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 2.6.18-8.1.10.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.
96 --------------------------------------------------------------------------------
98 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
100 * Support for kernels:
101 2.6.5-7.286 (SLES 9),
102 2.6.9-55.0.2.EL (RHEL 4),
103 2.6.16.46-0.14 (SLES 10),
104 2.6.18-8.1.8.el5 (RHEL 5),
105 2.6.18.8 vanilla (kernel.org)
106 * Client support for unpatched kernels:
107 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
108 2.6.16 - 2.6.22 vanilla (kernel.org)
109 * Due to recently discovered recovery problems, we do not recommend
110 using patchless RHEL 4 clients with this or any earlier release.
111 * Recommended e2fsprogs version: 1.40.2-cfs1
112 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
116 Description: Fix errors in lfs documentation
117 Details : Fixes man pages
119 Severity : enhancement
121 Description: Adaptive timeouts
122 Details : RPC timeouts adapt to changing server load and network
123 conditions to reduce resend attempts and improve recovery time.
125 Severity : enhancement
127 Description: llapi_file_create() does not allow some changes
128 Details : add llapi_file_open() that allows specifying the file creation
129 mode and open flags, and also returns an open file handle.
131 Severity : enhancement
133 Description: df doesn't work properly if diskfs blocksize != 4K
134 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
137 Frequency : always on directories with default striping set
139 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
140 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
142 Severity : enhancement
144 Description: merge and cleanup kernel patches.
145 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
147 Severity : enhancement
149 Description: RedHat Update kernel for RHEL5
150 Details : Kernel config file for RHEL5.
152 Severity : enhancement
154 Description: OSS needs mutliple precreate threads
155 Details : Add ability to start more than one create thread per OSS.
158 Frequency : only with quota on the root user
160 Description: mds_obd_create error creating tmp object
161 Details : When the user sets quota on root, llog will be affected and can't
162 create files and write files.
165 Frequency : Always on ia64 patchless client, and possibly others.
167 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
168 Details : This allows the patchless client to be loaded on architectures
174 Description: disorder of journal start and llog_add cause deadlock.
175 Details : in llog_origin_connect, journal start should happen before
176 llog_add keep the same order as other functions to avoid
179 Severity : enhancement
181 Description: RedHat Update kernel for RHEL5
182 Details : Modify the kernel config file more closer RHEL5.
184 Severity : enhancement
186 Description: Build failure against Centos5 (RHEL5)
187 Details : Define PAGE_SIZE when it isn't present.
190 Frequency : occasionally when using NFS
192 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
193 Details : with NFS, the anon dentry's parent was set to itself in
194 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
195 corresponding dentry other than use rec->ur_name.
197 Severity : enhancement
199 Description: enable data checksumming by default
200 Details : enable checksum by default, allow --disable-checksum
201 configure option and "-o nochecksum" mount option. Checksums
202 can also be disabled at runtime via $LPROC/llite/*/checksum_pages.
205 Frequency : Occasionally with failover
207 Description: Client eviction due to failover config
208 Details : after a connection loss, the lustre client should attempt to
209 reconnect to the last active server first before trying the
210 other potential connections.
212 Severity : enhancement
214 Description: client-side metadata stat-ahead during readdir(directory readahead)
215 Details : perform client-side metadata stat-ahead when the client detects
216 readdir and sequential stat of dir entries therein
221 Description: when mds and osts use different quota unit(32bit and 64bit),
222 quota will be released repeatly.
223 Details : void sending multiple quota reqs to mds, which will keep the
224 status between the reqs.
227 Frequency : only with liblustre clients on XT3
229 Description: evictions taking too long
230 Details : allow llrd to evict clients directly on OSTs
234 Description: osts not allocated evenly to files
235 Details : change the condition to increase offset_idx
237 Severity : enhancement
239 Description: self-adjustable client's lru lists
240 Details : use adaptive algorithm for managing client cached locks lru
241 lists according to current server load, other client's work
242 pattern, memory activities, etc. Both, server and client
243 side namespaces provide number of proc tunables for controlling
248 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
249 Details : rewrite inherited code (uuid parsing code from ext2 utils and
250 readdir code from ext3) from scratch preserving functionality.
254 Description: Only those disconnect error should be returned by rq_status.
255 Details : In open/enqueue processs, Some errors, which will cause client
256 disconnected, should be returned by rq_status, while other
257 errors should still be returned by intent, then mdc or llite will
260 Severity : enhancement
262 Description: Tune the kernel for good SCSI performance.
263 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
264 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
268 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
269 Bugzilla : 13547, 13627
270 Description: Data corruption for OSTs that are formatted larger than 2TB
272 Details : When generating the bio request for lustre file writes the
273 sector number would overflow a temporary variable before being
274 used for the IO. The data reads correctly from Lustre (which
275 will overflow in a similar manner) but other file data or
276 filesystem metadata may be corrupted in some cases.
280 Description: "lfs find -obd UUID" prints directories
281 Details : "lfs find -obd UUID" will return all directory names instead
282 of just file names. It is incorrect because the directories
283 do not reside on the OSTs.
287 Description: MDS hang after unclean shutdown of lots of clients
288 Details : Never resend AST requests.
292 Frequency : Always, for kernels after 2.6.16
293 Description: Fix warning idr_remove called for id=.. which is not allocated.
294 Details : Last kernels save old s_dev before kill super and not allow
295 to restore from callback - restore it before call kill_anon_super.
299 Description: client sometimes tried revalidate root dentry.
300 Details : not revalidate dentry if it root dentry.
302 --------------------------------------------------------------------------------
304 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
306 * Support for kernels:
307 2.6.5-7.286 (SLES 9),
308 2.6.9-55.0.2.EL (RHEL 4),
309 2.6.16.46-0.14 (SLES 10),
310 2.6.18-8.1.8.el5 (RHEL 5),
311 2.6.18.8 vanilla (kernel.org)
312 * Client support for unpatched kernels:
313 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
314 2.6.16 - 2.6.22 vanilla (kernel.org)
315 * Due to recently discovered recovery problems, we do not recommend
316 using patchless RHEL 4 clients with this or any earlier release.
317 * Recommended e2fsprogs version: 1.39.cfs8
318 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
323 Description: block reactivating mgc import until all deactivates complete
324 Details : Fix race when failing back MDT/MGS to itself (testing)
326 Severity : enhancement
328 Description: lfs setstripe enhancement
329 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
332 Frequency : mds/oss recovery
334 Description: llog ctxt is refrenced after it has been freed.
335 Details : llog ctxt refcount was added to avoide the race between ctxt free
336 and llog recovery process. Each llog user must hold ctxt refcount
337 before it access the llog. And the llog ctxt can only be freed
338 when its refcount is zero.
340 Severity : enhancement
342 Description: randomly memory allocation failure util
343 Details : Make lustre randomly failed allocating memory for testing purpose.
346 Frequency : only for SLES10
348 Description: Update kernel patch for SLES10 SP1
349 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
351 Severity : enhancement
353 Description: omit set fsid for export NFS
354 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
355 as FS_REQUIRES_DEV for avoid problems with generate fsid.
358 Frequency : after network failures
360 Description: Add sync option to mount_lustre.c
361 Details : Client loses data written to lustre after a network interruption.
363 Severity : enhancement
365 Description: Error message improvement.
366 Details : Merging of two LCONSOLE_ERROR_MSG into one.
368 Severity : enhancement
370 Description: don't use GFP_* in generic Lustre code.
371 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
373 Severity : enhancement
375 Description: obdclass is limited by single OBD_ALLOC(idarray)
376 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
378 Severity : enhancement
380 Description: updated patchess for new RHEL4 kernel
381 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
383 Severity : enhancement
385 Description: warnings with build patchless client with vanila 2.6.19 and up
386 Details : change old ctl_table style and replace ctl_table/ctl_table_header
387 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
389 Severity : enhancement
391 Description: O_DIRECT bypasses client statistics.
392 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
393 read_ahead_stats were updated. Copied stats section from
394 osc_send_oap_rpc() into async_internal().
397 Frequency : only for Cray XT3
399 Description: peer credits not enough on many OST per OSS systems.
400 Details : Use new lnet way to add credits as we need those for pings and ASTs
403 Frequency : only with liblustre
405 Description: Liblustre is not releasing flock locks on file close.
406 Details : Release flock locks on file close.
409 Frequency : only for RHEL4
411 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
412 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
413 Update target file and kernel config.
417 Description: lustre support for RHEL5
418 Details : Add support for RHEL5.
423 Description: ASSERTION(export != NULL) failed in target_handle_connect
424 Details : Assetion hit is result of rare race between disconnect and connet
425 to same nid. target_handle_connect found old connect cockie and
426 tried to reconnect, but can't find export for this cockie.
431 Description: umount blocks forever on error
432 Details : In result of wrong using obd_no_recov and obd_force flags client
433 can hand if cancel or some other requests is lost.
436 Frequency : Only for SLES
438 Description: sanity_quota fail test_1
439 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
440 file, which makes TSTID[2] inunique.
442 Severity : enhancement
444 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
445 Details : Update target/ChangeLog/which_patch .
447 Severity : enhancement
449 Description: jbd statistics
450 Details : Port older jbd statistics patch for sles10
455 Description: lvbo_init failed for resource with missing objects.
456 Details : Fix returning error if we do stat for file with missing/corrupted
457 objects and i_size set to all sum of size of all avaible objects.
458 if we to truncate/write to missing object - it is recreated.
463 Description: Oops in read and write path when failing to allocate lock.
464 Details : Check if lock allocation failed and return error back.
467 Frequency : When flocks are used.
469 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
470 Details : Flock locks might destroy just granted lock if it could be merged
471 with another existing flock, this is done in completion handler,
472 so teach ldlm_cli_enquque_fini that this is a valid case for
478 Description: reply_lock_interpret crash due to race with it and lock cancel.
479 Details : Do not replay locks that are being cancelled. Do not reference
480 locks by their address during replay, just by their handle.
483 Frequency : only with deactivated OSTs
485 Description: lstripe command fails for valid OST index
486 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
487 instead of lov->desc.ld_active_tgt_count.
489 Severity : enhancement
491 Description: Build failure against Centos5 (RHEL5)
492 Details : Use getpagesize() instead of PAGE_SIZE.
494 --------------------------------------------------------------------------------
496 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
498 * Support for kernels:
499 2.6.5-7.283 (SLES 9),
500 2.6.9-55.EL (RHEL 4),
501 2.6.16.46-0.14 (SLES 10),
502 2.6.18.8 vanilla (kernel.org)
503 * Client support for unpatched kernels:
504 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
505 2.6.16 - 2.6.22 vanilla (kernel.org)
506 * Due to recently discovered recovery problems, we do not recommend
507 using patchless RHEL 4 clients with this or any earlier release.
508 * Recommended e2fsprogs version: 1.39.cfs8
509 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
510 * Starting with this release, the ldiskfs backing filesystem required
511 by Lustre is now in its own package, lustre-ldiskfs. This package
512 should be installed. It is versioned separately from Lustre and
513 may be released separately in future.
518 Description: server LBUG when shutting down
519 Details : Block umount forever until the mount refcount is zero rather
520 than giving up after an arbitrary timeout.
522 Severity : enhancement
524 Description: add optional extra BUILD_VERSION info
525 Details : add a new environment variable (namely LUSTRE_VERS) which allows
526 to override the lustre version.
529 Frequency : 2.6.18 servers only
531 Description: ll_kern_mount() doesn't release the module reference
532 Details : The ldiskfs module reference count never drops down to 0
533 because ll_kern_mount() doesn't release the module reference.
538 Description: server LBUG when using old ost_num_threads parameter
539 Details : Accept the old ost_num_threads parameter but warn that it
540 is deprecated, and fix an off-by-one error that caused an LBUG.
545 Description: Transient SCSI error results in persistent IO issue
546 Details : iobuf->dr_error is not reinitialized to 0 between two
550 Frequency : sometimes when underlying device returns I/O errors
552 Description: OSTs not going read-only during write failures
553 Details : OSTs are not remounted read-only when the journal commit threads
554 get I/O errors because fsfilt_ext3 calls journal_start/stop()
555 instead of the ext3 wrappers.
558 Frequency : only with 10000 clients or more
560 Description: poor connect scaling with increasing client count
561 Details : Don't run filter_grant_sanity_check for more than 100 exports
562 to improve scaling for large numbers of clients.
565 Frequency : SLES10 only
567 Description: sanity-quota.sh quotacheck failed: rc = -22
568 Details : Quotas cannot be enabled on SLES10.
571 Frequency : liblustre clients only
573 Description: getdirentries does not give error when run on compute nodes
574 Details : getdirentries does not fail when the size specified as an argument
575 is too small to contain at least one entry
577 Severity : enhancement
579 Description: Add LNET router traceability for debug purposes
580 Details : If a checksum failure occurs with a router as part of the
581 IO path, the NID of the last router that forwarded the bulk data
582 is printed so it can be identified.
587 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
588 Details : Due to a race condition, liblustre clients were occasionally
591 Severity : enhancement
593 Description: lfs setstripe use optional parameters instead of postional
596 Severity : enhancement
598 Description: Nanosecond timestamp support for ldiskfs
599 Details : The on-disk ldiskfs filesystem has added support for nanosecond
600 resolution timestamps. There is not yet support for this at
601 the Lustre filesystem level.
604 Frequency : during server recovery
606 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
607 Details : request with rq_no_resend flag not awake l_wait_event if they get a
611 Frequency : nfs export on patchless client
613 Description: connectathon hang when test nfs export over patchless client
614 Details : Disconnected dentry cannot be found with lookup, so we do not need
615 to unhash it or make it invalid
618 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
619 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
620 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
625 Description: open req refcounting wrong on reconnect
626 Details : If reconnect happened between getting open reply from server and
627 call to mdc_set_replay_data in ll_file_open, we will schedule
628 replay for unreferenced request that we are about to free.
629 Subsequent close will crash in variety of ways.
630 Check that request is still eligible for replay in
631 mdc_set_replay_data().
636 Description: disable writes to filesystem when reading health_check file
637 Details : the default for reading the health_check proc file has changed
638 to NOT do a journal transaction and write to disk, because this
639 can cause reads of the /proc file to hang and block HA state
640 checking on a healthy but otherwise heavily loaded system. It
641 is possible to return to the previous behaviour during configure
642 with --enable-health-write.
644 Severity : enhancement
646 Description: 64-bit inode version
647 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
648 to the inode. This will be required for version-based recovery.
653 Description: MDS fails to start if a duplicate client export is detected
654 Details : in some rare cases it was possible for a client to connect to
655 an MDS multiple times. Upon recovery the MDS would detect this
656 and fail during startup. Handle this more gracefully.
658 Severity : enhancement
660 Description: Add -o localflock option to simulate old noflock behaviour.
661 Details : This will achieve local-only flock/fcntl locks coherentness.
666 Description: log_commit_thread vs filter_destroy race leads to crash
667 Details : Take import reference before releasing llog record semaphore
672 Description: Wrong request locking in request set processing
673 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
674 imp_delayed_list, in this place should be used imp_lock.
677 Frequency : when reconnecting
679 Description: Grant leak when OSC reconnect to OST
680 Details : When osc reconnect ost, OST(filter) should check whether it
681 should grant more space to client by comparing fed_grant and
682 cl_avail_grant, and return the granted space to client instead
683 of "new granted" space, because client will call osc_init_grant
684 to update the client grant space info.
687 Frequency : when client reconnects to OST
689 Description: Grant leak when OSC does a resend and replays bulk write
690 Details : When osc reconnect to OST, OST(filter) should clear grant info of
691 bulk write request, because the grant info while be sync between
692 OSC and OST when reconnect, and we should ignore the grant info
693 these of resend/replay write req.
698 Description: Grant space more than avaiable space sometimes.
699 Details : When then OST is about to be full, if two bulk writing from
700 different clients came to OST. Accord the avaliable space of the
701 OST, the first req should be permitted, and the second one
702 should be denied by ENOSPC. But if the seconde arrived before
703 the first one is commited. The OST might wrongly permit second
704 writing, which will cause grant space > avaiable space.
707 Frequency : when client is evicted
709 Description: Grant might be wrongly erased when osc is evicted by OST
710 Details : when the import is evicted by server, it will fork another
711 thread ptlrpc_invalidate_import_thread to invalidate the
712 import, where the grant will be set to 0. While the original
713 thread will update the grant it got when connecting. So if
714 the former happened latter, the grant will be wrongly errased
715 because of this race.
720 Description: Checking Stale with correct fid
721 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
722 to check whether it is stale, because sometimes, we want the
723 enqueue happened anyway, and op_data.fid2 will not be initialized.
725 Severity : enhancement
727 Description: update patchless client
728 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
731 Frequency : only with 2.4 kernel
733 Description: random memory corruption
734 Details : size of struct ll_inode_info is to big for union inode.u and this
735 can be cause of random memory corruption.
740 Description: Memory leak in recovery
741 Details : Lov_mds_md was not free in an error handler in mds_create_object.
742 It should also check obd_fail before fsfilt_start, otherwise if
743 fsfilt_start return -EROFS,(failover mds during mds recovery).
744 then the req will return with repmsg->transno = 0 and rc = EROFS.
745 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
746 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
747 be freed no matter whether fsfilt_commit success or not.
750 Frequency : only with huge numbers of clients
752 Description: Prevents from taking the superblock lock in llap_from_page for
754 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
755 for llap_from_page call in ll_removepage() prevents from taking
756 the superblock lock for a soon died page.
761 Description: Not check open intent error before release open handle
762 Details : in some rare cases, the open intent error is not checked before
763 release open handle, which may cause
764 ASSERTION(open_req->rq_transno != 0), because it tries to release
765 the failed open handle.
770 Description: Set cat log bitmap only after create log success.
771 Details : in some rare cases, the cat log bitmap is set too early. and it
772 should be set only after create log success.
776 Description: Accessing a block bevice can re-enable I/O when Lustre is
777 tearing down a device.
778 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
783 Description: service threads may hog cpus when there are a lot of requests
784 Details : Insert cond_resched to give other threads a chance to use some CPU
789 Description: the cat log was not initialized in recovery
790 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
791 unlink log on mds will not be initialized until mds post
792 recovery. And also in mds post recovery, the unlink log will
793 initialization will be done asynchronausly, so there will be race
794 between add unlink log and unlink log initialization.
798 Description: brw_stats were being printed incorrectly
799 Details : brw_stats were being printed as log2 but all of them were not
800 recorded as log2. Also remove some code duplication arising from
801 filter_tally_{read,write}.
805 Frequency : rare, only in recovery.
806 Description: ASSERTION(req->rq_type != LI_POISON) failed
807 Details : imp_lock should be held while iterating over imp_sending_list for
808 prevent destroy request after get timeout in ptlrpc_queue_wait.
812 Description: replay-single.sh test 52 fails
813 Details : A lock's skiplist need to be cleanup when it being unlinked
814 from its resource list.
819 Description: Short directio read returns full requested size rather than
821 Details : Direct I/O operations should return actual amount of bytes
822 transferred rather than requested size.
824 Severity : enhancement
826 Description: metadata RPC reduction (e.g. for rm performance)
827 Details : decrease the amount of synchronous RPC between clients and servers
828 by canceling conflicing lock before the operation on the client
829 and packing thier handles into the main operation RPC to server.
831 Severity : enhancement
833 Description: add #ifdef HAVE_KERNEL_CONFIG_H
834 Details : kernels from 2.6.19 not need include linux/config.h, but add
835 include linux/autoconf.h in commpiler command line.
837 Severity : enhancement
839 Description: patchless client support for 2.6.22 kernel
840 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
841 removed. In this case we need drop using os depended interface to
842 kmem_cache and use cfs_mem_cache API.
847 Description: fix mal-formatted messages
848 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
852 Frequency : always in liblustre
853 Description: wrong IS_ERR implementation in liblustre.h
854 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
859 Description: Correct condition for output debug message.
860 Details : inode i_nlink equal zero is not enough for output message about
861 disk corruption, i_ctime and i_mode should be also checked.
865 Frequency : always in patchless client
866 Description: add configure check for truncate_complete_page
867 Details : improve checks for exported symbols. This allow run check without
868 sources, but with Module.symvers shipped with kernel distribution.
869 add check for truncate_complete_page used by patchless client.
873 Description: sanity.sh test_77h fails with "test_77h file compare failed"
874 Details : test_77h uses a file which was messed by other test case.
878 Description: Not Check whether lov_tgts is NULL in some lov functions
879 Details : Checking whether lov_tgts is NULL in some functions.
883 Description: replace obdo_alloc() with OBDO_ALLOC macro
884 Details : nothing special is done in obdo_alloc() function, and for
885 debugging purpose, it needs to be replaced with macros.
889 Description: bad return value and errno from fcntl call
890 Details : In liblustre API, errno should be a negative value if error
895 Description: ptlrpc_check_set() LBUG
896 Details : In case of positive reply from server and failed client bulk
897 callback after bulk transfer shouldn't LBUG, but process this
898 request as erroneous.
900 Severity : enhancement
902 Description: VFS operations stats tool.
903 Details : Tool which collects stats by tracking value written in pid,
904 ppid, gid and uses llstat to generate output to plot graph using
906 Updated lustre/utils/Makefile.am
907 Added lustre/utils/ltrack_stats.c
909 Severity : enhancement
911 Description: 2.6.18 server support (lustre 1.6.1)
912 Details : Support for 2.6.18 kernels on the server side.
917 Description: ASSERTION(imp->imp_conn_current) failed
918 Details : an assertion failure is hit if a client node boots and attempts to
919 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
922 Frequency : only for i686
924 Description: 1.4.11 RC1 build fails for RHEL 4, i686
925 Details : Fixed config variable for build.
930 Description: Updated patchess for new RHEL4 kernel
931 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
932 Updated patch jbd-stats-2.6.9.patch
933 Updated patch qsnet-rhel4-2.6.patch
934 Updated patch quota-deadlock-on-pagelock-core.patch
935 Updated patch vfs_intent-2.6-rhel4.patch
936 Updated patch vfs_races-2.6-rhel4.patch
937 Updated series file 2.6-rhel4-titech.series
938 Updated series file 2.6-rhel4.series
939 Updated kernel config files
944 Description: lquota slave complains LBUG when reconnecting with mds
946 Details : quota slave depends qctxt->lqc_import to send its quota request.
947 This pointer will be invalid if mds did failover or broke its
948 connect to osts, which leads to LBUG.
951 Frequency : when qunit size is too small(less than 20M)
953 Description: write is stopped by improper -EDQUOT
954 Details : If the master is busy and qunit size is small enough(let's say 1M),
955 the slave can not get quota from master on time, which will lead
956 slave to trigger a -EQUOTA to client.
961 Description: Deadlock during metadata tests
962 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
963 called with the per-dentry lock held.
966 Frequency : SLES9 only
968 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
969 Details : set CONFIG_NR_CPUS to 128 instead of 8.
971 Severity : enhancement
973 Description: remove fs_prep_san_write operation and related patches
974 Details : remove the ext3-san-jdike patches which are no longer useful.
979 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
980 Details : a disk corruption can cause the mballoc code to assert on a
981 double free or other extent corruptions. Handle these with
982 ext3_error() instead of assertions.
985 Frequency : only with mballoc3 code and deep extent trees
987 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
988 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
989 can cause the filesystem to be remounted read-only.
994 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
995 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
996 few external users of this ioctl number is lfsck's call to
997 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
998 Add the old ioctl number to the handler so both old and new
1003 Description: parallel lock callbacks
1004 Details : Instead of sending blocking and completion callbacks as separated
1005 requests, adding them to a set and sending in parallel.
1009 Description: Disable most debugging by default
1010 Details : To improve performance, disable most logging (for debug purposes)
1011 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
1012 default, and HA includes fewer messages.
1016 Frequency : only run on patchless client.
1017 Description: use do_facet on sanity.sh for test handling recoverables errors
1018 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
1022 Description: hash tables for lists of nids, connections and uuids
1023 Details : Hash tables noticeably help when a lot of clients connect to a
1024 server, to faster identify duplicate connections or reconnects,
1025 also to faster find export to evict in manual eviction case.
1029 Description: Sometimes, when the server evict a client, and the client will
1030 not be evicted as soon as possible.
1031 Details : In enqueue req, the error was returned by intent, instead of
1032 rq_status which make ptlrpc layer not detect this error, and
1033 does not evict the client. So enqueue error should be returned
1037 Frequency : only at startup
1039 Description: Delay client connections to MDT until fist MDT->OST connect
1040 Details : If a client tried to create a new file before the MDT had
1041 connected to any OSTs, the create would return EIO. Now
1042 the client will simply block until the MDT connects to the
1043 first OST and the create can succeed.
1046 Frequency : only for SLES9
1048 Description: Routinely utilize latest Quadrics drivers in CFS releases
1049 Details : Update patch qsnet-suse-2.6.patch.
1052 Frequency : only for sles10
1054 Description: Update patches for SLES 10 SP1 kernel.
1055 Details : Update the patch vfs_intent-2.6-sles10.patch.
1060 Description: Routinely utilize latest Quadrics drivers in CFS releases
1061 Details : Update patch qsnet-rhel4-2.6.patch.
1064 Frequency : at statup only
1066 Description: mds_lov_synchronize race leads to various problems
1067 Details : simultaneous MDT->OST connections at startup can cause the
1068 sync to abort, leaving the OSC in a bad state.
1073 Description: Using wrong pointer in osc_brw_prep_request
1074 Details : Access to array[-1] can produce panic if kernel compiled with
1075 CONFIG_PAGE_ALLOC enabled
1077 Severity : enhancement
1079 Description: Async OSC create to avoid the blocking unnecessarily.
1080 Details : If a OST has no remain object, system will block on the creating
1081 when need to create a new object on this OST. Now, ways use
1082 pre-created objects when available, instead of blocking on an
1083 empty osc while others are not empty. If we must block, we block
1084 for the shortest possible period of time.
1088 Frequency : only in recovery
1089 Description: Mark OST as early accessible if his start SYNC.
1090 Details : osc_precreate return flag early accessible if oscc marked as
1091 OSCC_FLAG_SYNC_IN_PROGRESS.
1096 Description: Sometimes precreate code can triger create object on wrong ost
1097 Details : Wrong protected or not not restored variables aftre precreate loop
1098 can produce creation object on wrong ost.
1101 Frequency : oss recovery
1103 Description: llog_commit_thread cleanup should sync with llog_commit_thread
1105 Details : llog_commit_thread_count should be synced between llog_commit
1106 start and cleanup, so new llog_commit thread should not be started
1107 when llog_commit threads being stopped to avoid accessing some
1110 Severity : enhancement
1112 Description: Add printing inode info into message about error in writepage.
1114 --------------------------------------------------------------------------------
1116 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1121 Frequency : on some architectures
1123 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1124 Details : Uninitialized flags sometimes cause configuration commands to
1128 Frequency : patchless clients only
1130 Description: missing __iget() symbol export
1131 Details : The __iget() symbol export is missing. To avoid the need for
1132 this on patchless clients the deathrow inode reaper is turned
1133 off, and we depend on the VM to clean up old inodes. This
1134 dependency was during via the fix for bug 12181.
1139 Description: sanity.sh fail: test_52b
1140 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
1141 incorrect inode's flags to client.
1143 --------------------------------------------------------------------------------
1145 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1147 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1148 INTEROPERATE with older versions automatically. In many cases a
1149 special upgrade step is needed. Please read the
1150 user documentation before upgrading any part of a 1.4.x system.
1151 * WARNING: Lustre configuration and startup changes are required with
1152 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1154 * Support for kernels:
1155 2.4.21-47.0.1.EL (RHEL 3),
1156 2.6.5-7.283 (SLES 9),
1157 2.6.9-42.0.10.EL (RHEL 4),
1158 2.6.12.6 vanilla (kernel.org),
1159 2.6.16.27-0.9 (SLES10)
1160 * Client support for unpatched kernels:
1161 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
1162 2.6.16 - 2.6.19 vanilla (kernel.org),
1163 2.6.9-42.0.8EL (RHEL 4)
1164 * Recommended e2fsprogs version: 1.39.cfs6
1165 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1168 Severity : enhancement
1170 Description: Async OSC create to avoid the blocking unnecessarily.
1171 Details : If a OST has no remain object, system will block on the creating
1172 when need to create a new object on this OST. Now, ways use
1173 pre-created objects when available, instead of blocking on an
1174 empty osc while others are not empty. If we must block, we block
1175 for the shortest possible period of time.
1177 Severity : enhancement
1179 Description: MountConf
1180 Details : Lustre configuration is now managed via mkfs and mount
1181 commands instead of lmc and lconf. New obd types (MGS, MGC)
1182 are added for dynamic configuration management. See
1183 https://mail.clusterfs.com/wikis/lustre/MountConf for
1186 Severity : enhancement
1188 Description: dynamic OST addition
1189 Details : OSTs can now be added to a live filesystem
1191 Severity : enhancement
1193 Description: startup order invariance
1194 Details : MDTs and OSTs can be started in any order. Clients only
1195 require the MDT to complete startup.
1197 Severity : enhancement
1199 Description: parallel, asynchronous orphan cleanup
1200 Details : orphan cleanup is now performed in separate threads for each
1201 OST, allowing parallel non-blocking operation.
1203 Severity : enhancement
1205 Description: optimized stripe assignment
1206 Details : stripe assignments are now made based on ost space available,
1207 ost previous usage, and OSS previous usage, in order to try
1208 to optimize storage space and networking resources.
1210 Severity : enhancement
1212 Description: Permanently set tunables
1213 Details : All writable /proc/fs/lustre tunables can now be permanently
1214 set on a per-server basis, at mkfs time or on a live
1217 Severity : enhancement
1219 Description: Lustre message v2
1220 Details : Add lustre message format v2.
1222 Severity : enhancement
1224 Description: client OST exclusion list
1225 Details : Clients can be started with a list of OSTs that should be
1226 declared "inactive" for known non-responsive OSTs.
1230 Description: ENOENT returned for valid filehandle during dbench.
1231 Details : Check if a directory has children when invalidating dentries
1232 associated with an inode during lock cancellation. This fixes
1233 an incorrect ENOENT sometimes seen for valid filehandles during
1234 testing with dbench.
1237 Frequency : SFS test only (otherwise harmless)
1239 Description: SPEC SFS validation failure on NFS v2 over lustre.
1240 Details : Changes the blocksize for regular files to be 2x RPC size,
1241 and not depend on stripe size.
1243 Severity : enhancement
1245 Description: fine-grained SMP locking inside DLM
1246 Details : Improve DLM performance on SMP systems by removing the single
1247 per-namespace lock and replace it with per-resource locks.
1249 Severity : enhancement
1251 Description: don't hold multiple extent locks at one time
1252 Details : To avoid client eviction during large writes, locks are not
1253 held on multiple stripes at one time or for very large writes.
1254 Otherwise, clients can block waiting for a lock on a failed OST
1255 while holding locks on other OSTs and be evicted.
1257 Severity : enhancement
1259 Description: Multiple MD RPCs in flight.
1260 Details : Further unserialise some read-only MDT RPCs - learn about intents.
1261 To avoid overly-overloading MDT, introduce a limit on number of
1262 MDT RPCs in flight for a single client and add /proc controls
1263 to adjust this limit.
1265 Severity : enhancement
1267 Description: client read/write statistics
1268 Details : Add client read/write call usage stats for performance
1269 analysis of user processes.
1270 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1271 file access. extents_stats shows chunk size distribution.
1272 extents_stats_per_process show chunk size distribution per
1275 Severity : enhancement
1277 Description: per-client statistics on server
1278 Details : Add ldlm and operations statistics for each client in
1279 /proc/fs/lustre/mds|obdfilter/*/exports/
1281 Severity : enhancement
1283 Description: improved MDT statistics
1284 Details : Add detailed MDT operations statistics in
1285 /proc/fs/lustre/mds/*/stats
1287 Severity : enhancement
1289 Description: VFS operations stats
1290 Details : Add client VFS call stats, trackable by pid, ppid, or gid
1291 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
1296 Description: Fix client-side osc byte counters
1297 Details : The osc read/write byte counters in
1298 /proc/fs/lustre/osc/*/stats are now working
1301 Frequency : always as root on SLES
1303 Description: Failure of copying files with lustre special EAs.
1304 Details : Client side always return success for setxattr call for lustre
1305 special xattr (currently only "trusted.lov").
1310 Description: Refcount LNET uuids
1311 Details : The global LNET uuid list grew linearly with every startup;
1312 refcount repeated list entries instead of always adding to
1315 Severity : enhancement
1317 Description: Dynamic service threads
1318 Details : Within a small range, start extra service threads
1319 automatically when the request queue builds up.
1322 Frequency : mixed-endian client/server environments
1324 Description: mixed-endian crashes
1325 Details : The new msg_v2 system had some failures in mixed-endian
1328 Severity : enhancement
1330 Description: Easy OST removal
1331 Details : OSTs can be permanently deactivated with e.g. 'lctl
1332 conf_param lustre-OST0001.osc.active=0'
1334 Severity : enhancement
1336 Description: MGS proc entries
1337 Details : Added basic proc entries for the MGS showing what filesystems
1340 Severity : enhancement
1342 Description: provide MGS failover
1343 Details : Added config lock reacquisition after MGS server failover.
1345 Severity : enhancement
1347 Description: add Linux 2.4 support
1348 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1352 Description: a large application tries to do I/O to the same resource and dies
1353 in the middle of it.
1354 Details : Check the req->rq_arrival time after the call to
1355 ost_brw_lock_get(), but before we do anything about
1356 processing it & sending the BULK transfer request. This
1357 should help move old stale pending locks off the queue as
1358 quickly as obd_timeout.
1361 Frequency : when an incorrect nid is specified during startup
1363 Description: ptlrpc connect to non-existant node causes kernel crash
1364 Details : LNET can't be re-entered from an event callback, which
1365 happened when we expire a message after the export has been
1366 cleaned up. Instead, hand the zombie cleanup off to another
1369 Severity : enhancement
1371 Description: plain/inodebits lock performance improvement
1372 Details : Grouping plain/inodebits in granted list by their request modes
1373 and bits policy, thus improving the performance of search through
1377 Frequency : only if OST filesystem is corrupted
1379 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1380 Details : for a short time RPCs with bulk IO are in the replay list,
1381 but replay of bulk IOs is unimplemented. If the OST filesystem
1382 is corrupted due to disk cache incoherency and then replay is
1383 started it is possible to trip an assertion. Avoid putting
1384 committed RPCs into the replay list at all to avoid this issue.
1387 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1389 Description: System hang on startup
1390 Details : This bug allowed the liblustre (e.g. catamount) client to
1391 return to the app before handling all startup RPCs. This
1392 could leave the node unresponsive to lustre network traffic
1393 and manifested as a server ptllnd timeout.
1395 Severity : enhancement
1397 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1398 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1399 boolean to control whether to print peer debug info when a
1400 client's RPC times out.
1403 Frequency : only for kernels with patches from Lustre below 1.4.3
1405 Description: Remove old rdonly API
1406 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1409 Frequency : only for devices with external journals
1411 Description: Set external device read-only also
1412 Details : During a commanded failover stop, we set the disk device
1413 read-only while the server shuts down. We now also set any
1414 external journal device read-only at the same time.
1417 Frequency : when upgrading from 1.4 while trying to change parameters
1419 Description: The wrong (new) MDC name was used when setting parameters for
1420 upgraded MDT's. Also allows changing of OSC (and MDC)
1421 parameters if --writeconf is specified at tunefs upgrade time.
1424 Frequency : when setting specific ost indicies
1426 Description: QOS code breaks on skipped indicies
1427 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1428 created with --index need not be sequential.
1430 Severity : enhancement
1432 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1433 Details : The uninit_groups feature works in conjunction with the kernel
1434 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1435 up the pass1 processing of e2fsck. This is a read-only feature
1436 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1437 mount filesystems that have had this feature enabled.
1439 Severity : enhancement
1441 Description: Improve multi-block allocation algorithm to avoid fragmentation
1442 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1443 allocation locality and avoid filesystem fragmentation.
1445 ------------------------------------------------------------------------------
1447 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1449 * Support for kernels:
1450 2.4.21-47.0.1.EL (RHEL 3)
1451 2.6.5-7.283 (SLES 9)
1452 2.6.9-42.0.10.EL (RHEL 4)
1453 2.6.12.6 vanilla (kernel.org)
1454 2.6.16.27-0.9 (SLES 10)
1455 * Recommended e2fsprogs version: 1.39.cfs5
1457 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1461 Frequency : occasional, depends on client load and configuration
1462 Bugzilla : 12181, 12203
1463 Description: data loss for recently-modified files
1465 Details : In some cases it is possible that recently written or created
1466 files may not be written to disk in a timely manner (this should
1467 normally be within 30s unless client IO load is very high).
1468 The problem appears as zero-length files or files that are a
1469 multiple of 1MB in size after a client crash or client eviction
1470 that are missing data at the end of the file.
1472 This problem is more likely to be hit on clients where files are
1473 repeatedly created and unlinked in the same directory, clients
1474 have a large amount of RAM, have many CPUs, the filesystem has
1475 many OSTs, the clients are rebooted frequently, and/or the files
1476 are not accessed by other nodes after being written.
1478 The presence of the problem can be detected by looking at
1479 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1480 smaller than the second (nr_unused) then dirty files will not
1481 be flushed automatically to disk. "sync; sleep 10" should be
1482 run several times on the node before unmounting it to update
1483 Lustre (this is also safe to run on nodes without this problem).
1485 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1486 that can cause this same problem, so customers using that kernel
1487 also need to update the kernel in addition to Lustre. In order
1488 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1490 It is normal that files written just before a client crash (less
1491 than 30s) may not yet have been flushed to disk, even for local
1495 Frequency : frequent on thin XT3 nodes
1497 Description: UUID collision on thin XT3 Linux nodes
1498 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1499 randomly, since we relied on an insufficiently-seeded PRNG.
1504 Description: OSS hangs after "All ost request buffers busy"
1505 Details : A deadlock between quota and journal operations caused OSS
1506 hangs after printing "All ost request buffers busy."
1509 Frequency : always on liblustre builds
1511 Description: Cleanup compiler warnings on liblustre
1514 Frequency : always on liblustre builds on XT3
1516 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1520 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1521 > 32000 subdirectories
1522 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1523 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1524 subdirectory count crosses 32000. This will aid e2fsck to
1525 correctly handle more than 32000 subdirectories.
1528 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1530 Description: System hang on startup
1531 Details : This bug allowed the liblustre (e.g. catamount) client to
1532 return to the app before handling all startup RPCs. This
1533 could leave the node unresponsive to lustre network traffic
1534 and manifested as a server ptllnd timeout.
1536 Severity : enhancement
1538 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1539 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1540 boolean to control whether to print peer debug info when a
1541 client's RPC times out.
1546 Description: make O_SYNC working on 2.6 kernels
1547 Details : 2.6 kernels use different method for mark pages for write,
1548 so need add a code to lustre for O_SYNC work.
1553 Description: Failure to close file and release space on NFS
1554 Details : Put inode details into lock acquired in ll_intent_file_open.
1555 Use mdc_intent_lock in ll_intent_open to properly
1556 detect all kind of errors unhandled by mdc_enqueue.
1561 Description: proc file read during shutdown sometimes raced obd removal,
1563 Details : Add lock to prevent obd access after proc file removal.
1566 Frequency : Only for files larger than 4GB on 32-bit clients.
1568 Description: improperly doing page alignment of locks
1569 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1570 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1575 Description: RPCs being resent when they shouldn't be
1576 Details : Some RPCs that should not be resent are being resent. This
1577 can cause inconsistencies in the RPC state machine. Do not
1578 resend such requests.
1581 Frequency : rare, only with NFS export
1583 Description: Crash on NFS re-export node
1584 Details : under very unusual load conditions an assertion is hit in
1585 ll_intent_file_open()
1588 Frequency : only if OST filesystem is corrupted
1590 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1591 Details : for a short time RPCs with bulk IO are in the replay list,
1592 but replay of bulk IOs is unimplemented. If the OST filesystem
1593 is corrupted due to disk cache incoherency and then replay is
1594 started it is possible to trip an assertion. Avoid putting
1595 committed RPCs into the replay list at all to avoid this issue.
1600 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1601 Details : Large single O_DIRECT read and write calls can fail to allocate
1602 a sufficiently large buffer to process the request. In case of
1603 allocation failure the allocation is retried with a smaller
1604 buffer and broken into smaller requests.
1606 Severity : enhancement
1608 Description: Add -o localflock option to simulate old noflock behaviour.
1609 Details : This will achieve local-only flock/fcntl locks coherentness.
1614 Description: versioning check is incomplete
1615 Details : Checking the version difference of client vs. server, report
1616 error if the gap is too big.
1621 Description: add support PG_writeback bit
1622 Details : add support for PG_writeback bit for Lustre, for more carefull
1623 work with page cache in 2.6 kernel. This also fix some deadlocks
1624 and remove hack for work O_SYNC with 2.6 kernel.
1626 Severity : enhancement
1628 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1629 Details : The uninit_groups feature works in conjunction with the kernel
1630 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1631 up the pass1 processing of e2fsck. This is a read-only feature
1632 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1633 mount filesystems that have had this feature enabled.
1635 Severity : enhancement
1637 Description: Improve multi-block allocation algorithm to avoid fragmentation
1638 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1639 allocation locality and avoid filesystem fragmentation.
1641 ------------------------------------------------------------------------------
1643 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1645 * Support for kernels:
1646 2.6.9-42.0.3.EL (RHEL 4)
1647 2.6.5-7.276 (SLES 9)
1648 2.4.21-47.0.1.EL (RHEL 3)
1649 2.6.12.6 vanilla (kernel.org)
1650 2.6.16.21-0.8 (SLES10)
1651 * Recommended e2fsprogs version: 1.39.cfs2-0
1653 * The backwards-compatible /proc/sys/portals symlink has been removed
1654 in this release. Before upgrading, please ensure that you change
1655 any configuration scripts or /etc/sysctl.conf files that access
1656 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1657 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1658 in advance of the upgrade on any system running Lustre 1.4.6 or
1659 newer, since /proc/sys/lnet was added in that version.
1660 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1664 Frequency : only when quota is used
1666 Description: avoid scanning export list for quota master
1667 Details : Change the algorithms to avoid scanning export list in order
1668 to improve the efficiency.
1671 Frequency : MDS failover only, very rarely
1673 Description: "went back in time" messages on mds failover
1674 Details : The greatest transno may be lost when the current operation
1675 finishes with an error (transno==0) and the client's last_rcvd
1676 record is over-written. Save the greatest transno in the
1677 mds_last_transno for this case.
1680 Frequency : always for specific kernels and striping counts
1682 Description: client may get "Matching packet too big" without ACL support
1683 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1684 when trying to access files in certain configurations. The
1685 clients should in fact be denied when mounting because they do
1686 not understand ACLs.
1689 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1691 Description: many clients connecting with IO in progress causes connect timeouts
1692 Details : Avoid synchronous journal commits to avoid delays caused by many
1693 clients connecting/disconnecting when bulk IO is in progress.
1694 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1695 OST_IO_PORTAL to avoid delays behind potentially many pending
1699 Frequency : occasionally with multiple writers to a single file
1701 Description: shared writes to file may result in wrong size reported by stat()
1702 Details : Allow growing of kms when extent lock is cancelled
1705 Frequency : always with random mmap IO to multi-striped file
1707 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1708 Details : Only if the hole is at the end of OST object so that kms is too
1709 small. Fix is to increase kms accordingly in ll_nopage.
1712 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1714 Description: writes to a missing object would leak memory on the OST
1715 Details : If there is an inconsistency between the MDS and OST filesystems,
1716 such that the MDS references an object that doesn't exist, writes
1717 to that object will leak memory due to incorrect cleanup in the
1718 error handling path, eventually running out of memory on the OST.
1723 Description: Creating too long symlink causes lustre errors
1724 Details : Check symlink and name lengths before sending requests to MDS.
1727 Frequency : only if flock is enabled (not on by default)
1729 Description: posix locks not released on fd closure on 2.6.9+
1730 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1731 kernels, this caused such locks not to be released on fd close and
1732 then assertions on fs unmount about still used locks.
1735 Frequency : MDS failover only, very rarely
1737 Description: clients may get ASSERTION(granted_lock != NULL)
1738 Details : When request was taking a long time, and a client was resending
1739 a getattr by name lock request. The were multiple lock requests
1740 with the same client lock handle and
1741 mds_getattr_name->fixup_handle_for_resent_request found one of the
1742 lock handles but later failed with ASSERTION(granted_lock != NULL).
1747 Description: handle->h_buffer_credits > 0, assertion failure
1748 Details : h_buffer_credits is zero after truncate, causing assertion
1749 failure. This patch extends the transaction or creates a new
1753 Frequency : NFS re-export or patchless client
1754 Bugzilla : 11179, 10796
1755 Description: Crash on NFS re-export node (__d_move)
1756 Details : We do not want to hash the dentry if we don't have a lock.
1757 But if this dentry is later used in d_move, we'd hit uninitialised
1758 list head d_hash, so we just do this to init d_hash field but
1759 leave dentry unhashed.
1762 Frequency : NFS re-export or patchless client
1764 Description: NFS exports has problem with symbolic link
1765 Details : lustre client didn't properly install dentry when re-exported
1766 to NFS or running patchless client.
1769 Frequency : NFS re-export or patchless client
1771 Description: Various nfs/patchless fixes.
1772 Details : fixes reuse disconected alias for lookup process - this fixes
1773 warning "find_exported_dentry: npd != pd",
1774 fix permission error with open files at nfs.
1775 fix apply umask when do revalidate.
1778 Frequency : occasional
1780 Description: Crash on NFS re-export node
1781 Details : calling clear_page() on the wrong pointer triggered oops in
1782 generic_mapping_read().
1785 Frequency : rarely, using O_DIRECT IO
1787 Description: unaligned directio crashes client with LASSERT
1788 Details : check for unaligned buffers before trying any requests.
1791 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1793 Description: stale data returned from RAID cache
1794 Details : If only a small amount of IO is done to the RAID device before
1795 reading it again it is possible to get stale data from the RAID
1796 cache instead of reading it from disk.
1799 Frequency : always for sles10 kernel
1801 Description: sles10 support
1802 Details : ll_follow_link: compile fixes and using of nd_set_link
1803 under newer kernels.
1806 Frequency : depends on arch, kernel and compiler version, always on sles10
1809 Description: recursive or deep enough symlinks cause stack overflow
1810 Details : getting rid of large stack-allocated variable in
1814 Frequency : depends on hardware
1816 Description: lustre write performance loss in the SLES10 kernel
1817 Details : the performance loss is caused by using of write barriers in the
1818 ext3 code. The SLES10 kernel turns barrier support on by
1819 default. The fix is to undo that change for ldiskfs.
1821 ------------------------------------------------------------------------------
1823 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1825 * Support for kernels:
1826 2.6.9-42.0.3EL (RHEL 4)
1827 2.6.5-7.276 (SLES 9)
1828 2.4.21-47.0.1.EL (RHEL 3)
1829 2.6.12.6 vanilla (kernel.org)
1833 Frequency : quota enabled and large files being deleted
1835 Description: releasing more than 4GB of quota at once hangs OST
1836 Details : If a user deletes more than 4GB of files on a single OST it
1837 will cause the OST to spin in an infinite loop. Release
1838 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1843 Description: statfs data retrieved from /proc may be stale or zero
1844 Details : When reading per-device statfs data from /proc, in the
1845 {kbytes,files}_{total,free,avail} files, it may appear
1846 as zero or be out of date.
1849 Frequency : systems with MD RAID1 external journal devices
1851 Description: lconf's call to blkid is confused by RAID1 journal devices
1852 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1853 of returning all block devices that match the journal UUID.
1856 Frequency : always, for aggregate stripe size over 4GB
1858 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1859 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1860 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1863 Frequency : always if "lfs find" used on a local file/directory
1865 Description: "lfs find" segfaults if used on a local file/directory
1866 Details : The case where a directory component was not specified wasn't
1867 handled correctly. Handle this properly.
1870 Frequency : always on ppc64
1872 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1873 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1874 perform endianness conversion.
1877 Frequency : rarely (truncate to non-zero file size after write under load)
1878 Bugzilla : 10730, 10687
1879 Description: Files padded with zeros to next 4K multiple
1880 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1881 it is possible that files that are truncated to a non-zero size
1882 immediately after being written are filled with zero bytes beyond
1883 the truncated size. No file data is lost.
1885 Severity : enhancement
1887 Description: Allow recovery/failover for liblustre clients.
1888 Details : liblustre clients were unaware of failover configurations until
1891 Severity : enhancement
1893 Description: user file locks should fail when not mounting with flock option
1894 Details : Set up an error-returning stub in ll_file_operations.lock field
1895 to prevent incorrect behaviour when client is mounted without
1896 flock option. Also, set up properly f_op->flock field for
1900 Frequency : always on ia64
1902 Description: "lfs df" loops on printing out MDS statfs information
1903 Details : The obd_ioctl_data was not initialized and in some systems
1904 this caused a failure during the ioctl that did not return
1905 an error. Initialize the struct and return an error on failure.
1908 Frequency : SLES 9 only
1910 Description: Error of copying files with lustre special EAs as root
1911 Details : Client side always return success for setxattr call for lustre
1912 special xattr (currently only "trusted.lov").
1915 Frequency : rarely on clusters with both ia64+i386 clients
1917 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1918 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1919 clients are doing concurrent writes to the same file it is
1920 possible that the ia64 clients may LASSERT because the OST
1921 extent locks are not PAGE_SIZE aligned. Ensure that grown
1922 locks are always aligned on the request boundary.
1925 Frequency : specific use, occasional
1927 Description: Overwriting in use executable truncates on-disk binary image
1928 Details : If one node attempts to overwrite an executable in use by
1929 another node, we now correctly return ETXTBSY instead of
1930 truncating the file.
1932 Severity : enhancement
1934 Description: Async OSC create to avoid the blocking unnecessarily.
1935 Details : If a OST has no remain object, system will block on the creating
1936 when need to create a new object on this OST. Now, ways use
1937 pre-created objects when available, instead of blocking on an
1938 empty osc while others are not empty. If we must block, we block
1939 for the shortest possible period of time.
1944 Description: chmod on Lustre root is propagated to other clients
1945 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1946 invalid by the follow_mount time.
1949 Frequency : liblustre clients only
1951 Description: Race in 'instant cancel' lock handling could lead to such locks
1952 never to be granted in case of SMP MDS
1953 Details : Do not destroy not yet granted but cbpending locks in
1957 Frequency : replay/resend of open
1959 Description: non null lock assetion failure in mds_intent_policy
1960 Details : Trying to replay/resend lockless open requests resulted in
1961 mds_open() returning 0 with no lock. Now it sets a flag if
1962 a lock is going to be returned.
1964 Severity : enhancement
1966 Description: Checksum enhancements
1967 Details : New checksum enhancements allow for resending RPCs that failed
1970 Severity : enhancement
1972 Description: Tunables on number of dirty pages in cacche
1973 Details : Allow to set limit on number of dirty pages cached.
1978 Description: client crash on unmount - lock still has references
1979 Details : In some error handling cases it was possible to leak a lock
1980 reference on a client while accessing a file. This was not
1981 harmful to the client during operation, but would cause the
1982 client to crash when the filesystem is unmounted.
1985 Frequency : specific case, rare
1987 Description: ETXTBSY on mds though file not in use by client
1988 Details : ETXTBSY is no longer incorrectly returned when attempting to
1989 chmod or chown a directory that the user previously tried to
1990 execute or a currently-executing binary.
1993 Frequency : extremely rare except on liblustre-based clients
1995 Description: Lustre space not freed when files are deleted
1996 Details : Clean up open-unlinked files after client eviction. Previously
1997 the unlink was skipped and the files remained as orphans.
2002 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2003 Details : In some cases it was possible to send a blocking callback to
2004 a client doing a glimpse, even though that client didn't get
2005 a lock granted. When the glimpse lock is cancelled on the OST
2006 the freed lock is left on the waiting list and corrupted the list.
2009 Frequency : all core dumps
2011 Description: Broke core dumps to lustre
2012 Details : Negative dentry may be unhashed if parent does not have UPDATE
2013 lock, but some callers, e.g. do_coredump, expect dentry to be
2014 hashed after successful create, hash it in ll_create_it.
2016 ------------------------------------------------------------------------------
2018 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2020 * Support for kernels:
2021 2.6.9-42.0.2.EL (RHEL 4)
2022 2.6.5-7.276 (SLES 9)
2023 2.4.21-40.EL (RHEL 3)
2024 2.6.12.6 vanilla (kernel.org)
2028 Frequency : always on RHEL 3
2030 Description: Number of open files grows over time
2031 Details : The number of open files grows over time, whether or not
2032 Lustre is started. This was due to a filp leak introduced
2033 by one of our kernel patches.
2035 ------------------------------------------------------------------------------
2037 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
2039 * Support for kernels:
2040 2.6.9-42.EL (RHEL 4)
2041 2.6.5-7.267 (SLES 9)
2042 2.4.21-40.EL (RHEL 3)
2043 2.6.12.6 vanilla (kernel.org)
2048 Bugzilla : 5719, 9635, 9792, 9684
2049 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2050 Details : If a server is under heavy load and cannot reply to new
2051 connection requests before the client resends the (re)connect,
2052 the connection handling code can behave badly if two service
2053 threads are concurrently handing separate (re)connections from
2054 the same client. Add better locking to the connection handling
2055 code, and ensure that only a single connection will be processed
2056 for a given client UUID, even if the lock is dropped.
2058 Severity : enhancement
2060 Description: add TCP zero-copy support to kernel
2061 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2062 sends if the hardware supports scatter-gather and checksumming.
2063 This allows socklnd to do client-write and server-read more
2064 efficiently and reduce CPU utilization from skbuf copying.
2067 Frequency : only if NFS exporting from client
2069 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2070 Details : If setting attributes on a file created under NFS that had
2071 never been opened it would be possible to oops the client
2072 if the file had no objects.
2076 Bugzilla : 9326, 10402, 10897
2077 Description: client crash in ptlrpcd_wake() thread when sending async RPC
2078 Details : It is possible that ptlrpcd_wake() dereferences a freed async
2079 RPC. In rare cases the ptlrpcd thread alread processed the RPC
2080 before ptlrpcd_wake() was called and the request was freed.
2083 Frequency : always for liblustre
2085 Description: liblustre client does MDS+OSTs setattr RPC for each write
2086 Details : When doing a write from a liblustre client, the client
2087 incorrectly issued an RPC to the MDS and each OST the file was
2088 striped over in order to update the timestamps. When writing
2089 with small chunks and many clients this could overwhelm the MDS
2090 with RPCs. In all cases it would slow down the write because
2091 these RPCs are unnecessary.
2093 Severity : enhancement
2095 Description: allow number of MDS service threads to be changed at module load
2096 Details : It is now possible to change the number of MDS service threads
2097 running. Adding "options mds mds_num_threads={N}" to the MDS's
2098 /etc/modprobe.conf will set the number of threads for the next
2099 time Lustre is restarted (assuming the "mds" module is also
2100 reloaded at that time). The default number of threads will
2101 stay the same, 32 for most systems.
2106 Description: OST crash if filesystem is unformatted or corrupt
2107 Details : If an OST is started on a device that has never been formatted
2108 or if the filesystem is corrupt and cannot even mount then the
2109 error handling cleanup routines would dereference a NULL pointer.
2114 Description: NULL pointer deref in llap_from_page.
2115 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2116 incorrect) mapping if the page was truncated/reclaimed while it was
2117 searched for. Check for this condition and skip such pages when
2118 doing readahead. Introduce extra check to llap_from_page() to
2119 verify page->mapping->host is non-NULL (so page is not anonymous).
2122 Frequency : Sometimes when using sys_sendfile
2124 Description: "page not covered by a lock" warnings from ll_readpage
2125 Details : sendfile called ll_readpage without right page locks present.
2126 Now we introduced ll_file_sendfile that does necessary locking
2127 around call to generic_file_sendfile() much like we do in
2131 Frequency : with certain MDS communication failures at client mount time
2133 Description: NULL pointer deref after failed client mount
2134 Details : a client connection request may delayed by the network layer
2135 and not be sent until after the PTLRPC layer has timed out the
2136 request. If the client fails the mount immediately it will try
2137 to clean up before the network times out the request. Add a
2138 reference from the request import to the obd device and delay
2139 the cleanup until the network drops the request.
2142 Frequency : occasionally during client (re)connect
2144 Description: assertion failure during client (re)connect
2145 Details : processing a client connection request may be delayed by the
2146 client or server longer than the client connect timeout. This
2147 causes the client to resend the connection request. If the
2148 original connection request is replied in this interval, the
2149 client may trip an assertion failure in ptlrpc_connect_interpret()
2150 which thought it would be the only running connect process.
2153 Frequency : only with obd_echo servers and clients that are rebooted
2155 Description: kernel BUG accessing uninitialized data structure
2156 Details : When running an obd_echo server it did not start the ping_evictor
2157 thread, and when a client was evicted an uninitialized data
2158 structure was accessed. Start the ping_evictor in the RPC
2159 service startup instead of the OBD startup.
2161 Severity : enhancement
2162 Bugzilla : 10193 (patchless)
2163 Description: Remove dependency on various unexported kernel interfaces.
2164 Details : No longer need reparent_to_init, exit_mm, exit_files,
2165 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2168 Frequency : rare (only users of deprecated and unsupported LDAP config)
2170 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2171 Details : LDAP apparently contains 'lustreName' attributes instead of
2172 'name'. A simple remapping of the name is sufficient.
2175 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2177 Description: waiting_locks_callback trips kernel BUG if client is evicted
2178 Details : Running with the dump_on_timeout debug flag turned on makes
2179 it possible that the waiting_locks_callback() can try to dump
2180 the Lustre kernel debug logs from an interrupt handler. Defer
2181 this log dumping to the expired_lock_main() thread.
2183 Severity : enhancement
2185 Description: Support NFS exporting on 2.6 kernels.
2186 Details : Implement non-rawops metadata methods for NFS server to use without
2187 changing NFS server code.
2190 Frequency : very rare (synthetic metadata workload only)
2192 Description: two racing renames might cause an MDS thread to deadlock
2193 Details : Running the "racer" program may cause one MDS thread to rename
2194 a file from being the source of a rename to being the target of
2195 a rename at exactly the same time that another thread is doing
2196 so, and the second thread has already enqueued these locks after
2197 doing a lookup of the target and is trying to relock them in
2198 order. Ensure that we don't try to re-lock the same resource.
2201 Frequency : only very large systems with liblustre clients
2203 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2204 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2205 all OSTs in parallel. This allows the eviction of stale liblustre
2206 clients to proceed much faster than if they were done in series,
2207 and also offers similar improvements for other set_info RPCs.
2212 Description: excessive CPU usage during initial read phase on client
2213 Details : During the initial read phase on a client, it would agressively
2214 retry readahead on the file, consuming too much CPU and impacting
2215 performance (since 1.4.5.8). Improve the readahead algorithm
2216 to avoid this, and also improve some other common cases (read
2217 of small files in particular, where "small" is files smaller than
2218 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2223 Description: MDS crash when receiving packet with unknown intent.
2224 Details : Do not LBUG in unknown intent case, just return -EFAULT
2226 Severity : enhancement
2227 Bugzilla : 9293, 9385
2228 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2229 Details : Do not serialize getattr (non-intent version) and statfs.
2232 Frequency : occasional, when OST network is overloaded/intermittent
2234 Description: client evicted by OST after bulk IO timeout
2235 Details : If a client sends a bulk IO request (read or write) the OST
2236 may evict the client if it is unresposive to its data GET/PUT
2237 request. This is incorrect if the network is overloaded (takes
2238 too long to transfer the RPC data) or dropped the OST GET/PUT
2239 request. There is no need to evict the client at all, since
2240 the pinger and/or lock callbacks will handle this, and the
2241 client can restart the bulk request.
2244 Frequency : Always when mmapping file with no objects
2246 Description: client crashes when mmapping file with no objects
2247 Details : Check that we actually have objects in a file before doing any
2248 operations on objects in ll_vm_open, ll_vm_close and
2254 Description: Request leak when working with deleted CWD
2255 Details : Introduce advanced request refcount tracking for requests
2256 referenced from lustre intent.
2258 Severity : Enhancement
2260 Description: Cache open file handles on client.
2261 Details : MDS now will return special lock along with openhandle, if
2262 requested and client is allowed to hold openhandle, even if unused,
2263 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2264 closing files for every read/write openration.
2266 Severity : Enhancement
2268 Description: Cache open negative dentries on client when possible.
2269 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2270 negative dentries on lock revocation.
2275 Description: Remounting a client read-only wasn't possible with a zconf mount
2276 Details : It wasn't possible to remount a client read-only with llmount.
2278 Severity : enhancement
2279 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2280 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2281 MPICH is included to add Lustre-specific ADIO interfaces.
2282 This is based closely on the UFS ADIO layer and only differs
2283 in file creation, in order to allow the OST striping to be set.
2284 This is user-contributed code and not supported by CFS.
2289 Description: extended inode attributes (immutable, append-only) work improperly
2290 when 2.4 and 2.6 kernels are used on client/server or vice versa
2291 Details : Introduce kernel-independent values for these flags.
2293 Severity : enhancement
2296 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2297 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2298 granularity. Now, Fractional MB granularity is be supported,
2299 this is very useful for low memory system.
2301 Severity : enhancement
2303 Description: Getattr by fid
2304 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2305 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2306 have locks and avoids getattr rpc requests for it on every stat.
2309 Frequency : Always, for filesystems larger than 2TB
2311 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2312 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2313 filesystems larger than 2TB to be mounted with mballoc enabled.
2316 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2318 Description: filesystem corruption for non-standard kernels and very large OSTs
2319 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2320 filesystems larger than 2TB is mounted then the kernel will
2321 silently corrupt the start of the filesystem. CONFIG_LBD is
2322 enabled for all CFS-supported kernels, but the possibility of
2323 this happening with a modified kernel config exists.
2325 Severity : enhancement
2327 Description: add client O_DIRECT support for 2.6 kernels
2328 Details : It is now possible to do O_DIRECT reads and writes to files
2329 in the Lustre client mountpoint on 2.6 kernel clients.
2331 Severity : enhancement
2333 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2334 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2335 parallel, not waiting for response from every OST before sending
2336 a rpc to the next OST.
2341 Description: setattr vs write race when updating file timestamps
2342 Details : Client processes that update a file timestamp into the past
2343 right after writing to the file (e.g. tar) it is possible that
2344 the updated file modification time can be reset to the current
2345 time due to a race between processing the setattr and write RPC.
2347 Severity : enhancement
2349 Description: Bring 'lfs find' closer in line with regular Linux find.
2350 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2351 -print0 options and obtains all the needed info through the lustre
2354 Severity : enhancement
2356 Description: support up to 1024 configured devices on one node
2357 Details : change obd_dev array from statically allocated to dynamically
2358 allocated structs as they are first used to reduce memory usage
2363 Description: Flush dirty partially truncated pages during truncate
2364 Details : Immediatelly flush partially truncated pages in filter_setattr,
2365 this way we completely avoid having any pages in page cache on OST
2366 and can retire ugly workarounds during writes to flush such pages.
2371 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2372 Details : protect inode from truncation within vfs_unlink() context
2373 just take a reference before calling vfs_unlink() and release it
2374 when parent's i_sem is free.
2377 Frequency : always, if extents are used on OSTs
2379 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2380 Details : OSTs using the extents format would not zero the high 16 bits of
2381 the index physical block number. This is not a problem for any
2382 OST filesystems smaller than 16TB, and no kernels support ext3
2383 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2384 new/modified files) and can be fixed for existing filesystems
2385 with e2fsprogs-1.39-cfs1.
2390 Description: import connection selection may be incorrect if timer wraps
2391 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2392 import connections to be ignored if the 32-bit jiffies counter
2393 wraps. Use a 64-bit jiffies counter.
2396 Frequency : during server recovery
2398 Description: crash after server is denying duplicate export
2399 Details : If clients are resending connect requests to the server, the
2400 server refuses to allow a client to connect multiple times.
2401 Fixed a bug in the handling of this case.
2404 Frequency : very large clusters immediately after boot
2406 Description: LNET request buffers exhausted under heavy short-term load
2407 Details : If a large number of client requests are generated on a service
2408 that has previously never seen so many requests it is possible
2409 that the request buffer growth cannot keep up with the spike in
2410 demand. Instead of dropping incoming requests, they are held in
2411 the LND until the RPC service can accept more requests.
2414 Frequency : Sometimes during replay
2416 Description: Assertion failure in ll_local_open after replay.
2417 Details : If replay happened on an open request reply before we were able
2418 to set replay handler, reply will become not swabbed tripping the
2419 assertion in ll_local_open. Now we set the handler right after
2420 recognising of open request
2423 Frequency : very rare
2425 Description: kernel reports "badness in vsnprintf"
2426 Details : Reading from the "recovery_status" /proc file in small chunks
2427 may cause a negative length in lprocfs_obd_rd_recovery_status()
2428 call to vsnprintf() (which is otherwise harmless). Exit early
2429 if there is no more space in the output buffer.
2431 Severity : enhancement
2433 Description: clear OBD RPC statistics by writing to them
2434 Details : It is now possible to clear the OBD RPC statistics by writing
2435 to the "stats" file.
2440 Description: Client mtime is not the same on different clients after utimes
2441 Details : In some cases, the client was using the utimes() syscall on
2442 a file cached on another node. The clients now validate the
2443 ctime from the MDS + OSTs to determine which one is right.
2448 Description: Inability to activate failout mode
2449 Details : lconf script incorrectly assumed that in python string's numeric
2450 value is used in comparisons.
2453 Frequency : always with multiple stripes per file
2455 Description: Inefficient object allocation for mutli-stripe files
2456 Details : When selecting which OSTs to stripe files over, for files with
2457 a stripe count that divides evenly into the number of OSTs,
2458 the MDS is always picking the same starting OST for each file.
2459 Return the OST selection heuristic to the original design.
2464 Description: mount failures may take full timeout to return an error
2465 Details : Under some heavy load conditions it is possible that a
2466 failed mount can wait for the full obd_timeout interval,
2467 possibly several minutes, before reporting an error.
2468 Instead return an error as soon as the status is known.
2470 ------------------------------------------------------------------------------
2472 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
2474 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2475 INTEROPERATE with older versions automatically. Please read the
2476 user documentation before upgrading any part of a live system.
2477 * WARNING: Lustre networking configuration changes are required with
2478 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2481 * Support for kernels:
2482 2.6.9-22.0.2.EL (RHEL 4)
2483 2.6.5-7.244 (SLES 9)
2484 2.6.12.6 vanilla (kernel.org)
2487 Severity : enhancement
2488 Bugzilla : 7981/8208
2489 Description: Introduced Lustre Networking (LNET)
2490 Details : LNET is new networking infrastructure for Lustre, it includes
2491 a reorganized network configuration mode (see the user
2492 documentation for full details) as well as support for routing
2493 between different network fabrics. Lustre Networking Devices
2494 (LNDs) for the supported network fabrics have also been
2495 created for this new infrastructure.
2497 Severity : enhancement
2498 Description: Introduced Access control lists
2499 Details : clients can set ACLs on files and directories in order to have
2500 more fine-grained permissions than the standard Unix UGO+RWX.
2501 The MDS must be started with the "-o acl" mount option.
2503 Severity : enhancement
2504 Description: Introduced filesystem quotas
2505 Details : Administrators may now establish per-user quotas on the
2508 Severity : enhancement
2510 Description: Configuration change for the XT3
2511 The PTLLND is now used to run Lustre over Portals on the XT3
2512 The configure option(s) --with-cray-portals are no longer used.
2513 Rather --with-portals=<path-to-portals-includes> is used to
2514 enable building on the XT3. In addition to enable XT3 specific
2515 features the option --enable-cray-xt3 must be used.
2520 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2521 Details : A race between allocating a new llcd and re-getting the llcd_lock
2522 allowed another thread to grab newly-allocated llcd.
2524 Severity : enhancement
2526 Description: 2.6 OST async journal commit and locking fix to improve performance
2527 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2528 2.6 kernels are now async as they already were in 2.4 kernels so
2529 that they can commit concurrently with the network bulk transfer.
2530 For block-allocated files the filter allocation semaphore is held
2531 to avoid filesystem fragmentation during allocation. BKL lock
2532 removed for 2.6 xattr operations where it is no longer needed.
2537 Description: lconf incorrectly determined whether two IP networks could talk
2538 Details : In some more complicated routing and multiple-network
2539 configurations, lconf will avoid trying to make a network
2540 connection to a disjoint part of the IP space. It was doing the
2541 math incorrectly for one set of cases.
2546 Description: Fix for potential infinite loop processing records in an llog.
2547 Details : If an llog record is corrupted/zeroed, it is possible to loop
2548 forever in llog_process(). Validate the llog record length
2549 and skip the remainder of the block on error.
2552 Frequency : occasional (liblustre only)
2554 Description: liblustre could not open files whose last component is a symlink
2555 Details : sysio_path_walk() would incorrectly pass the open intent to
2556 intermediate path components.
2559 Frequency : rare (liblustre only with non-standard tuning)
2560 Bugzilla : 7201 (7350)
2561 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2562 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2563 after it has been released, only do this for referenced locks
2565 Severity : enhancement
2567 Description: specifying an (invalid) directory default stripe_size of -1
2568 would reset the directory default striping
2569 Details : stripe_size -1 was used internally to signal directory stripe
2570 removal, now use "all default" to signal dir stripe removal
2571 as a directory striping of "all default" is not useful
2574 Frequency : common for large clusters running liblustre clients
2576 Description: doing an ls when liblustre clients are running is slow
2577 Details : sending a glimpse AST to a liblustre client waits for every AST
2578 to time out, as liblustre clients will not respond. Since they
2579 cannot cache data we refresh the OST lock LVB from disk instead.
2581 Severity : enhancement
2583 Description: doing an ls at the same time as file IO can be slow
2584 Details : enqueue and other "small" requests can be blocked behind many
2585 large IO requests. Create a new OST IO portal for non-IO
2586 requests so they can be processed faster.
2589 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2591 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2592 created file with mode 000
2593 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2594 allow an ftruncate() as a non-root user to a file with mode 000.
2595 Lustre now respects this flag to disable mode checks when
2596 truncating a file owned by the user
2599 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2602 Description: Revoking locks from clients that went dead or catatonic might take
2604 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2605 cancellation of such locks instant on servers without waiting for
2606 any reply from clients. Clients drops these locks when cancel
2607 notification from server is received without replying.
2610 Frequency : liblustre-only, when liblustre client dies or becomes busy
2612 Description: Doing ls on Linux clients can take a long time with active
2614 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2615 granting such locks to it in the first place if possible. Locks
2616 are taken by proxy on the OST during the read or write and
2617 dropped immediately afterward. Add connect flags handling, do
2618 not grant locks to liblustre clients for glimpse ASTs.
2620 Severity : enhancement
2622 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2623 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2624 sequential reads and behaves badly with applications doing random
2625 reads. Improve it by reading ahead at least read region, and
2626 avoiding excessive large RPC for small reads.
2628 Severity : enhancement
2630 Description: Creating more than 1000 files for a single job may cause a load
2631 imbalance on the OSTs if there are also a large number of OSTs.
2632 Details : qos_prep_create() uses an OST index reseed value that is an
2633 even multiple of the number of available OSTs so that if the
2634 reseed happens in the middle of the object allocation it will
2635 still utilize the OSTs as uniformly as possible.
2640 Description: OST or MDS may oops in ping_evictor_main()
2641 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2642 but doesn't restart at beginning of obd_exports_timed list
2645 Severity : enhancement
2647 Description: improve by-nid export eviction on the MDS and OST
2648 Details : allow multiple exports with the same NID to be evicted at one
2649 time without re-searching the exports list.
2652 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2654 Description: MDS may oops in groups_free()
2655 Details : in rare race conditions a newly allocated group_info struct is
2656 freed again, and this can be NULL. The 2.4 compatibility code
2657 for groups_free() checked for a NULL pointer, but 2.6 did not.
2660 Frequency : common for liblustre clients doing little filesystem IO
2661 Bugzilla : 9352, 7313
2662 Description: server may evict liblustre clients accessing contended locks
2663 Details : if a client is granted a lock or receives a completion AST
2664 with a blocking AST already set it would not reply to the AST
2665 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2668 Frequency : lfs setstripe, only systems with more than 160 OSTs
2670 Description: unable to set striping with a starting offset beyond OST 160
2671 Details : llapi_create_file() incorrectly limited the starting stripe
2672 index to the maximum single-file stripe count.
2675 Frequency : LDAP users only
2677 Description: lconf did not handle in-kernel recovery with LDAP properly
2678 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2680 Severity : enhancement
2682 Description: bind OST threads to NUMA nodes to improve performance
2683 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2684 node and do their allocations there to localize memory access
2686 Severity : enhancement
2688 Description: llmount can determine client NID directly from Myrinet (GM)
2689 Details : the client NID code from gmnalnid was moved directly into
2690 llmount, removing the need to use this or specifying the
2691 client NID explicitly when mounting GM clients with zeroconf
2694 Frequency : if client is started with down MDS
2696 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2697 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2698 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2699 ptlrpc_import_delayed_req() because we didn't check if the
2700 request was interrupted, and we also didn't break out of the
2701 event loop if there was no timeout
2706 Description: data loss during non-page-aligned writes to a single file from
2707 both multiple nodes and multiple threads on one node at same time
2708 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2709 inconsistency led to false short-reads, that were cached and later
2710 used by ->prepare_write() to fill in partially written page,
2711 leading to data loss.
2714 Frequency : always, if lconf --abort_recovery used
2716 Description: lconf --abort_recovery fails with 'Operation not supported'
2717 Details : lconf was attempting to abort recovery on the MDT device and not
2720 Severity : enhancement
2722 Description: remove cleanup logs
2723 Details : replace lconf-generated cleanup logs with lustre internal
2724 cleanup routines. Eliminates the need for client-cleanup and
2727 Severity : enhancement
2729 Description: add support for EAs (user and system) on lustre filesystems
2730 Details : it is now possible to store extended attributes in the Lustre
2731 client filesystem, and with the user_xattr mount option it
2732 is possible to allow users to store EAs on their files also
2734 Severity : enhancement
2736 Description: Add possibility (config option) to show minimal available OST free
2738 Details : When compiled with --enable-mindf configure option, statfs(2)
2739 (and so, df) will return least minimal free space available from
2740 all OSTs as amount of free space on FS, instead of summary of
2741 free spaces of all OSTs.
2743 Severity : enhancement
2745 Description: do not expand extent locks acquired on OST-side
2746 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2747 by server: they are not cached anyway.
2750 Frequency : when mmap is used/binaries executed from Lustre
2752 Description: Unmmap pages before throwing them away from read cache.
2753 Details : llap_shrink cache now attempts to unmap pages before discarding
2754 them (if unmapping failed - do not discard). SLES9 kernel has
2755 extra checks that trigger if this unmapping is not done first.
2760 Description: lconf didn't resolve symlinks before checking to see whether a
2761 given mountpoint was already in use
2764 Frequency : when migrating failover services
2765 Bugzilla : 6395, 9514
2766 Description: When migrating a subset of services from a node (e.g. failback
2767 from a failover service node) the remaining services would
2768 time out and evict clients.
2769 Details : lconf --force (implied by --failover) sets the global obd_timeout
2770 to 5 seconds in order to quickly disconnect, but this caused
2771 other RPCs to time out too quickly. Do not change the global
2772 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2774 Severity : enhancement
2775 Frequency : if MDS is started with down OST
2776 Bugzilla : 9439,5706
2777 Description: Allow startup/shutdown of an MDS without depending on the
2778 availability of the OSTs.
2779 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2780 Add appropriate locking and lov-osc refcounts for safe
2781 cleaning. Add osc abort_inflight calls in case the
2782 synchronize never started.
2785 Frequency : occasional (Cray XT3 only)
2787 Description: root not authorized to access files in CRAY_PORTALS environment
2788 Details : The client process capabilities were not honoured on the MDS in
2789 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2790 been accessed by an authorized user then root was able to access
2791 the file on the local client also. The root user capabilities
2792 are now allowed on the MDS, as this environment has secure UID.
2795 Frequency : occasional
2797 Description: ldiskfs "too long searching" message happens too often
2798 Details : A debugging message (otherwise harmless) prints too often on
2799 the OST console. This has been reduced to only happen when
2800 there are fragmentation problems on the filesystem.
2805 Description: Division by zero in statfs when all OSCs are inactive
2806 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2807 lov_statfs divides by value returned by lov_get_stripecnt().
2811 Bugzilla : 9489, 3273
2812 Description: First write from each client to each OST was only 4kB in size,
2813 to initialize client writeback cache, which caused sub-optimal
2814 RPCs and poor layout on disk for the first writen file.
2815 Details : Clients now request an initial cache grant at (re)connect time
2816 and so that they can start streaming writes to the cache right
2817 away and always do full-sized RPCs if there is enough data.
2818 If the OST is rebooted the client also re-establishes its grant
2819 so that client cached writes will be honoured under the grant.
2824 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2825 Details : Now I/O RPCs go to different portal number and (presumably) fast
2826 lock requests (and glimses) and other RPCs get their own service
2827 threads pool that should be able to service those RPCs
2830 Severity : enhancement
2832 Description: Ability to exchange lustre version between client and servers and
2833 issue warnings at client side if client is too old. Also for
2834 liblustre clients there is ability to refuse connection of too old
2836 Details : New 'version' field is added to connect data structure that is
2837 filled with version info. That info is later checked by server and
2841 Frequency : rare, liblustre only.
2842 Bugzilla : 9296, 9581
2843 Description: Two simultaneous writes from liblustre at offset within same page
2844 might proceed at the same time overwriting eachother with stale
2846 Details : I/O lock withing llu_file_prwv was released too early, before data
2847 actually was hitting the wire. Extended lock-holding time until
2848 server acknowledges receiving data.
2851 Frequency : extremely rare. Never observed in practice.
2853 Description: avoid generating lustre_handle cookie of 0.
2854 Details : class_handle_hash() generates handle cookies by incrementing
2855 global counter, and can hit 0 occasionaly (this is unlikely, but
2856 not impossible, because initial value of cookie counter is
2857 selected randonly). Value of 0 is used as a sentinel meaning
2858 "unassigned handle" --- avoid it. Also coalesce two critical
2859 sections in this function into one.
2861 Severity : enhancement
2863 Description: allow liblustre clients to delegate truncate locking to OST
2864 Details : To avoid overhead of locking, liblustre client instructs OST to
2865 take extent lock in ost_punch() on client's behalf. New connection
2866 flag is added to handle backward compatibility.
2868 Severity : enhancement
2869 Bugzilla : 4928, 7341, 9758
2870 Description: allow number of OST service threads to be specified
2871 Details : a module parameter allows the number of OST service threads
2872 to be specified via "options ost ost_num_threads={N}" in the
2873 OSS's /etc/modules.conf or /etc/modprobe.conf.
2877 Bugzilla : 6146, 9635, 9895
2878 Description: servers crash with bad pointer in target_handle_connect()
2879 Details : In rare cases when a client is reconnecting it was possible that
2880 the connection request was the last reference for that export.
2881 We would temporarily drop the export reference and get a new
2882 one, but this may have been the last reference and the export
2883 was just destroyed. Get new reference before dropping old one.
2885 Severity : enhancement
2886 Frequency : if client is started with failover MDS
2888 Description: Allow multiple MDS hostnames in the mount command
2889 Details : Try to read the configuration from all specified MDS
2890 hostnames during a client mount in case the "primary"
2893 Severity : enhancement
2895 Description: Stop sending data to evicted clients as soon as possible.
2896 Details : Check if the client we are about to send or are sending data to
2897 was evicted already. (Check is done every second of waiting,
2898 for which l_wait_event interface was extended to allow checking
2899 of exit condition at specified intervals).
2902 Frequency : rare, normally only when NFS exporting is done from client
2904 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2906 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2909 Frequency : always, if srand() is called before liblustre initialization
2911 Description: Liblustre uses system PRNG disturbing its usage by user application
2912 Details : Introduce internal to lustre fast and high-quality PRNG for
2913 lustre usage and make liblustre and some other places in generic
2914 lustre code to use it.
2916 Severity : enhancement
2917 Bugzilla : 9477, 9557, 9870
2918 Description: Verify that the MDS configuration logs are updated when xml is
2919 Details : Check if the .xml configuration logs are newer than the config
2920 logs stored on the MDS and report an error if this is the case.
2921 Request --write-conf, or allow starting with --old_conf.
2923 Severity : enhancement
2925 Description: Handle symlinks in the path when checking if Lustre is mounted.
2926 Details : Resolve intermediate symlinks when checking if a client has
2927 mounted a filesystem to avoid duplicate client mounts.
2932 Description: lconf can hit an error exception but still return success.
2933 Details : The lconf command catches the Command error exception at the top
2934 level script context and will exit with the associated exit
2935 status, but doesn't ensure that this exit status is non-zero.
2940 Description: failure of ptlrpc thread startup can cause oops
2941 Details : Starting a ptlrpc service thread can fail if there are a large
2942 number of threads or the server memory is very fragmented.
2943 Handle this without oopsing.
2946 Frequency : always, only if liblustre and non-default acceptor port was used
2948 Description: liblustre cannot connect to servers with non-default acceptor port
2949 Details : tcpnal_set_default_params() was not called and was therefore
2950 ignoring the environment varaible TCPNAL_PORT, as well as other
2951 TCPNAL_ environment variables
2956 Description: two objects could be created on the same OST for a single file
2957 Details : If an OST is down, in some cases it was possible to create two
2958 objects on a single OST for a single file. No problems other
2959 than potential performance impact and spurious error messages.
2963 Bugzilla : 5681, 9562
2964 Description: Client may oops in ll_unhash_aliases
2965 Details : Client dcache may become inconsistent in race condition.
2966 In some cases "getcwd" can fail if the current directory is
2972 Description: Inode refcounting problems in NFS export code
2973 Details : link_raw functions used to call d_instantiate without obtaining
2974 extra inode reference first.
2978 Bugzilla : 9942, 9903
2979 Description: Referencing freed requests leading to crash, memleaks with NFS.
2980 Details : We used to require that call to ll_revalidate_it was always
2981 followed by ll_lookup_it. Also with revalidate_special() it is
2982 possible to call ll_revalidate_it() twice for the same dentry
2983 even if first occurence returned success. This fix changes semantic
2984 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2985 reference on a request referred from the intent.
2986 ll_intent_release() then releases such a request.
2989 Frequency : rare, normally benchmark loads only
2991 Description: unlinked inodes were kept in memory on the client
2992 Details : If a client is repeatedly creating and unlinking files it
2993 can accumulate a lot of stale inodes in the inode slab cache.
2994 If there is no other client load running this can cause the
2995 client node to run out of memory. Instead flush old inodes
2996 from client cache that have the same inode number as a new inode.
2999 Frequency : SLES9 2.6.5 kernel and long filenames only
3000 Bugzilla : 9969, 10379
3001 Description: utime reports stale NFS file handle
3002 Details : SLES9 uses out-of-dentry names in some cases, which confused
3003 the lustre dentry revalidation. Change it to always use the
3007 Frequency : rare, unless heavy write-truncate concurrency is continuous
3008 Bugzilla : 4180, 6984, 7171, 9963, 9331
3009 Description: OST becomes very slow and/or deadlocked during object unlink
3010 Details : filter_destroy() was holding onto the parent directory lock
3011 while truncating+unlinking objects. For very large objects this
3012 may block other threads for a long time and slow overall OST
3013 responsiveness. It may also be possible to get a lock ordering
3014 deadlock in this case, or run out of journal credits because of
3015 the combined truncate+unlink. Solution is to do object truncate
3016 first in one transaction without parent lock, and then do the
3017 final unlink in a new transaction with the parent lock. This
3018 reduces the lock hold time dramatically.
3021 Frequency : rare, 2.4 kernels only
3023 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3024 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3025 before unmounting their filesystem in order to avoid deadlock.
3026 The kernel_locked() function in 2.4 kernels only checks whether
3027 the kernel lock is held, not whether it is this process that is
3028 holding it as 2.6 kernels do.
3033 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3034 Details : The client ptlrpc code may be trying to reconnect to a down
3035 server before a previous connection attempt has timed out.
3036 Increase the reconnect interval to be longer than the connection
3037 timeout interval to avoid sending duplicate connections to
3041 Frequency : echo_client brw_test command
3043 Description: fix echo_client to work with OST preallocated code
3044 Details : OST preallocation code (5137) didn't take echo_client IO path
3045 into account: echo_client calls filter methods outside of any
3046 OST thread and, hence, there is no per-thread preallocated
3047 pages and buffers to use. Solution: hijack pga pages for IO. As
3048 a byproduct, this avoids unnecessary data copying.
3052 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3053 Description: Client can oops in mdc_commit_close() after open replay
3054 Details : It was possible for the MDS to return an open request with no
3055 transaction number in mds_finish_transno() if the client was
3056 evicted, but without actually returning an error. Clients
3057 would later try to replay that open and may trip an assertion
3058 Simplify the client close codepath, and always return an error
3059 from the MDS in case the open is not successful.
3062 Frequency : rare, 2.6 OSTs only
3064 Description: OST may deadlock under high load on fragmented files
3065 Details : If there was a heavy load and highly-fragmented OST filesystems
3066 it was possible to have all the OST threads deadlock waiting on
3067 allocation of biovecs, because the biovecs were not released
3068 until the entire RPC IO was completed. Instead, release biovecs
3069 as soon as they are complete to ensure forward IO progress.
3071 Severity : enhancement
3073 Description: Support for specifying external journal device at mount
3074 Details : If an OST or MDS device is formatted with an external journal
3075 device, this device major/minor is stored in the ext3 superblock
3076 and may not be valid for failover. Allow detecting and
3077 specifying the external journal at mount time.
3082 Description: Mounting an MDS with pending unlinked files may cause oops
3083 Details : target_finish_recovery() calls mds_postrecov() which returned
3084 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3085 considers this an error and immediately begins cleaning up the
3086 lov, just after starting the mds_lov process
3088 Severity : enhancement
3090 Description: Implement 'lfs df' to report actual free space on per-OST basis
3091 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3092 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3093 usage of objects; '-h' to report in human readable format.
3095 ------------------------------------------------------------------------------
3097 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
3104 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3105 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3106 at startup during orphan unlinking. Instead, skip buddy update
3107 before setup, as it will be regenerated after recovery is complete.
3112 Description: If an OST is inactive, its locks might reference stale inodes.
3113 Details : lov_change_cbdata() must iterate over all namespaces, even if
3114 they are inactive to clear inode references from the lock.
3116 Severity : enhancement
3117 Frequency : occasional, if non-standard max_dirty_mb used
3119 Description: Client will block write RPCs if not enough grant
3120 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3121 then the client will block writes while waiting for another RPC
3122 to complete instead of consuming its dirty limit. With change
3123 we get improved performance when max_dirty_mb is small.
3125 Severity : enhancement
3126 Bugzilla : 3389, 6253
3127 Description: Add support for supplementary groups on the MDS.
3128 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3129 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3130 lookups for user supplementary groups into a cache.
3134 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3135 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3136 for liblustre O_EXCL is enforced by MDS.
3140 Description: readdir never returns NULL in liblustre.
3141 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3142 dentry in struct dirent.
3146 Frequency : liblustre only, depends on application IO pattern
3147 Description: liblustre clients evicted if not contacting servers
3148 Details : Don't put liblustre clients into the ping_evictor list, so
3149 they will not be evicted by the pinger ever.
3151 Severity : enhancement
3153 Description: Add ability to evict clients by NID from MDS.
3154 Details : By echoing "nid:$NID" string into
3155 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3156 $NID would be instantly evicted from this MDS and from all active
3157 OSTs connected to it.
3161 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3162 Details : lookup_it_finish() used to query file size from OSTs that was not
3167 Description: service threads change working directory to that of init
3168 Details : Starting lustre service threads may pin the working directory
3169 of the parent thread, making that filesystem busy. Threads
3170 now change to the working directory of init to avoid this.
3174 Frequency : during shutdown only
3175 Description: shutdown with a failed MDS or OST can cause unmount to hang
3176 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3181 Frequency : 2.6 only
3182 Description: chmod/chown may include an extra supplementary group
3183 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3184 supplementary group and if none is specified this is used.
3187 Bugzilla : 5479 (6816)
3189 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3190 Details : If lookup is in progress on a file that is unlinked we might try
3191 to revalidate the inode and fail in revalidate after lookup is
3192 complete and ll_file_open() enqueues the open again but
3193 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3196 Frequency : always, if lconf --abort_recovery used
3198 Description: lconf --abort_recovery fails with 'Operation not supported'
3199 Details : lconf was attempting to abort recovery on the MDT device and not
3202 ------------------------------------------------------------------------------
3204 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3209 Frequency : rare (only unsupported configurations with a node running as an
3211 Bugzilla : 6514, 5137
3212 Description: Mounting a Lustre file system on a node running as an OST could
3214 Details : OSTs now preallocates memory needed to write out data at
3215 startup, instead of when needed, to avoid having to
3216 allocate memory in possibly low memory situations.
3217 Specifically, if the file system is mounted on on OST,
3218 memory pressure could force it to try to write out data,
3219 which it needed to allocate memory to do. Due to the low
3220 memory, it would be unable to do so and the node would
3221 become unresponsive.
3223 Severity : enhancement
3225 Description: Addition of lconf --service command line option
3226 Details : lconf now accepts a '--service <arg>' option, which is
3227 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3229 Severity : enhancement
3231 Description: Failover mode is now the default for OSTs.
3232 Details : By default, OSTs will now run in failover mode. To return to
3233 the old behaviour, add '--failout' to the lmc line for OSTs.
3235 Severity : enhancement
3237 Description: Health checks are now provided for MDS and OSTs
3238 Details : Additional detailed health check information on MSD and OSTs
3239 is now provided through the procfs health_check value.
3242 Frequency : occasional, depends on IO load
3244 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3245 after numerous create/delete cycles
3246 Details : The ext3 inode allocation policy would not allocate new inodes
3247 very well on the OSTs because there are no new directories
3248 being created. Instead we look for groups with free space if
3249 the parent directories are nearly full.
3254 Description: Network or server problems during mount may cause partially
3255 mounted clients instead of returning an error.
3256 Details : The config llog parsing code may overwrite the error return
3257 code during mount error handling, returning success instead
3263 Description: MDS can fail to allocate large reply buffers
3264 Details : After long uptimes the MDS can fail to allocate large reply
3265 buffers (e.g. zconf client mount config records) due to memory
3266 fragmentation or consumption by the buffer cache. Preallocate
3267 some large reply buffers so that these replies can be sent even
3268 under memory pressure.
3272 Frequency : rare (liblustre)
3273 Description: fsx running with liblustre complained that using truncate() to
3274 extend the file doesn't work. This patch corrects that issue.
3275 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3276 ATTR_SIZE and lsm use in llu_setattr_raw.
3280 Frequency : rare, only 2.6 kernels
3281 Description: Unusual file access patterns on the MDS may result in inode
3282 data being lost in very rare circumstances.
3283 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3284 code in the 2.6 kernel caused the inode and/or EA data not to
3285 be read from disk, causing single-file corruption.
3289 Frequency : rare, only 2.6 filesystems using extents
3290 Description: Heavy concurrent write and delete load may cause data corruption.
3291 Details : It was possible under high-load situations to have an extent
3292 metadata block in the block device cache from a just-unlinked
3293 file overwrite a newly-allocated data block. We now unmap any
3294 metadata buffers that alias just-allocated data blocks.
3298 Frequency : filesystems with default stripe_count larger than 77
3299 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3300 Details : lconf specifies an inode size of 4096 bytes when the default
3301 stripe_count is larger than 77. This conflicts with the default
3302 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3303 this case to avoid pinning too much memory for large EAs.
3305 ------------------------------------------------------------------------------
3307 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3312 Frequency : rare (extremely heavy IO load with hundreds of clients)
3314 Description: Client is evicted, gets IO error writing to file
3315 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3316 caused clients to be evicted for AST timeouts. The fixes in
3317 bug 5192 mean we no longer need to have such short AST timeouts
3318 so ldlm_timeout has been increased.
3321 Frequency : occasional during --force or --failover shutdown under load
3322 Bugzilla : 5949, 4834
3323 Description: Server oops/LBUG if stopped with --force or --failover under load
3324 Details : a collection of import/export refcount and cleanup ordering
3325 issues fixed for safer force cleanup
3328 Frequency : only filesystems larger than 120 OSTs
3329 Bugzilla : 5990, 6223
3330 Description: lfs getstripe would oops on a very large filesystem
3331 Details : lov_getconfig used kfree on vmalloc'd memory
3334 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3335 Bugzilla : 6242, 6243
3336 Description: reading from files that had been truncated to a non-zero size
3337 but never opened returned no data
3338 Details : ll_file_read() reads zeros from no-object files to EOF
3343 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3344 Details : There's little harm in aborting MDS/OSS recovery and letting it
3345 try again, so I removed the LASSERT and return an error instead.
3347 Severity : enhancement
3349 Description: New debugging infrastructure for tracking down data corruption
3350 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3351 (b) cover more of the client-side code path, and (c) try to narrow
3352 down where problems occurred
3356 Bugzilla : 3819, 4364, 4397, 6313
3357 Description: Racing close and eviction MDS could cause assertion in mds_close
3358 Details : It was possible to get multiple mfd references during close and
3359 client eviction, leading to one thread referencing a freed mfd.
3361 Severity: : enhancement
3362 Bugzilla : 3262, 6359
3363 Description: Attempts to reconnect to servers are now more aggressive.
3364 Details : This builds on the enhanced upcall-less recovery that was added
3365 in 1.4.2. When trying to reconnect to servers, clients will
3366 now try each server in the failover group every 10 seconds. By
3367 default, clients would previously try one server every 25 seconds.
3372 Description: After recovery, certain operations trigger a failed
3373 assertion on a client.
3374 Details : Failing over an mds, using lconf -d --failover, while a
3375 client was doing a readdir() call would cause the client to
3376 LBUG after recovery completed and the readdir() was resent.
3378 Severity : enhancement
3380 Description: Default groups are now added by lconf
3381 Details : You can now run lconf --group <servicename> without having to
3382 manually add groups with lmc.
3385 Frequency : occasional
3387 Description: Nodes with an elan id of 0 trigger a failed assertion
3390 Frequency : always when accessing e.g. tty/console device nodes
3392 Description: tty and some other devices nodes cannot be used on lustre
3393 Details : file's private_data field is used by device data and lustre
3394 values in there got lost. New field was added to struct file to
3395 store fs-specific private data.
3398 Frequency : when exporting Lustre via NFS
3400 Description: NFSD failed occasionally when looking up a path component
3401 Details : NFSD is looking up ".." which was broken in ext3 directories
3402 that had grown large enough to become hashed.
3405 Frequency : Clusters with multiple interfaces not on the same subnet
3407 Description: Nodes will repeatedly try to reconnect to an interface which it
3408 cannot reach and report an error to the log.
3409 Details : Extra peer list entries will be created by lconf with some peers
3410 unreachable. lconf now validates the peer before adding it.
3413 Frequency : Only if a default stripe is set on the filesystem root.
3415 Description: Setting a default stripe on the filesystem root prevented the
3416 filesystem from being remounted.
3417 Details : The client was sending extra request flags in the root getattr
3418 request and did not allocate a reply buffer for the dir EA.
3421 Frequency : occasional, higher if lots of files are accessed by one client
3422 Bugzilla : 6159, 6097
3423 Description: Client trips assertion regarding lsm mismatch/magic
3424 Details : While revalidating inodes the VFS looks up inodes with ifind()
3425 and in rare cases can find an inode that is being freed.
3426 The ll_test_inode() code will free the lsm during ifind()
3427 when it finds an existing inode and then the VFS later attaches
3428 this free lsm to a new inode.
3432 Bugzilla : 6422, 7030
3433 Description: MDS deadlock between mkdir and client eviction
3434 Details : Creating a new file via mkdir or mknod (starting a transaction
3435 and getting the ns lock) can deadlock with client eviction
3436 (gets ns lock and trying to finish a synchronous transaction).
3439 Frequency : occasional
3440 Description: While starting a server, the fsfilt_ext3 module could not be
3442 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3443 kernels. Previously, lconf would still use the ext3 name
3444 when trying to load modules. Now, it will correctly use
3445 ext3 on 2.4 and ldiskfs on 2.6.
3447 Severity : enhancement
3448 Description: The default stripe count has been changed to 1
3449 Details : The interpretation of the default stripe count (0, to lfs
3450 or lmc) has been changed to mean striping across a single
3451 OST, rather than all available. For general usage we have
3452 found a stripe count of 1 or 2 works best.
3454 Severity : enhancement
3455 Description: Add support for compiling against Cray portals.
3456 Details : Conditional compiling for some areas that are different
3460 Frequency : occasional
3461 Bugzilla : 6409, 6834
3462 Description: Creating files with an explicit stripe count may lead to
3463 a failed assertion on the MDS
3464 Details : If some OSTs are full or unavailable, creating files may
3465 trigger a failed assertion on the MDS. Now, Lustre will
3466 try to use other servers or return an error to the
3470 Frequency : occasional
3472 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3473 caused lock timeout during readahead (since 1.4.2).
3474 Details : Processes doing readahead might match a lock that hasn't been
3475 granted yet if there are overlapping and conflicting lock
3476 requests. The readahead process waits on ungranted lock
3477 (original lock is CBPENDING), while OST waits for that process
3478 to cancel CBPENDING read lock and eventually evicts client.
3480 Severity : enhancement
3482 Description: Initial enabling of flock support for clients
3483 Details : Implements fcntl advisory locking and file status functions.
3484 This feature is provided as an optional mount flag (default
3485 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3486 locking are implemented yet, and those that are are not guaranteed
3487 to be completely correct in production environments.
3488 mount -t lustre -o [flock|noflock] ...
3491 Frequency : occasional
3493 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3494 assertion in the ext3 JBD (journaling) layer.
3495 Details : The b_committed_data struct is protected by the big kernel lock
3496 in 2.4 kernels, serializing journal_commit_transaction() and
3497 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3498 access to this struct. In 2.6 kernels there is finer grained
3499 locking to improve SMP performance of the JBD layer.
3503 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3505 -----------------------------------------------------------------------------
3507 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3509 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3510 versions of Lustre. Please update all servers and clients to
3511 version 1.4.2 or later at the same time. You must also run
3512 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3513 to update the configuration logs.
3515 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3516 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3517 - swab llog records of type '0' so we get proper header size/idx (5861)
3518 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3519 - fix rename of one directory over another leaking an inode (5953)
3520 - avoid SetPageDirty on 2.6 (5981)
3521 - don't re-add just-being-destroyed locks to the waiting list (5653)
3522 - when creating new directories, inherit the parent's custom
3523 striping settings if present parent (3048)
3524 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3525 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3526 - don't hold client locks on temporary worklist from l_lru (5666)
3527 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3528 - automatically evict dead clients (5921)
3529 - Update file size properly in create+truncate+fstat case (6196)
3530 - Do not unhash mountpoint dentries, do not allow removal of
3532 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3533 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3534 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3535 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3536 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3537 - Correctly handle reads of files with no objects (6243)
3538 - lctl recover will also mark a device active if deactivate used (5933)
3540 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3541 - allow --write-conf on an MDS with different nettype than client (5619)
3542 - don't write config llogs to MDS for mounts not from that MDS (5617)
3543 - lconf should create multiple TCP connections from a client (5201)
3544 - init scripts are now turned off by default; run chkconfig --on
3545 lustre and chkconfig --on lustrefs to use them
3546 - upcalls are no longer needed for clients to recover to failover
3548 - add --abort-recovery option to lconf to abort recovery on device
3550 - add support for an arbitrary number of OSTs (3026)
3551 - Quota support protocol changes.
3552 - forward compatibility changes to wire structs (6007)
3553 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3554 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3555 - improved "lustre status" script
3556 - initialize blocksize for non-regular files (6062)
3557 - added --disable-server and --disable-client configure options (5782)
3558 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3559 - Vanilla 2.4.29 support
3560 - increase maximum number of obd devices to 520 (6242)
3561 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3562 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3563 SLES 9 2.6.5 kernels
3564 - limit stripes per file to 160 (the maximum EA size) (6093)
3566 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3569 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3570 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3571 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3572 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3573 - small ext3 extents cleanups and fixes (5733)
3574 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3575 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3576 - avoid needless client->OST connect, fix handle mismatch (5317)
3577 - fix DLM error path that led to out-of-sync client, long delays (5779)
3578 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3579 - fix several locking issues related to i_size (5492,5624,5654,5672)
3580 - don't move pending lock onto export if it is already evicted (5683)
3581 - fix kernel oops when creating .foo in unlinked directory (5548)
3582 - fix deadlock in obdfilter statistics vs. object create (5811)
3583 - use time_{before,after} to avoid timer jiffies wrap (5882)
3584 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3585 - Do not leak request if server was not able to process it (5154)
3586 - If mds_open unable to find parent dir, make that negative lookup(5154)
3587 - don't create new directories with extent-mapping (5909, 5936)
3589 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3590 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3591 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3592 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3593 - support for RHEL4 kernel on i686 (5773)
3594 - provide error messages when incompatible logs are encountered (5898)
3596 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3597 * version 1.4.0.10 (1.4.1 release candidate 1)
3599 - don't keep a lock reference when lock is not granted (4238)
3600 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3601 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3602 - drop import inflight refcount on signal_completed_replay error (5255)
3603 - unlock page after async write error during send (3677)
3604 - handle missing objects in filter_preprw_read properly (5265)
3605 - no transno return for symlink open, don't save no-trasno open (3440)
3606 - don't try to complete elan receive that already failed (4012)
3607 - free RPC server reply state on error (5406)
3608 - clean up thread from ptlrpc_start_thread() on error (5160)
3609 - readahead could read extra page into cache that wasn't ejected (5388)
3610 - prevent races in class_attach/setup/cleanup/detach (5260)
3611 - don't dereference de->d_inode after l_dput of de (5458)
3612 - use "int" for stripe value returned from lock_to_stripe (5544)
3613 - mballoc allocation and error-checking fixes in 2.6 (5504)
3614 - block device patches to fix I/O request sizes in 2.6 (5482)
3615 - look up hostnames for IB nals (5602)
3616 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3617 - don't start multiple acceptors for the same port (5277)
3618 - fix incorrect LASSERT in mds_getattr_name (5635)
3619 - export a proc file for general "ping" checking (5628)
3620 - fix "lfs check" to not block when the MDS is down (5628)
3622 - service request history (4965)
3623 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3624 - create an "evict_client" /proc entry on OSTs, like the MDS has
3625 - fix mount usage message, return errors per mount(8) (5168)
3626 - change grep [] to grep "[]" in tests so they work in more UMLs
3627 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3628 - remove ancient LOV_MAGIC_V0 EA support (5047)
3629 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3630 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3631 - allow readahead allocations to fail when low on memory (5383)
3632 - mmap locking landed again, after considerable improvement (2828)
3633 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3635 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3638 - send OST transaction number in read/write reply to free req (4966)
3639 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3640 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3641 - the watchdog thread now runs as interruptible (5246)
3642 - flock/lockf fixes (but it's still disabled, pending 5135)
3643 - don't use EXT3 constants in llite code (5094)
3644 - memory shortage at startup could cause assertion (5176)
3646 - reorganization of lov code
3647 - single portals codebase
3649 - add extents/mballoc support (5025)
3650 - direct I/O reads in the obdfilter (4048)
3651 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3653 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3656 - send OST transaction number in read/write reply to free req (4966)
3657 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3658 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3659 - the watchdog thread now runs as interruptible (5246)
3660 - handle missing objects in filter_preprw_read properly (5265)
3661 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3662 - drop import inflight refcount on signal_completed_replay error (5255)
3663 - unlock page after async write error during send (3677)
3664 - return original error code on reconstructed replies (3761)
3665 - no transno return for symlink open, don't save no-trasno open (3440)
3667 - add pid to ldlm debugging output (4922)
3668 - bump the watchdog timeouts -- we can't handle 30sec yet
3669 - extra debugging for orphan dentry/inode bug (5259)
3671 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3674 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3675 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3676 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3677 - fix NULL dereference in /proc/sys/portals/routes (4827)
3678 - allow failed mdc_close() operations to be interrupted (4561)
3679 - stop precreate on OST before MDS would time out on it (4778)
3680 - don't send partial-page writes before EOF from client (4410)
3681 - discard client grant for sub-page writes on large-page clients (4520)
3682 - don't free dentries not owned by NFS code, check generation (4806)
3683 - fix lsm leak if mds_create_objects() fails (4801)
3684 - limit debug_daemon file size, always print CERROR messages (4789)
3685 - use transno after validating reply (3892)
3686 - process timed out requests if import state changes (3754)
3687 - update mtime on OST during writes, return in glimpse (4829)
3688 - add mkfsoptions to LDAP (4679)
3689 - use ->max_readahead method instead of zapping global ra (5039)
3690 - don't interrupt __l_wait_event() during strace
3692 - add software watchdogs to catch hung threads quickly (4941)
3693 - make lustrefs init script start after nfs is mounted
3694 - fix CWARN/ERROR duplication (4930)
3695 - return async write errors to application if possible (2248)
3696 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3697 - print NAL number in %x format (4645)
3698 - update barely-supported suse-2.4.21-171 series (4842)
3699 - support for sles 9 %post scripts
3700 - support for building 2.6 kernel-source packages
3701 - support for sles km_* packages
3703 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3706 - ignore -ENOENT errors in osc_destroy (3639)
3707 - notify osc create thread that OSC is being cleaned up (4600)
3708 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3709 - reconstruct ost_handle() like mds_handle() (4657)
3710 - create a new thread to do import eviction to avoid deadlock (3969)
3711 - let lconf resolve symlinked-to devices (4629)
3712 - don't unlink "objects" from directory with default EA (4554)
3713 - hold socknal file ref over connect in case target is down (4394)
3714 - allow more than 32000 subdirectories in a single directory (3244)
3715 - fix blocks count for O_DIRECT writes (3751)
3716 - OST returns ENOSPC from object create when no space left (4539)
3717 - don't send truncate RPC if file size isn't changing (4410)
3718 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3719 - bind to privileged port in socknal and tcpnal (3689)
3721 - rate limit CERROR/CWARN console message to avoid overload (4519)
3722 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3723 - basic mmap support (3918)
3724 - kernel patch series update from b1_4 (4711)
3726 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3729 - avoid crash during MDS cleanup with OST shut down (2775)
3730 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3731 - don't use bad inodes on the MDS (3744)
3732 - dynamic object preallocation to improve recovery speed (4236)
3733 - don't hold spinlock over lock dumping or change debug flags (4401)
3734 - don't zero obd_dev when it is force cleaned (3651)
3735 - print grants to console if they go negative (4431)
3736 - "lctl deactivate" will stop automatic recovery attempts (3406)
3737 - look for existing locks in ldlm_handle_enqueue() (3764)
3738 - don't resolve lock handle twice in recovery avoiding race (4401)
3739 - revalidate should check working dir is a directory (4134)
3741 - don't always mark "slow" obdfilter messages as errors (4418)
3743 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3746 - don't close LustreDB during write_conf until it is done (3860)
3747 - fix typo in lconf for_each_profile (3821)
3748 - allow dumping logs from multiple threads at one time (3820)
3749 - don't allow multiple threads in OSC recovery (3812)
3750 - fix debug_size parameters (3864)
3751 - fix mds_postrecov to initialize import for llog ctxt (3121)
3752 - replace config semaphore with spinlock (3306)
3753 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3754 - don't allow enqueue to complete on a destroyed export (3822)
3755 - down write_lock before checking llog header bitmap (3825)
3756 - recover from lock replay timeout (3764)
3757 - up llog sem before sending rpc (3652)
3758 - reduce ns lock hold times when setting kms (3267)
3759 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3760 - fix NULL deref and obd_dev leak on setup error (3312)
3761 - replace some LBUG about llog ops with error handling (3841)
3762 - don't match INVALID dentries from d_lookup and spin (3784)
3763 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3764 - fix invalid assertion in ptlrpc_set_wait (3880)
3766 - add libwrap support for the TCP acceptor (3996)
3767 - add /proc/sys/portals/routes for non-root route listing (3994)
3768 - allow setting MDS UUID in .xml (2580)
3769 - print the stack of a process that LBUGs (4228)
3771 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3774 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3775 - make sure to unset replay flag from failed open requests (3440)
3776 - if default stripe count is 0, use OST count for inode size (3636)
3777 - update parent mtime/ctime on client for create/unlink (2611)
3778 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3779 - free recovery state on server during a forced cleanup (3571)
3780 - unregister_reply for resent reqs (3063)
3781 - loop back devices mounting and status check on 2.6 (3563)
3782 - fix resource-creation race that can provoke i_size == 0 (3513)
3783 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3784 - more debugging for page-accounting assertion (3746)
3785 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3786 - avoid deadlock after precreation failure (3758)
3787 - fix race and lock order deadlock in orphan handling (3450, 3750)
3788 - add validity checks when grabbing inodes from l_ast_data (3599)
3790 - add /proc/.../recovery_status to obdfilter (3428)
3791 - lightweight CDEBUG infrastructure, debug daemon (3668)
3792 - change default OSC RPC parameters to be better on small clusters
3793 - turn off OST read cache for files smaller than 32MB
3794 - install man pages and include them in rpms (3100)
3795 - add new init script for (un)mounting lustre filesystems (2593)
3796 - run chkconfig in %post for init scripts (3701)
3797 - drop scimac NAL (unmaintained)
3799 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3802 - clean kiobufs before and after use (3485)
3803 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3804 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3805 - change init script to fail more clearly if not run as root (1528)
3806 - allow clients to reconnect during replay (1742)
3807 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3808 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3809 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3810 - fix cond_resched (3554)
3811 - only evict unfinished clients after recovery (3515)
3812 - allow bulk resend, prevent data loss (3570)
3813 - dynamic ptlrpc request buffer allocation (2102)
3814 - don't allow unlinking open directory if it isn't empty (2904)
3815 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3816 - remove extraneous obd dereference causing LASSERT failure (3334)
3817 - don't use get_cycles() when creating temp. files on the mds (3156)
3818 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3819 - handle EEXIST for set-stripe, set proper directory name (3336)
3821 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3822 - fix ksocknal_fmb_callback() error messages (2918)
3824 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3827 - don't copy lvb into (possibly NULL) reply on error (2983)
3828 - don't deref dentry after dput, don't free lvb on error (2922)
3829 - use the kms to determine writeback rpc length (2947)
3830 - increment oti_logcookies when osc is inactive (2948)
3831 - update client's i_blocks count via lvb messages (2543)
3832 - handle intent open/close of special files properly (1557)
3833 - mount MDS with errors=remount-ro, like obdfilter (2009)
3834 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3835 - don't use cancelling-locks' kms values (2947)
3836 - use highest lock extent for kms, not last one (2925)
3837 - don't dereference ERR_PTR() dentry in error handling path (3107)
3838 - fix thread race in portals_debug_dumplog() (3122)
3839 - create lprocfs device entries at setup instead of at attach (1519)
3840 - common AST error handler, don't evict client on completion race (3145)
3841 - zero nameidata in detach_mnt in 2.6 (3118)
3842 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3843 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3844 - initialize RPC timeout timer earlier for 2.6 (3219)
3845 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3846 - print nal/nid for unknown nid (3258)
3847 - additional checks for oscc recovery before doing precreate (3284)
3848 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3849 - don't crash in mdc_close for bad permissions on open (3285)
3850 - zero i_rdev for non-device files (3147)
3851 - clear page->private before handing to FS, better assertion (3119)
3852 - tune the read pipeline (3236)
3853 - fix incorrect decref of invalidated dentry (2350)
3854 - provide read-ahead stats and refine rpc in flight stats (3328)
3855 - don't hold journal transaction open across create RPC (3313)
3856 - update atime on MDS at close time (3265)
3857 - close LDAP connection when recovering to avoid server load (3315)
3858 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3859 - don't leak open file on MDS after open resend (3325)
3860 - serialize filter_precreate and filter_destroy_precreated (3329)
3861 - loop device shouldn't call sync_dev() for nul device (3092)
3862 - clear page cache after eviction (2766)
3863 - resynchronize MDS->OST in background (2824)
3864 - refuse to mount the same filesystem twice on same mountpoint (3394)
3865 - allow llmount to create routes for mounting behind routers (3320)
3866 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3867 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3868 - fix rare problem with rename on htree directories (3417)
3870 - allow default OST striping configuration per directory (1414)
3871 - fix compilation for qswnal for 2.6 kernels (3125)
3872 - increase maximum number of MDS request buffers for large systems
3873 - change liblustreapi to be useful for external progs like lfsck (3098)
3874 - increase local configuration timeout for slow disks (3353)
3875 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3877 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3880 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3881 - don't overwrite extent policy data in reply if lock was blocked (2901)
3882 - drop filter export grants atomically with removal from device (2663)
3883 - del obd_self_export from work_list in class_disconnect_exports (2908)
3884 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3885 - swab reply message in mdc_close, other PPC fixes (2464)
3886 - fix destroying of named logs (2325)
3887 - overwrite old logs when running lconf --write_conf (2264)
3888 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3889 - fix race in target_handle_connect (2898)
3890 - mds_reint_create() should take same inode create lock (2926)
3891 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3892 - don't close files for self_export to avoid uninitialized obd (2936)
3893 - allow MDS with the same name as client node (2939)
3894 - hold dentry reference for closed log files for unlink (2325)
3895 - reserve space for all logs during transactions (2059)
3896 - don't evict page beyond end of stripe extent (2925)
3897 - don't oops on a deleted current working directory (2399)
3898 - handle hard links to targets without a parent properly (2517)
3899 - don't dereference NULL lock when racing during eviction (2867)
3900 - don't grow lock extents when lots of conflicting locks (2919)
3902 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3905 - account for cache space usage on clients to avoid data loss (974)
3906 - lfsck support in lustre kernel code (2349)
3907 - reduce journal credits needed for BRW writes (2370)
3908 - orphan handling to avoid losing space on client/server crashes
3909 - ptlrpcd can be blocked, stopping ALL progress (2477)
3910 - use lock value blocks to assist in proper KMS, faster stat (1021)
3911 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3912 - recovery for initial connections (2355)
3913 - fixes for mds_cleanup_orphans (1934)
3914 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3915 - block all file creations until orphan recovery completes (1901)
3916 - client remove rq_connection from request struct (2423)
3917 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3918 - recovery timer race (2670)
3919 - mdc_close recovey bug (2532)
3920 - ptlrpc cleanup bug (2710)
3921 - mds timeout on local locks (2588)
3922 - namespace lock held during RPCs (2431)
3923 - handle interrupted sync write properly (2503)
3924 - don't try to handle a message that hasn't been replied to (2699)
3925 - client assert failure during cleanup after abort recovery (2701)
3926 - leak mdc device after failed mount (2712)
3927 - ptlrpc_check_set allows timedout requests to complete (2714)
3928 - wait for inflight reqs when ptlrpcd finishes (2710)
3929 - make sure unregistered services are removed from the srv_list
3930 - reset bulk XID's when resending them (caught by 1138 test)
3931 - unregister_bulk after timeout
3932 - fix lconf error (2694)
3933 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3934 - readahead locks pages, leaves pending causing memory pressure (2673)
3935 - increase OST request buffers to 4096 on large machines (2729)
3936 - fix up permission of existing directories in simple_mkdir (2661)
3937 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3938 - don't assign transno to errored transactions (2742)
3939 - don't delete objects on OST if given a bogus objid from MDS (2751)
3940 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3941 - if rq_no_resend, then timeout request after recovery (2432)
3942 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3943 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3944 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3945 - MDS recovery shouldn't delete objects that it has given out (2730)
3946 - if enqueue arrives after completion, don't clobber LVB (2819)
3947 - don't unlock pages twice when trigger_group_io returns error (2814)
3948 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3949 - don't write pages to disk if there was an error (1450)
3950 - don't ping imports that have recovery disabled (2676)
3951 - take buffered bytes into account when balancing socknal conn (2817)
3952 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3953 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3954 - remove little-endian swabbing of llog records (1987)
3955 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3956 - retry reposting request buffers if they fail (1191)
3957 - grow extent at grant time to avoid granting a revoked lock (2809)
3958 - lock revoke doesn't evict page if covered by a second lock (2765)
3959 - disable VM readahead to avoid reading outside lock extents (2805)
3961 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3962 - updated kernel patches for hp-2.4.20 kernel (2681)
3964 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3967 - fix truncated write corruption (2366)
3968 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3970 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3971 - ensure len(uuid) < 37 in lmc (1171)
3972 - fix ia64 OOPS in llog_test (2255)
3973 - zero end of page at obdfilter for partial page writes (2648)
3974 - don't leave stale dentries around after renames (bug 2428)
3975 - fix timeouts when evicting a client with a single lock held (2642)
3976 - set deadline for the initial HELLO message to drain (2634)
3977 - print out dotted-quad IP addresses in the socknal (2302)
3979 - additional debugging for MDS client eviction problem (2443)
3980 - fix mkfsoptions support for osts (2603, 2604)
3982 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3985 - add series for the vanilla 2.6.0 kernel
3986 - add series for the vanilla 2.4.24 kernel
3987 - add series for a cray x86/64 UL kernel drop
3988 - fix xattr patches for the vanilla 2.4.19 series
3990 - generate true UUIDs in lmc (1171)
3991 - have portals stack dumping break in UML (2466)
3992 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3993 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3994 - wake callback waiting threads on client eviction (2460)
3995 - Add --ptldebug and --subsystem to lmc (1719)
3996 - update assertion to allow safe interrupt allocation
3997 - set rq_no_resend for cancel requests (2432)
3998 - recalculate ptlrpcd timeout after resend (2494)
3999 - call vfs_rmdir when removing pending directories (2368)
4000 - fix renaming a file to itself (2429)
4001 - lmc creates a default one-stripe lov (2454)
4002 - expand procfs space to handle large clusters (2326)
4003 - increase UML stack to avoid overflow
4004 - update lconf's list of debug and subsystem masks
4005 - fix lfs find --obd (2510)
4006 - /proc tunable for disabling filter read caching (2591)
4007 - stop rpm packages from altering slapd.conf (2301)
4008 - disable nagle in the socknal under 0conf (2578)
4009 - choose mds inode size based on stripe count (2572)
4010 - fix kernel-source rpm problems (2516)
4012 - add --disable-doc to avoid pdf generation (2421)
4013 - update documentation, tests, type-os, comments
4014 - avoid format warnings on ia64
4015 - remove the TOE NAL
4016 - tiny code cleanups by removing unused fields
4018 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4021 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4022 - remove the most blatant lies from BUILDING file (2371)
4023 - change default debug level to reasonable production setting
4024 - reduce client side cache size to reduce cache flush time
4025 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4026 - make TCP zerocopy and pinger support enabled by default (2476)
4027 - sync writes completed after process exits caused crashes (2319)
4028 - maintain correct mount count on the MDS (2356)
4029 - backout 1557, because 2316 wasn't really fixed
4030 - better file I/O statistics gathering in /proc
4031 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4032 - another kernel patch to fix zero-copy TCP function export
4033 - don't take duplicate lock when processing re-sent getattr (2420)
4034 - lctl uses obd_self_export instead of creating new conn (2353)
4035 - MDS/OST recovery case which requires object creation asserted (2425)
4036 - move lfs from /usr/sbin to /usr/bin in packages
4037 - fix race between mds_client_add and mds_client_free (2417)
4038 - use kmalloc instead of slabs in portals (2430)
4039 - don't create duplicate records when a failover MDS is present (2442)
4040 - remove unnecessary mount age check (2332)
4041 - don't remove directory inodes from locks prematurely (2451)
4042 - don't break if MDS service name is the same as hostname (2103)
4043 - fix races in client write RPC generation when cache full (2482)
4045 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4048 - remove now-unused request->rq_obd (278)
4049 - if an allocation fails, print out how much memory we've used (1933)
4050 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4051 - add the "configurable stack size" patch to most series files (1256)
4052 - ability to write large log records, for 100+ OST configs (2306)
4053 - fix NULL deref when filter_prep fails (2314)
4054 - fix operator precedence error in filter_sync
4055 - dynamic allocation of socknal TX descriptors (2315)
4056 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4057 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4058 - ensure that CWARN messages in Portals always get to the syslog
4059 - __init/__exit are not for prototype decls (ldlm_init/exit)
4060 - x86-64 compile warning fixes
4061 - fix gateway LMC keyword conflict (2318)
4062 - fix MDS lock inversions in getattr/reint paths (1844)
4063 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4064 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4065 - DLM race condition prevented some lock evictions (2328)
4066 - ENOMEM detection and retry on socknal sends (2230)
4067 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4068 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4069 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4070 - better error messages when a client is rejected during recovery (1505)
4071 - avoid cancelling locks which were never granted, after failure (2330)
4072 - fix i_sem/journal inversion in mds_client_add (2333)
4073 - fix truncate/getattr lock cycle deadlock (2334)
4074 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4075 - fix two rare exit paths which could leak an l_lock() ref (2321)
4076 - fencepost error in MDS/OST orphan recovery (2226)
4077 - make log record alignment 8 bytes (1988)
4078 - lstripe now fails when requested offset > ost_count (2237)
4079 - ensure that all kernel series have a complete list.h (1607)
4080 - fix crashes in special-file operations (2316)
4081 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4083 - allow configurable automake binary, for testing new versions
4084 - small update to the lfs documentation
4086 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4088 * fix negative export reference count in fsfilt_sync (2312)
4090 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4091 * release candidate 0.9.1
4093 - orphans are moved into the PENDING directory for possible recovery
4094 - replayed opens now open by fid for orphan/rename safety (1042)
4095 - last close of an orphan inode generates a transno (683)
4096 - chdir() and mount() now pin the directory entry (1020)
4097 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4098 - discard very old requests without processing them (1502)
4099 - remove some common, well-understood CERRORs (1505)
4100 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4101 - clear "grant" flags in OST replies until OST grant code lands (1644)
4102 - fix read performance by not clobbering i_blksize on client (1598)
4103 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4104 - make LRU size a /proc tunable, clears locks when reduced (707)
4105 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4106 - server locks take references on exports now (1558)
4107 - build fixes for 2.4.20-rh trees (1663)
4108 - return an error from lov_create if all OSCs are inactive (1751)
4109 - fix import levels when a reconnect happens without a timeout (1597)
4110 - exit early from mds_open if we get a lookup error (1749)
4111 - partial page read at EOF wouldn't wait for disk before sending (1642)
4112 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4113 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4114 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4115 - write/truncate lock inversion (1639)
4116 - Don't auto-load obdclass, portals modules during cleanup (1495)
4117 - fix timestamps from jumping to "now" (1763)
4118 - extra journal assertions (1648)
4119 - add an extra multiunlink test (1771)
4120 - fix read_record/write_record API (1776)
4121 - fix leak of offset_extent, possible incorrect i_size later (1772)
4122 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4123 - Debugging for the kqswnal_get_idle_tx problems (1820)
4124 - Allow recovery to be attempted multiple times (1536)
4125 - Write out MDS last_rcvd file after it is first created (1600)
4126 - Fix tx_descriptor leak in failed transmit situations (1827)
4127 - ext3 journaling fixes for assertion failure after IO error (1871)
4128 - class_export_put() on freed export after completion AST error (1896)
4129 - Fix revalidate looping in VFS (1322)
4130 - Don't access a freed export during MDS_REINT timeout (1521)
4131 - Add open-unlink recovery support on the MDS (1673,1764)
4132 - Return an error if no MDS data was read from last_rcvd (1946)
4133 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4134 - Don't setup a disk device that doesn't match exported UUID (317)
4135 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4136 - avoid committing NULL handle in force close
4137 - local.sh is now a one-stripe LOV configuration
4138 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4139 - fix ext3 htree duplicate directory entry corruption (1516)
4140 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4141 - update to new LOV EA format (2097)
4142 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4144 2003-06-15 Phil Schwan <phil@clusterfs.com>
4147 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4148 - per-import/export recovery handling (958, 931, 959)
4149 - multiple last-rcvd slots, for serving multiple FSes (949)
4150 - connections are again shared between multiple imp/exports (963, 964)
4151 - "umount -f" would hang if any requests needed to be sent (393, 978)
4152 - avoid pinning large req buffer by copying for queued messages (989)
4153 - add "uuid" to "lctl device" command to help upcalls (991)
4154 - "open" RPCs with transnos would confuse recovery counters (1037)
4155 - do proper endian conversion of all wire messages (288, 340, 891)
4156 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4157 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4158 - fix ptlrpc_connection leak in target_handle_connect (1174)
4159 - fix import refcounting bug in OST and MDS cleanup (1134)
4160 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4161 - fix very unlikely obd_types race condition (501)
4162 - remove osc_open hack for echo_client (1187)
4163 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4164 - a failure in read_inode2 leads to deadlock (1139)
4165 - cancel ack-locks as soon as transaction is committed (1072)
4166 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4167 - make sure to commitrw after any preprw to avoid deadlock (1162)
4168 - failing to execute a file in a lustre FS would lock inode (1203)
4169 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4170 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4171 - obd_uuid2tgt would walk off the end of the list (1255)
4172 - on IA64 the osc would give portals incorrect bulk size (1258)
4173 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4174 - fix lock inversion caused by new llite matching code (1282)
4175 - limit the number of dirty pages on a client to 10MB (1286)
4176 - timed out locks were not being corrected cancelled (1289)
4177 - fix O_DIRECT above 4GB on IA-32 (1292)
4178 * major user-visible changes
4179 - fail out/fail over policy now controlled by the upcall (993)
4181 - add OBD_PING to check server availability and failure (954)
4182 - lustre messages are now sent in sending host order (288, 340, 891)
4183 - add eadatalen to MDS getattr reply (340)
4184 - OST read replies may contain second buffer, with per-page status (593)
4186 2003-03-11 Phil Schwan <phil@clusterfs.com>
4189 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4190 - failed open()s could cause deadlock; fixed (867, 869)
4191 - stop cancelling OST locks when files are closed (481)
4192 - overlapping XID spaces caused network corruption (851, 853)
4193 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4194 - setattr_raw added, to do single-RPC, server-side setattrs
4195 - lmc/lconf syntax change for OST UUIDs
4196 - fix crashy race condition between ptlrpc_free_req and osc_close
4197 - don't use request in mdc_enqueue if we hit a timeout (889)
4198 - don't set the inode i_size for regular files from the MDS (896)
4199 - handle out of order completion AST (842)
4200 - don't LBUG if a lock request times out after receiving AST (913)
4201 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4202 - if a bad lock AST arrives, send an error instead of dropping entirely
4203 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4204 - fix leak in bulk IO when only partially completed (899, 900, 926)
4205 - fix O_DIRECT for ia64 (55)
4206 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4207 - C-z after timeout could hang a process forever; fixed (977)
4209 - client-side I/O cache (678, 924, 929, 941, 970)
4211 - READPAGE and SETATTRs which don't take server-side locks get
4214 2003-02-11 Phil Schwan <phil@clusterfs.com>
4217 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4218 - Fix fsfilt_extN_readpage() to read a full page of directory
4219 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4220 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4221 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4222 - Most intent operations were removed, in favour of a new RPC mode
4223 that does a single RPC to the server and bypasses most of the VFS
4224 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4225 - Aggressively cancel local locks on DLM servers
4226 - mds_reint_unlink sends EA to the client if it's the last nlink.
4227 client uses that EA to unlink OST objects.
4228 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4229 - recursive symlinks were fixed (439)
4230 - fixed NULL deref in DEBUG_REQ
4231 - filter_update_lastobjid no longer calls sync, which annoyed extN
4232 - fixed multi-client small-writes to a single file problem (445)
4233 - fixed mtime updates during file writes (607)
4234 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4235 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4236 - fixed Linux OST size reporting problem (444, 656)
4237 - OST now updates object mtime with writes or setattr (607, 619)
4238 - client verifies file size before zeroing page past EOF (445)
4239 - OST now writes last allocated objid to disk with allocation (108)
4240 - LOV on echo now works (409)
4242 - mds_reint_unlink sends a new buffer, with the EA included. this
4243 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4244 set if a regular file was being unlinked, and it was the last link
4245 - use PtlGet from the target for bulk writes (315)
4246 - OST now updates object mtime with writes or setattr (607, 619)
4247 - LDLM now has a grant-time callback to revalidate locked items, if
4249 - Many MDS operations were reorganized to combat race conditions
4251 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4252 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4254 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4257 - Fully reactivate OST imports after reconnection (512, others)
4258 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4259 - More graceful error handling for truncating on dead OST (515)
4260 - Don't error out unless we're actually accessing dead stripes (474)
4261 - Fix garbage sizes when stripes are missing (410)
4262 - LRU counters were broken, causing constant lock purge (433, 432)
4263 - garbage on read from stripes with failed OSTs (441)
4264 - mark OSCs as active before reconnecting during recovery (438)
4265 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4266 - lfind did not preserve OST order in output (443)
4267 - symlinks cause hung clients, incorrect data (439)
4268 - stop dereferencing request after dropping refcount (457)
4269 - don't LASSERT(spin_is_locked) on non-SMP (455)
4270 - fixes for many rename() bugs
4271 - fstat didn't correctly synchronize attributes (399)
4272 - server must handle lock cancellation during blocking AST prep (487)
4273 - bulk descriptors were free()d too soon (511)
4274 - fix paths in lconf, which would load incorrect modules (451, 507)
4275 - fix confusing lconf 'host not found' error message (386)
4276 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4277 - fix race condition in mdc_blocking_ast() for inode access (526)
4278 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4279 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4280 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4281 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4282 - fix lstripe in several ways (kernel side) (527)
4283 - fix request leak in ldlm_cli_enqueue (262)
4284 - incorrect OSC was marked inactive after OST failure
4285 - call mds_fs_cleanup before unmounting filesystem (524)
4286 - fix races between taking ns_lock and ldlm_lock_change_resource
4287 - fix races updating LOV export open file list
4288 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4289 - fix recovery NULL deref in ldlm_cli_cancel_unused
4290 - fix some DLM races by using new hash table for lock handles (419)
4291 - permit the client to specify desired inodes, at replay
4292 - duplicate requests when we queue them for replay reintegration
4293 - fix last_rcvd offset calculation
4294 - sync after each recovered transaction, so we always make progress
4295 - never, not always, ERESTART requests without transnos
4296 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4298 - skip replay if the MDS says that the client is already connected
4299 - don't check for a recovery-enabled export to match lctl's UUID
4300 - don't INC_USE_COUNT for phantom exports
4301 - don't crash when cleaning up phantom exports (567)
4302 - don't double-finish or set replay data for errored mdc_open requests
4303 - abort requests when they time out, so we don't get old replies
4304 - send/receive replies for AST messages again
4305 - if the client says that it doesn't have the lock, cancel it on the
4307 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4308 mark it as destroyed, it will all work out when decref is called
4309 - fix module use counts (22, 581)
4311 - ASTs now expect a reply (server cancels lock on error reply)
4313 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4316 - fix many simultaneous client startup (392)
4317 - fix dentry->d_it clobbering
4318 - credentials weren't being shipped for readdir/getattr operations
4319 - remove invalid assertions triggered during some concurrent MD
4321 - proper Lustre versions added (336, 389)
4322 - fix memory leak for create error case (398)
4323 - fix LOV locking bug that would get cli/srv out of sync
4324 - fix echo client over LOV (409)
4325 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4326 - fix double-O_EXCL intent crash (424)
4327 - avoid sending multiple lock CANCELs (352)
4329 - MDS can do multi-client recovery (modulo bugs in new code)
4331 - many updates, edits, cleanups
4333 2002-11-18 Phil Schwan <phil@clusterfs.com>
4336 - fix null d_it dereference (346)
4337 - fix full OST/dbench hang (333)
4338 - fix permission problem with file removal (286)
4339 - fix removal of OSCs from LOV when they fail
4340 - fix NULL deref during bulk timeout (214)
4341 - fix problems related to multiple filesystems on one MDS (241)
4342 - fixed serious subtle metadata locking bugs
4343 - free locks on clients when inodes are removed due to memory
4345 - fix inode pointer in lock data (285)
4346 - partial support for multiple MDS on a single host (241)
4347 - data locks weren't cancelled at clear_inode time (290, 311)
4348 - intent locks could lead to unbounded lock growth (205)
4349 - added a maximum lock count, an LRU list, and a flusher
4350 - fix multiple rename (365)
4351 - properly abstracted the echo client
4352 - OSC locked 1 byte too many; fixed
4353 - rewrote brw callback code:
4354 - fixed recovery bugs related to LOVs (306)
4355 - fixed too-many-pages-in-one-write crash (191)
4356 - fixed (again) crash in sync_io_timeout (214)
4357 - probably fixed callback-related race (385)
4359 - Add capability to MDS protocol
4360 - LDLM cancellations and callbacks on different portals
4362 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4365 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4366 - defer open object destruction to close time (601981, 138)
4367 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4368 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4369 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4370 - don't call obd_setattr in ll_file_release for destroyed objects
4371 * protocol change to lustre_msg: move |version| and add |flags|
4372 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4373 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4374 * added replay of create, unlink, link and rename operations during
4375 MDS failover; recovery should be much more robust now
4376 * remove failed OSCs from LOVs (only lov_create uses this so far)
4377 * the lustre-HOWTO was brought (more) up to date (582544)
4379 2002-10-23 Phil Schwan <phil@clusterfs.com>
4382 - in-use dentries weren't being reused properly (617851)
4383 - prevent multiple LDLM setup (599178)
4384 - fix LOV size calculations for truncate (617853)
4385 - fix client handling of MDS intent errors (POSIX)
4386 - fix permission bug in lovstripe.c test (624321)
4387 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4388 - truncate past end of file could corrupt data
4389 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4390 - a race in recovery could return ETIMEDOUT to apps (623947)
4391 - building outside the source directory was fixed
4392 * the lustre-HOWTO was brought (more) up to date (582544)
4393 * major progress was made on recovery functionality
4395 2002-10-10 Phil Schwan <phil@clusterfs.com>
4398 - recovery deadlock fix
4399 - rm -rf causes LBUG fix (617817)
4400 - file open by multiple tasks fix (618962)
4401 - directory permissions bugs (602707 and 620007)
4402 - journal_stop fixed with locking (611313)
4403 - O_APPEND failures resolved (618273, perhaps 614459)
4404 - lconf PATH fix (619770)
4405 - IA64 build fix (621450)
4406 - RPC buffer sizes scale with amount of memory
4408 2002-10-01 Phil Schwan <phil@clusterfs.com>
4411 - locks would be cancelled without throwing away data pages,
4412 resulting in inconsistent data (605627)
4413 - inode attributes were not always being refreshed (605627, 612449)
4414 - lconf now continues to cleanup after lctl reports an error
4415 - MDS now enforces user permissions (602707)
4416 - lprocfs cleanup fixed, but not yet enabled (614157)
4417 - fixed infinite server hang, should a client not respond to an AST
4418 - avoid going into recovery if user calls readlink() with a buffer
4419 that's too small (613941)
4420 - AST RPCs no longer require replies (614867) -- this may be changed
4421 - don't crash server if client sends an IOV that's too big (611336)
4422 - fixed lock conversion deadlock (611892)
4423 - fixed the following of symlinks (614622)
4424 * recovery: the server can remove locks from a client that dies, other
4425 clients can make progress
4426 * more extN patch fixes
4427 * compile-time configurable ptlrpc buffer allocations
4429 - collaborative read cache document
4430 - Lustre Lite Performance CDR document-in-progress
4432 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4435 - fix typo in patch-2.4.18
4437 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4440 - clear ptlrpc request each time in handle_incoming_request()
4441 - unlink of files now destroys the object on the OST
4443 2002-09-19 Peter Braam <braam@clusterfs.com>
4445 * add hard link support
4446 * change obdfile creation method
4447 * kernel patch changed
4449 2002-09-19 Peter Braam <braam@clusterfs.com>
4452 - stack overflow bug in extN fixed
4454 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4456 * documentation updates
4457 - add man pages for config tools
4458 - update tests/README to describe testing with new config tools
4459 - finish metadata API descriptions
4460 * bug fixes and cleanups
4461 - statfs workaround for 16TB limit
4462 - LOV stripe allocation improved, can stripe on subset of OSTs
4463 - LOV file size/IO offset was wrong for files > 4GB in size
4464 - object EA data was being dropped, caused files to be unreadable
4465 - memory overflow with non-LOV OST caused memory corruption
4466 - fixed regression tests to work with new config tools, obdfilter
4467 - fixed bug when directory size became larger than 1 block
4468 - fixed bug (for single client case) when PWD was deleted
4469 - invalidate local directory pages when doing intent-based ops
4470 - avoid LDLM oops when lock callback contained bad data
4472 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4474 * documentation updates
4475 * bug fixes and cleanups
4476 - configuration tools
4479 - 64-bit compile warnings
4480 - 64-bit internal statfs data
4482 * test_brw on persistent OST devices
4484 * lprocfs (disabled)
4486 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4488 * documentation updates
4489 * bug fixes and cleanups
4490 * configuration tools
4492 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4495 * many small fixes to 0_5_4
4496 * io/network handling
4497 * thinkos in MDS operations
4499 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4502 * crucial basic fixes to 0.5.3
4503 * IOR, Iozone work over Elan
4506 2002-08-07 Phil Schwan <phil@clusterfs.com>
4507 * version 0_5_3, our first alpha
4508 * we use the new Portals iovs
4509 * documentation updates
4510 * bug fixes and cleanups
4511 * small changes in the DLM wire protocol
4513 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4514 * version 0_5_1 with some initial stability,
4515 * locking on MD and file I/O.
4516 * documentation updates
4517 * several bug fixes since 0.5.0
4518 * small changes in wire protocol
4520 2002-07-18 Phil Schwan <phil@clusterfs.com>
4522 * delivered as Lustre Light Alpha
4523 * fixed a crash after handling invalid MDS requests
4524 * fixed directory pages for architectures with non-4k pages sizes
4526 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4527 * release version v0_4_4
4528 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4530 * Fixes a number of interruption problems with OST operations.
4531 * Update documentation for portals header changes
4532 * Move all wire protocol structs/defines to lustre_idl.h
4533 * Fixes symlink length bug.
4534 * Add tcpdump to repository.
4536 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4537 * release version v0_4_3
4538 * Fixes statfs for inodes on extN.
4539 * Fixes bug in runtests which would delete /etc/hosts.
4540 * Use 64-bit object IDs wherever possible (not into VFS though)
4541 Remove ost_get_info, which is unused by lustre, and out of date.
4543 2002-07-03 Peter Braam <braam@clusterfs.com>
4544 * release version v0_4_2 Fixes a lookup error (type not passed)
4545 * move forward to head of Portals
4546 * move forward to latest Lustre kernel
4548 2002-06-25 Peter Braam <braam@clusterfs.com>
4549 * release version v0_4_1. Hopefully stable on single node use.