5 2.6.16.46-0.14 (SLES 10),
6 2.6.18-8.1.8.EL5 (RHEL 5).
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: Fix errors in lfs documentation
18 Details : Fixes man pages
22 Description: don't use GFP_* in generic Lustre code.
23 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
27 Description: obdclass is limited by single OBD_ALLOC(idarray)
28 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
32 Description: warnings with build patchless client with vanila 2.6.19 and up
33 Details : change old ctl_table style and replace ctl_table/ctl_table_header
34 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
38 Frequency : Only for SLES
39 Description: sanity_quota fail test_1
40 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
41 file, which makes TSTID[2] inunique.
45 Frequency : Only for SLES9
46 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
47 Details : Update target/ChangeLog/which_patch .
51 Frequency : Only for test_10 in sanity-quota.sh
52 Description: a bug in quota test code
53 Details : checking a test flag in a wrong place
57 Frequency : Only for test_13 in sanity-quota.sh
58 Description: a bug in quota test script
59 Details : shouldn't check overfull allocated quota
63 Description: jbd statistics
64 Details : Port older jbd statistics patch for sles10
66 Severity : enhancement
68 Description: Add printing inode info into message about error in writepage.
73 Description: reply_lock_interpret crash due to race with it and lock cancel.
74 Details : Do not replay locks that are being cancelled. Do not reference
75 locks by their address during replay, just by their handle.
79 Frequency : When flocks are used.
80 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
81 Details : Flock locks might destroy just granted lock if it could be merged
82 with another existing flock, this is done in completion handler,
83 so teach ldlm_cli_enquque_fini that this is a valid case for
89 Description: Oops in read and write path when failing to allocate lock.
90 Details : Check if lock allocation failed and return error back.
94 Description: lstripe command fails for valid OST index
95 Details : The stripe offset is compared to lov->desc.ld_tgt_count
96 instead of lov->desc.ld_active_tgt_count.
100 Description: sanity.sh failed test 103
101 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
106 Description: df doesn't work properly if diskfs blocksize != 4K
107 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
111 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
112 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
116 Description: obd_health_check_timeout too short
117 Details : set obd_health_check_timeout as 1.5x of obd_timeout
121 Description: llapi_file_create() does not allow some changes
122 Details : add llapi_file_open() that allows specifying the mode and
123 open flags, and also returns an open file handle.
127 Description: merge and cleanup kernel patches.
128 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
132 Description: Add journal checksum support.(Kernel part)
133 Details : The journal checksum feature adds two new flags i.e
134 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
135 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
136 indicates that the commit block contains the checksum for
137 the blocks described by the descriptor blocks. Now commit
138 record can be sent to disk without waiting for descriptor
139 blocks to be written to disk. This behavior is controlled
140 using JBD2_FEATURE_ASYNC_COMMIT flag.
144 Description: OSS needs mutliple precreate threads
145 Details : Add ability to start more than one create thread per OSS.
149 Description: Sanity.sh test_65k failed: setstripe should have succeeded
150 Details : execute lfs setstripe on client
154 Description: mds_obd_create error creating tmp object
155 Details : When the user sets quota on root, llog will be affected and can't
156 create files and write files.
159 Frequency : Always on ia64 patchless client, and possibly others.
161 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
162 Details : This allows the patchless client to be loaded on architectures
167 Description: RedHat Update kernel for RHEL5
168 Details : Add the kernel config file for RHEL5.
172 Description: RedHat Update kernel for RHEL5
173 Details : Modify the kernel config file more closer RHEL5.
177 Description: Build failure against Centos5 (RHEL5)
178 Details : Define PAGE_SIZE when it isn't present.
182 Description: test_7 on sanity-quota.sh can't be used on separate machines
183 Details : add facet to handle it
187 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
188 Details : with NFS, the anon dentry's parent was set to itself in
189 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
190 corresponding dentry other than use rec->ur_name.
192 Severity : enhancement
194 Description: lfs setstripe enhancement
195 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
199 Description: testing performance impact of enabling checksumming
200 Details : enable checksum by default, allow --disable-checksum
201 configure option and "-o nochecksum" mount option
205 Description: lustre support for RHEL5
206 Details : Add support for RHEL5.
210 Description: Client eviction due to failover config
211 Details : after a connection loss, the lustre client should attempt to
212 reconnect to the last active server first before trying the
213 other potential connections.
217 Description: when mds and osts use different quota unit(32bit and 64bit),
218 quota will be released repeatly.
219 Details : void sending multiple quota reqs to mds, which will keep the
220 status between the reqs.
224 Description: osts not allocated evenly to files
225 Details : change the condition to increase offset_idx
228 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
230 Description: Data corruption for OSTs that are formatted larger than 2TB
232 Details : When generating the bio request for lustre file writes the
233 sector number would overflow a temporary variable before being
234 used for the IO. The data reads correctly from Lustre (which
235 will overflow in a similar manner) but other file data or
236 filesystem metadata may be corrupted in some cases.
240 Description: Tune the kernel for good SCSI performance.
241 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
242 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
247 Description: rewrite ext2-derived code in obdclass/uuid.c
248 Details : rewrite inherited code (uuid parsing code from ext2 utils)
249 from scratch preserving functionality.
253 Description: "lfs find -obd UUID" prints directories
254 Details : "lfs find -obd UUID" will return all directory names instead
255 of just file names. It is incorrect because the directories
256 do not reside on the OSTs.
260 Description: use i_size_read and i_size_write in 2.6 port
261 Details : replace inode->i_size access with i_size_read/write()
264 Frequency : when removing large files
266 Description: scheduling issue during removal of large Lustre files
267 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
268 It causes scheduling issues when removing large files (17TB in the
272 Frequency : only with liblustre clients on XT3
274 Description: evictions taking too long
275 Details : allow llrd to evict clients directly on OSTs
278 Frequency : only on ppc
280 Description: /proc/fs/lustre/devices broken on ppc
281 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
282 We should make sure the type of 'index' is loff_t and then cast
283 explicitly as needed below. Do not assign an explicitly cast
285 --------------------------------------------------------------------------------
287 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
289 * Support for kernels:
290 2.6.5-7.283 (SLES 9),
291 2.6.9-55.EL (RHEL 4),
292 2.6.16.46-0.14 (SLES 10),
293 2.6.18.8 vanilla (kernel.org)
294 * Client support for unpatched kernels:
295 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
296 2.6.16 - 2.6.22 vanilla (kernel.org)
297 * Due to recently discovered recovery problems, we do not recommend
298 using patchless RHEL 4 clients with this or any earlier release.
299 * Recommended e2fsprogs version: 1.39.cfs8
300 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
301 * Starting with this release, the ldiskfs backing filesystem required
302 by Lustre is now in its own package, lustre-ldiskfs. This package
303 should be installed. It is versioned separately from Lustre and
304 may be released separately in future.
308 Description: block reactivating mgc import until all deactivates complete
309 Details : Fix race when failing back MDT/MGS to itself (testing)
311 Severity : enhancement
313 Description: add optional extra BUILD_VERSION info
314 Details : add a new environment variable (namely LUSTRE_VERS) which allows
315 to override the lustre version.
318 Frequency : 2.6.18 servers only
320 Description: ll_kern_mount() doesn't release the module reference
321 Details : The ldiskfs module reference count never drops down to 0
322 because ll_kern_mount() doesn't release the module reference.
327 Description: server LBUG when using old ost_num_threads parameter
328 Details : Accept the old ost_num_threads parameter but warn that it
329 is deprecated, and fix an off-by-one error that caused an LBUG.
334 Description: Transient SCSI error results in persistent IO issue
335 Details : iobuf->dr_error is not reinitialized to 0 between two
339 Frequency : sometimes when underlying device returns I/O errors
341 Description: OSTs not going read-only during write failures
342 Details : OSTs are not remounted read-only when the journal commit threads
343 get I/O errors because fsfilt_ext3 calls journal_start/stop()
344 instead of the ext3 wrappers.
348 Description: poor connect scaling with increasing client count
349 Details : Don't run filter_grant_sanity_check for more than 100 exports
350 to improve scaling for large numbers of clients.
353 Frequency : SLES10 only
355 Description: sanity-quota.sh quotacheck failed: rc = -22
356 Details : Quotas cannot be enabled on SLES10.
359 Frequency : liblustre clients only
361 Description: getdirentries does not give error when run on compute nodes
362 Details : getdirentries does not fail when the size specified as an argument
363 is too small to contain at least one entry
365 Severity : enhancement
367 Description: Add LNET router traceability for debug purposes
368 Details : If a checksum failure occurs with a router as part of the
369 IO path, the NID of the last router that forwarded the bulk data
370 is printed so it can be identified.
375 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
376 Details : Due to a race condition, liblustre clients were occasionally
379 Severity : enhancement
381 Description: lfs setstripe use optional parameters instead of postional
384 Severity : enhancement
386 Description: Nanosecond timestamp support for ldiskfs
387 Details : The on-disk ldiskfs filesystem has added support for nanosecond
388 resolution timestamps. There is not yet support for this at
389 the Lustre filesystem level.
392 Frequency : during server recovery
394 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
395 Details : request with rq_no_resend flag not awake l_wait_event if they get a
399 Frequency : nfs export on patchless client
401 Description: connectathon hang when test nfs export over patchless client
402 Details : Disconnected dentry cannot be found with lookup, so we do not need
403 to unhash it or make it invalid
406 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
407 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
408 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
413 Description: open req refcounting wrong on reconnect
414 Details : If reconnect happened between getting open reply from server and
415 call to mdc_set_replay_data in ll_file_open, we will schedule
416 replay for unreferenced request that we are about to free.
417 Subsequent close will crash in variety of ways.
418 Check that request is still eligible for replay in
419 mdc_set_replay_data().
424 Description: disable writes to filesystem when reading health_check file
425 Details : the default for reading the health_check proc file has changed
426 to NOT do a journal transaction and write to disk, because this
427 can cause reads of the /proc file to hang and block HA state
428 checking on a healthy but otherwise heavily loaded system. It
429 is possible to return to the previous behaviour during configure
430 with --enable-health-write.
432 Severity : enhancement
434 Description: 64-bit inode version
435 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
436 to the inode. This will be required for version-based recovery.
441 Description: MDS fails to start if a duplicate client export is detected
442 Details : in some rare cases it was possible for a client to connect to
443 an MDS multiple times. Upon recovery the MDS would detect this
444 and fail during startup. Handle this more gracefully.
446 Severity : enhancement
448 Description: Add -o localflock option to simulate old noflock
450 Details : This will achieve local-only flock/fcntl locks
456 Description: log_commit_thread vs filter_destroy race leads to crash
457 Details : Take import reference before releasing llog record semaphore
462 Description: Wrong request locking in request set processing
463 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
464 imp_delayed_list, in this place should be used imp_lock.
467 Frequency : when reconnection
469 Description: Grant Leak when osc reconnect to OST
470 Details : When osc reconnect ost, OST(filter) should check whether it
471 should grant more space to client by comparing fed_grant and
472 cl_avail_grant, and return the granted space to client instead
473 of "new granted" space, because client will call osc_init_grant
474 to update the client grant space info.
477 Frequency : when client reconnect to OST
479 Description: Grant Leak when osc do resend and replay bulk write
480 Details : When osc reconnect to OST, OST(filter)should clear grant info of
481 bulk write request, because the grant info while be sync between
482 OSC and OST when reconnect, and we should ignore the grant info
483 these of resend/replay write req.
488 Description: Grant space more than avaiable left space sometimes.
489 Details : When then OST is about to be full, if two bulk writing from
490 different clients came to OST. Accord the avaliable space of the
491 OST, the first req should be permitted, and the second one
492 should be denied by ENOSPC. But if the seconde arrived before
493 the first one is commited. The OST might wrongly permit second
494 writing, which will cause grant space > avaiable space.
497 Frequency : when client is evicted
499 Description: Grant might be wrongly erased when osc is evicted by OST
500 Details : when the import is evicted by server, it will fork another
501 thread ptlrpc_invalidate_import_thread to invalidate the
502 import, where the grant will be set to 0. While the original
503 thread will update the grant it got when connecting. So if
504 the former happened latter, the grant will be wrongly errased
505 because of this race.
510 Description: Checking Stale with correct fid
511 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
512 to check whether it is stale, because sometimes, we want the
513 enqueue happened anyway, and op_data.fid2 will not be initialized.
515 Severity : enhancement
517 Description: update patchless client
518 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
521 Frequency : only with 2.4 kernel
523 Description: random memory corruption
524 Details : size of struct ll_inode_info is to big for union inode.u and this
525 can be cause of random memory corruption.
530 Description: Memory leak in recovery
531 Details : Lov_mds_md was not free in an error handler in mds_create_object.
532 It should also check obd_fail before fsfilt_start, otherwise if
533 fsfilt_start return -EROFS,(failover mds during mds recovery).
534 then the req will return with repmsg->transno = 0 and rc = EROFS.
535 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
536 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
537 be freed no matter whether fsfilt_commit success or not.
540 Frequency : only with huge count clients
542 Description: Prevents from taking the superblock lock in llap_from_page for
544 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
545 for llap_from_page call in ll_removepage prevents from taking the
546 superblock lock for a soon died page.
551 Description: Not check open intent error before release open handle
552 Details : in some rare cases, the open intent error is not checked before
553 release open handle, which may cause
554 ASSERTION(open_req->rq_transno != 0), because it tries to release
555 the failed open handle.
560 Description: Set cat log bitmap only after create log success.
561 Details : in some rare cases, the cat log bitmap is set too early. and it
562 should be set only after create log success.
566 Description: Accessing a block bevice can re-enable I/O when Lustre is
567 tearing down a device.
568 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
573 Description: service threads may hog cpus when there are a lot of requests
575 Details : Insert cond_resched to give other threads a chance to use some of
581 Description: the cat log was not initialized in recovery
582 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
583 unlink log on mds will not be initialized until mds post
584 recovery. And also in mds post recovery, the unlink log will
585 initialization will be done asynchronausly, so there will be race
586 between add unlink log and unlink log initialization.
590 Description: brw_stats were being printed incorrectly
591 Details : brw_stats were being printed as log2 but all of them were not
592 recorded as log2. Also remove some code duplication arising from
593 filter_tally_{read,write}.
597 Frequency : rare, only in recovery.
598 Description: ASSERTION(req->rq_type != LI_POISON) failed
599 Details : imp_lock should be held while iterating over imp_sending_list for
600 prevent destroy request after get timeout in ptlrpc_queue_wait.
604 Description: replay-single.sh test 52 fails
605 Details : A lock's skiplist need to be cleanup when it being unlinked
606 from its resource list.
610 Description: Short directio read returns full requested size rather than
612 Details : Direct I/O operations should return actual amount of bytes
613 transferred rather than requested size.
615 Severity : enhancement
617 Description: metadata RPC reduction (e.g. for rm performance)
618 Details : decrease the amount of synchronous RPC between clients and servers
619 by canceling conflicing lock before the operation on the client side
620 and packing thier handles into the main operation RPC to server.
622 Severity : enhancement
624 Description: Async OSC create to avoid the blocking unnecessarily.
625 Details : If a OST has no remain object, system will block on the creating
626 when need to create a new object on this OST. Now, ways use
627 pre-created objects when available, instead of blocking on an
628 empty osc while others are not empty. If we must block, we block
629 for the shortest possible period of time.
631 --------------------------------------------------------------------------------
633 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
638 Frequency : on some architectures
640 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
641 Details : Uninitialized flags sometimes cause configuration commands to
645 Frequency : patchless clients only
647 Description: missing __iget() symbol export
648 Details : The __iget() symbol export is missing. To avoid the need for
649 this on patchless clients the deathrow inode reaper is turned
650 off, and we depend on the VM to clean up old inodes. This
651 dependency was during via the fix for bug 12181.
653 --------------------------------------------------------------------------------
655 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
657 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
658 INTEROPERATE with older versions automatically. In many cases a
659 special upgrade step is needed. Please read the
660 user documentation before upgrading any part of a live system.
661 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
662 version will not interoperate with 1.6 betas before beta5 (1.5.95).
663 * WARNING: Lustre configuration and startup changes are required with
664 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
670 Severity : enhancement
672 Description: MountConf
673 Details : Lustre configuration is now managed via mkfs and mount
674 commands instead of lmc and lconf. New obd types (MGS, MGC)
675 are added for dynamic configuration management. See
676 https://mail.clusterfs.com/wikis/lustre/MountConf for
679 Severity : enhancement
681 Description: dynamic OST addition
682 Details : OSTs can now be added to a live filesystem
684 Severity : enhancement
686 Description: startup order invariance
687 Details : MDTs and OSTs can be started in any order. Clients only
688 require the MDT to complete startup.
690 Severity : enhancement
692 Description: parallel, asynchronous orphan cleanup
693 Details : orphan cleanup is now performed in separate threads for each
694 OST, allowing parallel non-blocking operation.
696 Severity : enhancement
698 Description: optimized stripe assignment
699 Details : stripe assignments are now made based on ost space available,
700 ost previous usage, and OSS previous usage, in order to try
701 to optimize storage space and networking resources.
703 Severity : enhancement
705 Description: Permanently set tunables
706 Details : All writable /proc/fs/lustre tunables can now be permanently
707 set on a per-server basis, at mkfs time or on a live
710 Severity : enhancement
712 Description: Lustre message v2
713 Details : Add lustre message format v2.
715 Severity : enhancement
717 Description: client OST exclusion list
718 Details : Clients can be started with a list of OSTs that should be
719 declared "inactive" for known non-responsive OSTs.
723 Description: SPEC SFS validation failure on NFS v2 over lustre.
724 Details : Changes the blocksize for regular files to be 2x RPC size,
725 and not depend on stripe size.
727 Severity : enhancement
729 Description: Multiple MD RPCs in flight.
730 Details : Further unserialise some read-only MDS RPCs - learn about intents.
731 To avoid overly-overloading MDS, introduce a limit on number of
732 MDS RPCs in flight for a single client and add /proc controls
733 to adjust this limit.
735 Severity : enhancement
737 Description: client read/write statistics
738 Details : Add client read/write call usage stats for performance
739 analysis of user processes.
740 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
741 file access. extents_stats shows chunk size distribution.
742 extents_stats_per_process show chunk size distribution per
745 Severity : enhancement
747 Description: mds statistics
748 Details : Add detailed mds operations statistics in
749 /proc/fs/lustre/mds/*/stats.
753 Description: Failure of copying files with lustre special EAs.
754 Details : Client side always return success for setxattr call for lustre
755 special xattr (currently only "trusted.lov").
760 Description: Refcount LNET uuids
761 Details : The global LNET uuid list grew linearly with every startup;
762 refcount repeated list entries instead of always adding to
765 Severity : enhancement
767 Description: Dynamic service threads
768 Details : Within a small range, start extra service threads
769 automatically when the request queue builds up.
772 Frequency : mixed-endian client/server environments
774 Description: mixed-endian crashes
775 Details : The new msg_v2 system had some failures in mixed-endian
778 Severity : enhancement
780 Description: Easy OST removal
781 Details : OSTs can be permanently deactivated with e.g. 'lctl
782 conf_param lustre-OST0001.osc.active=0'
784 Severity : enhancement
786 Description: MGS proc entries
787 Details : Added basic proc entries for the MGS showing what filesystems
790 Severity : enhancement
792 Description: provide MGS failover
793 Details : Added config lock reacquisition after MGS server failover.
795 Severity : enhancement
797 Description: add Linux 2.4 support
798 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
802 Description: a large application tries to do I/O to the same resource and dies
804 Details : Check the req->rq_arrival time after the call to
805 ost_brw_lock_get(), but before we do anything about
806 processing it & sending the BULK transfer request. This
807 should help move old stale pending locks off the queue as
808 quickly as obd_timeout.
811 Frequency : when an incorrect nid is specified during startup
813 Description: ptlrpc connect to non-existant node causes kernel crash
814 Details : LNET can't be re-entered from an event callback, which
815 happened when we expire a message after the export has been
816 cleaned up. Instead, hand the zombie cleanup off to another
819 Severity : enhancement
821 Description: plain/inodebits lock performance improvement
822 Details : Grouping plain/inodebits in granted list by their request modes
823 and bits policy, thus improving the performance of search through
827 Frequency : only if OST filesystem is corrupted
829 Description: client incorrectly hits assertion in ptlrpc_replay_req()
830 Details : for a short time RPCs with bulk IO are in the replay list,
831 but replay of bulk IOs is unimplemented. If the OST filesystem
832 is corrupted due to disk cache incoherency and then replay is
833 started it is possible to trip an assertion. Avoid putting
834 committed RPCs into the replay list at all to avoid this issue.
837 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
839 Description: System hang on startup
840 Details : This bug allowed the liblustre (e.g. catamount) client to
841 return to the app before handling all startup RPCs. This
842 could leave the node unresponsive to lustre network traffic
843 and manifested as a server ptllnd timeout.
845 Severity : enhancement
847 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
848 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
849 boolean to control whether to print peer debug info when a
850 client's RPC times out.
853 Frequency : only for kernels with patches from Lustre below 1.4.3
855 Description: Remove old rdonly API
856 Details : Remove old rdonly API which unused from at least lustre 1.4.3
859 Frequency : only for devices with external journals
861 Description: Set external device read-only also
862 Details : During a commanded failover stop, we set the disk device
863 read-only while the server shuts down. We now also set any
864 external journal device read-only at the same time.
867 Frequency : when upgrading from 1.4 while trying to change parameters
869 Description: The wrong (new) MDC name was used when setting parameters for
870 upgraded MDT's. Also allows changing of OSC (and MDC)
871 parameters if --writeconf is specified at tunefs upgrade time.
874 Frequency : when setting specific ost indicies
876 Description: QOS code breaks on skipped indicies
877 Details : Add checks for missing OST indicies in the QOS code, so OSTs
878 created with --index need not be sequential.
880 Severity : enhancement
882 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
883 Details : The uninit_groups feature works in conjunction with the kernel
884 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
885 up the pass1 processing of e2fsck. This is a read-only feature
886 in ldiskfs2 only, so older kernels and current ldiskfs cannot
887 mount filesystems that have had this feature enabled.
889 Severity : enhancement
891 Description: Improve multi-block allocation algorithm to avoid fragmentation
892 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
893 allocation locality and avoid filesystem fragmentation.
895 ------------------------------------------------------------------------------
897 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
899 * Support for kernels:
900 2.4.21-47.0.1.EL (RHEL 3)
902 2.6.9-42.0.10.EL (RHEL 4)
903 2.6.12.6 vanilla (kernel.org)
904 2.6.16.27-0.9 (SLES 10)
905 * Recommended e2fsprogs version: 1.39.cfs5
907 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
911 Frequency : occasional, depends on client load and configuration
912 Bugzilla : 12181, 12203
913 Description: data loss for recently-modified files
915 Details : In some cases it is possible that recently written or created
916 files may not be written to disk in a timely manner (this should
917 normally be within 30s unless client IO load is very high).
918 The problem appears as zero-length files or files that are a
919 multiple of 1MB in size after a client crash or client eviction
920 that are missing data at the end of the file.
922 This problem is more likely to be hit on clients where files are
923 repeatedly created and unlinked in the same directory, clients
924 have a large amount of RAM, have many CPUs, the filesystem has
925 many OSTs, the clients are rebooted frequently, and/or the files
926 are not accessed by other nodes after being written.
928 The presence of the problem can be detected by looking at
929 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
930 smaller than the second (nr_unused) then dirty files will not
931 be flushed automatically to disk. "sync; sleep 10" should be
932 run several times on the node before unmounting it to update
933 Lustre (this is also safe to run on nodes without this problem).
935 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
936 that can cause this same problem, so customers using that kernel
937 also need to update the kernel in addition to Lustre. In order
938 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
940 It is normal that files written just before a client crash (less
941 than 30s) may not yet have been flushed to disk, even for local
945 Frequency : frequent on thin XT3 nodes
947 Description: UUID collision on thin XT3 Linux nodes
948 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
949 randomly, since we relied on an insufficiently-seeded PRNG.
954 Description: OSS hangs after "All ost request buffers busy"
955 Details : A deadlock between quota and journal operations caused OSS
956 hangs after printing "All ost request buffers busy."
959 Frequency : always on liblustre builds
961 Description: Cleanup compiler warnings on liblustre
964 Frequency : always on liblustre builds on XT3
966 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
970 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
971 > 32000 subdirectories
972 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
973 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
974 subdirectory count crosses 32000. This will aid e2fsck to
975 correctly handle more than 32000 subdirectories.
978 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
980 Description: System hang on startup
981 Details : This bug allowed the liblustre (e.g. catamount) client to
982 return to the app before handling all startup RPCs. This
983 could leave the node unresponsive to lustre network traffic
984 and manifested as a server ptllnd timeout.
986 Severity : enhancement
988 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
989 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
990 boolean to control whether to print peer debug info when a
991 client's RPC times out.
996 Description: make O_SYNC working on 2.6 kernels
997 Details : 2.6 kernels use different method for mark pages for write,
998 so need add a code to lustre for O_SYNC work.
1003 Description: Failure to close file and release space on NFS
1004 Details : Put inode details into lock acquired in ll_intent_file_open.
1005 Use mdc_intent_lock in ll_intent_open to properly
1006 detect all kind of errors unhandled by mdc_enqueue.
1011 Description: proc file read during shutdown sometimes raced obd removal,
1013 Details : Add lock to prevent obd access after proc file removal.
1016 Frequency : Only for files larger than 4GB on 32-bit clients.
1018 Description: improperly doing page alignment of locks
1019 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1020 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1025 Description: RPCs being resent when they shouldn't be
1026 Details : Some RPCs that should not be resent are being resent. This
1027 can cause inconsistencies in the RPC state machine. Do not
1028 resend such requests.
1031 Frequency : rare, only with NFS export
1033 Description: Crash on NFS re-export node
1034 Details : under very unusual load conditions an assertion is hit in
1035 ll_intent_file_open()
1038 Frequency : only if OST filesystem is corrupted
1040 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1041 Details : for a short time RPCs with bulk IO are in the replay list,
1042 but replay of bulk IOs is unimplemented. If the OST filesystem
1043 is corrupted due to disk cache incoherency and then replay is
1044 started it is possible to trip an assertion. Avoid putting
1045 committed RPCs into the replay list at all to avoid this issue.
1050 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1051 Details : Large single O_DIRECT read and write calls can fail to allocate
1052 a sufficiently large buffer to process the request. In case of
1053 allocation failure the allocation is retried with a smaller
1054 buffer and broken into smaller requests.
1056 Severity : enhancement
1058 Description: Add -o localflock option to simulate old noflock behaviour.
1059 Details : This will achieve local-only flock/fcntl locks coherentness.
1064 Description: versioning check is incomplete
1065 Details : Checking the version difference of client vs. server, report
1066 error if the gap is too big.
1071 Description: add support PG_writeback bit
1072 Details : add support for PG_writeback bit for Lustre, for more carefull
1073 work with page cache in 2.6 kernel. This also fix some deadlocks
1074 and remove hack for work O_SYNC with 2.6 kernel.
1076 Severity : enhancement
1078 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1079 Details : The uninit_groups feature works in conjunction with the kernel
1080 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1081 up the pass1 processing of e2fsck. This is a read-only feature
1082 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1083 mount filesystems that have had this feature enabled.
1085 Severity : enhancement
1087 Description: Improve multi-block allocation algorithm to avoid fragmentation
1088 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1089 allocation locality and avoid filesystem fragmentation.
1091 ------------------------------------------------------------------------------
1093 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1095 * Support for kernels:
1096 2.6.9-42.0.3.EL (RHEL 4)
1097 2.6.5-7.276 (SLES 9)
1098 2.4.21-47.0.1.EL (RHEL 3)
1099 2.6.12.6 vanilla (kernel.org)
1100 2.6.16.21-0.8 (SLES10)
1101 * Recommended e2fsprogs version: 1.39.cfs2-0
1103 * The backwards-compatible /proc/sys/portals symlink has been removed
1104 in this release. Before upgrading, please ensure that you change
1105 any configuration scripts or /etc/sysctl.conf files that access
1106 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1107 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1108 in advance of the upgrade on any system running Lustre 1.4.6 or
1109 newer, since /proc/sys/lnet was added in that version.
1110 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1114 Frequency : only when quota is used
1116 Description: avoid scanning export list for quota master
1117 Details : Change the algorithms to avoid scanning export list in order
1118 to improve the efficiency.
1121 Frequency : MDS failover only, very rarely
1123 Description: "went back in time" messages on mds failover
1124 Details : The greatest transno may be lost when the current operation
1125 finishes with an error (transno==0) and the client's last_rcvd
1126 record is over-written. Save the greatest transno in the
1127 mds_last_transno for this case.
1130 Frequency : always for specific kernels and striping counts
1132 Description: client may get "Matching packet too big" without ACL support
1133 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1134 when trying to access files in certain configurations. The
1135 clients should in fact be denied when mounting because they do
1136 not understand ACLs.
1139 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1141 Description: many clients connecting with IO in progress causes connect timeouts
1142 Details : Avoid synchronous journal commits to avoid delays caused by many
1143 clients connecting/disconnecting when bulk IO is in progress.
1144 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1145 OST_IO_PORTAL to avoid delays behind potentially many pending
1149 Frequency : occasionally with multiple writers to a single file
1151 Description: shared writes to file may result in wrong size reported by stat()
1152 Details : Allow growing of kms when extent lock is cancelled
1155 Frequency : always with random mmap IO to multi-striped file
1157 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1158 Details : Only if the hole is at the end of OST object so that kms is too
1159 small. Fix is to increase kms accordingly in ll_nopage.
1162 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1164 Description: writes to a missing object would leak memory on the OST
1165 Details : If there is an inconsistency between the MDS and OST filesystems,
1166 such that the MDS references an object that doesn't exist, writes
1167 to that object will leak memory due to incorrect cleanup in the
1168 error handling path, eventually running out of memory on the OST.
1173 Description: Creating too long symlink causes lustre errors
1174 Details : Check symlink and name lengths before sending requests to MDS.
1177 Frequency : only if flock is enabled (not on by default)
1179 Description: posix locks not released on fd closure on 2.6.9+
1180 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1181 kernels, this caused such locks not to be released on fd close and
1182 then assertions on fs unmount about still used locks.
1185 Frequency : MDS failover only, very rarely
1187 Description: clients may get ASSERTION(granted_lock != NULL)
1188 Details : When request was taking a long time, and a client was resending
1189 a getattr by name lock request. The were multiple lock requests
1190 with the same client lock handle and
1191 mds_getattr_name->fixup_handle_for_resent_request found one of the
1192 lock handles but later failed with ASSERTION(granted_lock != NULL).
1197 Description: handle->h_buffer_credits > 0, assertion failure
1198 Details : h_buffer_credits is zero after truncate, causing assertion
1199 failure. This patch extends the transaction or creates a new
1203 Frequency : NFS re-export or patchless client
1204 Bugzilla : 11179, 10796
1205 Description: Crash on NFS re-export node (__d_move)
1206 Details : We do not want to hash the dentry if we don't have a lock.
1207 But if this dentry is later used in d_move, we'd hit uninitialised
1208 list head d_hash, so we just do this to init d_hash field but
1209 leave dentry unhashed.
1212 Frequency : NFS re-export or patchless client
1214 Description: NFS exports has problem with symbolic link
1215 Details : lustre client didn't properly install dentry when re-exported
1216 to NFS or running patchless client.
1219 Frequency : NFS re-export or patchless client
1221 Description: Various nfs/patchless fixes.
1222 Details : fixes reuse disconected alias for lookup process - this fixes
1223 warning "find_exported_dentry: npd != pd",
1224 fix permission error with open files at nfs.
1225 fix apply umask when do revalidate.
1228 Frequency : occasional
1230 Description: Crash on NFS re-export node
1231 Details : calling clear_page() on the wrong pointer triggered oops in
1232 generic_mapping_read().
1235 Frequency : rarely, using O_DIRECT IO
1237 Description: unaligned directio crashes client with LASSERT
1238 Details : check for unaligned buffers before trying any requests.
1241 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1243 Description: stale data returned from RAID cache
1244 Details : If only a small amount of IO is done to the RAID device before
1245 reading it again it is possible to get stale data from the RAID
1246 cache instead of reading it from disk.
1249 Frequency : always for sles10 kernel
1251 Description: sles10 support
1252 Details : ll_follow_link: compile fixes and using of nd_set_link
1253 under newer kernels.
1256 Frequency : depends on arch, kernel and compiler version, always on sles10
1259 Description: recursive or deep enough symlinks cause stack overflow
1260 Details : getting rid of large stack-allocated variable in
1264 Frequency : depends on hardware
1266 Description: lustre write performance loss in the SLES10 kernel
1267 Details : the performance loss is caused by using of write barriers in the
1268 ext3 code. The SLES10 kernel turns barrier support on by
1269 default. The fix is to undo that change for ldiskfs.
1271 ------------------------------------------------------------------------------
1273 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1275 * Support for kernels:
1276 2.6.9-42.0.3EL (RHEL 4)
1277 2.6.5-7.276 (SLES 9)
1278 2.4.21-47.0.1.EL (RHEL 3)
1279 2.6.12.6 vanilla (kernel.org)
1283 Frequency : quota enabled and large files being deleted
1285 Description: releasing more than 4GB of quota at once hangs OST
1286 Details : If a user deletes more than 4GB of files on a single OST it
1287 will cause the OST to spin in an infinite loop. Release
1288 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1293 Description: statfs data retrieved from /proc may be stale or zero
1294 Details : When reading per-device statfs data from /proc, in the
1295 {kbytes,files}_{total,free,avail} files, it may appear
1296 as zero or be out of date.
1299 Frequency : systems with MD RAID1 external journal devices
1301 Description: lconf's call to blkid is confused by RAID1 journal devices
1302 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1303 of returning all block devices that match the journal UUID.
1306 Frequency : always, for aggregate stripe size over 4GB
1308 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1309 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1310 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1313 Frequency : always if "lfs find" used on a local file/directory
1315 Description: "lfs find" segfaults if used on a local file/directory
1316 Details : The case where a directory component was not specified wasn't
1317 handled correctly. Handle this properly.
1320 Frequency : always on ppc64
1322 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1323 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1324 perform endianness conversion.
1327 Frequency : rarely (truncate to non-zero file size after write under load)
1328 Bugzilla : 10730, 10687
1329 Description: Files padded with zeros to next 4K multiple
1330 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1331 it is possible that files that are truncated to a non-zero size
1332 immediately after being written are filled with zero bytes beyond
1333 the truncated size. No file data is lost.
1335 Severity : enhancement
1337 Description: Allow recovery/failover for liblustre clients.
1338 Details : liblustre clients were unaware of failover configurations until
1341 Severity : enhancement
1343 Description: user file locks should fail when not mounting with flock option
1344 Details : Set up an error-returning stub in ll_file_operations.lock field
1345 to prevent incorrect behaviour when client is mounted without
1346 flock option. Also, set up properly f_op->flock field for
1350 Frequency : always on ia64
1352 Description: "lfs df" loops on printing out MDS statfs information
1353 Details : The obd_ioctl_data was not initialized and in some systems
1354 this caused a failure during the ioctl that did not return
1355 an error. Initialize the struct and return an error on failure.
1358 Frequency : SLES 9 only
1360 Description: Error of copying files with lustre special EAs as root
1361 Details : Client side always return success for setxattr call for lustre
1362 special xattr (currently only "trusted.lov").
1365 Frequency : rarely on clusters with both ia64+i386 clients
1367 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1368 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1369 clients are doing concurrent writes to the same file it is
1370 possible that the ia64 clients may LASSERT because the OST
1371 extent locks are not PAGE_SIZE aligned. Ensure that grown
1372 locks are always aligned on the request boundary.
1375 Frequency : specific use, occasional
1377 Description: Overwriting in use executable truncates on-disk binary image
1378 Details : If one node attempts to overwrite an executable in use by
1379 another node, we now correctly return ETXTBSY instead of
1380 truncating the file.
1382 Severity : enhancement
1384 Description: Async OSC create to avoid the blocking unnecessarily.
1385 Details : If a OST has no remain object, system will block on the creating
1386 when need to create a new object on this OST. Now, ways use
1387 pre-created objects when available, instead of blocking on an
1388 empty osc while others are not empty. If we must block, we block
1389 for the shortest possible period of time.
1394 Description: chmod on Lustre root is propagated to other clients
1395 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1396 invalid by the follow_mount time.
1399 Frequency : liblustre clients only
1401 Description: Race in 'instant cancel' lock handling could lead to such locks
1402 never to be granted in case of SMP MDS
1403 Details : Do not destroy not yet granted but cbpending locks in
1407 Frequency : replay/resend of open
1409 Description: non null lock assetion failure in mds_intent_policy
1410 Details : Trying to replay/resend lockless open requests resulted in
1411 mds_open() returning 0 with no lock. Now it sets a flag if
1412 a lock is going to be returned.
1414 Severity : enhancement
1416 Description: Checksum enhancements
1417 Details : New checksum enhancements allow for resending RPCs that failed
1420 Severity : enhancement
1422 Description: Tunables on number of dirty pages in cacche
1423 Details : Allow to set limit on number of dirty pages cached.
1428 Description: client crash on unmount - lock still has references
1429 Details : In some error handling cases it was possible to leak a lock
1430 reference on a client while accessing a file. This was not
1431 harmful to the client during operation, but would cause the
1432 client to crash when the filesystem is unmounted.
1435 Frequency : specific case, rare
1437 Description: ETXTBSY on mds though file not in use by client
1438 Details : ETXTBSY is no longer incorrectly returned when attempting to
1439 chmod or chown a directory that the user previously tried to
1440 execute or a currently-executing binary.
1443 Frequency : extremely rare except on liblustre-based clients
1445 Description: Lustre space not freed when files are deleted
1446 Details : Clean up open-unlinked files after client eviction. Previously
1447 the unlink was skipped and the files remained as orphans.
1452 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1453 Details : In some cases it was possible to send a blocking callback to
1454 a client doing a glimpse, even though that client didn't get
1455 a lock granted. When the glimpse lock is cancelled on the OST
1456 the freed lock is left on the waiting list and corrupted the list.
1459 Frequency : all core dumps
1461 Description: Broke core dumps to lustre
1462 Details : Negative dentry may be unhashed if parent does not have UPDATE
1463 lock, but some callers, e.g. do_coredump, expect dentry to be
1464 hashed after successful create, hash it in ll_create_it.
1466 ------------------------------------------------------------------------------
1468 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1470 * Support for kernels:
1471 2.6.9-42.0.2.EL (RHEL 4)
1472 2.6.5-7.276 (SLES 9)
1473 2.4.21-40.EL (RHEL 3)
1474 2.6.12.6 vanilla (kernel.org)
1478 Frequency : always on RHEL 3
1480 Description: Number of open files grows over time
1481 Details : The number of open files grows over time, whether or not
1482 Lustre is started. This was due to a filp leak introduced
1483 by one of our kernel patches.
1485 ------------------------------------------------------------------------------
1487 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1489 * Support for kernels:
1490 2.6.9-42.EL (RHEL 4)
1491 2.6.5-7.276 (SLES 9)
1492 2.4.21-40.EL (RHEL 3)
1493 2.6.12.6 vanilla (kernel.org)
1498 Bugzilla : 5719, 9635, 9792, 9684
1499 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1500 Details : If a server is under heavy load and cannot reply to new
1501 connection requests before the client resends the (re)connect,
1502 the connection handling code can behave badly if two service
1503 threads are concurrently handing separate (re)connections from
1504 the same client. Add better locking to the connection handling
1505 code, and ensure that only a single connection will be processed
1506 for a given client UUID, even if the lock is dropped.
1508 Severity : enhancement
1510 Description: add TCP zero-copy support to kernel
1511 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1512 sends if the hardware supports scatter-gather and checksumming.
1513 This allows socklnd to do client-write and server-read more
1514 efficiently and reduce CPU utilization from skbuf copying.
1517 Frequency : only if NFS exporting from client
1519 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1520 Details : If setting attributes on a file created under NFS that had
1521 never been opened it would be possible to oops the client
1522 if the file had no objects.
1525 Frequency : always for liblustre
1527 Description: liblustre client does MDS+OSTs setattr RPC for each write
1528 Details : When doing a write from a liblustre client, the client
1529 incorrectly issued an RPC to the MDS and each OST the file was
1530 striped over in order to update the timestamps. When writing
1531 with small chunks and many clients this could overwhelm the MDS
1532 with RPCs. In all cases it would slow down the write because
1533 these RPCs are unnecessary.
1535 Severity : enhancement
1537 Description: allow number of MDS service threads to be changed at module load
1538 Details : It is now possible to change the number of MDS service threads
1539 running. Adding "options mds mds_num_threads={N}" to the MDS's
1540 /etc/modprobe.conf will set the number of threads for the next
1541 time Lustre is restarted (assuming the "mds" module is also
1542 reloaded at that time). The default number of threads will
1543 stay the same, 32 for most systems.
1548 Description: OST crash if filesystem is unformatted or corrupt
1549 Details : If an OST is started on a device that has never been formatted
1550 or if the filesystem is corrupt and cannot even mount then the
1551 error handling cleanup routines would dereference a NULL pointer.
1556 Description: NULL pointer deref in llap_from_page.
1557 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1558 incorrect) mapping if the page was truncated/reclaimed while it was
1559 searched for. Check for this condition and skip such pages when
1560 doing readahead. Introduce extra check to llap_from_page() to
1561 verify page->mapping->host is non-NULL (so page is not anonymous).
1564 Frequency : Sometimes when using sys_sendfile
1566 Description: "page not covered by a lock" warnings from ll_readpage
1567 Details : sendfile called ll_readpage without right page locks present.
1568 Now we introduced ll_file_sendfile that does necessary locking
1569 around call to generic_file_sendfile() much like we do in
1573 Frequency : with certain MDS communication failures at client mount time
1575 Description: NULL pointer deref after failed client mount
1576 Details : a client connection request may delayed by the network layer
1577 and not be sent until after the PTLRPC layer has timed out the
1578 request. If the client fails the mount immediately it will try
1579 to clean up before the network times out the request. Add a
1580 reference from the request import to the obd device and delay
1581 the cleanup until the network drops the request.
1584 Frequency : occasionally during client (re)connect
1586 Description: assertion failure during client (re)connect
1587 Details : processing a client connection request may be delayed by the
1588 client or server longer than the client connect timeout. This
1589 causes the client to resend the connection request. If the
1590 original connection request is replied in this interval, the
1591 client may trip an assertion failure in ptlrpc_connect_interpret()
1592 which thought it would be the only running connect process.
1595 Frequency : only with obd_echo servers and clients that are rebooted
1597 Description: kernel BUG accessing uninitialized data structure
1598 Details : When running an obd_echo server it did not start the ping_evictor
1599 thread, and when a client was evicted an uninitialized data
1600 structure was accessed. Start the ping_evictor in the RPC
1601 service startup instead of the OBD startup.
1603 Severity : enhancement
1604 Bugzilla : 10193 (patchless)
1605 Description: Remove dependency on various unexported kernel interfaces.
1606 Details : No longer need reparent_to_init, exit_mm, exit_files,
1607 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1610 Frequency : rare (only users of deprecated and unsupported LDAP config)
1612 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1613 Details : LDAP apparently contains 'lustreName' attributes instead of
1614 'name'. A simple remapping of the name is sufficient.
1617 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1619 Description: waiting_locks_callback trips kernel BUG if client is evicted
1620 Details : Running with the dump_on_timeout debug flag turned on makes
1621 it possible that the waiting_locks_callback() can try to dump
1622 the Lustre kernel debug logs from an interrupt handler. Defer
1623 this log dumping to the expired_lock_main() thread.
1625 Severity : enhancement
1627 Description: Support NFS exporting on 2.6 kernels.
1628 Details : Implement non-rawops metadata methods for NFS server to use without
1629 changing NFS server code.
1632 Frequency : very rare (synthetic metadata workload only)
1634 Description: two racing renames might cause an MDS thread to deadlock
1635 Details : Running the "racer" program may cause one MDS thread to rename
1636 a file from being the source of a rename to being the target of
1637 a rename at exactly the same time that another thread is doing
1638 so, and the second thread has already enqueued these locks after
1639 doing a lookup of the target and is trying to relock them in
1640 order. Ensure that we don't try to re-lock the same resource.
1643 Frequency : only very large systems with liblustre clients
1645 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1646 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1647 all OSTs in parallel. This allows the eviction of stale liblustre
1648 clients to proceed much faster than if they were done in series,
1649 and also offers similar improvements for other set_info RPCs.
1654 Description: excessive CPU usage during initial read phase on client
1655 Details : During the initial read phase on a client, it would agressively
1656 retry readahead on the file, consuming too much CPU and impacting
1657 performance (since 1.4.5.8). Improve the readahead algorithm
1658 to avoid this, and also improve some other common cases (read
1659 of small files in particular, where "small" is files smaller than
1660 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1665 Description: MDS crash when receiving packet with unknown intent.
1666 Details : Do not LBUG in unknown intent case, just return -EFAULT
1668 Severity : enhancement
1669 Bugzilla : 9293, 9385
1670 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1671 Details : Do not serialize getattr (non-intent version) and statfs.
1674 Frequency : occasional, when OST network is overloaded/intermittent
1676 Description: client evicted by OST after bulk IO timeout
1677 Details : If a client sends a bulk IO request (read or write) the OST
1678 may evict the client if it is unresposive to its data GET/PUT
1679 request. This is incorrect if the network is overloaded (takes
1680 too long to transfer the RPC data) or dropped the OST GET/PUT
1681 request. There is no need to evict the client at all, since
1682 the pinger and/or lock callbacks will handle this, and the
1683 client can restart the bulk request.
1686 Frequency : Always when mmapping file with no objects
1688 Description: client crashes when mmapping file with no objects
1689 Details : Check that we actually have objects in a file before doing any
1690 operations on objects in ll_vm_open, ll_vm_close and
1696 Description: Request leak when working with deleted CWD
1697 Details : Introduce advanced request refcount tracking for requests
1698 referenced from lustre intent.
1700 Severity : Enhancement
1702 Description: Cache open file handles on client.
1703 Details : MDS now will return special lock along with openhandle, if
1704 requested and client is allowed to hold openhandle, even if unused,
1705 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1706 closing files for every read/write openration.
1708 Severity : Enhancement
1710 Description: Cache open negative dentries on client when possible.
1711 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1712 negative dentries on lock revocation.
1717 Description: Remounting a client read-only wasn't possible with a zconf mount
1718 Details : It wasn't possible to remount a client read-only with llmount.
1720 Severity : enhancement
1721 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1722 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1723 MPICH is included to add Lustre-specific ADIO interfaces.
1724 This is based closely on the UFS ADIO layer and only differs
1725 in file creation, in order to allow the OST striping to be set.
1726 This is user-contributed code and not supported by CFS.
1731 Description: extended inode attributes (immutable, append-only) work improperly
1732 when 2.4 and 2.6 kernels are used on client/server or vice versa
1733 Details : Introduce kernel-independent values for these flags.
1735 Severity : enhancement
1738 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1739 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1740 granularity. Now, Fractional MB granularity is be supported,
1741 this is very useful for low memory system.
1743 Severity : enhancement
1745 Description: Getattr by fid
1746 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1747 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1748 have locks and avoids getattr rpc requests for it on every stat.
1751 Frequency : Always, for filesystems larger than 2TB
1753 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1754 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1755 filesystems larger than 2TB to be mounted with mballoc enabled.
1758 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1760 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1761 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1762 filesystems larger than 2TB is mounted then the kernel will
1763 silently corrupt the start of the filesystem. CONFIG_LBD is
1764 enabled for all CFS-supported kernels, but the possibility of
1765 this happening with a modified kernel config exists.
1767 Severity : enhancement
1769 Description: add client O_DIRECT support for 2.6 kernels
1770 Details : It is now possible to do O_DIRECT reads and writes to files
1771 in the Lustre client mountpoint on 2.6 kernel clients.
1773 Severity : enhancement
1775 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1776 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1777 parallel, not waiting for response from every OST before sending
1778 a rpc to the next OST.
1783 Description: setattr vs write race when updating file timestamps
1784 Details : Client processes that update a file timestamp into the past
1785 right after writing to the file (e.g. tar) it is possible that
1786 the updated file modification time can be reset to the current
1787 time due to a race between processing the setattr and write RPC.
1789 Severity : enhancement
1791 Description: Bring 'lfs find' closer in line with regular Linux find.
1792 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1793 -print0 options and obtains all the needed info through the lustre
1796 Severity : enhancement
1798 Description: support up to 1024 configured devices on one node
1799 Details : change obd_dev array from statically allocated to dynamically
1800 allocated structs as they are first used to reduce memory usage
1805 Description: Flush dirty partially truncated pages during truncate
1806 Details : Immediatelly flush partially truncated pages in filter_setattr,
1807 this way we completely avoid having any pages in page cache on OST
1808 and can retire ugly workarounds during writes to flush such pages.
1813 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1814 Details : protect inode from truncation within vfs_unlink() context
1815 just take a reference before calling vfs_unlink() and release it
1816 when parent's i_sem is free.
1821 Description: last_id value checked outside lock on OST caused LASSERT failure
1822 Details : If there were multiple MDS->OST object precreate requests in
1823 flight, it was possible that the OST's last object id was checked
1824 outside a lock and incorrectly tripped an assertion. Move checks
1825 inside locks, and discard old precreate requests.
1828 Frequency : always, if extents are used on OSTs
1830 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1831 Details : OSTs using the extents format would not zero the high 16 bits of
1832 the index physical block number. This is not a problem for any
1833 OST filesystems smaller than 16TB, and no kernels support ext3
1834 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1835 new/modified files) and can be fixed for existing filesystems
1836 with e2fsprogs-1.39-cfs1.
1841 Description: import connection selection may be incorrect if timer wraps
1842 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1843 import connections to be ignored if the 32-bit jiffies counter
1844 wraps. Use a 64-bit jiffies counter.
1847 Frequency : very large clusters immediately after boot
1849 Description: LNET request buffers exhausted under heavy short-term load
1850 Details : If a large number of client requests are generated on a service
1851 that has previously never seen so many requests it is possible
1852 that the request buffer growth cannot keep up with the spike in
1853 demand. Instead of dropping incoming requests, they are held in
1854 the LND until the RPC service can accept more requests.
1857 Frequency : Sometimes during replay
1859 Description: Assertion failure in ll_local_open after replay.
1860 Details : If replay happened on an open request reply before we were able
1861 to set replay handler, reply will become not swabbed tripping the
1862 assertion in ll_local_open. Now we set the handler right after
1863 recognising of open request
1866 Frequency : very rare
1868 Description: kernel reports "badness in vsnprintf"
1869 Details : Reading from the "recovery_status" /proc file in small chunks
1870 may cause a negative length in lprocfs_obd_rd_recovery_status()
1871 call to vsnprintf() (which is otherwise harmless). Exit early
1872 if there is no more space in the output buffer.
1874 Severity : enhancement
1876 Description: clear OBD RPC statistics by writing to them
1877 Details : It is now possible to clear the OBD RPC statistics by writing
1878 to the "stats" file.
1883 Description: Client mtime is not the same on different clients after utimes
1884 Details : In some cases, the client was using the utimes() syscall on
1885 a file cached on another node. The clients now validate the
1886 ctime from the MDS + OSTs to determine which one is right.
1891 Description: Inability to activate failout mode
1892 Details : lconf script incorrectly assumed that in pythong string's numeric
1893 value is used in comparisons.
1896 Frequency : always with multiple stripes per file
1898 Description: Inefficient object allocation for mutli-stripe files
1899 Details : When selecting which OSTs to stripe files over, for files with
1900 a stripe count that divides evenly into the number of OSTs,
1901 the MDS is always picking the same starting OST for each file.
1902 Return the OST selection heuristic to the original design.
1907 Description: mount failures may take full timeout to return an error
1908 Details : Under some heavy load conditions it is possible that a
1909 failed mount can wait for the full obd_timeout interval,
1910 possibly several minutes, before reporting an error.
1911 Instead return an error as soon as the status is known.
1913 Frequency : quota enabled and large files being deleted
1915 Description: releasing more than 4GB of quota at once hangs OST
1916 Details : If a user deletes more than 4GB of files on a single OST it
1917 will cause the OST to spin in an infinite loop. Release
1918 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1923 Description: statfs data retrieved from /proc may be stale or zero
1924 Details : When reading per-device statfs data from /proc, in the
1925 {kbytes,files}_{total,free,avail} files, it may appear
1926 as zero or be out of date.
1929 Frequency : systems with MD RAID1 external journal devices
1931 Description: lconf's call to blkid is confused by RAID1 journal devices
1932 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1933 of returning all block devices that match the journal UUID.
1936 Frequency : always, for aggregate stripe size over 4GB
1938 Description: assertion fails when trying to use 4GB stripe size
1939 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1940 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1943 Frequency : always on ppc64
1945 Description: the first write on an ext3 filesystem with mballoc got stuck
1946 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1947 perform endianness conversion.
1949 ------------------------------------------------------------------------------
1951 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1953 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1954 INTEROPERATE with older versions automatically. Please read the
1955 user documentation before upgrading any part of a live system.
1956 * WARNING: Lustre networking configuration changes are required with
1957 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1960 * Support for kernels:
1961 2.6.9-22.0.2.EL (RHEL 4)
1962 2.6.5-7.244 (SLES 9)
1963 2.6.12.6 vanilla (kernel.org)
1966 Severity : enhancement
1967 Bugzilla : 7981/8208
1968 Description: Introduced Lustre Networking (LNET)
1969 Details : LNET is new networking infrastructure for Lustre, it includes
1970 a reorganized network configuration mode (see the user
1971 documentation for full details) as well as support for routing
1972 between different network fabrics. Lustre Networking Devices
1973 (LNDs) for the supported network fabrics have also been
1974 created for this new infrastructure.
1976 Severity : enhancement
1977 Description: Introduced Access control lists
1978 Details : clients can set ACLs on files and directories in order to have
1979 more fine-grained permissions than the standard Unix UGO+RWX.
1980 The MDS must be started with the "-o acl" mount option.
1982 Severity : enhancement
1983 Description: Introduced filesystem quotas
1984 Details : Administrators may now establish per-user quotas on the
1987 Severity : enhancement
1989 Description: Configuration change for the XT3
1990 The PTLLND is now used to run Lustre over Portals on the XT3
1991 The configure option(s) --with-cray-portals are no longer used.
1992 Rather --with-portals=<path-to-portals-includes> is used to
1993 enable building on the XT3. In addition to enable XT3 specific
1994 features the option --enable-cray-xt3 must be used.
1999 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2000 Details : A race between allocating a new llcd and re-getting the llcd_lock
2001 allowed another thread to grab newly-allocated llcd.
2003 Severity : enhancement
2005 Description: 2.6 OST async journal commit and locking fix to improve performance
2006 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2007 2.6 kernels are now async as they already were in 2.4 kernels so
2008 that they can commit concurrently with the network bulk transfer.
2009 For block-allocated files the filter allocation semaphore is held
2010 to avoid filesystem fragmentation during allocation. BKL lock
2011 removed for 2.6 xattr operations where it is no longer needed.
2016 Description: lconf incorrectly determined whether two IP networks could talk
2017 Details : In some more complicated routing and multiple-network
2018 configurations, lconf will avoid trying to make a network
2019 connection to a disjoint part of the IP space. It was doing the
2020 math incorrectly for one set of cases.
2025 Description: Fix for potential infinite loop processing records in an llog.
2026 Details : If an llog record is corrupted/zeroed, it is possible to loop
2027 forever in llog_process(). Validate the llog record length
2028 and skip the remainder of the block on error.
2031 Frequency : occasional (liblustre only)
2033 Description: liblustre could not open files whose last component is a symlink
2034 Details : sysio_path_walk() would incorrectly pass the open intent to
2035 intermediate path components.
2038 Frequency : rare (liblustre only with non-standard tuning)
2039 Bugzilla : 7201 (7350)
2040 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2041 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2042 after it has been released, only do this for referenced locks
2044 Severity : enhancement
2046 Description: specifying an (invalid) directory default stripe_size of -1
2047 would reset the directory default striping
2048 Details : stripe_size -1 was used internally to signal directory stripe
2049 removal, now use "all default" to signal dir stripe removal
2050 as a directory striping of "all default" is not useful
2053 Frequency : common for large clusters running liblustre clients
2055 Description: doing an ls when liblustre clients are running is slow
2056 Details : sending a glimpse AST to a liblustre client waits for every AST
2057 to time out, as liblustre clients will not respond. Since they
2058 cannot cache data we refresh the OST lock LVB from disk instead.
2060 Severity : enhancement
2062 Description: doing an ls at the same time as file IO can be slow
2063 Details : enqueue and other "small" requests can be blocked behind many
2064 large IO requests. Create a new OST IO portal for non-IO
2065 requests so they can be processed faster.
2068 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2070 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2071 created file with mode 000
2072 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2073 allow an ftruncate() as a non-root user to a file with mode 000.
2074 Lustre now respects this flag to disable mode checks when
2075 truncating a file owned by the user
2078 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2081 Description: Revoking locks from clients that went dead or catatonic might take
2083 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2084 cancellation of such locks instant on servers without waiting for
2085 any reply from clients. Clients drops these locks when cancel
2086 notification from server is received without replying.
2089 Frequency : liblustre-only, when liblustre client dies or becomes busy
2091 Description: Doing ls on Linux clients can take a long time with active
2093 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2094 granting such locks to it in the first place if possible. Locks
2095 are taken by proxy on the OST during the read or write and
2096 dropped immediately afterward. Add connect flags handling, do
2097 not grant locks to liblustre clients for glimpse ASTs.
2099 Severity : enhancement
2101 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2102 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2103 sequential reads and behaves badly with applications doing random
2104 reads. Improve it by reading ahead at least read region, and
2105 avoiding excessive large RPC for small reads.
2107 Severity : enhancement
2109 Description: Creating more than 1000 files for a single job may cause a load
2110 imbalance on the OSTs if there are also a large number of OSTs.
2111 Details : qos_prep_create() uses an OST index reseed value that is an
2112 even multiple of the number of available OSTs so that if the
2113 reseed happens in the middle of the object allocation it will
2114 still utilize the OSTs as uniformly as possible.
2119 Description: OST or MDS may oops in ping_evictor_main()
2120 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2121 but doesn't restart at beginning of obd_exports_timed list
2124 Severity : enhancement
2126 Description: improve by-nid export eviction on the MDS and OST
2127 Details : allow multiple exports with the same NID to be evicted at one
2128 time without re-searching the exports list.
2131 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2133 Description: MDS may oops in groups_free()
2134 Details : in rare race conditions a newly allocated group_info struct is
2135 freed again, and this can be NULL. The 2.4 compatibility code
2136 for groups_free() checked for a NULL pointer, but 2.6 did not.
2139 Frequency : common for liblustre clients doing little filesystem IO
2140 Bugzilla : 9352, 7313
2141 Description: server may evict liblustre clients accessing contended locks
2142 Details : if a client is granted a lock or receives a completion AST
2143 with a blocking AST already set it would not reply to the AST
2144 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2147 Frequency : lfs setstripe, only systems with more than 160 OSTs
2149 Description: unable to set striping with a starting offset beyond OST 160
2150 Details : llapi_create_file() incorrectly limited the starting stripe
2151 index to the maximum single-file stripe count.
2154 Frequency : LDAP users only
2156 Description: lconf did not handle in-kernel recovery with LDAP properly
2157 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2159 Severity : enhancement
2161 Description: bind OST threads to NUMA nodes to improve performance
2162 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2163 node and do their allocations there to localize memory access
2165 Severity : enhancement
2167 Description: llmount can determine client NID directly from Myrinet (GM)
2168 Details : the client NID code from gmnalnid was moved directly into
2169 llmount, removing the need to use this or specifying the
2170 client NID explicitly when mounting GM clients with zeroconf
2173 Frequency : if client is started with down MDS
2175 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2176 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2177 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2178 ptlrpc_import_delayed_req() because we didn't check if the
2179 request was interrupted, and we also didn't break out of the
2180 event loop if there was no timeout
2185 Description: data loss during non-page-aligned writes to a single file from
2186 both multiple nodes and multiple threads on one node at same time
2187 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2188 inconsistency led to false short-reads, that were cached and later
2189 used by ->prepare_write() to fill in partially written page,
2190 leading to data loss.
2193 Frequency : always, if lconf --abort_recovery used
2195 Description: lconf --abort_recovery fails with 'Operation not supported'
2196 Details : lconf was attempting to abort recovery on the MDT device and not
2199 Severity : enhancement
2201 Description: remove cleanup logs
2202 Details : replace lconf-generated cleanup logs with lustre internal
2203 cleanup routines. Eliminates the need for client-cleanup and
2206 Severity : enhancement
2208 Description: add support for EAs (user and system) on lustre filesystems
2209 Details : it is now possible to store extended attributes in the Lustre
2210 client filesystem, and with the user_xattr mount option it
2211 is possible to allow users to store EAs on their files also
2213 Severity : enhancement
2215 Description: Add possibility (config option) to show minimal available OST free
2217 Details : When compiled with --enable-mindf configure option, statfs(2)
2218 (and so, df) will return least minimal free space available from
2219 all OSTs as amount of free space on FS, instead of summary of
2220 free spaces of all OSTs.
2222 Severity : enhancement
2224 Description: do not expand extent locks acquired on OST-side
2225 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2226 by server: they are not cached anyway.
2229 Frequency : when mmap is used/binaries executed from Lustre
2231 Description: Unmmap pages before throwing them away from read cache.
2232 Details : llap_shrink cache now attempts to unmap pages before discarding
2233 them (if unmapping failed - do not discard). SLES9 kernel has
2234 extra checks that trigger if this unmapping is not done first.
2239 Description: lconf didn't resolve symlinks before checking to see whether a
2240 given mountpoint was already in use
2243 Frequency : when migrating failover services
2244 Bugzilla : 6395, 9514
2245 Description: When migrating a subset of services from a node (e.g. failback
2246 from a failover service node) the remaining services would
2247 time out and evict clients.
2248 Details : lconf --force (implied by --failover) sets the global obd_timeout
2249 to 5 seconds in order to quickly disconnect, but this caused
2250 other RPCs to time out too quickly. Do not change the global
2251 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2253 Severity : enhancement
2254 Frequency : if MDS is started with down OST
2255 Bugzilla : 9439,5706
2256 Description: Allow startup/shutdown of an MDS without depending on the
2257 availability of the OSTs.
2258 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2259 Add appropriate locking and lov-osc refcounts for safe
2260 cleaning. Add osc abort_inflight calls in case the
2261 synchronize never started.
2264 Frequency : occasional (Cray XT3 only)
2266 Description: root not authorized to access files in CRAY_PORTALS environment
2267 Details : The client process capabilities were not honoured on the MDS in
2268 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2269 been accessed by an authorized user then root was able to access
2270 the file on the local client also. The root user capabilities
2271 are now allowed on the MDS, as this environment has secure UID.
2274 Frequency : occasional
2276 Description: ldiskfs "too long searching" message happens too often
2277 Details : A debugging message (otherwise harmless) prints too often on
2278 the OST console. This has been reduced to only happen when
2279 there are fragmentation problems on the filesystem.
2284 Description: Division by zero in statfs when all OSCs are inactive
2285 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2286 lov_statfs divides by value returned by lov_get_stripecnt().
2290 Bugzilla : 9489, 3273
2291 Description: First write from each client to each OST was only 4kB in size,
2292 to initialize client writeback cache, which caused sub-optimal
2293 RPCs and poor layout on disk for the first writen file.
2294 Details : Clients now request an initial cache grant at (re)connect time
2295 and so that they can start streaming writes to the cache right
2296 away and always do full-sized RPCs if there is enough data.
2297 If the OST is rebooted the client also re-establishes its grant
2298 so that client cached writes will be honoured under the grant.
2303 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2304 Details : Now I/O RPCs go to different portal number and (presumably) fast
2305 lock requests (and glimses) and other RPCs get their own service
2306 threads pool that should be able to service those RPCs
2309 Severity : enhancement
2311 Description: Ability to exchange lustre version between client and servers and
2312 issue warnings at client side if client is too old. Also for
2313 liblustre clients there is ability to refuse connection of too old
2315 Details : New 'version' field is added to connect data structure that is
2316 filled with version info. That info is later checked by server and
2320 Frequency : rare, liblustre only.
2321 Bugzilla : 9296, 9581
2322 Description: Two simultaneous writes from liblustre at offset within same page
2323 might proceed at the same time overwriting eachother with stale
2325 Details : I/O lock withing llu_file_prwv was released too early, before data
2326 actually was hitting the wire. Extended lock-holding time until
2327 server acknowledges receiving data.
2330 Frequency : extremely rare. Never observed in practice.
2332 Description: avoid generating lustre_handle cookie of 0.
2333 Details : class_handle_hash() generates handle cookies by incrementing
2334 global counter, and can hit 0 occasionaly (this is unlikely, but
2335 not impossible, because initial value of cookie counter is
2336 selected randonly). Value of 0 is used as a sentinel meaning
2337 "unassigned handle" --- avoid it. Also coalesce two critical
2338 sections in this function into one.
2340 Severity : enhancement
2342 Description: allow liblustre clients to delegate truncate locking to OST
2343 Details : To avoid overhead of locking, liblustre client instructs OST to
2344 take extent lock in ost_punch() on client's behalf. New connection
2345 flag is added to handle backward compatibility.
2347 Severity : enhancement
2348 Bugzilla : 4928, 7341, 9758
2349 Description: allow number of OST service threads to be specified
2350 Details : a module parameter allows the number of OST service threads
2351 to be specified via "options ost ost_num_threads={N}" in the
2352 OSS's /etc/modules.conf or /etc/modprobe.conf.
2356 Bugzilla : 6146, 9635, 9895
2357 Description: servers crash with bad pointer in target_handle_connect()
2358 Details : In rare cases when a client is reconnecting it was possible that
2359 the connection request was the last reference for that export.
2360 We would temporarily drop the export reference and get a new
2361 one, but this may have been the last reference and the export
2362 was just destroyed. Get new reference before dropping old one.
2364 Severity : enhancement
2365 Frequency : if client is started with failover MDS
2367 Description: Allow multiple MDS hostnames in the mount command
2368 Details : Try to read the configuration from all specified MDS
2369 hostnames during a client mount in case the "primary"
2372 Severity : enhancement
2374 Description: Stop sending data to evicted clients as soon as possible.
2375 Details : Check if the client we are about to send or are sending data to
2376 was evicted already. (Check is done every second of waiting,
2377 for which l_wait_event interface was extended to allow checking
2378 of exit condition at specified intervals).
2381 Frequency : rare, normally only when NFS exporting is done from client
2383 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2385 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2388 Frequency : always, if srand() is called before liblustre initialization
2390 Description: Liblustre uses system PRNG disturbing its usage by user application
2391 Details : Introduce internal to lustre fast and high-quality PRNG for
2392 lustre usage and make liblustre and some other places in generic
2393 lustre code to use it.
2395 Severity : enhancement
2396 Bugzilla : 9477, 9557, 9870
2397 Description: Verify that the MDS configuration logs are updated when xml is
2398 Details : Check if the .xml configuration logs are newer than the config
2399 logs stored on the MDS and report an error if this is the case.
2400 Request --write-conf, or allow starting with --old_conf.
2402 Severity : enhancement
2404 Description: Handle symlinks in the path when checking if Lustre is mounted.
2405 Details : Resolve intermediate symlinks when checking if a client has
2406 mounted a filesystem to avoid duplicate client mounts.
2411 Description: lconf can hit an error exception but still return success.
2412 Details : The lconf command catches the Command error exception at the top
2413 level script context and will exit with the associated exit
2414 status, but doesn't ensure that this exit status is non-zero.
2419 Description: failure of ptlrpc thread startup can cause oops
2420 Details : Starting a ptlrpc service thread can fail if there are a large
2421 number of threads or the server memory is very fragmented.
2422 Handle this without oopsing.
2425 Frequency : always, only if liblustre and non-default acceptor port was used
2427 Description: liblustre cannot connect to servers with non-default acceptor port
2428 Details : tcpnal_set_default_params() was not called and was therefore
2429 ignoring the environment varaible TCPNAL_PORT, as well as other
2430 TCPNAL_ environment variables
2435 Description: two objects could be created on the same OST for a single file
2436 Details : If an OST is down, in some cases it was possible to create two
2437 objects on a single OST for a single file. No problems other
2438 than potential performance impact and spurious error messages.
2442 Bugzilla : 5681, 9562
2443 Description: Client may oops in ll_unhash_aliases
2444 Details : Client dcache may become inconsistent in race condition.
2445 In some cases "getcwd" can fail if the current directory is
2451 Description: Inode refcounting problems in NFS export code
2452 Details : link_raw functions used to call d_instantiate without obtaining
2453 extra inode reference first.
2457 Bugzilla : 9942, 9903
2458 Description: Referencing freed requests leading to crash, memleaks with NFS.
2459 Details : We used to require that call to ll_revalidate_it was always
2460 followed by ll_lookup_it. Also with revalidate_special() it is
2461 possible to call ll_revalidate_it() twice for the same dentry
2462 even if first occurence returned success. This fix changes semantic
2463 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2464 reference on a request referred from the intent.
2465 ll_intent_release() then releases such a request.
2468 Frequency : rare, normally benchmark loads only
2470 Description: unlinked inodes were kept in memory on the client
2471 Details : If a client is repeatedly creating and unlinking files it
2472 can accumulate a lot of stale inodes in the inode slab cache.
2473 If there is no other client load running this can cause the
2474 client node to run out of memory. Instead flush old inodes
2475 from client cache that have the same inode number as a new inode.
2478 Frequency : SLES9 2.6.5 kernel and long filenames only
2479 Bugzilla : 9969, 10379
2480 Description: utime reports stale NFS file handle
2481 Details : SLES9 uses out-of-dentry names in some cases, which confused
2482 the lustre dentry revalidation. Change it to always use the
2486 Frequency : rare, unless heavy write-truncate concurrency is continuous
2487 Bugzilla : 4180, 6984, 7171, 9963, 9331
2488 Description: OST becomes very slow and/or deadlocked during object unlink
2489 Details : filter_destroy() was holding onto the parent directory lock
2490 while truncating+unlinking objects. For very large objects this
2491 may block other threads for a long time and slow overall OST
2492 responsiveness. It may also be possible to get a lock ordering
2493 deadlock in this case, or run out of journal credits because of
2494 the combined truncate+unlink. Solution is to do object truncate
2495 first in one transaction without parent lock, and then do the
2496 final unlink in a new transaction with the parent lock. This
2497 reduces the lock hold time dramatically.
2500 Frequency : rare, 2.4 kernels only
2502 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2503 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2504 before unmounting their filesystem in order to avoid deadlock.
2505 The kernel_locked() function in 2.4 kernels only checks whether
2506 the kernel lock is held, not whether it is this process that is
2507 holding it as 2.6 kernels do.
2512 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2513 Details : The client ptlrpc code may be trying to reconnect to a down
2514 server before a previous connection attempt has timed out.
2515 Increase the reconnect interval to be longer than the connection
2516 timeout interval to avoid sending duplicate connections to
2520 Frequency : echo_client brw_test command
2522 Description: fix echo_client to work with OST preallocated code
2523 Details : OST preallocation code (5137) didn't take echo_client IO path
2524 into account: echo_client calls filter methods outside of any
2525 OST thread and, hence, there is no per-thread preallocated
2526 pages and buffers to use. Solution: hijack pga pages for IO. As
2527 a byproduct, this avoids unnecessary data copying.
2531 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2532 Description: Client can oops in mdc_commit_close() after open replay
2533 Details : It was possible for the MDS to return an open request with no
2534 transaction number in mds_finish_transno() if the client was
2535 evicted, but without actually returning an error. Clients
2536 would later try to replay that open and may trip an assertion
2537 Simplify the client close codepath, and always return an error
2538 from the MDS in case the open is not successful.
2541 Frequency : rare, 2.6 OSTs only
2543 Description: OST may deadlock under high load on fragmented files
2544 Details : If there was a heavy load and highly-fragmented OST filesystems
2545 it was possible to have all the OST threads deadlock waiting on
2546 allocation of biovecs, because the biovecs were not released
2547 until the entire RPC IO was completed. Instead, release biovecs
2548 as soon as they are complete to ensure forward IO progress.
2550 Severity : enhancement
2552 Description: Support for specifying external journal device at mount
2553 Details : If an OST or MDS device is formatted with an external journal
2554 device, this device major/minor is stored in the ext3 superblock
2555 and may not be valid for failover. Allow detecting and
2556 specifying the external journal at mount time.
2561 Description: Mounting an MDS with pending unlinked files may cause oops
2562 Details : target_finish_recovery() calls mds_postrecov() which returned
2563 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2564 considers this an error and immediately begins cleaning up the
2565 lov, just after starting the mds_lov process
2567 Severity : enhancement
2569 Description: Implement 'lfs df' to report actual free space on per-OST basis
2570 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2571 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2572 usage of objects; '-h' to report in human readable format.
2574 ------------------------------------------------------------------------------
2576 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2583 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2584 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2585 at startup during orphan unlinking. Instead, skip buddy update
2586 before setup, as it will be regenerated after recovery is complete.
2591 Description: If an OST is inactive, its locks might reference stale inodes.
2592 Details : lov_change_cbdata() must iterate over all namespaces, even if
2593 they are inactive to clear inode references from the lock.
2595 Severity : enhancement
2596 Frequency : occasional, if non-standard max_dirty_mb used
2598 Description: Client will block write RPCs if not enough grant
2599 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2600 then the client will block writes while waiting for another RPC
2601 to complete instead of consuming its dirty limit. With change
2602 we get improved performance when max_dirty_mb is small.
2604 Severity : enhancement
2605 Bugzilla : 3389, 6253
2606 Description: Add support for supplementary groups on the MDS.
2607 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2608 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2609 lookups for user supplementary groups into a cache.
2613 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2614 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2615 for liblustre O_EXCL is enforced by MDS.
2619 Description: readdir never returns NULL in liblustre.
2620 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2621 dentry in struct dirent.
2625 Frequency : liblustre only, depends on application IO pattern
2626 Description: liblustre clients evicted if not contacting servers
2627 Details : Don't put liblustre clients into the ping_evictor list, so
2628 they will not be evicted by the pinger ever.
2630 Severity : enhancement
2632 Description: Add ability to evict clients by NID from MDS.
2633 Details : By echoing "nid:$NID" string into
2634 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2635 $NID would be instantly evicted from this MDS and from all active
2636 OSTs connected to it.
2640 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2641 Details : lookup_it_finish() used to query file size from OSTs that was not
2646 Description: service threads change working directory to that of init
2647 Details : Starting lustre service threads may pin the working directory
2648 of the parent thread, making that filesystem busy. Threads
2649 now change to the working directory of init to avoid this.
2653 Frequency : during shutdown only
2654 Description: shutdown with a failed MDS or OST can cause unmount to hang
2655 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2660 Frequency : 2.6 only
2661 Description: chmod/chown may include an extra supplementary group
2662 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2663 supplementary group and if none is specified this is used.
2666 Bugzilla : 5479 (6816)
2668 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2669 Details : If lookup is in progress on a file that is unlinked we might try
2670 to revalidate the inode and fail in revalidate after lookup is
2671 complete and ll_file_open() enqueues the open again but
2672 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2675 Frequency : always, if lconf --abort_recovery used
2677 Description: lconf --abort_recovery fails with 'Operation not supported'
2678 Details : lconf was attempting to abort recovery on the MDT device and not
2681 ------------------------------------------------------------------------------
2683 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2688 Frequency : rare (only unsupported configurations with a node running as an
2690 Bugzilla : 6514, 5137
2691 Description: Mounting a Lustre file system on a node running as an OST could
2693 Details : OSTs now preallocates memory needed to write out data at
2694 startup, instead of when needed, to avoid having to
2695 allocate memory in possibly low memory situations.
2696 Specifically, if the file system is mounted on on OST,
2697 memory pressure could force it to try to write out data,
2698 which it needed to allocate memory to do. Due to the low
2699 memory, it would be unable to do so and the node would
2700 become unresponsive.
2702 Severity : enhancement
2704 Description: Addition of lconf --service command line option
2705 Details : lconf now accepts a '--service <arg>' option, which is
2706 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2708 Severity : enhancement
2710 Description: Failover mode is now the default for OSTs.
2711 Details : By default, OSTs will now run in failover mode. To return to
2712 the old behaviour, add '--failout' to the lmc line for OSTs.
2714 Severity : enhancement
2716 Description: Health checks are now provided for MDS and OSTs
2717 Details : Additional detailed health check information on MSD and OSTs
2718 is now provided through the procfs health_check value.
2721 Frequency : occasional, depends on IO load
2723 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2724 after numerous create/delete cycles
2725 Details : The ext3 inode allocation policy would not allocate new inodes
2726 very well on the OSTs because there are no new directories
2727 being created. Instead we look for groups with free space if
2728 the parent directories are nearly full.
2733 Description: Network or server problems during mount may cause partially
2734 mounted clients instead of returning an error.
2735 Details : The config llog parsing code may overwrite the error return
2736 code during mount error handling, returning success instead
2742 Description: MDS can fail to allocate large reply buffers
2743 Details : After long uptimes the MDS can fail to allocate large reply
2744 buffers (e.g. zconf client mount config records) due to memory
2745 fragmentation or consumption by the buffer cache. Preallocate
2746 some large reply buffers so that these replies can be sent even
2747 under memory pressure.
2751 Frequency : rare (liblustre)
2752 Description: fsx running with liblustre complained that using truncate() to
2753 extend the file doesn't work. This patch corrects that issue.
2754 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2755 ATTR_SIZE and lsm use in llu_setattr_raw.
2759 Frequency : rare, only 2.6 kernels
2760 Description: Unusual file access patterns on the MDS may result in inode
2761 data being lost in very rare circumstances.
2762 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2763 code in the 2.6 kernel caused the inode and/or EA data not to
2764 be read from disk, causing single-file corruption.
2768 Frequency : rare, only 2.6 filesystems using extents
2769 Description: Heavy concurrent write and delete load may cause data corruption.
2770 Details : It was possible under high-load situations to have an extent
2771 metadata block in the block device cache from a just-unlinked
2772 file overwrite a newly-allocated data block. We now unmap any
2773 metadata buffers that alias just-allocated data blocks.
2777 Frequency : filesystems with default stripe_count larger than 77
2778 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2779 Details : lconf specifies an inode size of 4096 bytes when the default
2780 stripe_count is larger than 77. This conflicts with the default
2781 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2782 this case to avoid pinning too much memory for large EAs.
2784 ------------------------------------------------------------------------------
2786 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2791 Frequency : rare (extremely heavy IO load with hundreds of clients)
2793 Description: Client is evicted, gets IO error writing to file
2794 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2795 caused clients to be evicted for AST timeouts. The fixes in
2796 bug 5192 mean we no longer need to have such short AST timeouts
2797 so ldlm_timeout has been increased.
2800 Frequency : occasional during --force or --failover shutdown under load
2801 Bugzilla : 5949, 4834
2802 Description: Server oops/LBUG if stopped with --force or --failover under load
2803 Details : a collection of import/export refcount and cleanup ordering
2804 issues fixed for safer force cleanup
2807 Frequency : only filesystems larger than 120 OSTs
2808 Bugzilla : 5990, 6223
2809 Description: lfs getstripe would oops on a very large filesystem
2810 Details : lov_getconfig used kfree on vmalloc'd memory
2813 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2814 Bugzilla : 6242, 6243
2815 Description: reading from files that had been truncated to a non-zero size
2816 but never opened returned no data
2817 Details : ll_file_read() reads zeros from no-object files to EOF
2822 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2823 Details : There's little harm in aborting MDS/OSS recovery and letting it
2824 try again, so I removed the LASSERT and return an error instead.
2826 Severity : enhancement
2828 Description: New debugging infrastructure for tracking down data corruption
2829 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2830 (b) cover more of the client-side code path, and (c) try to narrow
2831 down where problems occurred
2835 Bugzilla : 3819, 4364, 4397, 6313
2836 Description: Racing close and eviction MDS could cause assertion in mds_close
2837 Details : It was possible to get multiple mfd references during close and
2838 client eviction, leading to one thread referencing a freed mfd.
2840 Severity: : enhancement
2841 Bugzilla : 3262, 6359
2842 Description: Attempts to reconnect to servers are now more aggressive.
2843 Details : This builds on the enhanced upcall-less recovery that was added
2844 in 1.4.2. When trying to reconnect to servers, clients will
2845 now try each server in the failover group every 10 seconds. By
2846 default, clients would previously try one server every 25 seconds.
2851 Description: After recovery, certain operations trigger a failed
2852 assertion on a client.
2853 Details : Failing over an mds, using lconf -d --failover, while a
2854 client was doing a readdir() call would cause the client to
2855 LBUG after recovery completed and the readdir() was resent.
2857 Severity : enhancement
2859 Description: Default groups are now added by lconf
2860 Details : You can now run lconf --group <servicename> without having to
2861 manually add groups with lmc.
2864 Frequency : occasional
2866 Description: Nodes with an elan id of 0 trigger a failed assertion
2869 Frequency : always when accessing e.g. tty/console device nodes
2871 Description: tty and some other devices nodes cannot be used on lustre
2872 Details : file's private_data field is used by device data and lustre
2873 values in there got lost. New field was added to struct file to
2874 store fs-specific private data.
2877 Frequency : when exporting Lustre via NFS
2879 Description: NFSD failed occasionally when looking up a path component
2880 Details : NFSD is looking up ".." which was broken in ext3 directories
2881 that had grown large enough to become hashed.
2884 Frequency : Clusters with multiple interfaces not on the same subnet
2886 Description: Nodes will repeatedly try to reconnect to an interface which it
2887 cannot reach and report an error to the log.
2888 Details : Extra peer list entries will be created by lconf with some peers
2889 unreachable. lconf now validates the peer before adding it.
2892 Frequency : Only if a default stripe is set on the filesystem root.
2894 Description: Setting a default stripe on the filesystem root prevented the
2895 filesystem from being remounted.
2896 Details : The client was sending extra request flags in the root getattr
2897 request and did not allocate a reply buffer for the dir EA.
2900 Frequency : occasional, higher if lots of files are accessed by one client
2901 Bugzilla : 6159, 6097
2902 Description: Client trips assertion regarding lsm mismatch/magic
2903 Details : While revalidating inodes the VFS looks up inodes with ifind()
2904 and in rare cases can find an inode that is being freed.
2905 The ll_test_inode() code will free the lsm during ifind()
2906 when it finds an existing inode and then the VFS later attaches
2907 this free lsm to a new inode.
2911 Bugzilla : 6422, 7030
2912 Description: MDS deadlock between mkdir and client eviction
2913 Details : Creating a new file via mkdir or mknod (starting a transaction
2914 and getting the ns lock) can deadlock with client eviction
2915 (gets ns lock and trying to finish a synchronous transaction).
2918 Frequency : occasional
2919 Description: While starting a server, the fsfilt_ext3 module could not be
2921 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2922 kernels. Previously, lconf would still use the ext3 name
2923 when trying to load modules. Now, it will correctly use
2924 ext3 on 2.4 and ldiskfs on 2.6.
2926 Severity : enhancement
2927 Description: The default stripe count has been changed to 1
2928 Details : The interpretation of the default stripe count (0, to lfs
2929 or lmc) has been changed to mean striping across a single
2930 OST, rather than all available. For general usage we have
2931 found a stripe count of 1 or 2 works best.
2933 Severity : enhancement
2934 Description: Add support for compiling against Cray portals.
2935 Details : Conditional compiling for some areas that are different
2939 Frequency : occasional
2940 Bugzilla : 6409, 6834
2941 Description: Creating files with an explicit stripe count may lead to
2942 a failed assertion on the MDS
2943 Details : If some OSTs are full or unavailable, creating files may
2944 trigger a failed assertion on the MDS. Now, Lustre will
2945 try to use other servers or return an error to the
2949 Frequency : occasional
2951 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2952 caused lock timeout during readahead (since 1.4.2).
2953 Details : Processes doing readahead might match a lock that hasn't been
2954 granted yet if there are overlapping and conflicting lock
2955 requests. The readahead process waits on ungranted lock
2956 (original lock is CBPENDING), while OST waits for that process
2957 to cancel CBPENDING read lock and eventually evicts client.
2959 Severity : enhancement
2961 Description: Initial enabling of flock support for clients
2962 Details : Implements fcntl advisory locking and file status functions.
2963 This feature is provided as an optional mount flag (default
2964 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2965 locking are implemented yet, and those that are are not guaranteed
2966 to be completely correct in production environments.
2967 mount -t lustre -o [flock|noflock] ...
2970 Frequency : occasional
2972 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2973 assertion in the ext3 JBD (journaling) layer.
2974 Details : The b_committed_data struct is protected by the big kernel lock
2975 in 2.4 kernels, serializing journal_commit_transaction() and
2976 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2977 access to this struct. In 2.6 kernels there is finer grained
2978 locking to improve SMP performance of the JBD layer.
2982 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2987 Description: merge and cleanup kernel patches.
2990 -----------------------------------------------------------------------------
2992 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2994 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2995 versions of Lustre. Please update all servers and clients to
2996 version 1.4.2 or later at the same time. You must also run
2997 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2998 to update the configuration logs.
3000 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3001 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3002 - swab llog records of type '0' so we get proper header size/idx (5861)
3003 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3004 - fix rename of one directory over another leaking an inode (5953)
3005 - avoid SetPageDirty on 2.6 (5981)
3006 - don't re-add just-being-destroyed locks to the waiting list (5653)
3007 - when creating new directories, inherit the parent's custom
3008 striping settings if present parent (3048)
3009 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3010 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3011 - don't hold client locks on temporary worklist from l_lru (5666)
3012 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3013 - automatically evict dead clients (5921)
3014 - Update file size properly in create+truncate+fstat case (6196)
3015 - Do not unhash mountpoint dentries, do not allow removal of
3017 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3018 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3019 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3020 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3021 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3022 - Correctly handle reads of files with no objects (6243)
3023 - lctl recover will also mark a device active if deactivate used (5933)
3025 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3026 - allow --write-conf on an MDS with different nettype than client (5619)
3027 - don't write config llogs to MDS for mounts not from that MDS (5617)
3028 - lconf should create multiple TCP connections from a client (5201)
3029 - init scripts are now turned off by default; run chkconfig --on
3030 lustre and chkconfig --on lustrefs to use them
3031 - upcalls are no longer needed for clients to recover to failover
3033 - add --abort-recovery option to lconf to abort recovery on device
3035 - add support for an arbitrary number of OSTs (3026)
3036 - Quota support protocol changes.
3037 - forward compatibility changes to wire structs (6007)
3038 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3039 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3040 - improved "lustre status" script
3041 - initialize blocksize for non-regular files (6062)
3042 - added --disable-server and --disable-client configure options (5782)
3043 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3044 - Vanilla 2.4.29 support
3045 - increase maximum number of obd devices to 520 (6242)
3046 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3047 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3048 SLES 9 2.6.5 kernels
3049 - limit stripes per file to 160 (the maximum EA size) (6093)
3051 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3054 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3055 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3056 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3057 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3058 - small ext3 extents cleanups and fixes (5733)
3059 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3060 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3061 - avoid needless client->OST connect, fix handle mismatch (5317)
3062 - fix DLM error path that led to out-of-sync client, long delays (5779)
3063 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3064 - fix several locking issues related to i_size (5492,5624,5654,5672)
3065 - don't move pending lock onto export if it is already evicted (5683)
3066 - fix kernel oops when creating .foo in unlinked directory (5548)
3067 - fix deadlock in obdfilter statistics vs. object create (5811)
3068 - use time_{before,after} to avoid timer jiffies wrap (5882)
3069 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3070 - Do not leak request if server was not able to process it (5154)
3071 - If mds_open unable to find parent dir, make that negative lookup(5154)
3072 - don't create new directories with extent-mapping (5909, 5936)
3074 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3075 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3076 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3077 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3078 - support for RHEL4 kernel on i686 (5773)
3079 - provide error messages when incompatible logs are encountered (5898)
3081 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3082 * version 1.4.0.10 (1.4.1 release candidate 1)
3084 - don't keep a lock reference when lock is not granted (4238)
3085 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3086 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3087 - drop import inflight refcount on signal_completed_replay error (5255)
3088 - unlock page after async write error during send (3677)
3089 - handle missing objects in filter_preprw_read properly (5265)
3090 - no transno return for symlink open, don't save no-trasno open (3440)
3091 - don't try to complete elan receive that already failed (4012)
3092 - free RPC server reply state on error (5406)
3093 - clean up thread from ptlrpc_start_thread() on error (5160)
3094 - readahead could read extra page into cache that wasn't ejected (5388)
3095 - prevent races in class_attach/setup/cleanup/detach (5260)
3096 - don't dereference de->d_inode after l_dput of de (5458)
3097 - use "int" for stripe value returned from lock_to_stripe (5544)
3098 - mballoc allocation and error-checking fixes in 2.6 (5504)
3099 - block device patches to fix I/O request sizes in 2.6 (5482)
3100 - look up hostnames for IB nals (5602)
3101 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3102 - don't start multiple acceptors for the same port (5277)
3103 - fix incorrect LASSERT in mds_getattr_name (5635)
3104 - export a proc file for general "ping" checking (5628)
3105 - fix "lfs check" to not block when the MDS is down (5628)
3107 - service request history (4965)
3108 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3109 - create an "evict_client" /proc entry on OSTs, like the MDS has
3110 - fix mount usage message, return errors per mount(8) (5168)
3111 - change grep [] to grep "[]" in tests so they work in more UMLs
3112 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3113 - remove ancient LOV_MAGIC_V0 EA support (5047)
3114 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3115 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3116 - allow readahead allocations to fail when low on memory (5383)
3117 - mmap locking landed again, after considerable improvement (2828)
3118 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3120 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3123 - send OST transaction number in read/write reply to free req (4966)
3124 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3125 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3126 - the watchdog thread now runs as interruptible (5246)
3127 - flock/lockf fixes (but it's still disabled, pending 5135)
3128 - don't use EXT3 constants in llite code (5094)
3129 - memory shortage at startup could cause assertion (5176)
3131 - reorganization of lov code
3132 - single portals codebase
3134 - add extents/mballoc support (5025)
3135 - direct I/O reads in the obdfilter (4048)
3136 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3138 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3141 - send OST transaction number in read/write reply to free req (4966)
3142 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3143 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3144 - the watchdog thread now runs as interruptible (5246)
3145 - handle missing objects in filter_preprw_read properly (5265)
3146 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3147 - drop import inflight refcount on signal_completed_replay error (5255)
3148 - unlock page after async write error during send (3677)
3149 - return original error code on reconstructed replies (3761)
3150 - no transno return for symlink open, don't save no-trasno open (3440)
3152 - add pid to ldlm debugging output (4922)
3153 - bump the watchdog timeouts -- we can't handle 30sec yet
3154 - extra debugging for orphan dentry/inode bug (5259)
3156 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3159 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3160 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3161 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3162 - fix NULL dereference in /proc/sys/portals/routes (4827)
3163 - allow failed mdc_close() operations to be interrupted (4561)
3164 - stop precreate on OST before MDS would time out on it (4778)
3165 - don't send partial-page writes before EOF from client (4410)
3166 - discard client grant for sub-page writes on large-page clients (4520)
3167 - don't free dentries not owned by NFS code, check generation (4806)
3168 - fix lsm leak if mds_create_objects() fails (4801)
3169 - limit debug_daemon file size, always print CERROR messages (4789)
3170 - use transno after validating reply (3892)
3171 - process timed out requests if import state changes (3754)
3172 - update mtime on OST during writes, return in glimpse (4829)
3173 - add mkfsoptions to LDAP (4679)
3174 - use ->max_readahead method instead of zapping global ra (5039)
3175 - don't interrupt __l_wait_event() during strace
3177 - add software watchdogs to catch hung threads quickly (4941)
3178 - make lustrefs init script start after nfs is mounted
3179 - fix CWARN/ERROR duplication (4930)
3180 - return async write errors to application if possible (2248)
3181 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3182 - print NAL number in %x format (4645)
3183 - update barely-supported suse-2.4.21-171 series (4842)
3184 - support for sles 9 %post scripts
3185 - support for building 2.6 kernel-source packages
3186 - support for sles km_* packages
3188 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3191 - ignore -ENOENT errors in osc_destroy (3639)
3192 - notify osc create thread that OSC is being cleaned up (4600)
3193 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3194 - reconstruct ost_handle() like mds_handle() (4657)
3195 - create a new thread to do import eviction to avoid deadlock (3969)
3196 - let lconf resolve symlinked-to devices (4629)
3197 - don't unlink "objects" from directory with default EA (4554)
3198 - hold socknal file ref over connect in case target is down (4394)
3199 - allow more than 32000 subdirectories in a single directory (3244)
3200 - fix blocks count for O_DIRECT writes (3751)
3201 - OST returns ENOSPC from object create when no space left (4539)
3202 - don't send truncate RPC if file size isn't changing (4410)
3203 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3204 - bind to privileged port in socknal and tcpnal (3689)
3206 - rate limit CERROR/CWARN console message to avoid overload (4519)
3207 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3208 - basic mmap support (3918)
3209 - kernel patch series update from b1_4 (4711)
3211 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3214 - avoid crash during MDS cleanup with OST shut down (2775)
3215 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3216 - don't use bad inodes on the MDS (3744)
3217 - dynamic object preallocation to improve recovery speed (4236)
3218 - don't hold spinlock over lock dumping or change debug flags (4401)
3219 - don't zero obd_dev when it is force cleaned (3651)
3220 - print grants to console if they go negative (4431)
3221 - "lctl deactivate" will stop automatic recovery attempts (3406)
3222 - look for existing locks in ldlm_handle_enqueue() (3764)
3223 - don't resolve lock handle twice in recovery avoiding race (4401)
3224 - revalidate should check working dir is a directory (4134)
3226 - don't always mark "slow" obdfilter messages as errors (4418)
3228 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3231 - don't close LustreDB during write_conf until it is done (3860)
3232 - fix typo in lconf for_each_profile (3821)
3233 - allow dumping logs from multiple threads at one time (3820)
3234 - don't allow multiple threads in OSC recovery (3812)
3235 - fix debug_size parameters (3864)
3236 - fix mds_postrecov to initialize import for llog ctxt (3121)
3237 - replace config semaphore with spinlock (3306)
3238 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3239 - don't allow enqueue to complete on a destroyed export (3822)
3240 - down write_lock before checking llog header bitmap (3825)
3241 - recover from lock replay timeout (3764)
3242 - up llog sem before sending rpc (3652)
3243 - reduce ns lock hold times when setting kms (3267)
3244 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3245 - fix NULL deref and obd_dev leak on setup error (3312)
3246 - replace some LBUG about llog ops with error handling (3841)
3247 - don't match INVALID dentries from d_lookup and spin (3784)
3248 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3249 - fix invalid assertion in ptlrpc_set_wait (3880)
3251 - add libwrap support for the TCP acceptor (3996)
3252 - add /proc/sys/portals/routes for non-root route listing (3994)
3253 - allow setting MDS UUID in .xml (2580)
3254 - print the stack of a process that LBUGs (4228)
3256 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3259 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3260 - make sure to unset replay flag from failed open requests (3440)
3261 - if default stripe count is 0, use OST count for inode size (3636)
3262 - update parent mtime/ctime on client for create/unlink (2611)
3263 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3264 - free recovery state on server during a forced cleanup (3571)
3265 - unregister_reply for resent reqs (3063)
3266 - loop back devices mounting and status check on 2.6 (3563)
3267 - fix resource-creation race that can provoke i_size == 0 (3513)
3268 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3269 - more debugging for page-accounting assertion (3746)
3270 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3271 - avoid deadlock after precreation failure (3758)
3272 - fix race and lock order deadlock in orphan handling (3450, 3750)
3273 - add validity checks when grabbing inodes from l_ast_data (3599)
3275 - add /proc/.../recovery_status to obdfilter (3428)
3276 - lightweight CDEBUG infrastructure, debug daemon (3668)
3277 - change default OSC RPC parameters to be better on small clusters
3278 - turn off OST read cache for files smaller than 32MB
3279 - install man pages and include them in rpms (3100)
3280 - add new init script for (un)mounting lustre filesystems (2593)
3281 - run chkconfig in %post for init scripts (3701)
3282 - drop scimac NAL (unmaintained)
3284 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3287 - clean kiobufs before and after use (3485)
3288 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3289 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3290 - change init script to fail more clearly if not run as root (1528)
3291 - allow clients to reconnect during replay (1742)
3292 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3293 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3294 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3295 - fix cond_resched (3554)
3296 - only evict unfinished clients after recovery (3515)
3297 - allow bulk resend, prevent data loss (3570)
3298 - dynamic ptlrpc request buffer allocation (2102)
3299 - don't allow unlinking open directory if it isn't empty (2904)
3300 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3301 - remove extraneous obd dereference causing LASSERT failure (3334)
3302 - don't use get_cycles() when creating temp. files on the mds (3156)
3303 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3304 - handle EEXIST for set-stripe, set proper directory name (3336)
3306 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3307 - fix ksocknal_fmb_callback() error messages (2918)
3309 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3312 - don't copy lvb into (possibly NULL) reply on error (2983)
3313 - don't deref dentry after dput, don't free lvb on error (2922)
3314 - use the kms to determine writeback rpc length (2947)
3315 - increment oti_logcookies when osc is inactive (2948)
3316 - update client's i_blocks count via lvb messages (2543)
3317 - handle intent open/close of special files properly (1557)
3318 - mount MDS with errors=remount-ro, like obdfilter (2009)
3319 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3320 - don't use cancelling-locks' kms values (2947)
3321 - use highest lock extent for kms, not last one (2925)
3322 - don't dereference ERR_PTR() dentry in error handling path (3107)
3323 - fix thread race in portals_debug_dumplog() (3122)
3324 - create lprocfs device entries at setup instead of at attach (1519)
3325 - common AST error handler, don't evict client on completion race (3145)
3326 - zero nameidata in detach_mnt in 2.6 (3118)
3327 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3328 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3329 - initialize RPC timeout timer earlier for 2.6 (3219)
3330 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3331 - print nal/nid for unknown nid (3258)
3332 - additional checks for oscc recovery before doing precreate (3284)
3333 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3334 - don't crash in mdc_close for bad permissions on open (3285)
3335 - zero i_rdev for non-device files (3147)
3336 - clear page->private before handing to FS, better assertion (3119)
3337 - tune the read pipeline (3236)
3338 - fix incorrect decref of invalidated dentry (2350)
3339 - provide read-ahead stats and refine rpc in flight stats (3328)
3340 - don't hold journal transaction open across create RPC (3313)
3341 - update atime on MDS at close time (3265)
3342 - close LDAP connection when recovering to avoid server load (3315)
3343 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3344 - don't leak open file on MDS after open resend (3325)
3345 - serialize filter_precreate and filter_destroy_precreated (3329)
3346 - loop device shouldn't call sync_dev() for nul device (3092)
3347 - clear page cache after eviction (2766)
3348 - resynchronize MDS->OST in background (2824)
3349 - refuse to mount the same filesystem twice on same mountpoint (3394)
3350 - allow llmount to create routes for mounting behind routers (3320)
3351 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3352 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3353 - fix rare problem with rename on htree directories (3417)
3355 - allow default OST striping configuration per directory (1414)
3356 - fix compilation for qswnal for 2.6 kernels (3125)
3357 - increase maximum number of MDS request buffers for large systems
3358 - change liblustreapi to be useful for external progs like lfsck (3098)
3359 - increase local configuration timeout for slow disks (3353)
3360 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3362 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3365 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3366 - don't overwrite extent policy data in reply if lock was blocked (2901)
3367 - drop filter export grants atomically with removal from device (2663)
3368 - del obd_self_export from work_list in class_disconnect_exports (2908)
3369 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3370 - swab reply message in mdc_close, other PPC fixes (2464)
3371 - fix destroying of named logs (2325)
3372 - overwrite old logs when running lconf --write_conf (2264)
3373 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3374 - fix race in target_handle_connect (2898)
3375 - mds_reint_create() should take same inode create lock (2926)
3376 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3377 - don't close files for self_export to avoid uninitialized obd (2936)
3378 - allow MDS with the same name as client node (2939)
3379 - hold dentry reference for closed log files for unlink (2325)
3380 - reserve space for all logs during transactions (2059)
3381 - don't evict page beyond end of stripe extent (2925)
3382 - don't oops on a deleted current working directory (2399)
3383 - handle hard links to targets without a parent properly (2517)
3384 - don't dereference NULL lock when racing during eviction (2867)
3385 - don't grow lock extents when lots of conflicting locks (2919)
3387 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3390 - account for cache space usage on clients to avoid data loss (974)
3391 - lfsck support in lustre kernel code (2349)
3392 - reduce journal credits needed for BRW writes (2370)
3393 - orphan handling to avoid losing space on client/server crashes
3394 - ptlrpcd can be blocked, stopping ALL progress (2477)
3395 - use lock value blocks to assist in proper KMS, faster stat (1021)
3396 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3397 - recovery for initial connections (2355)
3398 - fixes for mds_cleanup_orphans (1934)
3399 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3400 - block all file creations until orphan recovery completes (1901)
3401 - client remove rq_connection from request struct (2423)
3402 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3403 - recovery timer race (2670)
3404 - mdc_close recovey bug (2532)
3405 - ptlrpc cleanup bug (2710)
3406 - mds timeout on local locks (2588)
3407 - namespace lock held during RPCs (2431)
3408 - handle interrupted sync write properly (2503)
3409 - don't try to handle a message that hasn't been replied to (2699)
3410 - client assert failure during cleanup after abort recovery (2701)
3411 - leak mdc device after failed mount (2712)
3412 - ptlrpc_check_set allows timedout requests to complete (2714)
3413 - wait for inflight reqs when ptlrpcd finishes (2710)
3414 - make sure unregistered services are removed from the srv_list
3415 - reset bulk XID's when resending them (caught by 1138 test)
3416 - unregister_bulk after timeout
3417 - fix lconf error (2694)
3418 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3419 - readahead locks pages, leaves pending causing memory pressure (2673)
3420 - increase OST request buffers to 4096 on large machines (2729)
3421 - fix up permission of existing directories in simple_mkdir (2661)
3422 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3423 - don't assign transno to errored transactions (2742)
3424 - don't delete objects on OST if given a bogus objid from MDS (2751)
3425 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3426 - if rq_no_resend, then timeout request after recovery (2432)
3427 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3428 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3429 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3430 - MDS recovery shouldn't delete objects that it has given out (2730)
3431 - if enqueue arrives after completion, don't clobber LVB (2819)
3432 - don't unlock pages twice when trigger_group_io returns error (2814)
3433 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3434 - don't write pages to disk if there was an error (1450)
3435 - don't ping imports that have recovery disabled (2676)
3436 - take buffered bytes into account when balancing socknal conn (2817)
3437 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3438 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3439 - remove little-endian swabbing of llog records (1987)
3440 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3441 - retry reposting request buffers if they fail (1191)
3442 - grow extent at grant time to avoid granting a revoked lock (2809)
3443 - lock revoke doesn't evict page if covered by a second lock (2765)
3444 - disable VM readahead to avoid reading outside lock extents (2805)
3446 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3447 - updated kernel patches for hp-2.4.20 kernel (2681)
3449 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3452 - fix truncated write corruption (2366)
3453 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3455 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3456 - ensure len(uuid) < 37 in lmc (1171)
3457 - fix ia64 OOPS in llog_test (2255)
3458 - zero end of page at obdfilter for partial page writes (2648)
3459 - don't leave stale dentries around after renames (bug 2428)
3460 - fix timeouts when evicting a client with a single lock held (2642)
3461 - set deadline for the initial HELLO message to drain (2634)
3462 - print out dotted-quad IP addresses in the socknal (2302)
3464 - additional debugging for MDS client eviction problem (2443)
3465 - fix mkfsoptions support for osts (2603, 2604)
3467 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3470 - add series for the vanilla 2.6.0 kernel
3471 - add series for the vanilla 2.4.24 kernel
3472 - add series for a cray x86/64 UL kernel drop
3473 - fix xattr patches for the vanilla 2.4.19 series
3475 - generate true UUIDs in lmc (1171)
3476 - have portals stack dumping break in UML (2466)
3477 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3478 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3479 - wake callback waiting threads on client eviction (2460)
3480 - Add --ptldebug and --subsystem to lmc (1719)
3481 - update assertion to allow safe interrupt allocation
3482 - set rq_no_resend for cancel requests (2432)
3483 - recalculate ptlrpcd timeout after resend (2494)
3484 - call vfs_rmdir when removing pending directories (2368)
3485 - fix renaming a file to itself (2429)
3486 - lmc creates a default one-stripe lov (2454)
3487 - expand procfs space to handle large clusters (2326)
3488 - increase UML stack to avoid overflow
3489 - update lconf's list of debug and subsystem masks
3490 - fix lfs find --obd (2510)
3491 - /proc tunable for disabling filter read caching (2591)
3492 - stop rpm packages from altering slapd.conf (2301)
3493 - disable nagle in the socknal under 0conf (2578)
3494 - choose mds inode size based on stripe count (2572)
3495 - fix kernel-source rpm problems (2516)
3497 - add --disable-doc to avoid pdf generation (2421)
3498 - update documentation, tests, type-os, comments
3499 - avoid format warnings on ia64
3500 - remove the TOE NAL
3501 - tiny code cleanups by removing unused fields
3503 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3506 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3507 - remove the most blatant lies from BUILDING file (2371)
3508 - change default debug level to reasonable production setting
3509 - reduce client side cache size to reduce cache flush time
3510 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3511 - make TCP zerocopy and pinger support enabled by default (2476)
3512 - sync writes completed after process exits caused crashes (2319)
3513 - maintain correct mount count on the MDS (2356)
3514 - backout 1557, because 2316 wasn't really fixed
3515 - better file I/O statistics gathering in /proc
3516 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3517 - another kernel patch to fix zero-copy TCP function export
3518 - don't take duplicate lock when processing re-sent getattr (2420)
3519 - lctl uses obd_self_export instead of creating new conn (2353)
3520 - MDS/OST recovery case which requires object creation asserted (2425)
3521 - move lfs from /usr/sbin to /usr/bin in packages
3522 - fix race between mds_client_add and mds_client_free (2417)
3523 - use kmalloc instead of slabs in portals (2430)
3524 - don't create duplicate records when a failover MDS is present (2442)
3525 - remove unnecessary mount age check (2332)
3526 - don't remove directory inodes from locks prematurely (2451)
3527 - don't break if MDS service name is the same as hostname (2103)
3528 - fix races in client write RPC generation when cache full (2482)
3530 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3533 - remove now-unused request->rq_obd (278)
3534 - if an allocation fails, print out how much memory we've used (1933)
3535 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3536 - add the "configurable stack size" patch to most series files (1256)
3537 - ability to write large log records, for 100+ OST configs (2306)
3538 - fix NULL deref when filter_prep fails (2314)
3539 - fix operator precedence error in filter_sync
3540 - dynamic allocation of socknal TX descriptors (2315)
3541 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3542 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3543 - ensure that CWARN messages in Portals always get to the syslog
3544 - __init/__exit are not for prototype decls (ldlm_init/exit)
3545 - x86-64 compile warning fixes
3546 - fix gateway LMC keyword conflict (2318)
3547 - fix MDS lock inversions in getattr/reint paths (1844)
3548 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3549 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3550 - DLM race condition prevented some lock evictions (2328)
3551 - ENOMEM detection and retry on socknal sends (2230)
3552 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3553 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3554 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3555 - better error messages when a client is rejected during recovery (1505)
3556 - avoid cancelling locks which were never granted, after failure (2330)
3557 - fix i_sem/journal inversion in mds_client_add (2333)
3558 - fix truncate/getattr lock cycle deadlock (2334)
3559 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3560 - fix two rare exit paths which could leak an l_lock() ref (2321)
3561 - fencepost error in MDS/OST orphan recovery (2226)
3562 - make log record alignment 8 bytes (1988)
3563 - lstripe now fails when requested offset > ost_count (2237)
3564 - ensure that all kernel series have a complete list.h (1607)
3565 - fix crashes in special-file operations (2316)
3566 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3568 - allow configurable automake binary, for testing new versions
3569 - small update to the lfs documentation
3571 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3573 * fix negative export reference count in fsfilt_sync (2312)
3575 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3576 * release candidate 0.9.1
3578 - orphans are moved into the PENDING directory for possible recovery
3579 - replayed opens now open by fid for orphan/rename safety (1042)
3580 - last close of an orphan inode generates a transno (683)
3581 - chdir() and mount() now pin the directory entry (1020)
3582 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3583 - discard very old requests without processing them (1502)
3584 - remove some common, well-understood CERRORs (1505)
3585 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3586 - clear "grant" flags in OST replies until OST grant code lands (1644)
3587 - fix read performance by not clobbering i_blksize on client (1598)
3588 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3589 - make LRU size a /proc tunable, clears locks when reduced (707)
3590 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3591 - server locks take references on exports now (1558)
3592 - build fixes for 2.4.20-rh trees (1663)
3593 - return an error from lov_create if all OSCs are inactive (1751)
3594 - fix import levels when a reconnect happens without a timeout (1597)
3595 - exit early from mds_open if we get a lookup error (1749)
3596 - partial page read at EOF wouldn't wait for disk before sending (1642)
3597 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3598 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3599 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3600 - write/truncate lock inversion (1639)
3601 - Don't auto-load obdclass, portals modules during cleanup (1495)
3602 - fix timestamps from jumping to "now" (1763)
3603 - extra journal assertions (1648)
3604 - add an extra multiunlink test (1771)
3605 - fix read_record/write_record API (1776)
3606 - fix leak of offset_extent, possible incorrect i_size later (1772)
3607 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3608 - Debugging for the kqswnal_get_idle_tx problems (1820)
3609 - Allow recovery to be attempted multiple times (1536)
3610 - Write out MDS last_rcvd file after it is first created (1600)
3611 - Fix tx_descriptor leak in failed transmit situations (1827)
3612 - ext3 journaling fixes for assertion failure after IO error (1871)
3613 - class_export_put() on freed export after completion AST error (1896)
3614 - Fix revalidate looping in VFS (1322)
3615 - Don't access a freed export during MDS_REINT timeout (1521)
3616 - Add open-unlink recovery support on the MDS (1673,1764)
3617 - Return an error if no MDS data was read from last_rcvd (1946)
3618 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3619 - Don't setup a disk device that doesn't match exported UUID (317)
3620 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3621 - avoid committing NULL handle in force close
3622 - local.sh is now a one-stripe LOV configuration
3623 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3624 - fix ext3 htree duplicate directory entry corruption (1516)
3625 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3626 - update to new LOV EA format (2097)
3627 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3629 2003-06-15 Phil Schwan <phil@clusterfs.com>
3632 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3633 - per-import/export recovery handling (958, 931, 959)
3634 - multiple last-rcvd slots, for serving multiple FSes (949)
3635 - connections are again shared between multiple imp/exports (963, 964)
3636 - "umount -f" would hang if any requests needed to be sent (393, 978)
3637 - avoid pinning large req buffer by copying for queued messages (989)
3638 - add "uuid" to "lctl device" command to help upcalls (991)
3639 - "open" RPCs with transnos would confuse recovery counters (1037)
3640 - do proper endian conversion of all wire messages (288, 340, 891)
3641 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3642 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3643 - fix ptlrpc_connection leak in target_handle_connect (1174)
3644 - fix import refcounting bug in OST and MDS cleanup (1134)
3645 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3646 - fix very unlikely obd_types race condition (501)
3647 - remove osc_open hack for echo_client (1187)
3648 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3649 - a failure in read_inode2 leads to deadlock (1139)
3650 - cancel ack-locks as soon as transaction is committed (1072)
3651 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3652 - make sure to commitrw after any preprw to avoid deadlock (1162)
3653 - failing to execute a file in a lustre FS would lock inode (1203)
3654 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3655 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3656 - obd_uuid2tgt would walk off the end of the list (1255)
3657 - on IA64 the osc would give portals incorrect bulk size (1258)
3658 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3659 - fix lock inversion caused by new llite matching code (1282)
3660 - limit the number of dirty pages on a client to 10MB (1286)
3661 - timed out locks were not being corrected cancelled (1289)
3662 - fix O_DIRECT above 4GB on IA-32 (1292)
3663 * major user-visible changes
3664 - fail out/fail over policy now controlled by the upcall (993)
3666 - add OBD_PING to check server availability and failure (954)
3667 - lustre messages are now sent in sending host order (288, 340, 891)
3668 - add eadatalen to MDS getattr reply (340)
3669 - OST read replies may contain second buffer, with per-page status (593)
3671 2003-03-11 Phil Schwan <phil@clusterfs.com>
3674 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3675 - failed open()s could cause deadlock; fixed (867, 869)
3676 - stop cancelling OST locks when files are closed (481)
3677 - overlapping XID spaces caused network corruption (851, 853)
3678 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3679 - setattr_raw added, to do single-RPC, server-side setattrs
3680 - lmc/lconf syntax change for OST UUIDs
3681 - fix crashy race condition between ptlrpc_free_req and osc_close
3682 - don't use request in mdc_enqueue if we hit a timeout (889)
3683 - don't set the inode i_size for regular files from the MDS (896)
3684 - handle out of order completion AST (842)
3685 - don't LBUG if a lock request times out after receiving AST (913)
3686 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3687 - if a bad lock AST arrives, send an error instead of dropping entirely
3688 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3689 - fix leak in bulk IO when only partially completed (899, 900, 926)
3690 - fix O_DIRECT for ia64 (55)
3691 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3692 - C-z after timeout could hang a process forever; fixed (977)
3694 - client-side I/O cache (678, 924, 929, 941, 970)
3696 - READPAGE and SETATTRs which don't take server-side locks get
3699 2003-02-11 Phil Schwan <phil@clusterfs.com>
3702 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3703 - Fix fsfilt_extN_readpage() to read a full page of directory
3704 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3705 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3706 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3707 - Most intent operations were removed, in favour of a new RPC mode
3708 that does a single RPC to the server and bypasses most of the VFS
3709 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3710 - Aggressively cancel local locks on DLM servers
3711 - mds_reint_unlink sends EA to the client if it's the last nlink.
3712 client uses that EA to unlink OST objects.
3713 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3714 - recursive symlinks were fixed (439)
3715 - fixed NULL deref in DEBUG_REQ
3716 - filter_update_lastobjid no longer calls sync, which annoyed extN
3717 - fixed multi-client small-writes to a single file problem (445)
3718 - fixed mtime updates during file writes (607)
3719 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3720 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3721 - fixed Linux OST size reporting problem (444, 656)
3722 - OST now updates object mtime with writes or setattr (607, 619)
3723 - client verifies file size before zeroing page past EOF (445)
3724 - OST now writes last allocated objid to disk with allocation (108)
3725 - LOV on echo now works (409)
3727 - mds_reint_unlink sends a new buffer, with the EA included. this
3728 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3729 set if a regular file was being unlinked, and it was the last link
3730 - use PtlGet from the target for bulk writes (315)
3731 - OST now updates object mtime with writes or setattr (607, 619)
3732 - LDLM now has a grant-time callback to revalidate locked items, if
3734 - Many MDS operations were reorganized to combat race conditions
3736 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3737 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3739 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3742 - Fully reactivate OST imports after reconnection (512, others)
3743 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3744 - More graceful error handling for truncating on dead OST (515)
3745 - Don't error out unless we're actually accessing dead stripes (474)
3746 - Fix garbage sizes when stripes are missing (410)
3747 - LRU counters were broken, causing constant lock purge (433, 432)
3748 - garbage on read from stripes with failed OSTs (441)
3749 - mark OSCs as active before reconnecting during recovery (438)
3750 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3751 - lfind did not preserve OST order in output (443)
3752 - symlinks cause hung clients, incorrect data (439)
3753 - stop dereferencing request after dropping refcount (457)
3754 - don't LASSERT(spin_is_locked) on non-SMP (455)
3755 - fixes for many rename() bugs
3756 - fstat didn't correctly synchronize attributes (399)
3757 - server must handle lock cancellation during blocking AST prep (487)
3758 - bulk descriptors were free()d too soon (511)
3759 - fix paths in lconf, which would load incorrect modules (451, 507)
3760 - fix confusing lconf 'host not found' error message (386)
3761 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3762 - fix race condition in mdc_blocking_ast() for inode access (526)
3763 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3764 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3765 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3766 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3767 - fix lstripe in several ways (kernel side) (527)
3768 - fix request leak in ldlm_cli_enqueue (262)
3769 - incorrect OSC was marked inactive after OST failure
3770 - call mds_fs_cleanup before unmounting filesystem (524)
3771 - fix races between taking ns_lock and ldlm_lock_change_resource
3772 - fix races updating LOV export open file list
3773 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3774 - fix recovery NULL deref in ldlm_cli_cancel_unused
3775 - fix some DLM races by using new hash table for lock handles (419)
3776 - permit the client to specify desired inodes, at replay
3777 - duplicate requests when we queue them for replay reintegration
3778 - fix last_rcvd offset calculation
3779 - sync after each recovered transaction, so we always make progress
3780 - never, not always, ERESTART requests without transnos
3781 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3783 - skip replay if the MDS says that the client is already connected
3784 - don't check for a recovery-enabled export to match lctl's UUID
3785 - don't INC_USE_COUNT for phantom exports
3786 - don't crash when cleaning up phantom exports (567)
3787 - don't double-finish or set replay data for errored mdc_open requests
3788 - abort requests when they time out, so we don't get old replies
3789 - send/receive replies for AST messages again
3790 - if the client says that it doesn't have the lock, cancel it on the
3792 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3793 mark it as destroyed, it will all work out when decref is called
3794 - fix module use counts (22, 581)
3796 - ASTs now expect a reply (server cancels lock on error reply)
3798 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3801 - fix many simultaneous client startup (392)
3802 - fix dentry->d_it clobbering
3803 - credentials weren't being shipped for readdir/getattr operations
3804 - remove invalid assertions triggered during some concurrent MD
3806 - proper Lustre versions added (336, 389)
3807 - fix memory leak for create error case (398)
3808 - fix LOV locking bug that would get cli/srv out of sync
3809 - fix echo client over LOV (409)
3810 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3811 - fix double-O_EXCL intent crash (424)
3812 - avoid sending multiple lock CANCELs (352)
3814 - MDS can do multi-client recovery (modulo bugs in new code)
3816 - many updates, edits, cleanups
3818 2002-11-18 Phil Schwan <phil@clusterfs.com>
3821 - fix null d_it dereference (346)
3822 - fix full OST/dbench hang (333)
3823 - fix permission problem with file removal (286)
3824 - fix removal of OSCs from LOV when they fail
3825 - fix NULL deref during bulk timeout (214)
3826 - fix problems related to multiple filesystems on one MDS (241)
3827 - fixed serious subtle metadata locking bugs
3828 - free locks on clients when inodes are removed due to memory
3830 - fix inode pointer in lock data (285)
3831 - partial support for multiple MDS on a single host (241)
3832 - data locks weren't cancelled at clear_inode time (290, 311)
3833 - intent locks could lead to unbounded lock growth (205)
3834 - added a maximum lock count, an LRU list, and a flusher
3835 - fix multiple rename (365)
3836 - properly abstracted the echo client
3837 - OSC locked 1 byte too many; fixed
3838 - rewrote brw callback code:
3839 - fixed recovery bugs related to LOVs (306)
3840 - fixed too-many-pages-in-one-write crash (191)
3841 - fixed (again) crash in sync_io_timeout (214)
3842 - probably fixed callback-related race (385)
3844 - Add capability to MDS protocol
3845 - LDLM cancellations and callbacks on different portals
3847 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3850 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3851 - defer open object destruction to close time (601981, 138)
3852 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3853 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3854 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3855 - don't call obd_setattr in ll_file_release for destroyed objects
3856 * protocol change to lustre_msg: move |version| and add |flags|
3857 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3858 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3859 * added replay of create, unlink, link and rename operations during
3860 MDS failover; recovery should be much more robust now
3861 * remove failed OSCs from LOVs (only lov_create uses this so far)
3862 * the lustre-HOWTO was brought (more) up to date (582544)
3864 2002-10-23 Phil Schwan <phil@clusterfs.com>
3867 - in-use dentries weren't being reused properly (617851)
3868 - prevent multiple LDLM setup (599178)
3869 - fix LOV size calculations for truncate (617853)
3870 - fix client handling of MDS intent errors (POSIX)
3871 - fix permission bug in lovstripe.c test (624321)
3872 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3873 - truncate past end of file could corrupt data
3874 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3875 - a race in recovery could return ETIMEDOUT to apps (623947)
3876 - building outside the source directory was fixed
3877 * the lustre-HOWTO was brought (more) up to date (582544)
3878 * major progress was made on recovery functionality
3880 2002-10-10 Phil Schwan <phil@clusterfs.com>
3883 - recovery deadlock fix
3884 - rm -rf causes LBUG fix (617817)
3885 - file open by multiple tasks fix (618962)
3886 - directory permissions bugs (602707 and 620007)
3887 - journal_stop fixed with locking (611313)
3888 - O_APPEND failures resolved (618273, perhaps 614459)
3889 - lconf PATH fix (619770)
3890 - IA64 build fix (621450)
3891 - RPC buffer sizes scale with amount of memory
3893 2002-10-01 Phil Schwan <phil@clusterfs.com>
3896 - locks would be cancelled without throwing away data pages,
3897 resulting in inconsistent data (605627)
3898 - inode attributes were not always being refreshed (605627, 612449)
3899 - lconf now continues to cleanup after lctl reports an error
3900 - MDS now enforces user permissions (602707)
3901 - lprocfs cleanup fixed, but not yet enabled (614157)
3902 - fixed infinite server hang, should a client not respond to an AST
3903 - avoid going into recovery if user calls readlink() with a buffer
3904 that's too small (613941)
3905 - AST RPCs no longer require replies (614867) -- this may be changed
3906 - don't crash server if client sends an IOV that's too big (611336)
3907 - fixed lock conversion deadlock (611892)
3908 - fixed the following of symlinks (614622)
3909 * recovery: the server can remove locks from a client that dies, other
3910 clients can make progress
3911 * more extN patch fixes
3912 * compile-time configurable ptlrpc buffer allocations
3914 - collaborative read cache document
3915 - Lustre Lite Performance CDR document-in-progress
3917 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3920 - fix typo in patch-2.4.18
3922 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3925 - clear ptlrpc request each time in handle_incoming_request()
3926 - unlink of files now destroys the object on the OST
3928 2002-09-19 Peter Braam <braam@clusterfs.com>
3930 * add hard link support
3931 * change obdfile creation method
3932 * kernel patch changed
3934 2002-09-19 Peter Braam <braam@clusterfs.com>
3937 - stack overflow bug in extN fixed
3939 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3941 * documentation updates
3942 - add man pages for config tools
3943 - update tests/README to describe testing with new config tools
3944 - finish metadata API descriptions
3945 * bug fixes and cleanups
3946 - statfs workaround for 16TB limit
3947 - LOV stripe allocation improved, can stripe on subset of OSTs
3948 - LOV file size/IO offset was wrong for files > 4GB in size
3949 - object EA data was being dropped, caused files to be unreadable
3950 - memory overflow with non-LOV OST caused memory corruption
3951 - fixed regression tests to work with new config tools, obdfilter
3952 - fixed bug when directory size became larger than 1 block
3953 - fixed bug (for single client case) when PWD was deleted
3954 - invalidate local directory pages when doing intent-based ops
3955 - avoid LDLM oops when lock callback contained bad data
3957 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3959 * documentation updates
3960 * bug fixes and cleanups
3961 - configuration tools
3964 - 64-bit compile warnings
3965 - 64-bit internal statfs data
3967 * test_brw on persistent OST devices
3969 * lprocfs (disabled)
3971 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3973 * documentation updates
3974 * bug fixes and cleanups
3975 * configuration tools
3977 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3980 * many small fixes to 0_5_4
3981 * io/network handling
3982 * thinkos in MDS operations
3984 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3987 * crucial basic fixes to 0.5.3
3988 * IOR, Iozone work over Elan
3991 2002-08-07 Phil Schwan <phil@clusterfs.com>
3992 * version 0_5_3, our first alpha
3993 * we use the new Portals iovs
3994 * documentation updates
3995 * bug fixes and cleanups
3996 * small changes in the DLM wire protocol
3998 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3999 * version 0_5_1 with some initial stability,
4000 * locking on MD and file I/O.
4001 * documentation updates
4002 * several bug fixes since 0.5.0
4003 * small changes in wire protocol
4005 2002-07-18 Phil Schwan <phil@clusterfs.com>
4007 * delivered as Lustre Light Alpha
4008 * fixed a crash after handling invalid MDS requests
4009 * fixed directory pages for architectures with non-4k pages sizes
4011 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4012 * release version v0_4_4
4013 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4015 * Fixes a number of interruption problems with OST operations.
4016 * Update documentation for portals header changes
4017 * Move all wire protocol structs/defines to lustre_idl.h
4018 * Fixes symlink length bug.
4019 * Add tcpdump to repository.
4021 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4022 * release version v0_4_3
4023 * Fixes statfs for inodes on extN.
4024 * Fixes bug in runtests which would delete /etc/hosts.
4025 * Use 64-bit object IDs wherever possible (not into VFS though)
4026 Remove ost_get_info, which is unused by lustre, and out of date.
4028 2002-07-03 Peter Braam <braam@clusterfs.com>
4029 * release version v0_4_2 Fixes a lookup error (type not passed)
4030 * move forward to head of Portals
4031 * move forward to latest Lustre kernel
4033 2002-06-25 Peter Braam <braam@clusterfs.com>
4034 * release version v0_4_1. Hopefully stable on single node use.