends an IOV that's too big (611336)
4449 - fixed lock conversion deadlock (611892)
4450 - fixed the following of symlinks (614622)
4451 * recovery: the server can remove locks from a client that dies, other
4452 clients can make progress
4453 * more extN patch fixes
4454 * compile-time configurable ptlrpc buffer allocations
4456 - collaborative read cache document
4457 - Lustre Lite Performance CDR document-in-progress
4459 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4462 - fix typo in patch-2.4.18
4464 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4467 - clear ptlrpc request each time in handle_incoming_request()
4468 - unlink of files now destroys the object on the OST
4470 2002-09-19 Peter Braam <braam@clusterfs.com>
4472 * add hard link support
4473 * change obdfile creation method
4474 * kernel patch changed
4476 2002-09-19 Peter Braam <braam@clusterfs.com>
4479 - stack overflow bug in extN fixed
4481 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4483 * documentation updates
4484 - add man pages for config tools
4485 - update tests/README to describe testing with new config tools
4486 - finish metadata API descriptions
4487 * bug fixes and cleanups
4488 - statfs workaround for 16TB limit
4489 - LOV stripe allocation improved, can stripe on subset of OSTs
4490 - LOV file size/IO offset was wrong for files > 4GB in size
4491 - object EA data was being dropped, caused files to be unreadable
4492 - memory overflow with non-LOV OST caused memory corruption
4493 - fixed regression tests to work with new config tools, obdfilter
4494 - fixed bug when directory size became larger than 1 block
4495 - fixed bug (for single client case) when PWD was deleted
4496 - invalidate local directory pages when doing intent-based ops
4497 - avoid LDLM oops when lock callback contained bad data
4499 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4501 * documentation updates
4502 * bug fixes and cleanups
4503 - configuration tools
4506 - 64-bit compile warnings
4507 - 64-bit internal statfs data
4509 * test_brw on persistent OST devices
4511 * lprocfs (disabled)
4513 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4515 * documentation updates
4516 * bug fixes and cleanups
4517 * configuration tools
4519 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4522 * many small fixes to 0_5_4
4523 * io/network handling
4524 * thinkos in MDS operations
4526 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4529 * crucial basic fixes to 0.5.3
4530 * IOR, Iozone work over Elan
4533 2002-08-07 Phil Schwan <phil@clusterfs.com>
4534 * version 0_5_3, our first alpha
4535 * we use the new Portals iovs
4536 * documentation updates
4537 * bug fixes and cleanups
4538 * small changes in the DLM wire protocol
4540 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4541 * version 0_5_1 with some initial stability,
4542 * locking on MD and file I/O.
4543 * documentation updates
4544 * several bug fixes since 0.5.0
4545 * small changes in wire protocol
4547 2002-07-18 Phil Schwan <phil@clusterfs.com>
4549 * delivered as Lustre Light Alpha
4550 * fixed a crash after handling invalid MDS requests
4551 * fixed directory pages for architectures with non-4k pages sizes
4553 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4554 * release version v0_4_4
4555 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4557 * Fixes a number of interruption problems with OST operations.
4558 * Update documentation for portals header changes
4559 * Move all wire protocol structs/defines to lustre_idl.h
4560 * Fixes symlink length bug.
4561 * Add tcpdump to repository.
4563 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4564 * release version v0_4_3
4565 * Fixes statfs for inodes on extN.
4566 * Fixes bug in runtests which would delete /etc/hosts.
4567 * Use 64-bit object IDs wherever possible (not into VFS though)
4568 Remove ost_get_info, which is unused by lustre, and out of date.
4570 2002-07-03 Peter Braam <braam@clusterfs.com>
4571 * release version v0_4_2 Fixes a lookup error (type not passed)
4572 * move forward to head of Portals
4573 * move forward to latest Lustre kernel
4575 2002-06-25 Peter Braam <braam@clusterfs.com>
4576 * release version v0_4_1. Hopefully stable on single node use.