5 2.6.16.46-0.14 (SLES 10),
6 2.6.18-8.1.10.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.
15 Severity : enhancement
17 Description: light-weight GSS support
18 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
19 and reduce performance overhead.
21 Severity : enhancement
23 Description: Linux keyring support
24 Details : Support using service of Linux keyring for Lustre GSS internal
25 context refresh/cache mechanism.
29 Description: Fix errors in lfs documentation
30 Details : Fixes man pages
34 Description: don't use GFP_* in generic Lustre code.
35 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
39 Description: obdclass is limited by single OBD_ALLOC(idarray)
40 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
44 Description: warnings with build patchless client with vanila 2.6.19 and up
45 Details : change old ctl_table style and replace ctl_table/ctl_table_header
46 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
50 Frequency : Only for SLES
51 Description: sanity_quota fail test_1
52 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
53 file, which makes TSTID[2] inunique.
57 Frequency : Only for SLES9
58 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
59 Details : Update target/ChangeLog/which_patch .
63 Frequency : Only for test_10 in sanity-quota.sh
64 Description: a bug in quota test code
65 Details : checking a test flag in a wrong place
69 Frequency : Only for test_13 in sanity-quota.sh
70 Description: a bug in quota test script
71 Details : shouldn't check overfull allocated quota
75 Description: jbd statistics
76 Details : Port older jbd statistics patch for sles10
78 Severity : enhancement
80 Description: Add printing inode info into message about error in writepage.
85 Description: reply_lock_interpret crash due to race with it and lock cancel.
86 Details : Do not replay locks that are being cancelled. Do not reference
87 locks by their address during replay, just by their handle.
91 Frequency : When flocks are used.
92 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
93 Details : Flock locks might destroy just granted lock if it could be merged
94 with another existing flock, this is done in completion handler,
95 so teach ldlm_cli_enquque_fini that this is a valid case for
101 Description: Oops in read and write path when failing to allocate lock.
102 Details : Check if lock allocation failed and return error back.
106 Description: lstripe command fails for valid OST index
107 Details : The stripe offset is compared to lov->desc.ld_tgt_count
108 instead of lov->desc.ld_active_tgt_count.
112 Description: sanity.sh failed test 103
113 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
118 Description: df doesn't work properly if diskfs blocksize != 4K
119 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
123 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
124 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
128 Description: obd_health_check_timeout too short
129 Details : set obd_health_check_timeout as 1.5x of obd_timeout
133 Description: llapi_file_create() does not allow some changes
134 Details : add llapi_file_open() that allows specifying the mode and
135 open flags, and also returns an open file handle.
139 Description: merge and cleanup kernel patches.
140 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
144 Description: Add journal checksum support.(Kernel part)
145 Details : The journal checksum feature adds two new flags i.e
146 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
147 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
148 indicates that the commit block contains the checksum for
149 the blocks described by the descriptor blocks. Now commit
150 record can be sent to disk without waiting for descriptor
151 blocks to be written to disk. This behavior is controlled
152 using JBD2_FEATURE_ASYNC_COMMIT flag.
156 Description: OSS needs mutliple precreate threads
157 Details : Add ability to start more than one create thread per OSS.
161 Description: Sanity.sh test_65k failed: setstripe should have succeeded
162 Details : execute lfs setstripe on client
166 Description: mds_obd_create error creating tmp object
167 Details : When the user sets quota on root, llog will be affected and can't
168 create files and write files.
171 Frequency : Always on ia64 patchless client, and possibly others.
173 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
174 Details : This allows the patchless client to be loaded on architectures
179 Description: RedHat Update kernel for RHEL5
180 Details : Add the kernel config file for RHEL5.
184 Description: RedHat Update kernel for RHEL5
185 Details : Modify the kernel config file more closer RHEL5.
189 Description: Build failure against Centos5 (RHEL5)
190 Details : Define PAGE_SIZE when it isn't present.
194 Description: test_7 on sanity-quota.sh can't be used on separate machines
195 Details : add facet to handle it
199 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
200 Details : with NFS, the anon dentry's parent was set to itself in
201 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
202 corresponding dentry other than use rec->ur_name.
204 Severity : enhancement
206 Description: lfs setstripe enhancement
207 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
211 Description: testing performance impact of enabling checksumming
212 Details : enable checksum by default, allow --disable-checksum
213 configure option and "-o nochecksum" mount option
217 Description: lustre support for RHEL5
218 Details : Add support for RHEL5.
222 Description: Client eviction due to failover config
223 Details : after a connection loss, the lustre client should attempt to
224 reconnect to the last active server first before trying the
225 other potential connections.
229 Description: when mds and osts use different quota unit(32bit and 64bit),
230 quota will be released repeatly.
231 Details : void sending multiple quota reqs to mds, which will keep the
232 status between the reqs.
236 Description: osts not allocated evenly to files
237 Details : change the condition to increase offset_idx
240 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
242 Description: Data corruption for OSTs that are formatted larger than 2TB
244 Details : When generating the bio request for lustre file writes the
245 sector number would overflow a temporary variable before being
246 used for the IO. The data reads correctly from Lustre (which
247 will overflow in a similar manner) but other file data or
248 filesystem metadata may be corrupted in some cases.
252 Description: Tune the kernel for good SCSI performance.
253 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
254 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
259 Description: rewrite ext2-derived code in obdclass/uuid.c
260 Details : rewrite inherited code (uuid parsing code from ext2 utils)
261 from scratch preserving functionality.
265 Description: "lfs find -obd UUID" prints directories
266 Details : "lfs find -obd UUID" will return all directory names instead
267 of just file names. It is incorrect because the directories
268 do not reside on the OSTs.
272 Description: use i_size_read and i_size_write in 2.6 port
273 Details : replace inode->i_size access with i_size_read/write()
276 Frequency : when removing large files
278 Description: scheduling issue during removal of large Lustre files
279 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
280 It causes scheduling issues when removing large files (17TB in the
284 Frequency : only with liblustre clients on XT3
286 Description: evictions taking too long
287 Details : allow llrd to evict clients directly on OSTs
290 Frequency : only on ppc
292 Description: /proc/fs/lustre/devices broken on ppc
293 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
294 We should make sure the type of 'index' is loff_t and then cast
295 explicitly as needed below. Do not assign an explicitly cast
299 Frequency : only for rhel5
301 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
302 Details : Modify the target file & which_kernel.
304 --------------------------------------------------------------------------------
306 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
308 * Support for kernels:
309 2.6.5-7.283 (SLES 9),
310 2.6.9-55.EL (RHEL 4),
311 2.6.16.46-0.14 (SLES 10),
312 2.6.18.8 vanilla (kernel.org)
313 * Client support for unpatched kernels:
314 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
315 2.6.16 - 2.6.22 vanilla (kernel.org)
316 * Due to recently discovered recovery problems, we do not recommend
317 using patchless RHEL 4 clients with this or any earlier release.
318 * Recommended e2fsprogs version: 1.39.cfs8
319 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
320 * Starting with this release, the ldiskfs backing filesystem required
321 by Lustre is now in its own package, lustre-ldiskfs. This package
322 should be installed. It is versioned separately from Lustre and
323 may be released separately in future.
327 Description: block reactivating mgc import until all deactivates complete
328 Details : Fix race when failing back MDT/MGS to itself (testing)
330 Severity : enhancement
332 Description: add optional extra BUILD_VERSION info
333 Details : add a new environment variable (namely LUSTRE_VERS) which allows
334 to override the lustre version.
337 Frequency : 2.6.18 servers only
339 Description: ll_kern_mount() doesn't release the module reference
340 Details : The ldiskfs module reference count never drops down to 0
341 because ll_kern_mount() doesn't release the module reference.
346 Description: server LBUG when using old ost_num_threads parameter
347 Details : Accept the old ost_num_threads parameter but warn that it
348 is deprecated, and fix an off-by-one error that caused an LBUG.
353 Description: Transient SCSI error results in persistent IO issue
354 Details : iobuf->dr_error is not reinitialized to 0 between two
358 Frequency : sometimes when underlying device returns I/O errors
360 Description: OSTs not going read-only during write failures
361 Details : OSTs are not remounted read-only when the journal commit threads
362 get I/O errors because fsfilt_ext3 calls journal_start/stop()
363 instead of the ext3 wrappers.
367 Description: poor connect scaling with increasing client count
368 Details : Don't run filter_grant_sanity_check for more than 100 exports
369 to improve scaling for large numbers of clients.
372 Frequency : SLES10 only
374 Description: sanity-quota.sh quotacheck failed: rc = -22
375 Details : Quotas cannot be enabled on SLES10.
378 Frequency : liblustre clients only
380 Description: getdirentries does not give error when run on compute nodes
381 Details : getdirentries does not fail when the size specified as an argument
382 is too small to contain at least one entry
384 Severity : enhancement
386 Description: Add LNET router traceability for debug purposes
387 Details : If a checksum failure occurs with a router as part of the
388 IO path, the NID of the last router that forwarded the bulk data
389 is printed so it can be identified.
394 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
395 Details : Due to a race condition, liblustre clients were occasionally
398 Severity : enhancement
400 Description: lfs setstripe use optional parameters instead of postional
403 Severity : enhancement
405 Description: Nanosecond timestamp support for ldiskfs
406 Details : The on-disk ldiskfs filesystem has added support for nanosecond
407 resolution timestamps. There is not yet support for this at
408 the Lustre filesystem level.
411 Frequency : during server recovery
413 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
414 Details : request with rq_no_resend flag not awake l_wait_event if they get a
418 Frequency : nfs export on patchless client
420 Description: connectathon hang when test nfs export over patchless client
421 Details : Disconnected dentry cannot be found with lookup, so we do not need
422 to unhash it or make it invalid
425 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
426 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
427 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
432 Description: open req refcounting wrong on reconnect
433 Details : If reconnect happened between getting open reply from server and
434 call to mdc_set_replay_data in ll_file_open, we will schedule
435 replay for unreferenced request that we are about to free.
436 Subsequent close will crash in variety of ways.
437 Check that request is still eligible for replay in
438 mdc_set_replay_data().
443 Description: disable writes to filesystem when reading health_check file
444 Details : the default for reading the health_check proc file has changed
445 to NOT do a journal transaction and write to disk, because this
446 can cause reads of the /proc file to hang and block HA state
447 checking on a healthy but otherwise heavily loaded system. It
448 is possible to return to the previous behaviour during configure
449 with --enable-health-write.
451 Severity : enhancement
453 Description: 64-bit inode version
454 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
455 to the inode. This will be required for version-based recovery.
460 Description: MDS fails to start if a duplicate client export is detected
461 Details : in some rare cases it was possible for a client to connect to
462 an MDS multiple times. Upon recovery the MDS would detect this
463 and fail during startup. Handle this more gracefully.
465 Severity : enhancement
467 Description: Add -o localflock option to simulate old noflock
469 Details : This will achieve local-only flock/fcntl locks
475 Description: log_commit_thread vs filter_destroy race leads to crash
476 Details : Take import reference before releasing llog record semaphore
481 Description: Wrong request locking in request set processing
482 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
483 imp_delayed_list, in this place should be used imp_lock.
486 Frequency : when reconnection
488 Description: Grant Leak when osc reconnect to OST
489 Details : When osc reconnect ost, OST(filter) should check whether it
490 should grant more space to client by comparing fed_grant and
491 cl_avail_grant, and return the granted space to client instead
492 of "new granted" space, because client will call osc_init_grant
493 to update the client grant space info.
496 Frequency : when client reconnect to OST
498 Description: Grant Leak when osc do resend and replay bulk write
499 Details : When osc reconnect to OST, OST(filter)should clear grant info of
500 bulk write request, because the grant info while be sync between
501 OSC and OST when reconnect, and we should ignore the grant info
502 these of resend/replay write req.
507 Description: Grant space more than avaiable left space sometimes.
508 Details : When then OST is about to be full, if two bulk writing from
509 different clients came to OST. Accord the avaliable space of the
510 OST, the first req should be permitted, and the second one
511 should be denied by ENOSPC. But if the seconde arrived before
512 the first one is commited. The OST might wrongly permit second
513 writing, which will cause grant space > avaiable space.
516 Frequency : when client is evicted
518 Description: Grant might be wrongly erased when osc is evicted by OST
519 Details : when the import is evicted by server, it will fork another
520 thread ptlrpc_invalidate_import_thread to invalidate the
521 import, where the grant will be set to 0. While the original
522 thread will update the grant it got when connecting. So if
523 the former happened latter, the grant will be wrongly errased
524 because of this race.
529 Description: Checking Stale with correct fid
530 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
531 to check whether it is stale, because sometimes, we want the
532 enqueue happened anyway, and op_data.fid2 will not be initialized.
534 Severity : enhancement
536 Description: update patchless client
537 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
540 Frequency : only with 2.4 kernel
542 Description: random memory corruption
543 Details : size of struct ll_inode_info is to big for union inode.u and this
544 can be cause of random memory corruption.
549 Description: Memory leak in recovery
550 Details : Lov_mds_md was not free in an error handler in mds_create_object.
551 It should also check obd_fail before fsfilt_start, otherwise if
552 fsfilt_start return -EROFS,(failover mds during mds recovery).
553 then the req will return with repmsg->transno = 0 and rc = EROFS.
554 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
555 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
556 be freed no matter whether fsfilt_commit success or not.
559 Frequency : only with huge count clients
561 Description: Prevents from taking the superblock lock in llap_from_page for
563 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
564 for llap_from_page call in ll_removepage prevents from taking the
565 superblock lock for a soon died page.
570 Description: Not check open intent error before release open handle
571 Details : in some rare cases, the open intent error is not checked before
572 release open handle, which may cause
573 ASSERTION(open_req->rq_transno != 0), because it tries to release
574 the failed open handle.
579 Description: Set cat log bitmap only after create log success.
580 Details : in some rare cases, the cat log bitmap is set too early. and it
581 should be set only after create log success.
585 Description: Accessing a block bevice can re-enable I/O when Lustre is
586 tearing down a device.
587 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
592 Description: service threads may hog cpus when there are a lot of requests
594 Details : Insert cond_resched to give other threads a chance to use some of
600 Description: the cat log was not initialized in recovery
601 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
602 unlink log on mds will not be initialized until mds post
603 recovery. And also in mds post recovery, the unlink log will
604 initialization will be done asynchronausly, so there will be race
605 between add unlink log and unlink log initialization.
609 Description: brw_stats were being printed incorrectly
610 Details : brw_stats were being printed as log2 but all of them were not
611 recorded as log2. Also remove some code duplication arising from
612 filter_tally_{read,write}.
616 Frequency : rare, only in recovery.
617 Description: ASSERTION(req->rq_type != LI_POISON) failed
618 Details : imp_lock should be held while iterating over imp_sending_list for
619 prevent destroy request after get timeout in ptlrpc_queue_wait.
623 Description: replay-single.sh test 52 fails
624 Details : A lock's skiplist need to be cleanup when it being unlinked
625 from its resource list.
629 Description: Short directio read returns full requested size rather than
631 Details : Direct I/O operations should return actual amount of bytes
632 transferred rather than requested size.
634 Severity : enhancement
636 Description: metadata RPC reduction (e.g. for rm performance)
637 Details : decrease the amount of synchronous RPC between clients and servers
638 by canceling conflicing lock before the operation on the client side
639 and packing thier handles into the main operation RPC to server.
641 Severity : enhancement
643 Description: Async OSC create to avoid the blocking unnecessarily.
644 Details : If a OST has no remain object, system will block on the creating
645 when need to create a new object on this OST. Now, ways use
646 pre-created objects when available, instead of blocking on an
647 empty osc while others are not empty. If we must block, we block
648 for the shortest possible period of time.
650 --------------------------------------------------------------------------------
652 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
657 Frequency : on some architectures
659 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
660 Details : Uninitialized flags sometimes cause configuration commands to
664 Frequency : patchless clients only
666 Description: missing __iget() symbol export
667 Details : The __iget() symbol export is missing. To avoid the need for
668 this on patchless clients the deathrow inode reaper is turned
669 off, and we depend on the VM to clean up old inodes. This
670 dependency was during via the fix for bug 12181.
672 --------------------------------------------------------------------------------
674 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
676 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
677 INTEROPERATE with older versions automatically. In many cases a
678 special upgrade step is needed. Please read the
679 user documentation before upgrading any part of a live system.
680 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
681 version will not interoperate with 1.6 betas before beta5 (1.5.95).
682 * WARNING: Lustre configuration and startup changes are required with
683 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
689 Severity : enhancement
691 Description: MountConf
692 Details : Lustre configuration is now managed via mkfs and mount
693 commands instead of lmc and lconf. New obd types (MGS, MGC)
694 are added for dynamic configuration management. See
695 https://mail.clusterfs.com/wikis/lustre/MountConf for
698 Severity : enhancement
700 Description: dynamic OST addition
701 Details : OSTs can now be added to a live filesystem
703 Severity : enhancement
705 Description: startup order invariance
706 Details : MDTs and OSTs can be started in any order. Clients only
707 require the MDT to complete startup.
709 Severity : enhancement
711 Description: parallel, asynchronous orphan cleanup
712 Details : orphan cleanup is now performed in separate threads for each
713 OST, allowing parallel non-blocking operation.
715 Severity : enhancement
717 Description: optimized stripe assignment
718 Details : stripe assignments are now made based on ost space available,
719 ost previous usage, and OSS previous usage, in order to try
720 to optimize storage space and networking resources.
722 Severity : enhancement
724 Description: Permanently set tunables
725 Details : All writable /proc/fs/lustre tunables can now be permanently
726 set on a per-server basis, at mkfs time or on a live
729 Severity : enhancement
731 Description: Lustre message v2
732 Details : Add lustre message format v2.
734 Severity : enhancement
736 Description: client OST exclusion list
737 Details : Clients can be started with a list of OSTs that should be
738 declared "inactive" for known non-responsive OSTs.
742 Description: SPEC SFS validation failure on NFS v2 over lustre.
743 Details : Changes the blocksize for regular files to be 2x RPC size,
744 and not depend on stripe size.
746 Severity : enhancement
748 Description: Multiple MD RPCs in flight.
749 Details : Further unserialise some read-only MDS RPCs - learn about intents.
750 To avoid overly-overloading MDS, introduce a limit on number of
751 MDS RPCs in flight for a single client and add /proc controls
752 to adjust this limit.
754 Severity : enhancement
756 Description: client read/write statistics
757 Details : Add client read/write call usage stats for performance
758 analysis of user processes.
759 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
760 file access. extents_stats shows chunk size distribution.
761 extents_stats_per_process show chunk size distribution per
764 Severity : enhancement
766 Description: mds statistics
767 Details : Add detailed mds operations statistics in
768 /proc/fs/lustre/mds/*/stats.
772 Description: Failure of copying files with lustre special EAs.
773 Details : Client side always return success for setxattr call for lustre
774 special xattr (currently only "trusted.lov").
779 Description: Refcount LNET uuids
780 Details : The global LNET uuid list grew linearly with every startup;
781 refcount repeated list entries instead of always adding to
784 Severity : enhancement
786 Description: Dynamic service threads
787 Details : Within a small range, start extra service threads
788 automatically when the request queue builds up.
791 Frequency : mixed-endian client/server environments
793 Description: mixed-endian crashes
794 Details : The new msg_v2 system had some failures in mixed-endian
797 Severity : enhancement
799 Description: Easy OST removal
800 Details : OSTs can be permanently deactivated with e.g. 'lctl
801 conf_param lustre-OST0001.osc.active=0'
803 Severity : enhancement
805 Description: MGS proc entries
806 Details : Added basic proc entries for the MGS showing what filesystems
809 Severity : enhancement
811 Description: provide MGS failover
812 Details : Added config lock reacquisition after MGS server failover.
814 Severity : enhancement
816 Description: add Linux 2.4 support
817 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
821 Description: a large application tries to do I/O to the same resource and dies
823 Details : Check the req->rq_arrival time after the call to
824 ost_brw_lock_get(), but before we do anything about
825 processing it & sending the BULK transfer request. This
826 should help move old stale pending locks off the queue as
827 quickly as obd_timeout.
830 Frequency : when an incorrect nid is specified during startup
832 Description: ptlrpc connect to non-existant node causes kernel crash
833 Details : LNET can't be re-entered from an event callback, which
834 happened when we expire a message after the export has been
835 cleaned up. Instead, hand the zombie cleanup off to another
838 Severity : enhancement
840 Description: plain/inodebits lock performance improvement
841 Details : Grouping plain/inodebits in granted list by their request modes
842 and bits policy, thus improving the performance of search through
846 Frequency : only if OST filesystem is corrupted
848 Description: client incorrectly hits assertion in ptlrpc_replay_req()
849 Details : for a short time RPCs with bulk IO are in the replay list,
850 but replay of bulk IOs is unimplemented. If the OST filesystem
851 is corrupted due to disk cache incoherency and then replay is
852 started it is possible to trip an assertion. Avoid putting
853 committed RPCs into the replay list at all to avoid this issue.
856 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
858 Description: System hang on startup
859 Details : This bug allowed the liblustre (e.g. catamount) client to
860 return to the app before handling all startup RPCs. This
861 could leave the node unresponsive to lustre network traffic
862 and manifested as a server ptllnd timeout.
864 Severity : enhancement
866 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
867 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
868 boolean to control whether to print peer debug info when a
869 client's RPC times out.
872 Frequency : only for kernels with patches from Lustre below 1.4.3
874 Description: Remove old rdonly API
875 Details : Remove old rdonly API which unused from at least lustre 1.4.3
878 Frequency : only for devices with external journals
880 Description: Set external device read-only also
881 Details : During a commanded failover stop, we set the disk device
882 read-only while the server shuts down. We now also set any
883 external journal device read-only at the same time.
886 Frequency : when upgrading from 1.4 while trying to change parameters
888 Description: The wrong (new) MDC name was used when setting parameters for
889 upgraded MDT's. Also allows changing of OSC (and MDC)
890 parameters if --writeconf is specified at tunefs upgrade time.
893 Frequency : when setting specific ost indicies
895 Description: QOS code breaks on skipped indicies
896 Details : Add checks for missing OST indicies in the QOS code, so OSTs
897 created with --index need not be sequential.
899 Severity : enhancement
901 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
902 Details : The uninit_groups feature works in conjunction with the kernel
903 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
904 up the pass1 processing of e2fsck. This is a read-only feature
905 in ldiskfs2 only, so older kernels and current ldiskfs cannot
906 mount filesystems that have had this feature enabled.
908 Severity : enhancement
910 Description: Improve multi-block allocation algorithm to avoid fragmentation
911 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
912 allocation locality and avoid filesystem fragmentation.
914 ------------------------------------------------------------------------------
916 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
918 * Support for kernels:
919 2.4.21-47.0.1.EL (RHEL 3)
921 2.6.9-42.0.10.EL (RHEL 4)
922 2.6.12.6 vanilla (kernel.org)
923 2.6.16.27-0.9 (SLES 10)
924 * Recommended e2fsprogs version: 1.39.cfs5
926 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
930 Frequency : occasional, depends on client load and configuration
931 Bugzilla : 12181, 12203
932 Description: data loss for recently-modified files
934 Details : In some cases it is possible that recently written or created
935 files may not be written to disk in a timely manner (this should
936 normally be within 30s unless client IO load is very high).
937 The problem appears as zero-length files or files that are a
938 multiple of 1MB in size after a client crash or client eviction
939 that are missing data at the end of the file.
941 This problem is more likely to be hit on clients where files are
942 repeatedly created and unlinked in the same directory, clients
943 have a large amount of RAM, have many CPUs, the filesystem has
944 many OSTs, the clients are rebooted frequently, and/or the files
945 are not accessed by other nodes after being written.
947 The presence of the problem can be detected by looking at
948 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
949 smaller than the second (nr_unused) then dirty files will not
950 be flushed automatically to disk. "sync; sleep 10" should be
951 run several times on the node before unmounting it to update
952 Lustre (this is also safe to run on nodes without this problem).
954 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
955 that can cause this same problem, so customers using that kernel
956 also need to update the kernel in addition to Lustre. In order
957 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
959 It is normal that files written just before a client crash (less
960 than 30s) may not yet have been flushed to disk, even for local
964 Frequency : frequent on thin XT3 nodes
966 Description: UUID collision on thin XT3 Linux nodes
967 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
968 randomly, since we relied on an insufficiently-seeded PRNG.
973 Description: OSS hangs after "All ost request buffers busy"
974 Details : A deadlock between quota and journal operations caused OSS
975 hangs after printing "All ost request buffers busy."
978 Frequency : always on liblustre builds
980 Description: Cleanup compiler warnings on liblustre
983 Frequency : always on liblustre builds on XT3
985 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
989 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
990 > 32000 subdirectories
991 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
992 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
993 subdirectory count crosses 32000. This will aid e2fsck to
994 correctly handle more than 32000 subdirectories.
997 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
999 Description: System hang on startup
1000 Details : This bug allowed the liblustre (e.g. catamount) client to
1001 return to the app before handling all startup RPCs. This
1002 could leave the node unresponsive to lustre network traffic
1003 and manifested as a server ptllnd timeout.
1005 Severity : enhancement
1007 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1008 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1009 boolean to control whether to print peer debug info when a
1010 client's RPC times out.
1015 Description: make O_SYNC working on 2.6 kernels
1016 Details : 2.6 kernels use different method for mark pages for write,
1017 so need add a code to lustre for O_SYNC work.
1022 Description: Failure to close file and release space on NFS
1023 Details : Put inode details into lock acquired in ll_intent_file_open.
1024 Use mdc_intent_lock in ll_intent_open to properly
1025 detect all kind of errors unhandled by mdc_enqueue.
1030 Description: proc file read during shutdown sometimes raced obd removal,
1032 Details : Add lock to prevent obd access after proc file removal.
1035 Frequency : Only for files larger than 4GB on 32-bit clients.
1037 Description: improperly doing page alignment of locks
1038 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1039 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1044 Description: RPCs being resent when they shouldn't be
1045 Details : Some RPCs that should not be resent are being resent. This
1046 can cause inconsistencies in the RPC state machine. Do not
1047 resend such requests.
1050 Frequency : rare, only with NFS export
1052 Description: Crash on NFS re-export node
1053 Details : under very unusual load conditions an assertion is hit in
1054 ll_intent_file_open()
1057 Frequency : only if OST filesystem is corrupted
1059 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1060 Details : for a short time RPCs with bulk IO are in the replay list,
1061 but replay of bulk IOs is unimplemented. If the OST filesystem
1062 is corrupted due to disk cache incoherency and then replay is
1063 started it is possible to trip an assertion. Avoid putting
1064 committed RPCs into the replay list at all to avoid this issue.
1069 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1070 Details : Large single O_DIRECT read and write calls can fail to allocate
1071 a sufficiently large buffer to process the request. In case of
1072 allocation failure the allocation is retried with a smaller
1073 buffer and broken into smaller requests.
1075 Severity : enhancement
1077 Description: Add -o localflock option to simulate old noflock behaviour.
1078 Details : This will achieve local-only flock/fcntl locks coherentness.
1083 Description: versioning check is incomplete
1084 Details : Checking the version difference of client vs. server, report
1085 error if the gap is too big.
1090 Description: add support PG_writeback bit
1091 Details : add support for PG_writeback bit for Lustre, for more carefull
1092 work with page cache in 2.6 kernel. This also fix some deadlocks
1093 and remove hack for work O_SYNC with 2.6 kernel.
1095 Severity : enhancement
1097 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1098 Details : The uninit_groups feature works in conjunction with the kernel
1099 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1100 up the pass1 processing of e2fsck. This is a read-only feature
1101 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1102 mount filesystems that have had this feature enabled.
1104 Severity : enhancement
1106 Description: Improve multi-block allocation algorithm to avoid fragmentation
1107 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1108 allocation locality and avoid filesystem fragmentation.
1110 ------------------------------------------------------------------------------
1112 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1114 * Support for kernels:
1115 2.6.9-42.0.3.EL (RHEL 4)
1116 2.6.5-7.276 (SLES 9)
1117 2.4.21-47.0.1.EL (RHEL 3)
1118 2.6.12.6 vanilla (kernel.org)
1119 2.6.16.21-0.8 (SLES10)
1120 * Recommended e2fsprogs version: 1.39.cfs2-0
1122 * The backwards-compatible /proc/sys/portals symlink has been removed
1123 in this release. Before upgrading, please ensure that you change
1124 any configuration scripts or /etc/sysctl.conf files that access
1125 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1126 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1127 in advance of the upgrade on any system running Lustre 1.4.6 or
1128 newer, since /proc/sys/lnet was added in that version.
1129 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1133 Frequency : only when quota is used
1135 Description: avoid scanning export list for quota master
1136 Details : Change the algorithms to avoid scanning export list in order
1137 to improve the efficiency.
1140 Frequency : MDS failover only, very rarely
1142 Description: "went back in time" messages on mds failover
1143 Details : The greatest transno may be lost when the current operation
1144 finishes with an error (transno==0) and the client's last_rcvd
1145 record is over-written. Save the greatest transno in the
1146 mds_last_transno for this case.
1149 Frequency : always for specific kernels and striping counts
1151 Description: client may get "Matching packet too big" without ACL support
1152 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1153 when trying to access files in certain configurations. The
1154 clients should in fact be denied when mounting because they do
1155 not understand ACLs.
1158 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1160 Description: many clients connecting with IO in progress causes connect timeouts
1161 Details : Avoid synchronous journal commits to avoid delays caused by many
1162 clients connecting/disconnecting when bulk IO is in progress.
1163 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1164 OST_IO_PORTAL to avoid delays behind potentially many pending
1168 Frequency : occasionally with multiple writers to a single file
1170 Description: shared writes to file may result in wrong size reported by stat()
1171 Details : Allow growing of kms when extent lock is cancelled
1174 Frequency : always with random mmap IO to multi-striped file
1176 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1177 Details : Only if the hole is at the end of OST object so that kms is too
1178 small. Fix is to increase kms accordingly in ll_nopage.
1181 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1183 Description: writes to a missing object would leak memory on the OST
1184 Details : If there is an inconsistency between the MDS and OST filesystems,
1185 such that the MDS references an object that doesn't exist, writes
1186 to that object will leak memory due to incorrect cleanup in the
1187 error handling path, eventually running out of memory on the OST.
1192 Description: Creating too long symlink causes lustre errors
1193 Details : Check symlink and name lengths before sending requests to MDS.
1196 Frequency : only if flock is enabled (not on by default)
1198 Description: posix locks not released on fd closure on 2.6.9+
1199 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1200 kernels, this caused such locks not to be released on fd close and
1201 then assertions on fs unmount about still used locks.
1204 Frequency : MDS failover only, very rarely
1206 Description: clients may get ASSERTION(granted_lock != NULL)
1207 Details : When request was taking a long time, and a client was resending
1208 a getattr by name lock request. The were multiple lock requests
1209 with the same client lock handle and
1210 mds_getattr_name->fixup_handle_for_resent_request found one of the
1211 lock handles but later failed with ASSERTION(granted_lock != NULL).
1216 Description: handle->h_buffer_credits > 0, assertion failure
1217 Details : h_buffer_credits is zero after truncate, causing assertion
1218 failure. This patch extends the transaction or creates a new
1222 Frequency : NFS re-export or patchless client
1223 Bugzilla : 11179, 10796
1224 Description: Crash on NFS re-export node (__d_move)
1225 Details : We do not want to hash the dentry if we don't have a lock.
1226 But if this dentry is later used in d_move, we'd hit uninitialised
1227 list head d_hash, so we just do this to init d_hash field but
1228 leave dentry unhashed.
1231 Frequency : NFS re-export or patchless client
1233 Description: NFS exports has problem with symbolic link
1234 Details : lustre client didn't properly install dentry when re-exported
1235 to NFS or running patchless client.
1238 Frequency : NFS re-export or patchless client
1240 Description: Various nfs/patchless fixes.
1241 Details : fixes reuse disconected alias for lookup process - this fixes
1242 warning "find_exported_dentry: npd != pd",
1243 fix permission error with open files at nfs.
1244 fix apply umask when do revalidate.
1247 Frequency : occasional
1249 Description: Crash on NFS re-export node
1250 Details : calling clear_page() on the wrong pointer triggered oops in
1251 generic_mapping_read().
1254 Frequency : rarely, using O_DIRECT IO
1256 Description: unaligned directio crashes client with LASSERT
1257 Details : check for unaligned buffers before trying any requests.
1260 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1262 Description: stale data returned from RAID cache
1263 Details : If only a small amount of IO is done to the RAID device before
1264 reading it again it is possible to get stale data from the RAID
1265 cache instead of reading it from disk.
1268 Frequency : always for sles10 kernel
1270 Description: sles10 support
1271 Details : ll_follow_link: compile fixes and using of nd_set_link
1272 under newer kernels.
1275 Frequency : depends on arch, kernel and compiler version, always on sles10
1278 Description: recursive or deep enough symlinks cause stack overflow
1279 Details : getting rid of large stack-allocated variable in
1283 Frequency : depends on hardware
1285 Description: lustre write performance loss in the SLES10 kernel
1286 Details : the performance loss is caused by using of write barriers in the
1287 ext3 code. The SLES10 kernel turns barrier support on by
1288 default. The fix is to undo that change for ldiskfs.
1290 ------------------------------------------------------------------------------
1292 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1294 * Support for kernels:
1295 2.6.9-42.0.3EL (RHEL 4)
1296 2.6.5-7.276 (SLES 9)
1297 2.4.21-47.0.1.EL (RHEL 3)
1298 2.6.12.6 vanilla (kernel.org)
1302 Frequency : quota enabled and large files being deleted
1304 Description: releasing more than 4GB of quota at once hangs OST
1305 Details : If a user deletes more than 4GB of files on a single OST it
1306 will cause the OST to spin in an infinite loop. Release
1307 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1312 Description: statfs data retrieved from /proc may be stale or zero
1313 Details : When reading per-device statfs data from /proc, in the
1314 {kbytes,files}_{total,free,avail} files, it may appear
1315 as zero or be out of date.
1318 Frequency : systems with MD RAID1 external journal devices
1320 Description: lconf's call to blkid is confused by RAID1 journal devices
1321 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1322 of returning all block devices that match the journal UUID.
1325 Frequency : always, for aggregate stripe size over 4GB
1327 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1328 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1329 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1332 Frequency : always if "lfs find" used on a local file/directory
1334 Description: "lfs find" segfaults if used on a local file/directory
1335 Details : The case where a directory component was not specified wasn't
1336 handled correctly. Handle this properly.
1339 Frequency : always on ppc64
1341 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1342 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1343 perform endianness conversion.
1346 Frequency : rarely (truncate to non-zero file size after write under load)
1347 Bugzilla : 10730, 10687
1348 Description: Files padded with zeros to next 4K multiple
1349 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1350 it is possible that files that are truncated to a non-zero size
1351 immediately after being written are filled with zero bytes beyond
1352 the truncated size. No file data is lost.
1354 Severity : enhancement
1356 Description: Allow recovery/failover for liblustre clients.
1357 Details : liblustre clients were unaware of failover configurations until
1360 Severity : enhancement
1362 Description: user file locks should fail when not mounting with flock option
1363 Details : Set up an error-returning stub in ll_file_operations.lock field
1364 to prevent incorrect behaviour when client is mounted without
1365 flock option. Also, set up properly f_op->flock field for
1369 Frequency : always on ia64
1371 Description: "lfs df" loops on printing out MDS statfs information
1372 Details : The obd_ioctl_data was not initialized and in some systems
1373 this caused a failure during the ioctl that did not return
1374 an error. Initialize the struct and return an error on failure.
1377 Frequency : SLES 9 only
1379 Description: Error of copying files with lustre special EAs as root
1380 Details : Client side always return success for setxattr call for lustre
1381 special xattr (currently only "trusted.lov").
1384 Frequency : rarely on clusters with both ia64+i386 clients
1386 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1387 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1388 clients are doing concurrent writes to the same file it is
1389 possible that the ia64 clients may LASSERT because the OST
1390 extent locks are not PAGE_SIZE aligned. Ensure that grown
1391 locks are always aligned on the request boundary.
1394 Frequency : specific use, occasional
1396 Description: Overwriting in use executable truncates on-disk binary image
1397 Details : If one node attempts to overwrite an executable in use by
1398 another node, we now correctly return ETXTBSY instead of
1399 truncating the file.
1401 Severity : enhancement
1403 Description: Async OSC create to avoid the blocking unnecessarily.
1404 Details : If a OST has no remain object, system will block on the creating
1405 when need to create a new object on this OST. Now, ways use
1406 pre-created objects when available, instead of blocking on an
1407 empty osc while others are not empty. If we must block, we block
1408 for the shortest possible period of time.
1413 Description: chmod on Lustre root is propagated to other clients
1414 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1415 invalid by the follow_mount time.
1418 Frequency : liblustre clients only
1420 Description: Race in 'instant cancel' lock handling could lead to such locks
1421 never to be granted in case of SMP MDS
1422 Details : Do not destroy not yet granted but cbpending locks in
1426 Frequency : replay/resend of open
1428 Description: non null lock assetion failure in mds_intent_policy
1429 Details : Trying to replay/resend lockless open requests resulted in
1430 mds_open() returning 0 with no lock. Now it sets a flag if
1431 a lock is going to be returned.
1433 Severity : enhancement
1435 Description: Checksum enhancements
1436 Details : New checksum enhancements allow for resending RPCs that failed
1439 Severity : enhancement
1441 Description: Tunables on number of dirty pages in cacche
1442 Details : Allow to set limit on number of dirty pages cached.
1447 Description: client crash on unmount - lock still has references
1448 Details : In some error handling cases it was possible to leak a lock
1449 reference on a client while accessing a file. This was not
1450 harmful to the client during operation, but would cause the
1451 client to crash when the filesystem is unmounted.
1454 Frequency : specific case, rare
1456 Description: ETXTBSY on mds though file not in use by client
1457 Details : ETXTBSY is no longer incorrectly returned when attempting to
1458 chmod or chown a directory that the user previously tried to
1459 execute or a currently-executing binary.
1462 Frequency : extremely rare except on liblustre-based clients
1464 Description: Lustre space not freed when files are deleted
1465 Details : Clean up open-unlinked files after client eviction. Previously
1466 the unlink was skipped and the files remained as orphans.
1471 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1472 Details : In some cases it was possible to send a blocking callback to
1473 a client doing a glimpse, even though that client didn't get
1474 a lock granted. When the glimpse lock is cancelled on the OST
1475 the freed lock is left on the waiting list and corrupted the list.
1478 Frequency : all core dumps
1480 Description: Broke core dumps to lustre
1481 Details : Negative dentry may be unhashed if parent does not have UPDATE
1482 lock, but some callers, e.g. do_coredump, expect dentry to be
1483 hashed after successful create, hash it in ll_create_it.
1485 ------------------------------------------------------------------------------
1487 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1489 * Support for kernels:
1490 2.6.9-42.0.2.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)
1497 Frequency : always on RHEL 3
1499 Description: Number of open files grows over time
1500 Details : The number of open files grows over time, whether or not
1501 Lustre is started. This was due to a filp leak introduced
1502 by one of our kernel patches.
1504 ------------------------------------------------------------------------------
1506 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1508 * Support for kernels:
1509 2.6.9-42.EL (RHEL 4)
1510 2.6.5-7.276 (SLES 9)
1511 2.4.21-40.EL (RHEL 3)
1512 2.6.12.6 vanilla (kernel.org)
1517 Bugzilla : 5719, 9635, 9792, 9684
1518 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1519 Details : If a server is under heavy load and cannot reply to new
1520 connection requests before the client resends the (re)connect,
1521 the connection handling code can behave badly if two service
1522 threads are concurrently handing separate (re)connections from
1523 the same client. Add better locking to the connection handling
1524 code, and ensure that only a single connection will be processed
1525 for a given client UUID, even if the lock is dropped.
1527 Severity : enhancement
1529 Description: add TCP zero-copy support to kernel
1530 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1531 sends if the hardware supports scatter-gather and checksumming.
1532 This allows socklnd to do client-write and server-read more
1533 efficiently and reduce CPU utilization from skbuf copying.
1536 Frequency : only if NFS exporting from client
1538 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1539 Details : If setting attributes on a file created under NFS that had
1540 never been opened it would be possible to oops the client
1541 if the file had no objects.
1544 Frequency : always for liblustre
1546 Description: liblustre client does MDS+OSTs setattr RPC for each write
1547 Details : When doing a write from a liblustre client, the client
1548 incorrectly issued an RPC to the MDS and each OST the file was
1549 striped over in order to update the timestamps. When writing
1550 with small chunks and many clients this could overwhelm the MDS
1551 with RPCs. In all cases it would slow down the write because
1552 these RPCs are unnecessary.
1554 Severity : enhancement
1556 Description: allow number of MDS service threads to be changed at module load
1557 Details : It is now possible to change the number of MDS service threads
1558 running. Adding "options mds mds_num_threads={N}" to the MDS's
1559 /etc/modprobe.conf will set the number of threads for the next
1560 time Lustre is restarted (assuming the "mds" module is also
1561 reloaded at that time). The default number of threads will
1562 stay the same, 32 for most systems.
1567 Description: OST crash if filesystem is unformatted or corrupt
1568 Details : If an OST is started on a device that has never been formatted
1569 or if the filesystem is corrupt and cannot even mount then the
1570 error handling cleanup routines would dereference a NULL pointer.
1575 Description: NULL pointer deref in llap_from_page.
1576 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1577 incorrect) mapping if the page was truncated/reclaimed while it was
1578 searched for. Check for this condition and skip such pages when
1579 doing readahead. Introduce extra check to llap_from_page() to
1580 verify page->mapping->host is non-NULL (so page is not anonymous).
1583 Frequency : Sometimes when using sys_sendfile
1585 Description: "page not covered by a lock" warnings from ll_readpage
1586 Details : sendfile called ll_readpage without right page locks present.
1587 Now we introduced ll_file_sendfile that does necessary locking
1588 around call to generic_file_sendfile() much like we do in
1592 Frequency : with certain MDS communication failures at client mount time
1594 Description: NULL pointer deref after failed client mount
1595 Details : a client connection request may delayed by the network layer
1596 and not be sent until after the PTLRPC layer has timed out the
1597 request. If the client fails the mount immediately it will try
1598 to clean up before the network times out the request. Add a
1599 reference from the request import to the obd device and delay
1600 the cleanup until the network drops the request.
1603 Frequency : occasionally during client (re)connect
1605 Description: assertion failure during client (re)connect
1606 Details : processing a client connection request may be delayed by the
1607 client or server longer than the client connect timeout. This
1608 causes the client to resend the connection request. If the
1609 original connection request is replied in this interval, the
1610 client may trip an assertion failure in ptlrpc_connect_interpret()
1611 which thought it would be the only running connect process.
1614 Frequency : only with obd_echo servers and clients that are rebooted
1616 Description: kernel BUG accessing uninitialized data structure
1617 Details : When running an obd_echo server it did not start the ping_evictor
1618 thread, and when a client was evicted an uninitialized data
1619 structure was accessed. Start the ping_evictor in the RPC
1620 service startup instead of the OBD startup.
1622 Severity : enhancement
1623 Bugzilla : 10193 (patchless)
1624 Description: Remove dependency on various unexported kernel interfaces.
1625 Details : No longer need reparent_to_init, exit_mm, exit_files,
1626 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1629 Frequency : rare (only users of deprecated and unsupported LDAP config)
1631 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1632 Details : LDAP apparently contains 'lustreName' attributes instead of
1633 'name'. A simple remapping of the name is sufficient.
1636 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1638 Description: waiting_locks_callback trips kernel BUG if client is evicted
1639 Details : Running with the dump_on_timeout debug flag turned on makes
1640 it possible that the waiting_locks_callback() can try to dump
1641 the Lustre kernel debug logs from an interrupt handler. Defer
1642 this log dumping to the expired_lock_main() thread.
1644 Severity : enhancement
1646 Description: Support NFS exporting on 2.6 kernels.
1647 Details : Implement non-rawops metadata methods for NFS server to use without
1648 changing NFS server code.
1651 Frequency : very rare (synthetic metadata workload only)
1653 Description: two racing renames might cause an MDS thread to deadlock
1654 Details : Running the "racer" program may cause one MDS thread to rename
1655 a file from being the source of a rename to being the target of
1656 a rename at exactly the same time that another thread is doing
1657 so, and the second thread has already enqueued these locks after
1658 doing a lookup of the target and is trying to relock them in
1659 order. Ensure that we don't try to re-lock the same resource.
1662 Frequency : only very large systems with liblustre clients
1664 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1665 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1666 all OSTs in parallel. This allows the eviction of stale liblustre
1667 clients to proceed much faster than if they were done in series,
1668 and also offers similar improvements for other set_info RPCs.
1673 Description: excessive CPU usage during initial read phase on client
1674 Details : During the initial read phase on a client, it would agressively
1675 retry readahead on the file, consuming too much CPU and impacting
1676 performance (since 1.4.5.8). Improve the readahead algorithm
1677 to avoid this, and also improve some other common cases (read
1678 of small files in particular, where "small" is files smaller than
1679 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1684 Description: MDS crash when receiving packet with unknown intent.
1685 Details : Do not LBUG in unknown intent case, just return -EFAULT
1687 Severity : enhancement
1688 Bugzilla : 9293, 9385
1689 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1690 Details : Do not serialize getattr (non-intent version) and statfs.
1693 Frequency : occasional, when OST network is overloaded/intermittent
1695 Description: client evicted by OST after bulk IO timeout
1696 Details : If a client sends a bulk IO request (read or write) the OST
1697 may evict the client if it is unresposive to its data GET/PUT
1698 request. This is incorrect if the network is overloaded (takes
1699 too long to transfer the RPC data) or dropped the OST GET/PUT
1700 request. There is no need to evict the client at all, since
1701 the pinger and/or lock callbacks will handle this, and the
1702 client can restart the bulk request.
1705 Frequency : Always when mmapping file with no objects
1707 Description: client crashes when mmapping file with no objects
1708 Details : Check that we actually have objects in a file before doing any
1709 operations on objects in ll_vm_open, ll_vm_close and
1715 Description: Request leak when working with deleted CWD
1716 Details : Introduce advanced request refcount tracking for requests
1717 referenced from lustre intent.
1719 Severity : Enhancement
1721 Description: Cache open file handles on client.
1722 Details : MDS now will return special lock along with openhandle, if
1723 requested and client is allowed to hold openhandle, even if unused,
1724 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1725 closing files for every read/write openration.
1727 Severity : Enhancement
1729 Description: Cache open negative dentries on client when possible.
1730 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1731 negative dentries on lock revocation.
1736 Description: Remounting a client read-only wasn't possible with a zconf mount
1737 Details : It wasn't possible to remount a client read-only with llmount.
1739 Severity : enhancement
1740 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1741 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1742 MPICH is included to add Lustre-specific ADIO interfaces.
1743 This is based closely on the UFS ADIO layer and only differs
1744 in file creation, in order to allow the OST striping to be set.
1745 This is user-contributed code and not supported by CFS.
1750 Description: extended inode attributes (immutable, append-only) work improperly
1751 when 2.4 and 2.6 kernels are used on client/server or vice versa
1752 Details : Introduce kernel-independent values for these flags.
1754 Severity : enhancement
1757 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1758 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1759 granularity. Now, Fractional MB granularity is be supported,
1760 this is very useful for low memory system.
1762 Severity : enhancement
1764 Description: Getattr by fid
1765 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1766 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1767 have locks and avoids getattr rpc requests for it on every stat.
1770 Frequency : Always, for filesystems larger than 2TB
1772 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1773 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1774 filesystems larger than 2TB to be mounted with mballoc enabled.
1777 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1779 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1780 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1781 filesystems larger than 2TB is mounted then the kernel will
1782 silently corrupt the start of the filesystem. CONFIG_LBD is
1783 enabled for all CFS-supported kernels, but the possibility of
1784 this happening with a modified kernel config exists.
1786 Severity : enhancement
1788 Description: add client O_DIRECT support for 2.6 kernels
1789 Details : It is now possible to do O_DIRECT reads and writes to files
1790 in the Lustre client mountpoint on 2.6 kernel clients.
1792 Severity : enhancement
1794 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1795 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1796 parallel, not waiting for response from every OST before sending
1797 a rpc to the next OST.
1802 Description: setattr vs write race when updating file timestamps
1803 Details : Client processes that update a file timestamp into the past
1804 right after writing to the file (e.g. tar) it is possible that
1805 the updated file modification time can be reset to the current
1806 time due to a race between processing the setattr and write RPC.
1808 Severity : enhancement
1810 Description: Bring 'lfs find' closer in line with regular Linux find.
1811 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1812 -print0 options and obtains all the needed info through the lustre
1815 Severity : enhancement
1817 Description: support up to 1024 configured devices on one node
1818 Details : change obd_dev array from statically allocated to dynamically
1819 allocated structs as they are first used to reduce memory usage
1824 Description: Flush dirty partially truncated pages during truncate
1825 Details : Immediatelly flush partially truncated pages in filter_setattr,
1826 this way we completely avoid having any pages in page cache on OST
1827 and can retire ugly workarounds during writes to flush such pages.
1832 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1833 Details : protect inode from truncation within vfs_unlink() context
1834 just take a reference before calling vfs_unlink() and release it
1835 when parent's i_sem is free.
1840 Description: last_id value checked outside lock on OST caused LASSERT failure
1841 Details : If there were multiple MDS->OST object precreate requests in
1842 flight, it was possible that the OST's last object id was checked
1843 outside a lock and incorrectly tripped an assertion. Move checks
1844 inside locks, and discard old precreate requests.
1847 Frequency : always, if extents are used on OSTs
1849 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
1850 Details : OSTs using the extents format would not zero the high 16 bits of
1851 the index physical block number. This is not a problem for any
1852 OST filesystems smaller than 16TB, and no kernels support ext3
1853 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
1854 new/modified files) and can be fixed for existing filesystems
1855 with e2fsprogs-1.39-cfs1.
1860 Description: import connection selection may be incorrect if timer wraps
1861 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
1862 import connections to be ignored if the 32-bit jiffies counter
1863 wraps. Use a 64-bit jiffies counter.
1866 Frequency : very large clusters immediately after boot
1868 Description: LNET request buffers exhausted under heavy short-term load
1869 Details : If a large number of client requests are generated on a service
1870 that has previously never seen so many requests it is possible
1871 that the request buffer growth cannot keep up with the spike in
1872 demand. Instead of dropping incoming requests, they are held in
1873 the LND until the RPC service can accept more requests.
1876 Frequency : Sometimes during replay
1878 Description: Assertion failure in ll_local_open after replay.
1879 Details : If replay happened on an open request reply before we were able
1880 to set replay handler, reply will become not swabbed tripping the
1881 assertion in ll_local_open. Now we set the handler right after
1882 recognising of open request
1885 Frequency : very rare
1887 Description: kernel reports "badness in vsnprintf"
1888 Details : Reading from the "recovery_status" /proc file in small chunks
1889 may cause a negative length in lprocfs_obd_rd_recovery_status()
1890 call to vsnprintf() (which is otherwise harmless). Exit early
1891 if there is no more space in the output buffer.
1893 Severity : enhancement
1895 Description: clear OBD RPC statistics by writing to them
1896 Details : It is now possible to clear the OBD RPC statistics by writing
1897 to the "stats" file.
1902 Description: Client mtime is not the same on different clients after utimes
1903 Details : In some cases, the client was using the utimes() syscall on
1904 a file cached on another node. The clients now validate the
1905 ctime from the MDS + OSTs to determine which one is right.
1910 Description: Inability to activate failout mode
1911 Details : lconf script incorrectly assumed that in pythong string's numeric
1912 value is used in comparisons.
1915 Frequency : always with multiple stripes per file
1917 Description: Inefficient object allocation for mutli-stripe files
1918 Details : When selecting which OSTs to stripe files over, for files with
1919 a stripe count that divides evenly into the number of OSTs,
1920 the MDS is always picking the same starting OST for each file.
1921 Return the OST selection heuristic to the original design.
1926 Description: mount failures may take full timeout to return an error
1927 Details : Under some heavy load conditions it is possible that a
1928 failed mount can wait for the full obd_timeout interval,
1929 possibly several minutes, before reporting an error.
1930 Instead return an error as soon as the status is known.
1932 Frequency : quota enabled and large files being deleted
1934 Description: releasing more than 4GB of quota at once hangs OST
1935 Details : If a user deletes more than 4GB of files on a single OST it
1936 will cause the OST to spin in an infinite loop. Release
1937 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1942 Description: statfs data retrieved from /proc may be stale or zero
1943 Details : When reading per-device statfs data from /proc, in the
1944 {kbytes,files}_{total,free,avail} files, it may appear
1945 as zero or be out of date.
1948 Frequency : systems with MD RAID1 external journal devices
1950 Description: lconf's call to blkid is confused by RAID1 journal devices
1951 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1952 of returning all block devices that match the journal UUID.
1955 Frequency : always, for aggregate stripe size over 4GB
1957 Description: assertion fails when trying to use 4GB stripe size
1958 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
1959 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
1962 Frequency : always on ppc64
1964 Description: the first write on an ext3 filesystem with mballoc got stuck
1965 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1966 perform endianness conversion.
1968 ------------------------------------------------------------------------------
1970 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1972 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1973 INTEROPERATE with older versions automatically. Please read the
1974 user documentation before upgrading any part of a live system.
1975 * WARNING: Lustre networking configuration changes are required with
1976 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1979 * Support for kernels:
1980 2.6.9-22.0.2.EL (RHEL 4)
1981 2.6.5-7.244 (SLES 9)
1982 2.6.12.6 vanilla (kernel.org)
1985 Severity : enhancement
1986 Bugzilla : 7981/8208
1987 Description: Introduced Lustre Networking (LNET)
1988 Details : LNET is new networking infrastructure for Lustre, it includes
1989 a reorganized network configuration mode (see the user
1990 documentation for full details) as well as support for routing
1991 between different network fabrics. Lustre Networking Devices
1992 (LNDs) for the supported network fabrics have also been
1993 created for this new infrastructure.
1995 Severity : enhancement
1996 Description: Introduced Access control lists
1997 Details : clients can set ACLs on files and directories in order to have
1998 more fine-grained permissions than the standard Unix UGO+RWX.
1999 The MDS must be started with the "-o acl" mount option.
2001 Severity : enhancement
2002 Description: Introduced filesystem quotas
2003 Details : Administrators may now establish per-user quotas on the
2006 Severity : enhancement
2008 Description: Configuration change for the XT3
2009 The PTLLND is now used to run Lustre over Portals on the XT3
2010 The configure option(s) --with-cray-portals are no longer used.
2011 Rather --with-portals=<path-to-portals-includes> is used to
2012 enable building on the XT3. In addition to enable XT3 specific
2013 features the option --enable-cray-xt3 must be used.
2018 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2019 Details : A race between allocating a new llcd and re-getting the llcd_lock
2020 allowed another thread to grab newly-allocated llcd.
2022 Severity : enhancement
2024 Description: 2.6 OST async journal commit and locking fix to improve performance
2025 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2026 2.6 kernels are now async as they already were in 2.4 kernels so
2027 that they can commit concurrently with the network bulk transfer.
2028 For block-allocated files the filter allocation semaphore is held
2029 to avoid filesystem fragmentation during allocation. BKL lock
2030 removed for 2.6 xattr operations where it is no longer needed.
2035 Description: lconf incorrectly determined whether two IP networks could talk
2036 Details : In some more complicated routing and multiple-network
2037 configurations, lconf will avoid trying to make a network
2038 connection to a disjoint part of the IP space. It was doing the
2039 math incorrectly for one set of cases.
2044 Description: Fix for potential infinite loop processing records in an llog.
2045 Details : If an llog record is corrupted/zeroed, it is possible to loop
2046 forever in llog_process(). Validate the llog record length
2047 and skip the remainder of the block on error.
2050 Frequency : occasional (liblustre only)
2052 Description: liblustre could not open files whose last component is a symlink
2053 Details : sysio_path_walk() would incorrectly pass the open intent to
2054 intermediate path components.
2057 Frequency : rare (liblustre only with non-standard tuning)
2058 Bugzilla : 7201 (7350)
2059 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2060 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2061 after it has been released, only do this for referenced locks
2063 Severity : enhancement
2065 Description: specifying an (invalid) directory default stripe_size of -1
2066 would reset the directory default striping
2067 Details : stripe_size -1 was used internally to signal directory stripe
2068 removal, now use "all default" to signal dir stripe removal
2069 as a directory striping of "all default" is not useful
2072 Frequency : common for large clusters running liblustre clients
2074 Description: doing an ls when liblustre clients are running is slow
2075 Details : sending a glimpse AST to a liblustre client waits for every AST
2076 to time out, as liblustre clients will not respond. Since they
2077 cannot cache data we refresh the OST lock LVB from disk instead.
2079 Severity : enhancement
2081 Description: doing an ls at the same time as file IO can be slow
2082 Details : enqueue and other "small" requests can be blocked behind many
2083 large IO requests. Create a new OST IO portal for non-IO
2084 requests so they can be processed faster.
2087 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2089 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2090 created file with mode 000
2091 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2092 allow an ftruncate() as a non-root user to a file with mode 000.
2093 Lustre now respects this flag to disable mode checks when
2094 truncating a file owned by the user
2097 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2100 Description: Revoking locks from clients that went dead or catatonic might take
2102 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2103 cancellation of such locks instant on servers without waiting for
2104 any reply from clients. Clients drops these locks when cancel
2105 notification from server is received without replying.
2108 Frequency : liblustre-only, when liblustre client dies or becomes busy
2110 Description: Doing ls on Linux clients can take a long time with active
2112 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2113 granting such locks to it in the first place if possible. Locks
2114 are taken by proxy on the OST during the read or write and
2115 dropped immediately afterward. Add connect flags handling, do
2116 not grant locks to liblustre clients for glimpse ASTs.
2118 Severity : enhancement
2120 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2121 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2122 sequential reads and behaves badly with applications doing random
2123 reads. Improve it by reading ahead at least read region, and
2124 avoiding excessive large RPC for small reads.
2126 Severity : enhancement
2128 Description: Creating more than 1000 files for a single job may cause a load
2129 imbalance on the OSTs if there are also a large number of OSTs.
2130 Details : qos_prep_create() uses an OST index reseed value that is an
2131 even multiple of the number of available OSTs so that if the
2132 reseed happens in the middle of the object allocation it will
2133 still utilize the OSTs as uniformly as possible.
2138 Description: OST or MDS may oops in ping_evictor_main()
2139 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2140 but doesn't restart at beginning of obd_exports_timed list
2143 Severity : enhancement
2145 Description: improve by-nid export eviction on the MDS and OST
2146 Details : allow multiple exports with the same NID to be evicted at one
2147 time without re-searching the exports list.
2150 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2152 Description: MDS may oops in groups_free()
2153 Details : in rare race conditions a newly allocated group_info struct is
2154 freed again, and this can be NULL. The 2.4 compatibility code
2155 for groups_free() checked for a NULL pointer, but 2.6 did not.
2158 Frequency : common for liblustre clients doing little filesystem IO
2159 Bugzilla : 9352, 7313
2160 Description: server may evict liblustre clients accessing contended locks
2161 Details : if a client is granted a lock or receives a completion AST
2162 with a blocking AST already set it would not reply to the AST
2163 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2166 Frequency : lfs setstripe, only systems with more than 160 OSTs
2168 Description: unable to set striping with a starting offset beyond OST 160
2169 Details : llapi_create_file() incorrectly limited the starting stripe
2170 index to the maximum single-file stripe count.
2173 Frequency : LDAP users only
2175 Description: lconf did not handle in-kernel recovery with LDAP properly
2176 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2178 Severity : enhancement
2180 Description: bind OST threads to NUMA nodes to improve performance
2181 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2182 node and do their allocations there to localize memory access
2184 Severity : enhancement
2186 Description: llmount can determine client NID directly from Myrinet (GM)
2187 Details : the client NID code from gmnalnid was moved directly into
2188 llmount, removing the need to use this or specifying the
2189 client NID explicitly when mounting GM clients with zeroconf
2192 Frequency : if client is started with down MDS
2194 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2195 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2196 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2197 ptlrpc_import_delayed_req() because we didn't check if the
2198 request was interrupted, and we also didn't break out of the
2199 event loop if there was no timeout
2204 Description: data loss during non-page-aligned writes to a single file from
2205 both multiple nodes and multiple threads on one node at same time
2206 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2207 inconsistency led to false short-reads, that were cached and later
2208 used by ->prepare_write() to fill in partially written page,
2209 leading to data loss.
2212 Frequency : always, if lconf --abort_recovery used
2214 Description: lconf --abort_recovery fails with 'Operation not supported'
2215 Details : lconf was attempting to abort recovery on the MDT device and not
2218 Severity : enhancement
2220 Description: remove cleanup logs
2221 Details : replace lconf-generated cleanup logs with lustre internal
2222 cleanup routines. Eliminates the need for client-cleanup and
2225 Severity : enhancement
2227 Description: add support for EAs (user and system) on lustre filesystems
2228 Details : it is now possible to store extended attributes in the Lustre
2229 client filesystem, and with the user_xattr mount option it
2230 is possible to allow users to store EAs on their files also
2232 Severity : enhancement
2234 Description: Add possibility (config option) to show minimal available OST free
2236 Details : When compiled with --enable-mindf configure option, statfs(2)
2237 (and so, df) will return least minimal free space available from
2238 all OSTs as amount of free space on FS, instead of summary of
2239 free spaces of all OSTs.
2241 Severity : enhancement
2243 Description: do not expand extent locks acquired on OST-side
2244 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2245 by server: they are not cached anyway.
2248 Frequency : when mmap is used/binaries executed from Lustre
2250 Description: Unmmap pages before throwing them away from read cache.
2251 Details : llap_shrink cache now attempts to unmap pages before discarding
2252 them (if unmapping failed - do not discard). SLES9 kernel has
2253 extra checks that trigger if this unmapping is not done first.
2258 Description: lconf didn't resolve symlinks before checking to see whether a
2259 given mountpoint was already in use
2262 Frequency : when migrating failover services
2263 Bugzilla : 6395, 9514
2264 Description: When migrating a subset of services from a node (e.g. failback
2265 from a failover service node) the remaining services would
2266 time out and evict clients.
2267 Details : lconf --force (implied by --failover) sets the global obd_timeout
2268 to 5 seconds in order to quickly disconnect, but this caused
2269 other RPCs to time out too quickly. Do not change the global
2270 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2272 Severity : enhancement
2273 Frequency : if MDS is started with down OST
2274 Bugzilla : 9439,5706
2275 Description: Allow startup/shutdown of an MDS without depending on the
2276 availability of the OSTs.
2277 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2278 Add appropriate locking and lov-osc refcounts for safe
2279 cleaning. Add osc abort_inflight calls in case the
2280 synchronize never started.
2283 Frequency : occasional (Cray XT3 only)
2285 Description: root not authorized to access files in CRAY_PORTALS environment
2286 Details : The client process capabilities were not honoured on the MDS in
2287 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2288 been accessed by an authorized user then root was able to access
2289 the file on the local client also. The root user capabilities
2290 are now allowed on the MDS, as this environment has secure UID.
2293 Frequency : occasional
2295 Description: ldiskfs "too long searching" message happens too often
2296 Details : A debugging message (otherwise harmless) prints too often on
2297 the OST console. This has been reduced to only happen when
2298 there are fragmentation problems on the filesystem.
2303 Description: Division by zero in statfs when all OSCs are inactive
2304 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2305 lov_statfs divides by value returned by lov_get_stripecnt().
2309 Bugzilla : 9489, 3273
2310 Description: First write from each client to each OST was only 4kB in size,
2311 to initialize client writeback cache, which caused sub-optimal
2312 RPCs and poor layout on disk for the first writen file.
2313 Details : Clients now request an initial cache grant at (re)connect time
2314 and so that they can start streaming writes to the cache right
2315 away and always do full-sized RPCs if there is enough data.
2316 If the OST is rebooted the client also re-establishes its grant
2317 so that client cached writes will be honoured under the grant.
2322 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2323 Details : Now I/O RPCs go to different portal number and (presumably) fast
2324 lock requests (and glimses) and other RPCs get their own service
2325 threads pool that should be able to service those RPCs
2328 Severity : enhancement
2330 Description: Ability to exchange lustre version between client and servers and
2331 issue warnings at client side if client is too old. Also for
2332 liblustre clients there is ability to refuse connection of too old
2334 Details : New 'version' field is added to connect data structure that is
2335 filled with version info. That info is later checked by server and
2339 Frequency : rare, liblustre only.
2340 Bugzilla : 9296, 9581
2341 Description: Two simultaneous writes from liblustre at offset within same page
2342 might proceed at the same time overwriting eachother with stale
2344 Details : I/O lock withing llu_file_prwv was released too early, before data
2345 actually was hitting the wire. Extended lock-holding time until
2346 server acknowledges receiving data.
2349 Frequency : extremely rare. Never observed in practice.
2351 Description: avoid generating lustre_handle cookie of 0.
2352 Details : class_handle_hash() generates handle cookies by incrementing
2353 global counter, and can hit 0 occasionaly (this is unlikely, but
2354 not impossible, because initial value of cookie counter is
2355 selected randonly). Value of 0 is used as a sentinel meaning
2356 "unassigned handle" --- avoid it. Also coalesce two critical
2357 sections in this function into one.
2359 Severity : enhancement
2361 Description: allow liblustre clients to delegate truncate locking to OST
2362 Details : To avoid overhead of locking, liblustre client instructs OST to
2363 take extent lock in ost_punch() on client's behalf. New connection
2364 flag is added to handle backward compatibility.
2366 Severity : enhancement
2367 Bugzilla : 4928, 7341, 9758
2368 Description: allow number of OST service threads to be specified
2369 Details : a module parameter allows the number of OST service threads
2370 to be specified via "options ost ost_num_threads={N}" in the
2371 OSS's /etc/modules.conf or /etc/modprobe.conf.
2375 Bugzilla : 6146, 9635, 9895
2376 Description: servers crash with bad pointer in target_handle_connect()
2377 Details : In rare cases when a client is reconnecting it was possible that
2378 the connection request was the last reference for that export.
2379 We would temporarily drop the export reference and get a new
2380 one, but this may have been the last reference and the export
2381 was just destroyed. Get new reference before dropping old one.
2383 Severity : enhancement
2384 Frequency : if client is started with failover MDS
2386 Description: Allow multiple MDS hostnames in the mount command
2387 Details : Try to read the configuration from all specified MDS
2388 hostnames during a client mount in case the "primary"
2391 Severity : enhancement
2393 Description: Stop sending data to evicted clients as soon as possible.
2394 Details : Check if the client we are about to send or are sending data to
2395 was evicted already. (Check is done every second of waiting,
2396 for which l_wait_event interface was extended to allow checking
2397 of exit condition at specified intervals).
2400 Frequency : rare, normally only when NFS exporting is done from client
2402 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2404 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2407 Frequency : always, if srand() is called before liblustre initialization
2409 Description: Liblustre uses system PRNG disturbing its usage by user application
2410 Details : Introduce internal to lustre fast and high-quality PRNG for
2411 lustre usage and make liblustre and some other places in generic
2412 lustre code to use it.
2414 Severity : enhancement
2415 Bugzilla : 9477, 9557, 9870
2416 Description: Verify that the MDS configuration logs are updated when xml is
2417 Details : Check if the .xml configuration logs are newer than the config
2418 logs stored on the MDS and report an error if this is the case.
2419 Request --write-conf, or allow starting with --old_conf.
2421 Severity : enhancement
2423 Description: Handle symlinks in the path when checking if Lustre is mounted.
2424 Details : Resolve intermediate symlinks when checking if a client has
2425 mounted a filesystem to avoid duplicate client mounts.
2430 Description: lconf can hit an error exception but still return success.
2431 Details : The lconf command catches the Command error exception at the top
2432 level script context and will exit with the associated exit
2433 status, but doesn't ensure that this exit status is non-zero.
2438 Description: failure of ptlrpc thread startup can cause oops
2439 Details : Starting a ptlrpc service thread can fail if there are a large
2440 number of threads or the server memory is very fragmented.
2441 Handle this without oopsing.
2444 Frequency : always, only if liblustre and non-default acceptor port was used
2446 Description: liblustre cannot connect to servers with non-default acceptor port
2447 Details : tcpnal_set_default_params() was not called and was therefore
2448 ignoring the environment varaible TCPNAL_PORT, as well as other
2449 TCPNAL_ environment variables
2454 Description: two objects could be created on the same OST for a single file
2455 Details : If an OST is down, in some cases it was possible to create two
2456 objects on a single OST for a single file. No problems other
2457 than potential performance impact and spurious error messages.
2461 Bugzilla : 5681, 9562
2462 Description: Client may oops in ll_unhash_aliases
2463 Details : Client dcache may become inconsistent in race condition.
2464 In some cases "getcwd" can fail if the current directory is
2470 Description: Inode refcounting problems in NFS export code
2471 Details : link_raw functions used to call d_instantiate without obtaining
2472 extra inode reference first.
2476 Bugzilla : 9942, 9903
2477 Description: Referencing freed requests leading to crash, memleaks with NFS.
2478 Details : We used to require that call to ll_revalidate_it was always
2479 followed by ll_lookup_it. Also with revalidate_special() it is
2480 possible to call ll_revalidate_it() twice for the same dentry
2481 even if first occurence returned success. This fix changes semantic
2482 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2483 reference on a request referred from the intent.
2484 ll_intent_release() then releases such a request.
2487 Frequency : rare, normally benchmark loads only
2489 Description: unlinked inodes were kept in memory on the client
2490 Details : If a client is repeatedly creating and unlinking files it
2491 can accumulate a lot of stale inodes in the inode slab cache.
2492 If there is no other client load running this can cause the
2493 client node to run out of memory. Instead flush old inodes
2494 from client cache that have the same inode number as a new inode.
2497 Frequency : SLES9 2.6.5 kernel and long filenames only
2498 Bugzilla : 9969, 10379
2499 Description: utime reports stale NFS file handle
2500 Details : SLES9 uses out-of-dentry names in some cases, which confused
2501 the lustre dentry revalidation. Change it to always use the
2505 Frequency : rare, unless heavy write-truncate concurrency is continuous
2506 Bugzilla : 4180, 6984, 7171, 9963, 9331
2507 Description: OST becomes very slow and/or deadlocked during object unlink
2508 Details : filter_destroy() was holding onto the parent directory lock
2509 while truncating+unlinking objects. For very large objects this
2510 may block other threads for a long time and slow overall OST
2511 responsiveness. It may also be possible to get a lock ordering
2512 deadlock in this case, or run out of journal credits because of
2513 the combined truncate+unlink. Solution is to do object truncate
2514 first in one transaction without parent lock, and then do the
2515 final unlink in a new transaction with the parent lock. This
2516 reduces the lock hold time dramatically.
2519 Frequency : rare, 2.4 kernels only
2521 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2522 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2523 before unmounting their filesystem in order to avoid deadlock.
2524 The kernel_locked() function in 2.4 kernels only checks whether
2525 the kernel lock is held, not whether it is this process that is
2526 holding it as 2.6 kernels do.
2531 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2532 Details : The client ptlrpc code may be trying to reconnect to a down
2533 server before a previous connection attempt has timed out.
2534 Increase the reconnect interval to be longer than the connection
2535 timeout interval to avoid sending duplicate connections to
2539 Frequency : echo_client brw_test command
2541 Description: fix echo_client to work with OST preallocated code
2542 Details : OST preallocation code (5137) didn't take echo_client IO path
2543 into account: echo_client calls filter methods outside of any
2544 OST thread and, hence, there is no per-thread preallocated
2545 pages and buffers to use. Solution: hijack pga pages for IO. As
2546 a byproduct, this avoids unnecessary data copying.
2550 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2551 Description: Client can oops in mdc_commit_close() after open replay
2552 Details : It was possible for the MDS to return an open request with no
2553 transaction number in mds_finish_transno() if the client was
2554 evicted, but without actually returning an error. Clients
2555 would later try to replay that open and may trip an assertion
2556 Simplify the client close codepath, and always return an error
2557 from the MDS in case the open is not successful.
2560 Frequency : rare, 2.6 OSTs only
2562 Description: OST may deadlock under high load on fragmented files
2563 Details : If there was a heavy load and highly-fragmented OST filesystems
2564 it was possible to have all the OST threads deadlock waiting on
2565 allocation of biovecs, because the biovecs were not released
2566 until the entire RPC IO was completed. Instead, release biovecs
2567 as soon as they are complete to ensure forward IO progress.
2569 Severity : enhancement
2571 Description: Support for specifying external journal device at mount
2572 Details : If an OST or MDS device is formatted with an external journal
2573 device, this device major/minor is stored in the ext3 superblock
2574 and may not be valid for failover. Allow detecting and
2575 specifying the external journal at mount time.
2580 Description: Mounting an MDS with pending unlinked files may cause oops
2581 Details : target_finish_recovery() calls mds_postrecov() which returned
2582 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2583 considers this an error and immediately begins cleaning up the
2584 lov, just after starting the mds_lov process
2586 Severity : enhancement
2588 Description: Implement 'lfs df' to report actual free space on per-OST basis
2589 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2590 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2591 usage of objects; '-h' to report in human readable format.
2593 ------------------------------------------------------------------------------
2595 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2602 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2603 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2604 at startup during orphan unlinking. Instead, skip buddy update
2605 before setup, as it will be regenerated after recovery is complete.
2610 Description: If an OST is inactive, its locks might reference stale inodes.
2611 Details : lov_change_cbdata() must iterate over all namespaces, even if
2612 they are inactive to clear inode references from the lock.
2614 Severity : enhancement
2615 Frequency : occasional, if non-standard max_dirty_mb used
2617 Description: Client will block write RPCs if not enough grant
2618 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2619 then the client will block writes while waiting for another RPC
2620 to complete instead of consuming its dirty limit. With change
2621 we get improved performance when max_dirty_mb is small.
2623 Severity : enhancement
2624 Bugzilla : 3389, 6253
2625 Description: Add support for supplementary groups on the MDS.
2626 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2627 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2628 lookups for user supplementary groups into a cache.
2632 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2633 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2634 for liblustre O_EXCL is enforced by MDS.
2638 Description: readdir never returns NULL in liblustre.
2639 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2640 dentry in struct dirent.
2644 Frequency : liblustre only, depends on application IO pattern
2645 Description: liblustre clients evicted if not contacting servers
2646 Details : Don't put liblustre clients into the ping_evictor list, so
2647 they will not be evicted by the pinger ever.
2649 Severity : enhancement
2651 Description: Add ability to evict clients by NID from MDS.
2652 Details : By echoing "nid:$NID" string into
2653 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2654 $NID would be instantly evicted from this MDS and from all active
2655 OSTs connected to it.
2659 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2660 Details : lookup_it_finish() used to query file size from OSTs that was not
2665 Description: service threads change working directory to that of init
2666 Details : Starting lustre service threads may pin the working directory
2667 of the parent thread, making that filesystem busy. Threads
2668 now change to the working directory of init to avoid this.
2672 Frequency : during shutdown only
2673 Description: shutdown with a failed MDS or OST can cause unmount to hang
2674 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2679 Frequency : 2.6 only
2680 Description: chmod/chown may include an extra supplementary group
2681 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2682 supplementary group and if none is specified this is used.
2685 Bugzilla : 5479 (6816)
2687 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2688 Details : If lookup is in progress on a file that is unlinked we might try
2689 to revalidate the inode and fail in revalidate after lookup is
2690 complete and ll_file_open() enqueues the open again but
2691 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2694 Frequency : always, if lconf --abort_recovery used
2696 Description: lconf --abort_recovery fails with 'Operation not supported'
2697 Details : lconf was attempting to abort recovery on the MDT device and not
2700 ------------------------------------------------------------------------------
2702 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2707 Frequency : rare (only unsupported configurations with a node running as an
2709 Bugzilla : 6514, 5137
2710 Description: Mounting a Lustre file system on a node running as an OST could
2712 Details : OSTs now preallocates memory needed to write out data at
2713 startup, instead of when needed, to avoid having to
2714 allocate memory in possibly low memory situations.
2715 Specifically, if the file system is mounted on on OST,
2716 memory pressure could force it to try to write out data,
2717 which it needed to allocate memory to do. Due to the low
2718 memory, it would be unable to do so and the node would
2719 become unresponsive.
2721 Severity : enhancement
2723 Description: Addition of lconf --service command line option
2724 Details : lconf now accepts a '--service <arg>' option, which is
2725 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2727 Severity : enhancement
2729 Description: Failover mode is now the default for OSTs.
2730 Details : By default, OSTs will now run in failover mode. To return to
2731 the old behaviour, add '--failout' to the lmc line for OSTs.
2733 Severity : enhancement
2735 Description: Health checks are now provided for MDS and OSTs
2736 Details : Additional detailed health check information on MSD and OSTs
2737 is now provided through the procfs health_check value.
2740 Frequency : occasional, depends on IO load
2742 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2743 after numerous create/delete cycles
2744 Details : The ext3 inode allocation policy would not allocate new inodes
2745 very well on the OSTs because there are no new directories
2746 being created. Instead we look for groups with free space if
2747 the parent directories are nearly full.
2752 Description: Network or server problems during mount may cause partially
2753 mounted clients instead of returning an error.
2754 Details : The config llog parsing code may overwrite the error return
2755 code during mount error handling, returning success instead
2761 Description: MDS can fail to allocate large reply buffers
2762 Details : After long uptimes the MDS can fail to allocate large reply
2763 buffers (e.g. zconf client mount config records) due to memory
2764 fragmentation or consumption by the buffer cache. Preallocate
2765 some large reply buffers so that these replies can be sent even
2766 under memory pressure.
2770 Frequency : rare (liblustre)
2771 Description: fsx running with liblustre complained that using truncate() to
2772 extend the file doesn't work. This patch corrects that issue.
2773 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2774 ATTR_SIZE and lsm use in llu_setattr_raw.
2778 Frequency : rare, only 2.6 kernels
2779 Description: Unusual file access patterns on the MDS may result in inode
2780 data being lost in very rare circumstances.
2781 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2782 code in the 2.6 kernel caused the inode and/or EA data not to
2783 be read from disk, causing single-file corruption.
2787 Frequency : rare, only 2.6 filesystems using extents
2788 Description: Heavy concurrent write and delete load may cause data corruption.
2789 Details : It was possible under high-load situations to have an extent
2790 metadata block in the block device cache from a just-unlinked
2791 file overwrite a newly-allocated data block. We now unmap any
2792 metadata buffers that alias just-allocated data blocks.
2796 Frequency : filesystems with default stripe_count larger than 77
2797 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2798 Details : lconf specifies an inode size of 4096 bytes when the default
2799 stripe_count is larger than 77. This conflicts with the default
2800 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2801 this case to avoid pinning too much memory for large EAs.
2803 ------------------------------------------------------------------------------
2805 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2810 Frequency : rare (extremely heavy IO load with hundreds of clients)
2812 Description: Client is evicted, gets IO error writing to file
2813 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2814 caused clients to be evicted for AST timeouts. The fixes in
2815 bug 5192 mean we no longer need to have such short AST timeouts
2816 so ldlm_timeout has been increased.
2819 Frequency : occasional during --force or --failover shutdown under load
2820 Bugzilla : 5949, 4834
2821 Description: Server oops/LBUG if stopped with --force or --failover under load
2822 Details : a collection of import/export refcount and cleanup ordering
2823 issues fixed for safer force cleanup
2826 Frequency : only filesystems larger than 120 OSTs
2827 Bugzilla : 5990, 6223
2828 Description: lfs getstripe would oops on a very large filesystem
2829 Details : lov_getconfig used kfree on vmalloc'd memory
2832 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2833 Bugzilla : 6242, 6243
2834 Description: reading from files that had been truncated to a non-zero size
2835 but never opened returned no data
2836 Details : ll_file_read() reads zeros from no-object files to EOF
2841 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2842 Details : There's little harm in aborting MDS/OSS recovery and letting it
2843 try again, so I removed the LASSERT and return an error instead.
2845 Severity : enhancement
2847 Description: New debugging infrastructure for tracking down data corruption
2848 Details : The I/O checksum code was replaced to: (a) control it at runtime,
2849 (b) cover more of the client-side code path, and (c) try to narrow
2850 down where problems occurred
2854 Bugzilla : 3819, 4364, 4397, 6313
2855 Description: Racing close and eviction MDS could cause assertion in mds_close
2856 Details : It was possible to get multiple mfd references during close and
2857 client eviction, leading to one thread referencing a freed mfd.
2859 Severity: : enhancement
2860 Bugzilla : 3262, 6359
2861 Description: Attempts to reconnect to servers are now more aggressive.
2862 Details : This builds on the enhanced upcall-less recovery that was added
2863 in 1.4.2. When trying to reconnect to servers, clients will
2864 now try each server in the failover group every 10 seconds. By
2865 default, clients would previously try one server every 25 seconds.
2870 Description: After recovery, certain operations trigger a failed
2871 assertion on a client.
2872 Details : Failing over an mds, using lconf -d --failover, while a
2873 client was doing a readdir() call would cause the client to
2874 LBUG after recovery completed and the readdir() was resent.
2876 Severity : enhancement
2878 Description: Default groups are now added by lconf
2879 Details : You can now run lconf --group <servicename> without having to
2880 manually add groups with lmc.
2883 Frequency : occasional
2885 Description: Nodes with an elan id of 0 trigger a failed assertion
2888 Frequency : always when accessing e.g. tty/console device nodes
2890 Description: tty and some other devices nodes cannot be used on lustre
2891 Details : file's private_data field is used by device data and lustre
2892 values in there got lost. New field was added to struct file to
2893 store fs-specific private data.
2896 Frequency : when exporting Lustre via NFS
2898 Description: NFSD failed occasionally when looking up a path component
2899 Details : NFSD is looking up ".." which was broken in ext3 directories
2900 that had grown large enough to become hashed.
2903 Frequency : Clusters with multiple interfaces not on the same subnet
2905 Description: Nodes will repeatedly try to reconnect to an interface which it
2906 cannot reach and report an error to the log.
2907 Details : Extra peer list entries will be created by lconf with some peers
2908 unreachable. lconf now validates the peer before adding it.
2911 Frequency : Only if a default stripe is set on the filesystem root.
2913 Description: Setting a default stripe on the filesystem root prevented the
2914 filesystem from being remounted.
2915 Details : The client was sending extra request flags in the root getattr
2916 request and did not allocate a reply buffer for the dir EA.
2919 Frequency : occasional, higher if lots of files are accessed by one client
2920 Bugzilla : 6159, 6097
2921 Description: Client trips assertion regarding lsm mismatch/magic
2922 Details : While revalidating inodes the VFS looks up inodes with ifind()
2923 and in rare cases can find an inode that is being freed.
2924 The ll_test_inode() code will free the lsm during ifind()
2925 when it finds an existing inode and then the VFS later attaches
2926 this free lsm to a new inode.
2930 Bugzilla : 6422, 7030
2931 Description: MDS deadlock between mkdir and client eviction
2932 Details : Creating a new file via mkdir or mknod (starting a transaction
2933 and getting the ns lock) can deadlock with client eviction
2934 (gets ns lock and trying to finish a synchronous transaction).
2937 Frequency : occasional
2938 Description: While starting a server, the fsfilt_ext3 module could not be
2940 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2941 kernels. Previously, lconf would still use the ext3 name
2942 when trying to load modules. Now, it will correctly use
2943 ext3 on 2.4 and ldiskfs on 2.6.
2945 Severity : enhancement
2946 Description: The default stripe count has been changed to 1
2947 Details : The interpretation of the default stripe count (0, to lfs
2948 or lmc) has been changed to mean striping across a single
2949 OST, rather than all available. For general usage we have
2950 found a stripe count of 1 or 2 works best.
2952 Severity : enhancement
2953 Description: Add support for compiling against Cray portals.
2954 Details : Conditional compiling for some areas that are different
2958 Frequency : occasional
2959 Bugzilla : 6409, 6834
2960 Description: Creating files with an explicit stripe count may lead to
2961 a failed assertion on the MDS
2962 Details : If some OSTs are full or unavailable, creating files may
2963 trigger a failed assertion on the MDS. Now, Lustre will
2964 try to use other servers or return an error to the
2968 Frequency : occasional
2970 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2971 caused lock timeout during readahead (since 1.4.2).
2972 Details : Processes doing readahead might match a lock that hasn't been
2973 granted yet if there are overlapping and conflicting lock
2974 requests. The readahead process waits on ungranted lock
2975 (original lock is CBPENDING), while OST waits for that process
2976 to cancel CBPENDING read lock and eventually evicts client.
2978 Severity : enhancement
2980 Description: Initial enabling of flock support for clients
2981 Details : Implements fcntl advisory locking and file status functions.
2982 This feature is provided as an optional mount flag (default
2983 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2984 locking are implemented yet, and those that are are not guaranteed
2985 to be completely correct in production environments.
2986 mount -t lustre -o [flock|noflock] ...
2989 Frequency : occasional
2991 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2992 assertion in the ext3 JBD (journaling) layer.
2993 Details : The b_committed_data struct is protected by the big kernel lock
2994 in 2.4 kernels, serializing journal_commit_transaction() and
2995 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2996 access to this struct. In 2.6 kernels there is finer grained
2997 locking to improve SMP performance of the JBD layer.
3001 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3006 Description: merge and cleanup kernel patches.
3009 -----------------------------------------------------------------------------
3011 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3013 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3014 versions of Lustre. Please update all servers and clients to
3015 version 1.4.2 or later at the same time. You must also run
3016 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3017 to update the configuration logs.
3019 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3020 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3021 - swab llog records of type '0' so we get proper header size/idx (5861)
3022 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3023 - fix rename of one directory over another leaking an inode (5953)
3024 - avoid SetPageDirty on 2.6 (5981)
3025 - don't re-add just-being-destroyed locks to the waiting list (5653)
3026 - when creating new directories, inherit the parent's custom
3027 striping settings if present parent (3048)
3028 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3029 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3030 - don't hold client locks on temporary worklist from l_lru (5666)
3031 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3032 - automatically evict dead clients (5921)
3033 - Update file size properly in create+truncate+fstat case (6196)
3034 - Do not unhash mountpoint dentries, do not allow removal of
3036 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3037 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3038 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3039 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3040 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3041 - Correctly handle reads of files with no objects (6243)
3042 - lctl recover will also mark a device active if deactivate used (5933)
3044 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3045 - allow --write-conf on an MDS with different nettype than client (5619)
3046 - don't write config llogs to MDS for mounts not from that MDS (5617)
3047 - lconf should create multiple TCP connections from a client (5201)
3048 - init scripts are now turned off by default; run chkconfig --on
3049 lustre and chkconfig --on lustrefs to use them
3050 - upcalls are no longer needed for clients to recover to failover
3052 - add --abort-recovery option to lconf to abort recovery on device
3054 - add support for an arbitrary number of OSTs (3026)
3055 - Quota support protocol changes.
3056 - forward compatibility changes to wire structs (6007)
3057 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3058 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3059 - improved "lustre status" script
3060 - initialize blocksize for non-regular files (6062)
3061 - added --disable-server and --disable-client configure options (5782)
3062 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3063 - Vanilla 2.4.29 support
3064 - increase maximum number of obd devices to 520 (6242)
3065 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3066 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3067 SLES 9 2.6.5 kernels
3068 - limit stripes per file to 160 (the maximum EA size) (6093)
3070 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3073 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3074 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3075 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3076 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3077 - small ext3 extents cleanups and fixes (5733)
3078 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3079 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3080 - avoid needless client->OST connect, fix handle mismatch (5317)
3081 - fix DLM error path that led to out-of-sync client, long delays (5779)
3082 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3083 - fix several locking issues related to i_size (5492,5624,5654,5672)
3084 - don't move pending lock onto export if it is already evicted (5683)
3085 - fix kernel oops when creating .foo in unlinked directory (5548)
3086 - fix deadlock in obdfilter statistics vs. object create (5811)
3087 - use time_{before,after} to avoid timer jiffies wrap (5882)
3088 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3089 - Do not leak request if server was not able to process it (5154)
3090 - If mds_open unable to find parent dir, make that negative lookup(5154)
3091 - don't create new directories with extent-mapping (5909, 5936)
3093 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3094 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3095 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3096 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3097 - support for RHEL4 kernel on i686 (5773)
3098 - provide error messages when incompatible logs are encountered (5898)
3100 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3101 * version 1.4.0.10 (1.4.1 release candidate 1)
3103 - don't keep a lock reference when lock is not granted (4238)
3104 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3105 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3106 - drop import inflight refcount on signal_completed_replay error (5255)
3107 - unlock page after async write error during send (3677)
3108 - handle missing objects in filter_preprw_read properly (5265)
3109 - no transno return for symlink open, don't save no-trasno open (3440)
3110 - don't try to complete elan receive that already failed (4012)
3111 - free RPC server reply state on error (5406)
3112 - clean up thread from ptlrpc_start_thread() on error (5160)
3113 - readahead could read extra page into cache that wasn't ejected (5388)
3114 - prevent races in class_attach/setup/cleanup/detach (5260)
3115 - don't dereference de->d_inode after l_dput of de (5458)
3116 - use "int" for stripe value returned from lock_to_stripe (5544)
3117 - mballoc allocation and error-checking fixes in 2.6 (5504)
3118 - block device patches to fix I/O request sizes in 2.6 (5482)
3119 - look up hostnames for IB nals (5602)
3120 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3121 - don't start multiple acceptors for the same port (5277)
3122 - fix incorrect LASSERT in mds_getattr_name (5635)
3123 - export a proc file for general "ping" checking (5628)
3124 - fix "lfs check" to not block when the MDS is down (5628)
3126 - service request history (4965)
3127 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3128 - create an "evict_client" /proc entry on OSTs, like the MDS has
3129 - fix mount usage message, return errors per mount(8) (5168)
3130 - change grep [] to grep "[]" in tests so they work in more UMLs
3131 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3132 - remove ancient LOV_MAGIC_V0 EA support (5047)
3133 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3134 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3135 - allow readahead allocations to fail when low on memory (5383)
3136 - mmap locking landed again, after considerable improvement (2828)
3137 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3139 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3142 - send OST transaction number in read/write reply to free req (4966)
3143 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3144 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3145 - the watchdog thread now runs as interruptible (5246)
3146 - flock/lockf fixes (but it's still disabled, pending 5135)
3147 - don't use EXT3 constants in llite code (5094)
3148 - memory shortage at startup could cause assertion (5176)
3150 - reorganization of lov code
3151 - single portals codebase
3153 - add extents/mballoc support (5025)
3154 - direct I/O reads in the obdfilter (4048)
3155 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3157 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3160 - send OST transaction number in read/write reply to free req (4966)
3161 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3162 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3163 - the watchdog thread now runs as interruptible (5246)
3164 - handle missing objects in filter_preprw_read properly (5265)
3165 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3166 - drop import inflight refcount on signal_completed_replay error (5255)
3167 - unlock page after async write error during send (3677)
3168 - return original error code on reconstructed replies (3761)
3169 - no transno return for symlink open, don't save no-trasno open (3440)
3171 - add pid to ldlm debugging output (4922)
3172 - bump the watchdog timeouts -- we can't handle 30sec yet
3173 - extra debugging for orphan dentry/inode bug (5259)
3175 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3178 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3179 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3180 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3181 - fix NULL dereference in /proc/sys/portals/routes (4827)
3182 - allow failed mdc_close() operations to be interrupted (4561)
3183 - stop precreate on OST before MDS would time out on it (4778)
3184 - don't send partial-page writes before EOF from client (4410)
3185 - discard client grant for sub-page writes on large-page clients (4520)
3186 - don't free dentries not owned by NFS code, check generation (4806)
3187 - fix lsm leak if mds_create_objects() fails (4801)
3188 - limit debug_daemon file size, always print CERROR messages (4789)
3189 - use transno after validating reply (3892)
3190 - process timed out requests if import state changes (3754)
3191 - update mtime on OST during writes, return in glimpse (4829)
3192 - add mkfsoptions to LDAP (4679)
3193 - use ->max_readahead method instead of zapping global ra (5039)
3194 - don't interrupt __l_wait_event() during strace
3196 - add software watchdogs to catch hung threads quickly (4941)
3197 - make lustrefs init script start after nfs is mounted
3198 - fix CWARN/ERROR duplication (4930)
3199 - return async write errors to application if possible (2248)
3200 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3201 - print NAL number in %x format (4645)
3202 - update barely-supported suse-2.4.21-171 series (4842)
3203 - support for sles 9 %post scripts
3204 - support for building 2.6 kernel-source packages
3205 - support for sles km_* packages
3207 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3210 - ignore -ENOENT errors in osc_destroy (3639)
3211 - notify osc create thread that OSC is being cleaned up (4600)
3212 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3213 - reconstruct ost_handle() like mds_handle() (4657)
3214 - create a new thread to do import eviction to avoid deadlock (3969)
3215 - let lconf resolve symlinked-to devices (4629)
3216 - don't unlink "objects" from directory with default EA (4554)
3217 - hold socknal file ref over connect in case target is down (4394)
3218 - allow more than 32000 subdirectories in a single directory (3244)
3219 - fix blocks count for O_DIRECT writes (3751)
3220 - OST returns ENOSPC from object create when no space left (4539)
3221 - don't send truncate RPC if file size isn't changing (4410)
3222 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3223 - bind to privileged port in socknal and tcpnal (3689)
3225 - rate limit CERROR/CWARN console message to avoid overload (4519)
3226 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3227 - basic mmap support (3918)
3228 - kernel patch series update from b1_4 (4711)
3230 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3233 - avoid crash during MDS cleanup with OST shut down (2775)
3234 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3235 - don't use bad inodes on the MDS (3744)
3236 - dynamic object preallocation to improve recovery speed (4236)
3237 - don't hold spinlock over lock dumping or change debug flags (4401)
3238 - don't zero obd_dev when it is force cleaned (3651)
3239 - print grants to console if they go negative (4431)
3240 - "lctl deactivate" will stop automatic recovery attempts (3406)
3241 - look for existing locks in ldlm_handle_enqueue() (3764)
3242 - don't resolve lock handle twice in recovery avoiding race (4401)
3243 - revalidate should check working dir is a directory (4134)
3245 - don't always mark "slow" obdfilter messages as errors (4418)
3247 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3250 - don't close LustreDB during write_conf until it is done (3860)
3251 - fix typo in lconf for_each_profile (3821)
3252 - allow dumping logs from multiple threads at one time (3820)
3253 - don't allow multiple threads in OSC recovery (3812)
3254 - fix debug_size parameters (3864)
3255 - fix mds_postrecov to initialize import for llog ctxt (3121)
3256 - replace config semaphore with spinlock (3306)
3257 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3258 - don't allow enqueue to complete on a destroyed export (3822)
3259 - down write_lock before checking llog header bitmap (3825)
3260 - recover from lock replay timeout (3764)
3261 - up llog sem before sending rpc (3652)
3262 - reduce ns lock hold times when setting kms (3267)
3263 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3264 - fix NULL deref and obd_dev leak on setup error (3312)
3265 - replace some LBUG about llog ops with error handling (3841)
3266 - don't match INVALID dentries from d_lookup and spin (3784)
3267 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3268 - fix invalid assertion in ptlrpc_set_wait (3880)
3270 - add libwrap support for the TCP acceptor (3996)
3271 - add /proc/sys/portals/routes for non-root route listing (3994)
3272 - allow setting MDS UUID in .xml (2580)
3273 - print the stack of a process that LBUGs (4228)
3275 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3278 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3279 - make sure to unset replay flag from failed open requests (3440)
3280 - if default stripe count is 0, use OST count for inode size (3636)
3281 - update parent mtime/ctime on client for create/unlink (2611)
3282 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3283 - free recovery state on server during a forced cleanup (3571)
3284 - unregister_reply for resent reqs (3063)
3285 - loop back devices mounting and status check on 2.6 (3563)
3286 - fix resource-creation race that can provoke i_size == 0 (3513)
3287 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3288 - more debugging for page-accounting assertion (3746)
3289 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3290 - avoid deadlock after precreation failure (3758)
3291 - fix race and lock order deadlock in orphan handling (3450, 3750)
3292 - add validity checks when grabbing inodes from l_ast_data (3599)
3294 - add /proc/.../recovery_status to obdfilter (3428)
3295 - lightweight CDEBUG infrastructure, debug daemon (3668)
3296 - change default OSC RPC parameters to be better on small clusters
3297 - turn off OST read cache for files smaller than 32MB
3298 - install man pages and include them in rpms (3100)
3299 - add new init script for (un)mounting lustre filesystems (2593)
3300 - run chkconfig in %post for init scripts (3701)
3301 - drop scimac NAL (unmaintained)
3303 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3306 - clean kiobufs before and after use (3485)
3307 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3308 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3309 - change init script to fail more clearly if not run as root (1528)
3310 - allow clients to reconnect during replay (1742)
3311 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3312 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3313 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3314 - fix cond_resched (3554)
3315 - only evict unfinished clients after recovery (3515)
3316 - allow bulk resend, prevent data loss (3570)
3317 - dynamic ptlrpc request buffer allocation (2102)
3318 - don't allow unlinking open directory if it isn't empty (2904)
3319 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3320 - remove extraneous obd dereference causing LASSERT failure (3334)
3321 - don't use get_cycles() when creating temp. files on the mds (3156)
3322 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3323 - handle EEXIST for set-stripe, set proper directory name (3336)
3325 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3326 - fix ksocknal_fmb_callback() error messages (2918)
3328 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3331 - don't copy lvb into (possibly NULL) reply on error (2983)
3332 - don't deref dentry after dput, don't free lvb on error (2922)
3333 - use the kms to determine writeback rpc length (2947)
3334 - increment oti_logcookies when osc is inactive (2948)
3335 - update client's i_blocks count via lvb messages (2543)
3336 - handle intent open/close of special files properly (1557)
3337 - mount MDS with errors=remount-ro, like obdfilter (2009)
3338 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3339 - don't use cancelling-locks' kms values (2947)
3340 - use highest lock extent for kms, not last one (2925)
3341 - don't dereference ERR_PTR() dentry in error handling path (3107)
3342 - fix thread race in portals_debug_dumplog() (3122)
3343 - create lprocfs device entries at setup instead of at attach (1519)
3344 - common AST error handler, don't evict client on completion race (3145)
3345 - zero nameidata in detach_mnt in 2.6 (3118)
3346 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3347 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3348 - initialize RPC timeout timer earlier for 2.6 (3219)
3349 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3350 - print nal/nid for unknown nid (3258)
3351 - additional checks for oscc recovery before doing precreate (3284)
3352 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3353 - don't crash in mdc_close for bad permissions on open (3285)
3354 - zero i_rdev for non-device files (3147)
3355 - clear page->private before handing to FS, better assertion (3119)
3356 - tune the read pipeline (3236)
3357 - fix incorrect decref of invalidated dentry (2350)
3358 - provide read-ahead stats and refine rpc in flight stats (3328)
3359 - don't hold journal transaction open across create RPC (3313)
3360 - update atime on MDS at close time (3265)
3361 - close LDAP connection when recovering to avoid server load (3315)
3362 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3363 - don't leak open file on MDS after open resend (3325)
3364 - serialize filter_precreate and filter_destroy_precreated (3329)
3365 - loop device shouldn't call sync_dev() for nul device (3092)
3366 - clear page cache after eviction (2766)
3367 - resynchronize MDS->OST in background (2824)
3368 - refuse to mount the same filesystem twice on same mountpoint (3394)
3369 - allow llmount to create routes for mounting behind routers (3320)
3370 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3371 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3372 - fix rare problem with rename on htree directories (3417)
3374 - allow default OST striping configuration per directory (1414)
3375 - fix compilation for qswnal for 2.6 kernels (3125)
3376 - increase maximum number of MDS request buffers for large systems
3377 - change liblustreapi to be useful for external progs like lfsck (3098)
3378 - increase local configuration timeout for slow disks (3353)
3379 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3381 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3384 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3385 - don't overwrite extent policy data in reply if lock was blocked (2901)
3386 - drop filter export grants atomically with removal from device (2663)
3387 - del obd_self_export from work_list in class_disconnect_exports (2908)
3388 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3389 - swab reply message in mdc_close, other PPC fixes (2464)
3390 - fix destroying of named logs (2325)
3391 - overwrite old logs when running lconf --write_conf (2264)
3392 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3393 - fix race in target_handle_connect (2898)
3394 - mds_reint_create() should take same inode create lock (2926)
3395 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3396 - don't close files for self_export to avoid uninitialized obd (2936)
3397 - allow MDS with the same name as client node (2939)
3398 - hold dentry reference for closed log files for unlink (2325)
3399 - reserve space for all logs during transactions (2059)
3400 - don't evict page beyond end of stripe extent (2925)
3401 - don't oops on a deleted current working directory (2399)
3402 - handle hard links to targets without a parent properly (2517)
3403 - don't dereference NULL lock when racing during eviction (2867)
3404 - don't grow lock extents when lots of conflicting locks (2919)
3406 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3409 - account for cache space usage on clients to avoid data loss (974)
3410 - lfsck support in lustre kernel code (2349)
3411 - reduce journal credits needed for BRW writes (2370)
3412 - orphan handling to avoid losing space on client/server crashes
3413 - ptlrpcd can be blocked, stopping ALL progress (2477)
3414 - use lock value blocks to assist in proper KMS, faster stat (1021)
3415 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3416 - recovery for initial connections (2355)
3417 - fixes for mds_cleanup_orphans (1934)
3418 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3419 - block all file creations until orphan recovery completes (1901)
3420 - client remove rq_connection from request struct (2423)
3421 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3422 - recovery timer race (2670)
3423 - mdc_close recovey bug (2532)
3424 - ptlrpc cleanup bug (2710)
3425 - mds timeout on local locks (2588)
3426 - namespace lock held during RPCs (2431)
3427 - handle interrupted sync write properly (2503)
3428 - don't try to handle a message that hasn't been replied to (2699)
3429 - client assert failure during cleanup after abort recovery (2701)
3430 - leak mdc device after failed mount (2712)
3431 - ptlrpc_check_set allows timedout requests to complete (2714)
3432 - wait for inflight reqs when ptlrpcd finishes (2710)
3433 - make sure unregistered services are removed from the srv_list
3434 - reset bulk XID's when resending them (caught by 1138 test)
3435 - unregister_bulk after timeout
3436 - fix lconf error (2694)
3437 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3438 - readahead locks pages, leaves pending causing memory pressure (2673)
3439 - increase OST request buffers to 4096 on large machines (2729)
3440 - fix up permission of existing directories in simple_mkdir (2661)
3441 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3442 - don't assign transno to errored transactions (2742)
3443 - don't delete objects on OST if given a bogus objid from MDS (2751)
3444 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3445 - if rq_no_resend, then timeout request after recovery (2432)
3446 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3447 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3448 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3449 - MDS recovery shouldn't delete objects that it has given out (2730)
3450 - if enqueue arrives after completion, don't clobber LVB (2819)
3451 - don't unlock pages twice when trigger_group_io returns error (2814)
3452 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3453 - don't write pages to disk if there was an error (1450)
3454 - don't ping imports that have recovery disabled (2676)
3455 - take buffered bytes into account when balancing socknal conn (2817)
3456 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3457 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3458 - remove little-endian swabbing of llog records (1987)
3459 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3460 - retry reposting request buffers if they fail (1191)
3461 - grow extent at grant time to avoid granting a revoked lock (2809)
3462 - lock revoke doesn't evict page if covered by a second lock (2765)
3463 - disable VM readahead to avoid reading outside lock extents (2805)
3465 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3466 - updated kernel patches for hp-2.4.20 kernel (2681)
3468 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3471 - fix truncated write corruption (2366)
3472 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3474 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3475 - ensure len(uuid) < 37 in lmc (1171)
3476 - fix ia64 OOPS in llog_test (2255)
3477 - zero end of page at obdfilter for partial page writes (2648)
3478 - don't leave stale dentries around after renames (bug 2428)
3479 - fix timeouts when evicting a client with a single lock held (2642)
3480 - set deadline for the initial HELLO message to drain (2634)
3481 - print out dotted-quad IP addresses in the socknal (2302)
3483 - additional debugging for MDS client eviction problem (2443)
3484 - fix mkfsoptions support for osts (2603, 2604)
3486 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3489 - add series for the vanilla 2.6.0 kernel
3490 - add series for the vanilla 2.4.24 kernel
3491 - add series for a cray x86/64 UL kernel drop
3492 - fix xattr patches for the vanilla 2.4.19 series
3494 - generate true UUIDs in lmc (1171)
3495 - have portals stack dumping break in UML (2466)
3496 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3497 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3498 - wake callback waiting threads on client eviction (2460)
3499 - Add --ptldebug and --subsystem to lmc (1719)
3500 - update assertion to allow safe interrupt allocation
3501 - set rq_no_resend for cancel requests (2432)
3502 - recalculate ptlrpcd timeout after resend (2494)
3503 - call vfs_rmdir when removing pending directories (2368)
3504 - fix renaming a file to itself (2429)
3505 - lmc creates a default one-stripe lov (2454)
3506 - expand procfs space to handle large clusters (2326)
3507 - increase UML stack to avoid overflow
3508 - update lconf's list of debug and subsystem masks
3509 - fix lfs find --obd (2510)
3510 - /proc tunable for disabling filter read caching (2591)
3511 - stop rpm packages from altering slapd.conf (2301)
3512 - disable nagle in the socknal under 0conf (2578)
3513 - choose mds inode size based on stripe count (2572)
3514 - fix kernel-source rpm problems (2516)
3516 - add --disable-doc to avoid pdf generation (2421)
3517 - update documentation, tests, type-os, comments
3518 - avoid format warnings on ia64
3519 - remove the TOE NAL
3520 - tiny code cleanups by removing unused fields
3522 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3525 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3526 - remove the most blatant lies from BUILDING file (2371)
3527 - change default debug level to reasonable production setting
3528 - reduce client side cache size to reduce cache flush time
3529 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3530 - make TCP zerocopy and pinger support enabled by default (2476)
3531 - sync writes completed after process exits caused crashes (2319)
3532 - maintain correct mount count on the MDS (2356)
3533 - backout 1557, because 2316 wasn't really fixed
3534 - better file I/O statistics gathering in /proc
3535 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3536 - another kernel patch to fix zero-copy TCP function export
3537 - don't take duplicate lock when processing re-sent getattr (2420)
3538 - lctl uses obd_self_export instead of creating new conn (2353)
3539 - MDS/OST recovery case which requires object creation asserted (2425)
3540 - move lfs from /usr/sbin to /usr/bin in packages
3541 - fix race between mds_client_add and mds_client_free (2417)
3542 - use kmalloc instead of slabs in portals (2430)
3543 - don't create duplicate records when a failover MDS is present (2442)
3544 - remove unnecessary mount age check (2332)
3545 - don't remove directory inodes from locks prematurely (2451)
3546 - don't break if MDS service name is the same as hostname (2103)
3547 - fix races in client write RPC generation when cache full (2482)
3549 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3552 - remove now-unused request->rq_obd (278)
3553 - if an allocation fails, print out how much memory we've used (1933)
3554 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3555 - add the "configurable stack size" patch to most series files (1256)
3556 - ability to write large log records, for 100+ OST configs (2306)
3557 - fix NULL deref when filter_prep fails (2314)
3558 - fix operator precedence error in filter_sync
3559 - dynamic allocation of socknal TX descriptors (2315)
3560 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3561 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3562 - ensure that CWARN messages in Portals always get to the syslog
3563 - __init/__exit are not for prototype decls (ldlm_init/exit)
3564 - x86-64 compile warning fixes
3565 - fix gateway LMC keyword conflict (2318)
3566 - fix MDS lock inversions in getattr/reint paths (1844)
3567 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3568 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3569 - DLM race condition prevented some lock evictions (2328)
3570 - ENOMEM detection and retry on socknal sends (2230)
3571 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3572 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3573 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3574 - better error messages when a client is rejected during recovery (1505)
3575 - avoid cancelling locks which were never granted, after failure (2330)
3576 - fix i_sem/journal inversion in mds_client_add (2333)
3577 - fix truncate/getattr lock cycle deadlock (2334)
3578 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3579 - fix two rare exit paths which could leak an l_lock() ref (2321)
3580 - fencepost error in MDS/OST orphan recovery (2226)
3581 - make log record alignment 8 bytes (1988)
3582 - lstripe now fails when requested offset > ost_count (2237)
3583 - ensure that all kernel series have a complete list.h (1607)
3584 - fix crashes in special-file operations (2316)
3585 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3587 - allow configurable automake binary, for testing new versions
3588 - small update to the lfs documentation
3590 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3592 * fix negative export reference count in fsfilt_sync (2312)
3594 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3595 * release candidate 0.9.1
3597 - orphans are moved into the PENDING directory for possible recovery
3598 - replayed opens now open by fid for orphan/rename safety (1042)
3599 - last close of an orphan inode generates a transno (683)
3600 - chdir() and mount() now pin the directory entry (1020)
3601 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3602 - discard very old requests without processing them (1502)
3603 - remove some common, well-understood CERRORs (1505)
3604 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3605 - clear "grant" flags in OST replies until OST grant code lands (1644)
3606 - fix read performance by not clobbering i_blksize on client (1598)
3607 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3608 - make LRU size a /proc tunable, clears locks when reduced (707)
3609 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3610 - server locks take references on exports now (1558)
3611 - build fixes for 2.4.20-rh trees (1663)
3612 - return an error from lov_create if all OSCs are inactive (1751)
3613 - fix import levels when a reconnect happens without a timeout (1597)
3614 - exit early from mds_open if we get a lookup error (1749)
3615 - partial page read at EOF wouldn't wait for disk before sending (1642)
3616 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3617 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3618 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3619 - write/truncate lock inversion (1639)
3620 - Don't auto-load obdclass, portals modules during cleanup (1495)
3621 - fix timestamps from jumping to "now" (1763)
3622 - extra journal assertions (1648)
3623 - add an extra multiunlink test (1771)
3624 - fix read_record/write_record API (1776)
3625 - fix leak of offset_extent, possible incorrect i_size later (1772)
3626 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3627 - Debugging for the kqswnal_get_idle_tx problems (1820)
3628 - Allow recovery to be attempted multiple times (1536)
3629 - Write out MDS last_rcvd file after it is first created (1600)
3630 - Fix tx_descriptor leak in failed transmit situations (1827)
3631 - ext3 journaling fixes for assertion failure after IO error (1871)
3632 - class_export_put() on freed export after completion AST error (1896)
3633 - Fix revalidate looping in VFS (1322)
3634 - Don't access a freed export during MDS_REINT timeout (1521)
3635 - Add open-unlink recovery support on the MDS (1673,1764)
3636 - Return an error if no MDS data was read from last_rcvd (1946)
3637 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3638 - Don't setup a disk device that doesn't match exported UUID (317)
3639 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3640 - avoid committing NULL handle in force close
3641 - local.sh is now a one-stripe LOV configuration
3642 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3643 - fix ext3 htree duplicate directory entry corruption (1516)
3644 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3645 - update to new LOV EA format (2097)
3646 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3648 2003-06-15 Phil Schwan <phil@clusterfs.com>
3651 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3652 - per-import/export recovery handling (958, 931, 959)
3653 - multiple last-rcvd slots, for serving multiple FSes (949)
3654 - connections are again shared between multiple imp/exports (963, 964)
3655 - "umount -f" would hang if any requests needed to be sent (393, 978)
3656 - avoid pinning large req buffer by copying for queued messages (989)
3657 - add "uuid" to "lctl device" command to help upcalls (991)
3658 - "open" RPCs with transnos would confuse recovery counters (1037)
3659 - do proper endian conversion of all wire messages (288, 340, 891)
3660 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3661 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3662 - fix ptlrpc_connection leak in target_handle_connect (1174)
3663 - fix import refcounting bug in OST and MDS cleanup (1134)
3664 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3665 - fix very unlikely obd_types race condition (501)
3666 - remove osc_open hack for echo_client (1187)
3667 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3668 - a failure in read_inode2 leads to deadlock (1139)
3669 - cancel ack-locks as soon as transaction is committed (1072)
3670 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3671 - make sure to commitrw after any preprw to avoid deadlock (1162)
3672 - failing to execute a file in a lustre FS would lock inode (1203)
3673 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3674 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3675 - obd_uuid2tgt would walk off the end of the list (1255)
3676 - on IA64 the osc would give portals incorrect bulk size (1258)
3677 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3678 - fix lock inversion caused by new llite matching code (1282)
3679 - limit the number of dirty pages on a client to 10MB (1286)
3680 - timed out locks were not being corrected cancelled (1289)
3681 - fix O_DIRECT above 4GB on IA-32 (1292)
3682 * major user-visible changes
3683 - fail out/fail over policy now controlled by the upcall (993)
3685 - add OBD_PING to check server availability and failure (954)
3686 - lustre messages are now sent in sending host order (288, 340, 891)
3687 - add eadatalen to MDS getattr reply (340)
3688 - OST read replies may contain second buffer, with per-page status (593)
3690 2003-03-11 Phil Schwan <phil@clusterfs.com>
3693 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3694 - failed open()s could cause deadlock; fixed (867, 869)
3695 - stop cancelling OST locks when files are closed (481)
3696 - overlapping XID spaces caused network corruption (851, 853)
3697 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3698 - setattr_raw added, to do single-RPC, server-side setattrs
3699 - lmc/lconf syntax change for OST UUIDs
3700 - fix crashy race condition between ptlrpc_free_req and osc_close
3701 - don't use request in mdc_enqueue if we hit a timeout (889)
3702 - don't set the inode i_size for regular files from the MDS (896)
3703 - handle out of order completion AST (842)
3704 - don't LBUG if a lock request times out after receiving AST (913)
3705 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3706 - if a bad lock AST arrives, send an error instead of dropping entirely
3707 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3708 - fix leak in bulk IO when only partially completed (899, 900, 926)
3709 - fix O_DIRECT for ia64 (55)
3710 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3711 - C-z after timeout could hang a process forever; fixed (977)
3713 - client-side I/O cache (678, 924, 929, 941, 970)
3715 - READPAGE and SETATTRs which don't take server-side locks get
3718 2003-02-11 Phil Schwan <phil@clusterfs.com>
3721 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3722 - Fix fsfilt_extN_readpage() to read a full page of directory
3723 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3724 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3725 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3726 - Most intent operations were removed, in favour of a new RPC mode
3727 that does a single RPC to the server and bypasses most of the VFS
3728 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3729 - Aggressively cancel local locks on DLM servers
3730 - mds_reint_unlink sends EA to the client if it's the last nlink.
3731 client uses that EA to unlink OST objects.
3732 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3733 - recursive symlinks were fixed (439)
3734 - fixed NULL deref in DEBUG_REQ
3735 - filter_update_lastobjid no longer calls sync, which annoyed extN
3736 - fixed multi-client small-writes to a single file problem (445)
3737 - fixed mtime updates during file writes (607)
3738 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3739 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3740 - fixed Linux OST size reporting problem (444, 656)
3741 - OST now updates object mtime with writes or setattr (607, 619)
3742 - client verifies file size before zeroing page past EOF (445)
3743 - OST now writes last allocated objid to disk with allocation (108)
3744 - LOV on echo now works (409)
3746 - mds_reint_unlink sends a new buffer, with the EA included. this
3747 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3748 set if a regular file was being unlinked, and it was the last link
3749 - use PtlGet from the target for bulk writes (315)
3750 - OST now updates object mtime with writes or setattr (607, 619)
3751 - LDLM now has a grant-time callback to revalidate locked items, if
3753 - Many MDS operations were reorganized to combat race conditions
3755 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3756 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3758 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3761 - Fully reactivate OST imports after reconnection (512, others)
3762 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3763 - More graceful error handling for truncating on dead OST (515)
3764 - Don't error out unless we're actually accessing dead stripes (474)
3765 - Fix garbage sizes when stripes are missing (410)
3766 - LRU counters were broken, causing constant lock purge (433, 432)
3767 - garbage on read from stripes with failed OSTs (441)
3768 - mark OSCs as active before reconnecting during recovery (438)
3769 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3770 - lfind did not preserve OST order in output (443)
3771 - symlinks cause hung clients, incorrect data (439)
3772 - stop dereferencing request after dropping refcount (457)
3773 - don't LASSERT(spin_is_locked) on non-SMP (455)
3774 - fixes for many rename() bugs
3775 - fstat didn't correctly synchronize attributes (399)
3776 - server must handle lock cancellation during blocking AST prep (487)
3777 - bulk descriptors were free()d too soon (511)
3778 - fix paths in lconf, which would load incorrect modules (451, 507)
3779 - fix confusing lconf 'host not found' error message (386)
3780 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3781 - fix race condition in mdc_blocking_ast() for inode access (526)
3782 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3783 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3784 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3785 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3786 - fix lstripe in several ways (kernel side) (527)
3787 - fix request leak in ldlm_cli_enqueue (262)
3788 - incorrect OSC was marked inactive after OST failure
3789 - call mds_fs_cleanup before unmounting filesystem (524)
3790 - fix races between taking ns_lock and ldlm_lock_change_resource
3791 - fix races updating LOV export open file list
3792 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3793 - fix recovery NULL deref in ldlm_cli_cancel_unused
3794 - fix some DLM races by using new hash table for lock handles (419)
3795 - permit the client to specify desired inodes, at replay
3796 - duplicate requests when we queue them for replay reintegration
3797 - fix last_rcvd offset calculation
3798 - sync after each recovered transaction, so we always make progress
3799 - never, not always, ERESTART requests without transnos
3800 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3802 - skip replay if the MDS says that the client is already connected
3803 - don't check for a recovery-enabled export to match lctl's UUID
3804 - don't INC_USE_COUNT for phantom exports
3805 - don't crash when cleaning up phantom exports (567)
3806 - don't double-finish or set replay data for errored mdc_open requests
3807 - abort requests when they time out, so we don't get old replies
3808 - send/receive replies for AST messages again
3809 - if the client says that it doesn't have the lock, cancel it on the
3811 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3812 mark it as destroyed, it will all work out when decref is called
3813 - fix module use counts (22, 581)
3815 - ASTs now expect a reply (server cancels lock on error reply)
3817 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3820 - fix many simultaneous client startup (392)
3821 - fix dentry->d_it clobbering
3822 - credentials weren't being shipped for readdir/getattr operations
3823 - remove invalid assertions triggered during some concurrent MD
3825 - proper Lustre versions added (336, 389)
3826 - fix memory leak for create error case (398)
3827 - fix LOV locking bug that would get cli/srv out of sync
3828 - fix echo client over LOV (409)
3829 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3830 - fix double-O_EXCL intent crash (424)
3831 - avoid sending multiple lock CANCELs (352)
3833 - MDS can do multi-client recovery (modulo bugs in new code)
3835 - many updates, edits, cleanups
3837 2002-11-18 Phil Schwan <phil@clusterfs.com>
3840 - fix null d_it dereference (346)
3841 - fix full OST/dbench hang (333)
3842 - fix permission problem with file removal (286)
3843 - fix removal of OSCs from LOV when they fail
3844 - fix NULL deref during bulk timeout (214)
3845 - fix problems related to multiple filesystems on one MDS (241)
3846 - fixed serious subtle metadata locking bugs
3847 - free locks on clients when inodes are removed due to memory
3849 - fix inode pointer in lock data (285)
3850 - partial support for multiple MDS on a single host (241)
3851 - data locks weren't cancelled at clear_inode time (290, 311)
3852 - intent locks could lead to unbounded lock growth (205)
3853 - added a maximum lock count, an LRU list, and a flusher
3854 - fix multiple rename (365)
3855 - properly abstracted the echo client
3856 - OSC locked 1 byte too many; fixed
3857 - rewrote brw callback code:
3858 - fixed recovery bugs related to LOVs (306)
3859 - fixed too-many-pages-in-one-write crash (191)
3860 - fixed (again) crash in sync_io_timeout (214)
3861 - probably fixed callback-related race (385)
3863 - Add capability to MDS protocol
3864 - LDLM cancellations and callbacks on different portals
3866 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
3869 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
3870 - defer open object destruction to close time (601981, 138)
3871 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
3872 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
3873 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
3874 - don't call obd_setattr in ll_file_release for destroyed objects
3875 * protocol change to lustre_msg: move |version| and add |flags|
3876 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
3877 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
3878 * added replay of create, unlink, link and rename operations during
3879 MDS failover; recovery should be much more robust now
3880 * remove failed OSCs from LOVs (only lov_create uses this so far)
3881 * the lustre-HOWTO was brought (more) up to date (582544)
3883 2002-10-23 Phil Schwan <phil@clusterfs.com>
3886 - in-use dentries weren't being reused properly (617851)
3887 - prevent multiple LDLM setup (599178)
3888 - fix LOV size calculations for truncate (617853)
3889 - fix client handling of MDS intent errors (POSIX)
3890 - fix permission bug in lovstripe.c test (624321)
3891 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
3892 - truncate past end of file could corrupt data
3893 - proper cleanup after timeouts, crashes, etc (592524, 550815)
3894 - a race in recovery could return ETIMEDOUT to apps (623947)
3895 - building outside the source directory was fixed
3896 * the lustre-HOWTO was brought (more) up to date (582544)
3897 * major progress was made on recovery functionality
3899 2002-10-10 Phil Schwan <phil@clusterfs.com>
3902 - recovery deadlock fix
3903 - rm -rf causes LBUG fix (617817)
3904 - file open by multiple tasks fix (618962)
3905 - directory permissions bugs (602707 and 620007)
3906 - journal_stop fixed with locking (611313)
3907 - O_APPEND failures resolved (618273, perhaps 614459)
3908 - lconf PATH fix (619770)
3909 - IA64 build fix (621450)
3910 - RPC buffer sizes scale with amount of memory
3912 2002-10-01 Phil Schwan <phil@clusterfs.com>
3915 - locks would be cancelled without throwing away data pages,
3916 resulting in inconsistent data (605627)
3917 - inode attributes were not always being refreshed (605627, 612449)
3918 - lconf now continues to cleanup after lctl reports an error
3919 - MDS now enforces user permissions (602707)
3920 - lprocfs cleanup fixed, but not yet enabled (614157)
3921 - fixed infinite server hang, should a client not respond to an AST
3922 - avoid going into recovery if user calls readlink() with a buffer
3923 that's too small (613941)
3924 - AST RPCs no longer require replies (614867) -- this may be changed
3925 - don't crash server if client sends an IOV that's too big (611336)
3926 - fixed lock conversion deadlock (611892)
3927 - fixed the following of symlinks (614622)
3928 * recovery: the server can remove locks from a client that dies, other
3929 clients can make progress
3930 * more extN patch fixes
3931 * compile-time configurable ptlrpc buffer allocations
3933 - collaborative read cache document
3934 - Lustre Lite Performance CDR document-in-progress
3936 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3939 - fix typo in patch-2.4.18
3941 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3944 - clear ptlrpc request each time in handle_incoming_request()
3945 - unlink of files now destroys the object on the OST
3947 2002-09-19 Peter Braam <braam@clusterfs.com>
3949 * add hard link support
3950 * change obdfile creation method
3951 * kernel patch changed
3953 2002-09-19 Peter Braam <braam@clusterfs.com>
3956 - stack overflow bug in extN fixed
3958 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3960 * documentation updates
3961 - add man pages for config tools
3962 - update tests/README to describe testing with new config tools
3963 - finish metadata API descriptions
3964 * bug fixes and cleanups
3965 - statfs workaround for 16TB limit
3966 - LOV stripe allocation improved, can stripe on subset of OSTs
3967 - LOV file size/IO offset was wrong for files > 4GB in size
3968 - object EA data was being dropped, caused files to be unreadable
3969 - memory overflow with non-LOV OST caused memory corruption
3970 - fixed regression tests to work with new config tools, obdfilter
3971 - fixed bug when directory size became larger than 1 block
3972 - fixed bug (for single client case) when PWD was deleted
3973 - invalidate local directory pages when doing intent-based ops
3974 - avoid LDLM oops when lock callback contained bad data
3976 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3978 * documentation updates
3979 * bug fixes and cleanups
3980 - configuration tools
3983 - 64-bit compile warnings
3984 - 64-bit internal statfs data
3986 * test_brw on persistent OST devices
3988 * lprocfs (disabled)
3990 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3992 * documentation updates
3993 * bug fixes and cleanups
3994 * configuration tools
3996 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3999 * many small fixes to 0_5_4
4000 * io/network handling
4001 * thinkos in MDS operations
4003 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4006 * crucial basic fixes to 0.5.3
4007 * IOR, Iozone work over Elan
4010 2002-08-07 Phil Schwan <phil@clusterfs.com>
4011 * version 0_5_3, our first alpha
4012 * we use the new Portals iovs
4013 * documentation updates
4014 * bug fixes and cleanups
4015 * small changes in the DLM wire protocol
4017 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4018 * version 0_5_1 with some initial stability,
4019 * locking on MD and file I/O.
4020 * documentation updates
4021 * several bug fixes since 0.5.0
4022 * small changes in wire protocol
4024 2002-07-18 Phil Schwan <phil@clusterfs.com>
4026 * delivered as Lustre Light Alpha
4027 * fixed a crash after handling invalid MDS requests
4028 * fixed directory pages for architectures with non-4k pages sizes
4030 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4031 * release version v0_4_4
4032 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4034 * Fixes a number of interruption problems with OST operations.
4035 * Update documentation for portals header changes
4036 * Move all wire protocol structs/defines to lustre_idl.h
4037 * Fixes symlink length bug.
4038 * Add tcpdump to repository.
4040 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4041 * release version v0_4_3
4042 * Fixes statfs for inodes on extN.
4043 * Fixes bug in runtests which would delete /etc/hosts.
4044 * Use 64-bit object IDs wherever possible (not into VFS though)
4045 Remove ost_get_info, which is unused by lustre, and out of date.
4047 2002-07-03 Peter Braam <braam@clusterfs.com>
4048 * release version v0_4_2 Fixes a lookup error (type not passed)
4049 * move forward to head of Portals
4050 * move forward to latest Lustre kernel
4052 2002-06-25 Peter Braam <braam@clusterfs.com>
4053 * release version v0_4_1. Hopefully stable on single node use.