3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-279.14.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-279.14.1.el6 (RHEL6)
9 2.6.18-238.19.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
11 * Recommended e2fsprogs version: 1.41.90.wc4
12 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
13 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
14 stack overflow. For more information, please refer to bugzilla 17630.
15 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
16 the re-exporting nodes to be mounted with "32bitapi" mount option
18 --------------------------------------------------------------------------------
20 9-30-2012 Whamcloud, Inc.
22 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
23 * Server support for kernels:
24 2.6.32-279.5.1.el6 (RHEL6)
25 * Client support for unpatched kernels:
26 2.6.32-279.5.1.el6 (RHEL6)
28 2.6.18-238.19.1.el5 (RHEL5)
29 2.6.16.60-0.69.1 (SLES10)
30 * Recommended e2fsprogs version: 1.41.90.wc4
31 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
32 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
33 stack overflow. For more information, please refer to bugzilla 17630.
34 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
35 the re-exporting nodes to be mounted with "32bitapi" mount option
37 Severity : enhancement
39 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
41 Severity : enhancement
43 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
45 Severity : enhancement
47 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
49 Severity : enhancement
51 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
53 Severity : enhancement
55 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
57 --------------------------------------------------------------------------------
59 04-30-2012 Whamcloud, Inc.
61 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
62 * Server support for kernels:
63 2.6.32-220.4.2.el6 (RHEL6)
64 * Client support for unpatched kernels:
65 2.6.32-220.4.2.el6 (RHEL6)
67 2.6.18-238.19.1.el5 (RHEL5)
68 2.6.16.60-0.69.1 (SLES10)
69 * Recommended e2fsprogs version: 1.41.90.wc4
70 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
71 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
72 stack overflow. For more information, please refer to bugzilla 17630.
73 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
74 the re-exporting nodes to be mounted with "32bitapi" mount option
76 Severity : enhancement
78 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
80 Severity : enhancement
82 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
84 Severity : enhancement
86 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
88 Severity : enhancement
90 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
92 --------------------------------------------------------------------------------
94 09-30-2011 Whamcloud, Inc.
96 * Server support for kernels:
97 2.6.18-238.19.1.el5 (RHEL5)
98 2.6.32-131.6.1.el6 (RHEL6)
99 * Client support for unpatched kernels:
100 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
101 2.6.18-238.19.1.el5 (RHEL5)
102 2.6.32-131.6.1.el6 (RHEL6)
103 2.6.29.4-167.fc11 (FC11)
104 2.6.16.60-0.69.1 (SLES10)
106 * Recommended e2fsprogs version: 1.41.90.wc3
107 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
108 Lustre filesystem with 4K stack may cause a stack overflow. For more
109 information, please refer to bugzilla 17630.
110 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
111 mounted with "32bitapi" mount option
113 Severity : enhancement
114 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
116 Severity : enhancement
117 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
119 Severity : enhancement
120 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
123 Frequency : only with SLES10
125 Description: Use OFED "KMP" provided by Novell
126 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
127 Lustre is now built against this package. That means you need to
128 install the ofed-kmp package from Novell for the patchless client.
129 Note that the ofed-kmp that Novell ships may not exactly match the
130 kernel version but should still be compatible.
132 Severity : enhancement
134 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
135 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
137 --------------------------------------------------------------------------------
139 2010-07-15 Oracle, Inc.
141 * Server support for kernels:
142 2.6.18-164.11.1.el5 (RHEL 5)
143 2.6.18-164.11.1.0.1.el5 (OEL 5)
144 * Client support for unpatched kernels:
145 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
146 2.6.18-164.11.1.el5 (RHEL 5),
147 2.6.18-164.11.1.0.1.el5 (OEL 5)
148 2.6.16.60-0.42.8 (SLES 10),
150 2.6.29.4-167.fc11 (FC11)
151 * Recommended e2fsprogs version: 1.41.10-sun2
153 Severity : enhancement
155 Description: Distribute lfs_migrate script with lustre
156 Details : The lfs_migrate mentioned in the manual is now included in the
157 lustre tree in lustre/scripts and will distributed in lustre RPMs.
159 Severity : enhancement
161 Description: Failover nid list added to import proc file.
163 Severity : enhancement
165 Description: "som_preview" mount option required to use SOM
166 Details : The clients will need to use som_preview mount option
170 Description: Lustre does not do 1MB IOs to HW RAID
171 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
173 Severity : enhancement
175 Description: lfs quota output is not very convenient for awk/sed-parsing
176 Details : Some positions in lfs quota output table could be empty or
177 non-empty which made it hard to parse it with scripts, now
178 a dash is put instead of space where there is not supposed
183 Description: Update to OFED 1.5.1
184 Details : OFED 1.5.1 will be used in environments that don't use
185 vendor supplied OFED.
189 Description: Remove lnet. prefix from lctl params display
190 Details : Also renamed lnet.memused to lnet_memused to
195 Description: Use conf_param -d to remove permanant settings.
199 Description: Add list_param -R
200 Details : lctl list_param now supports recursive mode.
203 Bugzilla : 21829, 22312
204 Description: Fix llobdstat.
205 Details : 1. Now updating the write_bytes counter
206 2. Port script fixes from 1.8.
210 Description: decrease the usage of memory on clients.
211 Details : 1. On clients, recycle dentries and inodes unused.
212 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
217 Description: llobdstat fix and enhancement
218 Details : add a counter to set a limit to how many samples will be returned
219 fix a wildcard in the path to limit to obdfilter stats only
223 Description: lfs check only allowed for root
224 Details : Code cleanups and added a regresion test of non-root use of
229 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
230 Details : Fix hr->hr_index overload.
234 Description: Fix FIEMAP functionality.
238 Description: FC11 patchless client support added
242 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
243 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
247 Description: New acc-sm script and various test-framework enhancments.
251 Description: Previously landed patch was reverted.
255 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
259 Description: Add ldiskfs maxdirsize mount option.
263 Description: Fix mkfs.lustre support for 16TB LUNs.
267 Description: Print mmp_check_interval
268 Details : print mmp_check_interval and make it possible to abort
269 mount operation in case it takes too long.
273 Description: Add cascading_rw to lustre mpi tests.
277 Description: Store SOM data in EA.
278 Details : Also several SOM related cleanups.
282 Description: Add refcount to md_open_data
286 Description: Use refcount on lov_request_set to prevent memory corruption.
290 Description: Statahead: use parent dir's i_mutex for synchronization.
294 Description: loadgen improvements.
298 Description: Enable peer health detection in ksocklnd and ko2iblnd.
302 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
306 Description: Fix race on obd_devcie:md_stats
310 Description: Fix SLES 11 client builds.
311 Details : Resolved uninitalized pointer gcc warnings.
315 Description: Allow non-root access for "lfs check".
316 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
320 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
322 Severity : enhancement
324 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
325 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
329 Description: Sanity test_27p fails.
330 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
331 Sanity tests 27 were cleaned up and improved.
335 Description: Unify req format on client/servers
336 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
337 buffers and swabbers approach..
342 Description: drop join file stuff.
344 Severity : enhancement
346 Description: Add 2.6.27 and SLES11 (patchless client) support
350 Description: Fix in ptlrpc_expire_one_request() to print the signed time
353 Severity : enhancement
355 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
358 Severity : enhancement
360 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
362 Severity : enhancement
364 Description: correctly handle big reply message.
365 Details : send LNet event if reply is bigger then buffer and adjust this buffer
370 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
372 Severity : enhancement
374 Description: Add support for OEL5.
378 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
380 Severity : enhancement
382 Description: "weak-modules" support
383 Details : Implement "weak-modules" support which enables kernel modules
384 to be used with any kernel that implements the same kABI. In
385 order to achieve this modules are now installed in
386 /lib/modules/$(uname -r)/updates/kernel on all distributions.
390 Description: client could not reconnect to OST because of an active request.
391 Details : The client now retries to reconnect to the same server, if a
392 connect request failed with EBUSY or -EAGAIN. Test case 35b
393 conf-sanity.sh has been created.
397 Description: Don't allow make backward step on assignin osc next id.
398 Details : race between allocation next id and ll_sync thread can be cause
399 of set wrong osc next id and can be kill valid ost objects.
403 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
407 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
411 Description: Error handling in osc_statfs_interpret() has been improved.
412 Details : Check in osc_statfs_interpret() for EBADR.
416 Description: Do not send statfs() requests to OSTs disabled by administrator.
417 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
419 Severity : enhancement
421 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
423 Severity : enhancement
425 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
427 Severity : enhancement
429 Description: Enhancement for lfs(1) command to use numeric uid/gid.
433 Description: Improved cleanup procedure for conf-sanity test_46a.
437 Description: Do not start bulk transfer, if client tried to reconnect.
438 Details : Do not start bulk transfer, if client tried to reconnect,
439 but failed due to active rpcs. Check for exp_abort_active_req
440 has been implemented.
444 Description: Prevent inconsistences between linux and lustre mount structures.
445 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
446 Make the sleep interruptible.
448 Severity : enhancement
450 Description: provide server to client comms path
451 Details : server to client communications path via new LDLM_SET_INFO rpc
455 Description: truncate starts GFP_FS allocation under transaction causing deadlock
456 Details : ldiskfs_truncate calls grab_cache_page which may start page
457 allocation under an open transaction. This may lead to
458 calling prune_icache with consequent lustre reentrance.
462 Description: send correctly lsm on open replay
463 Details : MDS is trust to LSM size on replay open, but client can set wrong size
468 Description: enable client interface failover
469 Details : When a child reconnects from another NID, properly update export
470 nid hash position and ldlm reverse import.
472 Severity : enhancement
474 Description: Communicate OST degraded/readonly state via statfs to MDS
475 Details : Flags in the statfs returned from OSTs indicate whether the
476 OST is in a degraded RAID state, or if the filesystem has
477 turned read-only after a filesystem error is detected.
482 Description: don't panic if EPROTO was hit when reading symlink
483 Details : correctly handling request reference in error cases.
485 Severity : enhancement
487 Description: Add LustreNetLink, a kernel-userspace communcation path.
489 Severity : enhancement
491 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
496 Description: access to llog context before init.
497 Details : move handling CATALOGS file at osc layer and forbid access to llog
501 Frequency : rare, only when client is writing to OST on local node
503 Description: Avoid deadlock for local client writes
504 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
505 memory freeing context. This allows OST threads to set the
506 PF_MEMALLOC flag on task structures in order to allocate memory
507 from reserved pools and complete IO.
508 Use GFP_HIGHUSER for OST allocations for non-local client writes,
509 so that the OST threads generate memory pressure and allow
510 inactive pages to be reclaimed.
512 Severity : enhancement
514 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
519 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
520 Details : move ldlm namespace creation in setup phase to avoid grab
521 _lprocfs_lock with cli_sem held.
525 Description: Temporarily disable grant shrink.
526 Details : Disable the feature for debugging.
529 Frequency : only when formatting test filesystems
531 Description: Unable to run several mkfs.lustre on loop devices at the same time
532 Details : mkfs.lustre returns error 256 on the concurrent loop devices
533 formatting. The solution is to proper handle the error.
535 Severity : enhancement
537 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
539 Severity : enhancement
541 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
546 Description: resolve race between obd_disconnect and class_disconnect_exports
547 Details : if obd_disconnect will be called to already disconnected export he
548 forget release one reference and osc module can't unloaded.
550 Severity : enhancement
552 Description: Allow tuning service thread via /proc
553 Details : For each service a new
554 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
555 created that can be used to set min/max thread counts, and get the
556 current number of running threads.
561 Descriptoin: don't return error if have partially created objects for file.
562 Details : lov_update_create_set uses set->set_success as index for created
563 objects, so if some requests will be failed, they will have hole
564 at end of array and qos_shrink_lsm for allocate correct lsm.
566 Severity : enhancement
568 Description: Update OFED support to 1.4.1
573 Description: Reduce small size read RPC
574 Details : Set read-ahead limit for every file and only do read-ahead when
575 available read-ahead pages are bigger than 1M to avoid small size
578 Severity : enhancement
580 Description: move AT tunable parameters for more consistent usage
581 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
583 Severity : enhancement
585 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
586 Details : allow skip disconnected ost for send statfs request and hide error
592 Description: fix racy locking of mballoc block bitmaps causing BUG
593 Details : The locking of the mballoc buddy bitmap and the in-memory
594 block bitmap was using two different spin locks in some
595 cases. This made it possible to incorrectly access the
596 mballoc bitmap while another process was modifying it,
597 causing a sanity assertion to fail. While no on-disk corruption
598 was reported, there was some risk of this happening.
601 Frequency : rare, on llog test 6
603 Descriptoin: don't allow connect to already connected import
604 Details : allowing connect to already connected import is hide connecting problem.
609 Descriptoin: allow kill process which wait statahead result
610 Details : for some reasons 'ls' can stick in waiting result from statahead,
611 in this case need way for kill this process.
613 Severity : enhancement
615 Description: Add state history info file, enhance import info file
616 Details : Track import connection state changes in a new osc/mdc proc file;
617 add overview-type data to the osc/mdc import proc file.
619 Severity : enhancement
621 Description: MDS create should not wait for statfs RPC while holding DLM lock.
623 Severity : enhancement
625 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
630 Descriptoin: OST grant too much space to client even there are not enough space.
631 Details : Client will shrink its grant cache to OST if there are no write
632 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
633 this grant cache if there are already not enough avaible space
634 (left_space < total_clients * 32M).
637 Frequency : start MDS on uncleanly shutdowned MDS device
639 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
640 Details : stay in waiting mds<>ost recovery finished produce random bugs
641 due race between two ll_sync thread for one lov target. send
642 ACTIVATE event only if connect realy finished and import have
646 Frequency : rare, connect and disconnect target at same time
648 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
649 Details : don't call obd_disconnect under lov_lock. this long time
650 operation and can block ptlrpcd which answer to connect request.
655 Descriptoin: don't lose wakeup for imp_recovery_waitq
656 Details : recover_import_no_retry or invalidate_import and import_close can
657 both sleep on imp_recovery_waitq, but we was send only one wakeup
661 Frequency : always with long access acl
663 Descriptoin: mds can't pack reply with long acl.
664 Details : mds don't control size of acl but they limited by reint/getattr
667 Severity : enhancement
669 Description: Update to SLES10 kernel-2.6.16.60-0.33.
671 Severity : enhancement
673 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
676 Frequency : start MDS on uncleanly shutdowned MDS device
678 Descriptoin: aborting recovery hang on MDS
679 Details : don't throttle destroy RPCs for the MDT.
682 Frequency : on remount
684 Description: external journal device not working after the remount
685 Details : clear dev_rdonly flag for external journal devices in
691 Description: shutdown vs evict race
692 Details : client_disconnect_export vs connect request race.
693 if client will evicted at this time - we start invalidate
694 thread without referece to import and import can be freed
698 Frequency : rare, need acl's on inode.
700 Description: client can't handle ost additional correctly
701 Details : if ost was added after client connected to mds client can have
702 hit lnet_try_match_md ... to big messages to wide striped files.
703 in this case need teach client to handle config events about add
704 lov target and update client max ea size at that event.
706 Severity : enhancement
708 Description: Changelogs
709 Details : Changelogs are a lightweight mechanism to track filesystem
710 metadata and namespace changes. The changelog is recorded
711 permanently on the MDTs, and is periodically "consumed" / purged
712 when records are no longer needed.
714 Severity : enhancement
716 Description: compact fld format with extents
717 Details : Store range of seq rather than every seq in FLD. Seq
718 controller update FLD rather than clients. In Case of CMD, mdt0
719 has FLD, all other metadata server act as non persistent proxy
720 for FLD queries and cache fld entries in fld cache.
725 Description: don't skip ost target if they assigned to file
726 Details : Drop slow OSCs if we can, but not for requested start idx.
727 This means "if OSC is slow and it is not the requested
728 start OST, then it can be skipped, otherwise skip it only
729 if it is inactive/recovering/out-of-space.
733 Description: more cleanup in mds_lov
734 Details : not send LOV EA under replay, we can't know about they size at this
735 time. Don't allow client connect to mds before any ost connected,
736 for avoid problems with LOV EA size and returning EIO to client.
738 Severity : enhancement
740 Description: Interoperability at server side (Disk interoperability)
742 Severity : enhancement
744 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
746 Severity : enhancement
748 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
750 Severity : enhancement
752 Description: New client IO stack (CLIO).
754 Severity : enhancement
756 Description: Commit on sharing. Eliminate inter-client dependencies between
757 uncommitted transactions by doing transaction commits.
758 Thereby clients may recovery independently.
761 Frequency : Create a symlink file with a very long name
763 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
764 Details : If there is no extra space in the request for early cancels,
765 ldlm_req_handles_avail() returns 0 instead of a negative value.
767 Severity : enhancement
769 Description: Add /proc entry for import status
770 Details : The mdc, osc, and mgc import directories now have
771 an import directory that contains useful import data for debugging
774 Severity : enhancement
776 Description: Re-disable certain /proc logging
777 Details : Enable and disable client's offset_stats, extents_stats and
778 extents_stats_per_process stats logging on the fly.
781 Frequency : Only on FC kernels 2.6.22+
783 Description: oops in statahead
784 Details : Do not drop reference count for the dentry from VFS when lookup,
785 VFS will do that by itself.
787 Severity : enhancement
789 Description: Generic /proc file permissions
790 Details : Set /Proc file permissions in a more generic way to enable non-
791 root users operate on some /proc files.
795 Description: Hitting mdc_commit_close() ASSERTION
796 Details : Properly handle request reference release in
797 ll_release_openhandle().
801 Description: quota recovery deadlock during mds failover
802 Details : This patch includes att18982, att18236, att18237 in bz14840.
804 1. fix osts hang when mds does failover with quotaon
805 2. prevent watchdog storm when osts threads wait for the
810 Frequency : only patchless client
811 Description: add workaround for race between add/remove dentry from hash
813 Severity : enhancement
815 Description: Allow OST glimpses to return PW locks
819 Description: LBUG when llog conf file is full
820 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
824 Description: Prevent import from entering FULL state when server in recovery
828 Description: service mount cannot take device name with ":"
829 Details : Only when device name contains ":/" will mount treat it as
835 Description: replace ptlrpcd with the statahead thread to interpret the async
836 statahead RPC callback
840 Frequency : on recovery
841 Description: I/O failures after umount during fail back
842 Details : if client reconnected to restarted server we need join to recovery
843 instead of find server handler is changed and process self
844 eviction with cancel all locks.
846 Severity : enhancement
848 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
850 Severity : enhancement
852 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
854 Severity : enhancement
856 Description: Upcall on Lustre log has been dumped
857 Details : Allow for a user mode script to be called once a Lustre log has
858 been dumped. It passes the filename of the dumped log to the
859 script, the location of the script can be specified via
860 /proc/sys/lnet/debug_log_upcall.
865 Description: avoid idr_remove called for id which is not allocated.
866 Details : Move assigment s_dev for clustered nfs to end of initialization,
867 to avoid problem with error handling.
872 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
873 Details : When connection is reused this not moved from CONN_UNUSED_HASH
874 into CONN_USED_HASH and this prodice warning when put connection
875 again in unused hash.
878 Severity : enhancement
880 Description: File striping can now be set to use an arbitrary pool of OSTs.
882 Severity : enhancement
884 Description: Export bytes_read/bytes_write count on OSC/OST.
888 Description: Early reply size mismatch, MGC loses connection
889 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
890 the connect flags are properly negotiated.
895 Description: quotas are not honored with O_DIRECT
896 Details : all writes with the flag O_DIRECT will use grants which leads to
897 this problem. Now using OBD_BRW_SYNC to guard this.
901 Description: add quota statistics
902 Details : 1. sort out quota proc entries and proc code.
903 2. add quota statistics
905 Severity : enhancement
907 Description: enable quota support for HEAD.
911 Description: Properly propagate oinfo flags from lov to osc for statfs
912 Details : restore missing copy oi_flags to lov requests.
914 Severity : enhancement
916 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
917 : l_getgroups(8), lst(8), routerstat(8)
918 Details : included man pages for llobdstat(8), llstat(8),
919 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
921 Severity : enhancement
923 Description: configure's --enable-quota should check the
924 : kernel .config for CONFIG_QUOTA
925 Details : configure is terminated if --enable-quota is passed but
926 : no quota support is in kernel
930 Description: Remove portals compatibility
931 Details : Remove portals compatibility, not interoperable with releases
936 Description: Resolve device initialization race
937 Details : Prevent proc handler from accessing devices added to the
938 obd_devs array but yet be intialized.
940 Severity : enhancement
942 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
944 Severity : enhancement
946 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
951 Description: Using wrong pointer in osc_brw_prep_request
952 Details : Access to array[-1] can produce panic if kernel compiled with
953 CONFIG_PAGE_ALLOC enabled
957 Description: Client runs out of low memory
958 Details : Consider only lowmem when counting initial number of llap pages
962 Description: *optional* service tags registration
963 Details : if the "service tags" package is installed on a Lustre node
964 When the filesystem is mounted, a local-node service tag will
965 be created. See http://inventory.sun.com/ for more information
966 about the Service Tags asset management system.
970 Description: Kernel BUG tries to release flock
971 Details : Lustre does not destroy flock lock before last reference goes
972 away. So always drop flock locks when client is evicted and
973 perform unlock regardless of successfulness of speaking to MDS.
977 Description: add refcount protection for osc callbacks, avoid panic on shutdown
981 Description: sanity test 65a fails if stripecount of -1 is set
982 Details : handle -1 striping on filesystem in ll_dirstripe_verify
987 Description: ASSERTION(CheckWriteback(page,cmd)) failed
988 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
991 Severity : enhancement
993 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
997 Description: do not process already freed flock
998 Details : flock can possibly be freed by another thread before it reaches
999 to ldlm_flock_completion_ast.
1003 Description: LBUG during stress test
1004 Details : Need properly lock accesses the flock deadlock detection list.
1008 Description: oops in page fault handler
1009 Details : kernel page fault handler can return two special 'pages' in error
1010 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1014 Description: timeout with invalidate import.
1015 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1016 should be handled by ptlrpcd. This produce long age waiting and
1017 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1019 Severity : enhancement
1021 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1025 Description: enable MGS and MDT services start separately
1026 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1027 a co-located MGS without starting the MGS, which is a complement
1028 to 'nosvc' mount option.
1032 Frequency : after recovery
1033 Description: precreate to many object's after del orphan.
1034 Details : del orphan st in oscc last_id == next_id and this triger growing
1035 count of precreated objects. Set flag LOW to skip increase count
1036 of precreated objects.
1040 Frequency : rare, on clear nid stats
1041 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1042 Details : when clean nid stats sometimes try destroy live entry,
1043 and this produce panic in free.
1047 Description: Stack overflow during MDS log replay
1048 ease stack pressure by using a thread dealing llog_process.
1052 Description: wait until IO finished before start new when do lock cancel.
1053 Details : VM protocol want old IO finished before start new, in this case
1054 need wait until PG_writeback is cleared until check dirty flag and
1055 call writepages in lock cancel callback.
1057 Severity : enhancement
1059 Description: using special macro for print time and cleanup in includes.
1063 Description: mds_mfd_close() ASSERTION(rc == 0)
1064 Details : In mds_mfd_close(), we need protect inode's writecount change
1065 within its orphan write semaphore to prevent possible races.
1069 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1073 Frequency : rare, on shutdown ost
1074 Description: don't hit live lock with umount ost.
1075 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1076 use shrink_dcache_sb instead.
1080 Description: don't panic with use echo client
1081 Details : echo client pass NULL as client nid pointer and this produce null
1082 pointer dereference.
1086 Description: fix build on ppc32
1087 Details : compile code with -m64 flag produce wrong object file for ppc32.
1091 Description: add message levels for liblustreapi
1095 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1096 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1097 will result in an -EAGAIN and in the exit call path we call
1098 alloc_rr() which will with increasing aggressiveness attempt to
1099 aquire precreated objects on the minimum number of required OSCs.
1103 Description: Use old size assignment to avoid deadlock
1104 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1105 the scheduling while holding a spinlock. We do not need locking
1106 for size in ll_update_inode() because size is only updated from
1107 the MDS for directories or files without objects, so there is no
1108 other place to do the update, and concurrent access to such inodes
1109 are protected by the inode lock.
1113 Description: resolve "_IOWR redefined" build error on SLES10
1117 Description: dump the memory debugging after all modules are unloaded to
1118 suppress false negative in conf_sanity test 39
1120 Severity : enhancement
1122 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1127 Description: fix SLES kernel versioning
1128 Details : the kernel version for our SLES 10 kernel did not include a "-"
1129 before the "smp" at the end. while this was not a problem in
1130 general, it did mean that software trying to use the kernel
1131 version to try to detect a vendor specific kernel would fail.
1132 this was most evident by the OFED build scripts.
1136 Description: Don't update lov_desc members until making sure they are valid
1137 Details : When updating lov_desc members via proc fs, need fix their
1138 validities before doing the real update.
1142 Description: don't put request into delay list while invalidate in flight.
1143 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1144 import in flight. this produce timeout for invalidate and sometimes
1145 can cause stale data.
1149 Frequency : on ppc only
1150 Description: not convert ost objects for directory because it's not exist.
1151 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1155 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1157 Severity : enhancement
1159 Description: Update to support 2.6.22.14 vanilla kernel.
1163 Frequency : rare, on recovery
1164 Description: read procfs can produce deadlock in some situation
1165 Details : Holding lprocfs lock which send rpc can produce block for destroy
1166 obd objects and this also block reconnect with -EALREADY. This
1167 isn't fix all lprocfs bugs - but make it rare.
1169 Severity : enhancement
1171 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1174 Frequency : frequent on X2 node
1176 Description: mdc_set_open_replay_data LBUG
1177 Details : Set replay data for requests that are eligible for replay.
1181 Description: lustre_mgs: operation 101 on unconnected MGS
1182 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1183 MGC, and late on MGC cannot successfully connect to MGS and a lot
1184 of the error messages complaining that MGS is not connected.
1188 Frequency : on network error
1189 Description: panic with double free request if network error
1190 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1191 but it's true only for synchronus enqueue, for async enqueue
1192 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1195 Severity : enhancement
1197 Description: client-side metadata stat-ahead during readdir(directory readahead)
1198 Details : perform client-side metadata stat-ahead when the client detects
1199 readdir and sequential stat of dir entries therein
1202 Frequency : on start mds
1204 Description: Implement get_info(last_id) in obdfilter.
1207 Frequency : occasional
1209 Description: Correctly check stale fid, not start epoch if ost not support SOM
1210 Details : open with flag O_CREATE need set old fid in op_fid3 because
1211 op_fid2 was overwritten with new generated fid, but mds can answer
1212 with one of these two fids and both is not stale. Setattr
1213 incorrectly started an epoch and assume will be called
1214 done_writing, but without SOM done_writing ever being called.
1217 Frequency : rare, depends on device drivers and load
1219 Description: MDS or OSS nodes crash due to stack overflow
1220 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1221 In some cases, in conjunction with device drivers that use a lot
1222 of stack the MDS (or possibly OSS) service threads could overflow
1223 the stack. One change which was identified to consume additional
1224 stack has been reworked to avoid the extra stack usage.
1227 Frequency : occasional
1229 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1230 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1231 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1232 oscc_has_objects and see OSC has no objests and call
1233 oscc_internal_create to resend create request.
1235 Severity : enhancement
1237 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1239 Severity : enhancement
1241 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1244 Frequency : very rare
1246 Description: Fix replay if there is an un-replied request and open
1247 Details : In some cases, older replay request will revert the
1248 mcd->mcd_last_xid on MDS which is used to record the client's
1249 latest sent request.
1251 Severity : enhancement
1253 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1255 Severity : enhancement
1257 Description: Add rhel5 support to HEAD.
1259 Serverity : enhancement
1261 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1266 Description: Don't allow skipping OSTs if index has been specified.
1267 Details : Don't allow skipping OSTs if index has been specified, make
1268 locking in internal create lots better.
1272 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1273 Details : access to bitfield in structure is always rounded to long
1274 and this produce problem with not atomic change any bit.
1278 Description: Lustre make rpms failed.
1279 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1280 builds Lustre rpms from tarball.
1283 Frequency : rare on shutdown ost
1285 Description: If llog cancel was not send before clean_exports phase, this can
1286 produce deadlock in llog code.
1287 Details : If llog thread has last reference to obd and call class_import_put
1288 this produce deadlock because llog_cleanup_commit_master wait when
1289 last llog_commit_thread exited, but this never success because was
1290 called from llog_commit_thread.
1294 Description: allow userland application know is lost one of stripes.
1295 Details : fill lvb_blocks with error code on ost and return it to
1296 application if error flag found.
1300 Description: NULL lov_tgts causing MDS oops
1301 Details : more safe checks for NULL lov_tgts for avoid oops.
1303 Severity : enhancement
1305 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1309 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1310 Details : If an OST is down the MDS will hang indefinitely in
1311 obd_statfs_rqset() waiting for the statfs data. While for
1312 MDS QOS usage of statfs, it should not stuck in waiting.
1314 Severity : enhancement
1316 Description: remote_acl support
1317 Details : Support ACL-based permission check for remote user.
1318 Support setfacl/getfacl for remote user with the utils
1319 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1320 the system "{s,g}etfacl" utils.
1322 Severity : enhancement
1324 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1326 Severity : enhancement
1328 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1332 Description: make mgs_setparam() handle fsname containing dash
1333 Details : fsname containing a dash does not work with lctl conf_param
1335 Severity : enhancement
1337 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1339 Severity : enhancement
1341 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1345 Frequency : rare, at shutdown
1346 Description: access already free / zero obd_namespace.
1347 Details : if client_disconnect_export was called without force flag set,
1348 and exist connect request in flight, this can produce access to
1349 NULL pointer (or already free pointer) when connect_interpret
1350 store ocd flags in obd_namespace.
1354 Frequency : only at startup
1355 Description: not alloc memory with spinlock held.
1356 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1357 if any spinlock held.
1359 Severity : enhancement
1361 Description: make lustre randomly fail allocating memory
1362 Details : Make lustre randomly failed allocating memory for testing purpose.
1364 Severity : enhancement
1366 Description: lost problems with lov objid file
1367 Details : Fixes some scability and access to not inited memory problems
1368 in work with lov objdid file.
1373 Description: lfs find does not continue on file error
1374 Details : Continue other files processing when a file/dir is absent.
1378 Description: Inconsistent usage of lustre_pack_reply()
1379 Details : Standardize the usage of lustre_pack_reply() such that it
1380 always generate a CERROR on failure.
1383 Frequency : occasional
1385 Description: MDS hang or stay in waiting lock
1386 Details : If client receive lock with CBPENDING flag ldlm need send lock
1387 cancel as separate rpc, to avoid situation when cancel request
1388 can't processed due all i/o threads stay in wait lock.
1392 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1396 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1400 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1404 Description: Update to RHEL4 latest kernel.
1406 Severity : enhancement
1408 Description: Build SLES10 patchless client fails
1409 Details : The configure was broken by run ./configure with
1410 --with-linux-obj=.... argument for patchless client. When the
1411 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1412 can't search header adequately. Use absolute path such as
1413 -I($LINUX)/include instead.
1417 Description: interrupt oig_wait produce painc on resend.
1418 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1419 private set, and this produce situation when rq_ptlrpcd_data not
1420 copyed to new allocated request and triggered LBUG on assert
1421 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1422 ptlrpcd set if request is changed and can be safety changed to use
1427 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1428 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1429 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1430 these macros and replace them with their difinitions instead.
1434 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1435 Details : Update which_patch & target file for SLES10 latest kernel.
1438 Description: Remove client patches from SLES 10 kernel.
1439 Details : This causes SLES 10 clients to behave as patchless clients
1440 even on a Lustre-patched (server) kernel.
1442 Severity : enhancement
1444 Description: self-adjustable client's lru lists
1445 Details : use adaptive algorithm for managing client cached locks lru
1446 lists according to current server load, other client's work
1447 pattern, memory activities, etc. Both, server and client
1448 side namespaces provide number of proc tunables for controlling
1451 Severity : enhancement
1453 Description: light-weight GSS support
1454 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1455 and reduce performance overhead.
1457 Severity : enhancement
1459 Description: Linux keyring support
1460 Details : Support using service of Linux keyring for Lustre GSS internal
1461 context refresh/cache mechanism.
1465 Description: Fix errors in lfs documentation
1466 Details : Fixes man pages
1470 Description: don't use GFP_* in generic Lustre code.
1471 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1475 Description: obdclass is limited by single OBD_ALLOC(idarray)
1476 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1480 Description: warnings with build patchless client with vanila 2.6.19 and up
1481 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1482 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1486 Frequency : Only for SLES
1487 Description: sanity_quota fail test_1
1488 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1489 file, which makes TSTID[2] inunique.
1493 Frequency : Only for SLES9
1494 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1495 Details : Update target/ChangeLog/which_patch .
1499 Frequency : Only for test_10 in sanity-quota.sh
1500 Description: a bug in quota test code
1501 Details : checking a test flag in a wrong place
1505 Frequency : Only for test_13 in sanity-quota.sh
1506 Description: a bug in quota test script
1507 Details : shouldn't check overfull allocated quota
1511 Description: jbd statistics
1512 Details : Port older jbd statistics patch for sles10
1514 Severity : enhancement
1516 Description: Add printing inode info into message about error in writepage.
1521 Description: reply_lock_interpret crash due to race with it and lock cancel.
1522 Details : Do not replay locks that are being cancelled. Do not reference
1523 locks by their address during replay, just by their handle.
1527 Frequency : When flocks are used.
1528 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1529 Details : Flock locks might destroy just granted lock if it could be merged
1530 with another existing flock, this is done in completion handler,
1531 so teach ldlm_cli_enquque_fini that this is a valid case for
1537 Description: Oops in read and write path when failing to allocate lock.
1538 Details : Check if lock allocation failed and return error back.
1542 Description: lstripe command fails for valid OST index
1543 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1544 instead of lov->desc.ld_active_tgt_count.
1548 Description: sanity.sh failed test 103
1549 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1554 Description: df doesn't work properly if diskfs blocksize != 4K
1555 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1559 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1560 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1562 Severity : enhancement
1564 Description: Adaptive timeouts
1565 Details : RPC timeouts adapt to changing server load and network
1566 conditions to reduce resend attempts and improve recovery time.
1570 Description: llapi_file_create() does not allow some changes
1571 Details : add llapi_file_open() that allows specifying the mode and
1572 open flags, and also returns an open file handle.
1576 Description: merge and cleanup kernel patches.
1577 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1581 Description: Add journal checksum support.(Kernel part)
1582 Details : The journal checksum feature adds two new flags i.e
1583 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1584 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1585 indicates that the commit block contains the checksum for
1586 the blocks described by the descriptor blocks. Now commit
1587 record can be sent to disk without waiting for descriptor
1588 blocks to be written to disk. This behavior is controlled
1589 using JBD2_FEATURE_ASYNC_COMMIT flag.
1593 Description: OSS needs mutliple precreate threads
1594 Details : Add ability to start more than one create thread per OSS.
1598 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1599 Details : execute lfs setstripe on client
1603 Description: mds_obd_create error creating tmp object
1604 Details : When the user sets quota on root, llog will be affected and can't
1605 create files and write files.
1608 Frequency : Always on ia64 patchless client, and possibly others.
1610 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1611 Details : This allows the patchless client to be loaded on architectures
1612 without this export.
1616 Description: RedHat Update kernel for RHEL5
1617 Details : Add the kernel config file for RHEL5.
1621 Description: RedHat Update kernel for RHEL5
1622 Details : Modify the kernel config file more closer RHEL5.
1626 Description: Build failure against Centos5 (RHEL5)
1627 Details : Define PAGE_SIZE when it isn't present.
1631 Description: test_7 on sanity-quota.sh can't be used on separate machines
1632 Details : add facet to handle it
1636 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1637 Details : with NFS, the anon dentry's parent was set to itself in
1638 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1639 corresponding dentry other than use rec->ur_name.
1641 Severity : enhancement
1643 Description: lfs setstripe enhancement
1644 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1648 Description: enable data checksumming by default
1649 Details : enable checksum by default, allow --disable-checksum
1650 configure option and "-o nochecksum" mount option. Checksums
1651 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1655 Description: lustre support for RHEL5
1656 Details : Add support for RHEL5.
1660 Description: Client eviction due to failover config
1661 Details : after a connection loss, the lustre client should attempt to
1662 reconnect to the last active server first before trying the
1663 other potential connections.
1667 Description: when mds and osts use different quota unit(32bit and 64bit),
1668 quota will be released repeatly.
1669 Details : void sending multiple quota reqs to mds, which will keep the
1670 status between the reqs.
1674 Description: osts not allocated evenly to files
1675 Details : change the condition to increase offset_idx
1678 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1680 Description: Data corruption for OSTs that are formatted larger than 2TB
1682 Details : When generating the bio request for lustre file writes the
1683 sector number would overflow a temporary variable before being
1684 used for the IO. The data reads correctly from Lustre (which
1685 will overflow in a similar manner) but other file data or
1686 filesystem metadata may be corrupted in some cases.
1690 Description: Tune the kernel for good SCSI performance.
1691 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1692 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1697 Description: rewrite ext2-derived code in obdclass/uuid.c
1698 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1699 from scratch preserving functionality.
1703 Description: "lfs find -obd UUID" prints directories
1704 Details : "lfs find -obd UUID" will return all directory names instead
1705 of just file names. It is incorrect because the directories
1706 do not reside on the OSTs.
1710 Description: use i_size_read and i_size_write in 2.6 port
1711 Details : replace inode->i_size access with i_size_read/write()
1714 Frequency : when removing large files
1716 Description: scheduling issue during removal of large Lustre files
1717 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1718 It causes scheduling issues when removing large files (17TB in the
1722 Frequency : only with liblustre clients on XT3
1724 Description: evictions taking too long
1725 Details : allow llrd to evict clients directly on OSTs
1728 Frequency : only on ppc
1730 Description: /proc/fs/lustre/devices broken on ppc
1731 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1732 We should make sure the type of 'index' is loff_t and then cast
1733 explicitly as needed below. Do not assign an explicitly cast
1737 Frequency : only for rhel5
1739 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1740 Details : Modify the target file & which_kernel.
1742 Severity : enhancement
1744 Description: omit set fsid for export NFS
1745 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1746 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1750 Frequency : Always, for kernels after 2.6.16
1751 Description: Fix warning idr_remove called for id=.. which is not allocated.
1752 Details : Last kernels save old s_dev before kill super and not allow
1753 to restore from callback - restore it before call kill_anon_super.
1757 Description: buffer overruns could theoretically occur
1758 Details : llapi_semantic_traverse() modifies the "path" argument by
1759 appending values to the end of the origin string, and a
1760 overrun may occur. Adding buffer overrun check in liblustreapi.
1764 Description: Fix error on 'ls .' at the top of the Lustre mount.
1765 Details : Don't revalidate dentry if it is a root dentry.
1769 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1770 Details : Modify vm-tunables-rhel4.patch.
1774 Description: Kernel config for 2.6.18-vanilla.
1775 Details : Modify targets/2.6-vanilla.target.in.
1776 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1777 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1778 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1779 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1783 Description: improve handling recoverable errors
1784 Details : if request processig with error which can be recoverable on server
1785 request should be resend, otherwise page released from cache and
1790 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1791 Details : Modify target file & which_patch.
1792 A flaw was found in the IA32 system call emulation provided
1793 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1794 value could be stored in the %RAX register, which could trigger an
1795 out-of-bounds system call table access. An untrusted local user
1796 could exploit this flaw to run code in the kernel
1797 (ie a root privilege escalation). (CVE-2007-4573).
1801 Description: O_DIRECT bypasses client statistics.
1802 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1803 read_ahead_stats were updated. copied the stats section from
1804 osc_send_oap_rpc() into async_internal().
1808 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1812 Description: change order of libsysio includes
1813 Details : '#include sysio.h' should always come before '#include xtio.h'
1817 Description: handle "serious error: objid * already exists" more gracefully
1818 Details : If LAST_ID value on disk is smaller than the objects existing in
1819 the O/0/d* directories, it indicates disk corruption and causes an
1820 LBUG(). If the object is 0-length, then we should use the existing
1821 object. This will help to avoid a full fsck in most cases.
1823 Severity : enhancement
1825 Description: adapt the lustre_config script to support the upgrade case
1826 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1827 server targets to 1.6 in parallel.
1831 Description: To avoid grant space > avaible space when the disk is almost
1832 full. Without this patch you might see the error "grant XXXX >
1833 available" or some LBUG about grant, when the disk is almost
1835 Details : In filter_check_grant, for non_grant cache write, we should
1836 check the left space by if (*left > ungranted + bytes), instead
1837 of (*left > ungranted), because only we are sure the left space
1838 is enough for another "bytes", then the ungrant space should be
1839 increase. In client, we should update cl_avail_grant only there
1840 is OBD_MD_FLGRANT in the reply.
1844 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1845 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1846 described in CVE-2007-4573. This problem could allow untrusted
1847 local users to gain root access.
1850 Frequency : when using O_DIRECT and quotas
1852 Description: Incorrect file ownership on O_DIRECT output files
1853 Details : block usage reported by 'lfs quota' does not take into account
1854 files that have been written with O_DIRECT.
1859 Description: touch file failed when fs is not full
1860 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1861 otherwise we can get ENOSP while fs is not full.
1865 Description: parallel lock callbacks
1866 Details : Instead of sending blocking and completion callbacks as separated
1867 requests, adding them to a set and sending in parallel.
1870 Frequency : only for Cray XT3
1871 Bugzilla : 12829/13455
1872 Description: Changing primary group doesn't change the group lustre assigns to
1874 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1875 overridden with the primary group provided by the group upcall,
1876 whereas the supplied fsgid can be trusted if it is in the list of
1877 supplementary groups returned by the group upcall.
1879 Severity : enhancement
1881 Description: Allow masking D_WARNING, D_ERROR messages from console
1882 Details : Console messages can now be disabled via lnet.printk.
1886 Description: User code with malformed file open parameter crashes client node
1887 Details : Before packing join_file req, all the related reference should be
1888 checked carefully in case some malformed flags cause fake
1889 join_file req on client.
1893 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1894 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1895 mode is equal to the granted mode. However, it should also check
1896 that LDLM_FL_CP_REQD is not set.
1900 Description: Heavy nfs access might result in deadlocks
1901 Details : After ELC code landed, it is now improper to enqueue any mds
1902 locks under och_sem, because enqueue might want to decide to
1903 cancel open locks for same inode we are holding och_sem for.
1907 Description: Client eviction while running blogbench
1908 Details : A lot of unlink operations with concurrent I/O can lead to a
1909 deadlock causing evictions. To address the problem, the number of
1910 oustanding OST_DESTROY requests is now throttled to
1911 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1912 callbacks are processed in priority.
1916 Description: enable ACLs on MDS by default
1917 Details : ACLs must be enabled on MDS by default.
1920 Frequency : PPC/PPC64 only
1922 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1923 Details : fix duplicated definitions between asm-ppc64/types.h and
1924 lustre_types.h on PPC.
1927 Frequency : PPC/PPC64 only
1929 Description: asm-ppc/segment.h does not exist
1930 Details : fix compile issue on PPC.
1934 Description: better handle error messages in extents code
1937 Frequency : RHEL4 only
1939 Description: mkfs is very slow on IA64/RHEL4
1940 Details : A performance regression has been discovered in the MPT Fusion
1941 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1942 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1943 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1945 Severity : enhancement
1947 Description: SNMP support enhancement
1948 Details : Adding total number of sampled request for an MDS node in snmp
1951 Severity : enhancement
1953 Description: Optimize ldlm waiting list processing for PR extent locks
1954 Details : When processing waiting list for read extent lock and meeting read
1955 lock that is same or wider to it that is not contended, skip
1956 processing rest of the list and immediatelly return current
1957 status of conflictness, since we are guaranteed there are no
1958 conflicting locks in the rest of the list.
1962 Description: Time out and refuse to reconnect
1963 Details : When the failover node is the primary node, it is possible
1964 to have two identical connections in imp_conn_list. We must
1965 compare not conn's pointers but NIDs, otherwise we can defeat
1966 connection throttling.
1970 Description: port llog fixes from b1_6 into HEAD
1971 Details : Port llog reference couting and some llog cleanups from b1_6
1972 (bug 10800) into HEAD, for protect from panic and access to already
1973 free llog structures.
1977 Description: Detect stride IO mode in read-ahead
1978 Details : When a client does stride read, read-ahead should detect that and
1979 read-ahead pages according to the detected stride pattern.
1983 Description: data checksumming impacts single node performance
1984 Details : add support for several checksum algorithm. Currently, only CRC32
1985 and Adler-32 are supported. The checksum type can be changed on
1986 the fly via /proc/fs/lustre/osc/*/checksum_type.
1990 Description: use adler32 for page checksums
1991 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1996 Description: build for x2 fails
1997 Details : fix compile issue on Cray systems.
2001 Description: Properly match for duplicate locks
2002 Details : Due to different lock order from skiplists code, we need to
2003 traverse entire list for now
2006 Frequency : only on PPC/SLES10
2008 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2009 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2010 be defined. Add a hack in mkfs_lustre.c to work around this compile
2015 Description: LASSERT on MDS when client holding flock lock dies
2016 Details : ldlm pool logic depends on number of granted locks equal to
2017 number of released locks which is not true for flock locks, so
2018 just exclude such locks from consideration.
2022 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2023 Details : Use fsfilt_sync() for both the whole filesystem sync and
2024 individual file sync to eliminate dangerous inode locking
2025 with I_LOCK that can lead to a deadlock.
2029 Description: performance in 1.6.3
2030 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2031 sure that 1MB requests are not fragmented by the block layer.
2033 Severity : enhancement
2035 Description: organize the server-side client stats on per-nid basis
2036 Details : Change the structure of stats under obdfilter and mds to
2044 The "uuid"s file would list the uuids of _active_ exports.
2045 And the clear entry is to clear all stats and stale nids.
2047 Severity : enhancement
2049 Description: eliminate client locks in face of contention
2050 Details : file contention detection and lockless i/o implementation
2051 for contended files.
2055 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2058 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2060 Description: MDS may lose file striping (and hence file data) in some cases
2061 Details : If there are additional extended attributes stored on the MDS,
2062 in particular ACLs, SELinux, or user attributes (if user_xattr
2063 is specified for the client mount options) then there is a risk
2064 of attribute loss. Additionally, the Lustre file striping
2065 needs to be larger than default (e.g. striped over all OSTs),
2066 and an additional attribute must be stored initially in the
2067 inode and then increase in size enough to be moved to the
2068 external attribute block (e.g. ACL growing in size) for file
2073 Description: skiplist implementation simplification
2074 Details : skiplists are used to group compatible locks on granted list
2075 that was implemented as tracking first and last lock of each
2076 lock group the patch changes that to using doubly linked lists
2080 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2081 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2082 because LASSERT on both the data supplied by a client, and the
2083 data on disk is dangerous and incorrect.
2085 Severity : enhancement
2087 Description: Slow truncate/writes to huge files at high offsets.
2088 Details : Directly associate cached pages to lock that protect those pages,
2089 this allows us to quickly find what pages to write and remove
2090 once lock callback is received.
2094 Description: more ldlm soft lockups
2095 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2096 starve other threads from the resource lock for a long time in
2097 case of long waiting queue, so change the debug level from
2098 D_OTHER to the less frequently used D_INFO.
2100 Severity : enhancement
2102 Description: add -gid, -group, -uid, -user options to lfs find
2106 Description: Hung threads in invalidate_inode_pages2_range
2107 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2108 once one is completed. As a result, some RPCs are stuck in the
2109 queue and are never sent.
2113 Description: filter threads hungs on waiting journal commit
2114 Details : Cleanup filter group llog code, then only filter group llog will
2115 be only created in the MDS/OST syncing process.
2119 Description: Procfs and llog threads access destoryed import sometimes.
2120 Details : Sync the import destoryed process with procfs and llog threads by
2121 the import refcount and semaphore.
2123 Severity : enhancement
2125 Description: openlock cache of b1_6 port to HEAD
2130 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2131 Details : If insertion of an extent fails, then discard the inode
2132 preallocation and free data blocks else it can lead to duplicate
2137 Description: don't always update ctime in ext3_xattr_set_handle()
2138 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2139 In some cases the ctime should not be updated, for example for
2140 2.0->1.8 compatibility it is necessary to delete an xattr and it
2141 should not update the ctime.
2145 Bugzilla : 15713/16362
2146 Description: Assertion in iopen_connect_dentry in 1.6.3
2147 Details : looking up an inode via iopen with the wrong generation number can
2148 populate the dcache with a disconneced dentry while the inode
2149 number is in the process of being reallocated. This causes an
2150 assertion failure in iopen since the inode's dentry list contains
2151 both a connected and disconnected dentry.
2155 Description: assertion failure in ldlm_handle2lock()
2156 Details : fix a race between class_handle_unhash() and class_handle2object()
2157 introduced in lustre 1.6.5 by bug 13622.
2162 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2163 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2168 Description: oops in sd_iostats_seq_show()
2169 Details : unloading/reloading the scsi low level driver triggers a kernel
2170 bug when trying to access the sd iostat file.
2175 Description: Kernel panics during QLogic driver reload
2176 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2177 patch, causing memory corruption.
2182 Description: journal_dev option does not work in b1_6
2183 Details : pass mount option during pre-mount.
2185 Severity : enhancement
2187 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2188 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2189 extent information of a file. It can be used to map logical blocks
2190 in a file to physical blocks in the block device.
2194 Description: LDLM soft lockups - improvement
2195 Details : It is be possible to send the lock handle along with each read
2196 or write request because the client is already doing a lock match
2197 itself so there isn't any reason the OST should have to re-do that
2201 Frequency : only on Cray X2
2203 Description: X2 build failures
2204 Details : fix build failures on Cray X2.
2208 Description: xid & resent requests
2209 Details : Initialize RPC XID from clock at startup (randomly if clock is
2212 Severity : enhancement
2214 Description: Add lustre_start utility to start or stop multiple Lustre servers
2219 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2220 Details : In case of memory pressure, list_del() can be called twice on
2221 req->rq_history_list, causing a kernel oops.
2225 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2226 Details : fix stack overflow in the distributed lock manager by defering
2227 export eviction after a failed AST to the elt thread instead of
2228 handling it in the dlm interpret routine.
2232 Description: Convert some comments to new format.
2233 Details : Update documenting comments to match doxygen conventions.
2237 Description: Grammar fixes.
2238 Details : A couple of trivial spelling fixes.
2242 Description: OSD_COUNTERS-mandatory
2243 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2244 to simplify the code.
2248 Description: simplify cmm_device freeing logic.
2249 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2253 Description: Add lockdep support to dt_object_operations locking interface.
2254 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2255 parameter indicating lock ordering. Update mdd code to use new
2260 Description: Introduce failloc constants for lockless IO tests.
2261 Details : Add two new failloc constants to test lockless IO. Only one of
2262 them in implemented---another is checked in yet to be landed
2267 Description: Add lockdep support for inode mutex.
2268 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2269 in the situations where Lustre has to lock more than one inode
2274 Description: Add optional invariants checking support.
2275 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2276 configure switch. This macro is to be used for consistency and
2277 sanity checks that are too expensive to be left in `production'
2282 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2283 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2288 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2289 Details : Use spin_lock_nested() in (the only) situation where more than
2290 one ldlm_lock is locked simultaneously. Also, fix possible
2291 dead-lock in ldlm_lock_change_resource() by enforcing particular
2296 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2297 Details : Instead of specifying each ldlm_lock call-back through separate
2298 parameter, wrap them into struct ldlm_callback_suite.
2302 Description: Kill join_lru obd method and its callers.
2303 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2304 in memory---a requirement implemented through ->o_join_lru() obd
2305 method in HEAD. Remove this method and its users.
2309 Description: Add asynchronous ldlm ENQUEUE completion handler.
2310 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2311 case that stock ldlm_completion_ast() cannot handle as it waits
2312 until lock is granted. Introduce new ldlm_completion_ast_async()
2313 for this. Also comment ldlm_completion_ast().
2317 Description: ldlm_error <-> errno conversion.
2318 Details : Add functions to map (rather arbitrary) between LDLM error codes
2319 and standard errno values. CLIO needs this to prevent LDLM specific
2320 constants from escaping ldlm and osc.
2324 Description: Kill unused ldlm_handle2lock_ns() function.
2325 Details : Kill unused ldlm_handle2lock_ns() function.
2329 Description: Add lu_ref support to ldlm_lock
2330 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2331 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2332 and ldlm_resource. LDLM interface has to be changed, because code that
2333 releases a reference on a lock, has to "know" what reference this is.
2334 In the most frequent case
2336 lock = ldlm_handle2lock(handle);
2338 LDLM_LOCK_PUT(lock);
2340 no changes are required. When any other reference (received _not_ from
2341 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2342 instead of LDLM_LOCK_PUT().
2344 Arguably, changes are pervasive, and interface requires some discipline
2345 for proper use. On the other hand, it was very instrumental in finding
2346 a few leaked lock references.
2350 Description: Add ldlm_lock_addref_try().
2351 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2352 attempts to addref a lock that might be being canceled
2357 Description: Add ldlm_weigh_callback().
2358 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2359 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2360 hard-coded `number of pages' logic.
2364 Description: Add lockdep annotations to llog code.
2365 Details : Use appropriately tagged _nested() locking calls in the places
2366 where llog takes more than one ->lgh_lock lock.
2370 Description: Add loi_kms_set().
2371 Details : Wrap kms updates into a helper function.
2375 Description: Constify instances of struct lsm_operations.
2376 Details : Constify instances of struct lsm_operations.
2380 Description: lu_conf support.
2381 Details : On a server, a file system object is uniquely identified
2382 by a fid, which is sufficient to locate and load all object
2383 state (inode). On a client, on the other hand, more data are
2384 necessary instantiate an object. Change lu_object_find() and
2385 friends to take additional `lu_conf' argument describing object.
2386 Typically this includes layout information.
2390 Description: lu_context fixes.
2391 Details : Introduce new lu_context functions that are needed on the client
2392 side, where some system threads (ptlrpcd) are shared by multiple
2393 modules, and so cannot be stopped during module shutdown.
2397 Description: Add start and stop methods to lu_device_type_operations.
2398 Details : Introduce two new methods in lu_device_type_operations, that are
2399 invoked when first instance of a given type is created and last one
2400 is destroyed respectively. This is need by CLIO.
2404 Description: Add lu_ref support to struct lu_device.
2405 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2406 track leaked references.
2410 Description: Introduce lu_kmem_descr.
2411 Details : lu_kmem_descr and its companion interface allow to create
2412 and destroy a number of kmem caches at once.
2416 Description: Fix lu_object finalization race.
2417 Details : Fix a race between lu_object_find() finding an object and its
2418 concurrent finalization. This race is (most likely) not possible
2419 on the server, but might happen on the client.
2423 Description: Introduce lu_ref interface.
2424 Details : lu_ref is a debugging module allowing to track references to
2425 a given object. It is quite cpu expensive, and has to be
2426 explicitly enabled with --enable-lu_ref. See usage description
2431 Description: Factor lu_site procfs stats into a separate function.
2432 Details : Separate lu_site stats printing code into a separate function
2433 that can be reused on a client.
2437 Description: Constify instances of struct {lu,dt,md}_device_operations.
2438 Details : Constify instances of struct {lu,dt,md}_device_operations.
2442 Description: Introduce struct md_site and move meta-data specific parts of
2443 struct lu_site here.
2444 Details : Move md-specific fields out of struct lu_site into special struct
2445 md_site, so that lu_site can be used on a client.
2449 Description: Kill mdd_lov_destroy().
2450 Details : Remove unused mdd code.
2454 Description: Add st_block checking to multistat.c.
2455 Details : Add st_block checking to multistat.c.
2459 Description: Add lu_ref support to struct obd_device.
2460 Details : Add lu_ref tracking to obd_device.
2464 Description: Kill obd_set_fail_loc().
2465 Details : Remove unused code.
2469 Description: Add special type for ptlrpc_request interpret functions.
2470 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2471 there. Actual usage will be in CLIO.
2475 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2476 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2477 doesn't work with lockdep.
2481 Description: Add rwv.c test program.
2482 Details : New testing program exercising readv(2) and writev(2) (Qian).
2486 Description: Add sendfile.c test program.
2487 Details : New testing program exercising sendfile(2) (Jay).
2491 Description: Ratelimit a message that can be very frequent.
2492 Details : Ratelimit a memory allocation failure message that can
2497 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2498 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2502 Description: Liblustre build fixes.
2503 Details : Liblustre build fixes.
2507 Description: libcfs: add cfs_{need,cond}_resched() interface.
2508 Details : libcfs: add cfs_{need,cond}_resched() definition and
2509 implementations for Linux, NT, and liblustre.
2511 Severity : enhancement
2513 Description: More exported tunables for mballoc
2514 Details : Add support for tunable preallocation window and new tunables for
2515 large/small requests
2519 Description: Detect corruption of block bitmap and checking for preallocations
2520 Details : Checks validity of on-disk block bitmap. Also it does better
2521 checking of number of applied preallocations. When corruption is
2522 found, it turns filesystem readonly to prevent further corruptions.
2526 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2527 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2528 window to 0. If it does hit cache miss, and read-pattern is still
2529 stride-io mode, does not reset the stride window, but also does not
2530 increase the stride window length in this case.
2534 Frequency : only for big-endian servers
2535 Description: Check if system is big-endian while mounting fs with extents feature
2536 Details : Mounting a filesystem with extents feature will fail on big-endian
2537 systems since ext3-based ldiskfs is not supported on big-endian
2538 systems. This can be over-riden with "bigendian_extents" mount option.
2540 Severity : enhancement
2542 Description: The root squash functionality
2543 Details : A security feature, which is to prevent users from being able
2544 to mount lustre on their desktop, run as root, and delete
2545 all of the files in the filesystem. The goal is accomplished by
2546 remapping user id (UID) and group id (GID) of the root user to
2547 a UID and GID specified by the system administartor via Lustre
2548 configuration management server (MGS). The functionality also
2549 allows to specify sets of clients for which the remapping does
2554 Description: Excessive recovery window
2555 Details : With AT enabled, the recovery window can be excessively long (6000+
2556 seconds). To address this problem, we no longer use
2557 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2558 timeout no longer depends on the service time, it is set to
2559 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2560 time via pb_service_time.
2564 Description: Watchdog triggered on MDS failover
2565 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2566 the OSTs know that the MDS "UUID" can be reused for the same export
2567 from a different NID, so we do not need to wait for the export to be
2571 Frequency : rare, only if using MMP with Linux RAID
2573 Description: MMP doesn't work with Linux RAID
2574 Details : While using HA for Lustre servers with Linux RAID, it is possible
2575 that MMP will not detect multiple mounts. To make this work we
2576 need to unplug the device queue in RAID when the MMP block is being
2577 written. Also while reading the MMP block, we should read it from
2578 disk and not the cached one.
2580 Severity : enhancement
2582 Description: open file using fid
2583 Details : A file can be opened using just its fid, like
2584 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2587 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2590 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2591 Details : add proper locking for creation of jbd_4k slab cache
2595 Description: MMP check in ext3_remount() fails without displaying any error
2596 Details : When multiple mount protection fails during remount, proper error
2599 Severity : enhancement
2601 Description: Allow stripe size to be up to 4G-64k
2602 Details : Fix math logic to allow large stripe sizes.
2606 Description: add check for >8TB ldiskfs filesystems
2607 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2608 Don't allow >8TB ldiskfs filesystems to be mounted without
2609 force_over_8tb mount option
2614 Descriptoin: OSS complains "too many blocks" and crashes
2615 Details : In filter_commitrw_write():
2616 1. lquota_chkquota();
2617 2. obdo_from_inode();
2618 3. lquota_pending_commit();
2619 "2." has changed uid and gid in this case, then caused
2620 lquota_chkquota() and lquota_pending_commit() to use
2621 different uid and gid. That is the root of the bug.
2625 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2626 Details : Detect and print out the 'stripe count = -1' case properly.
2630 Description: Out or order replies might be lost on replay
2631 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2632 request with tid smaller then one currently being added, add it
2633 to the start, not end of the list.
2635 --------------------------------------------------------------------------------
2637 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2639 * Support for kernels:
2640 2.6.5-7.283 (SLES 9),
2641 2.6.9-55.EL (RHEL 4),
2642 2.6.16.46-0.14 (SLES 10),
2643 2.6.18.8 vanilla (kernel.org)
2644 * Client support for unpatched kernels:
2645 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2646 2.6.16 - 2.6.22 vanilla (kernel.org)
2647 * Due to recently discovered recovery problems, we do not recommend
2648 using patchless RHEL 4 clients with this or any earlier release.
2649 * Recommended e2fsprogs version: 1.39.cfs8
2650 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2651 * Starting with this release, the ldiskfs backing filesystem required
2652 by Lustre is now in its own package, lustre-ldiskfs. This package
2653 should be installed. It is versioned separately from Lustre and
2654 may be released separately in future.
2658 Description: block reactivating mgc import until all deactivates complete
2659 Details : Fix race when failing back MDT/MGS to itself (testing)
2662 Frequency : at statup only
2664 Description: mds_lov_synchronize race leads to various problems
2665 Details : simultaneous MDT->OST connections at startup can cause the
2666 sync to abort, leaving the OSC in a bad state.
2668 Severity : enhancement
2670 Description: add optional extra BUILD_VERSION info
2671 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2672 to override the lustre version.
2675 Frequency : 2.6.18 servers only
2677 Description: ll_kern_mount() doesn't release the module reference
2678 Details : The ldiskfs module reference count never drops down to 0
2679 because ll_kern_mount() doesn't release the module reference.
2684 Description: server LBUG when using old ost_num_threads parameter
2685 Details : Accept the old ost_num_threads parameter but warn that it
2686 is deprecated, and fix an off-by-one error that caused an LBUG.
2691 Description: Transient SCSI error results in persistent IO issue
2692 Details : iobuf->dr_error is not reinitialized to 0 between two
2696 Frequency : sometimes when underlying device returns I/O errors
2698 Description: OSTs not going read-only during write failures
2699 Details : OSTs are not remounted read-only when the journal commit threads
2700 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2701 instead of the ext3 wrappers.
2705 Description: poor connect scaling with increasing client count
2706 Details : Don't run filter_grant_sanity_check for more than 100 exports
2707 to improve scaling for large numbers of clients.
2710 Frequency : SLES10 only
2712 Description: sanity-quota.sh quotacheck failed: rc = -22
2713 Details : Quotas cannot be enabled on SLES10.
2716 Frequency : liblustre clients only
2718 Description: getdirentries does not give error when run on compute nodes
2719 Details : getdirentries does not fail when the size specified as an argument
2720 is too small to contain at least one entry
2722 Severity : enhancement
2724 Description: Add LNET router traceability for debug purposes
2725 Details : If a checksum failure occurs with a router as part of the
2726 IO path, the NID of the last router that forwarded the bulk data
2727 is printed so it can be identified.
2732 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2733 Details : Due to a race condition, liblustre clients were occasionally
2734 evicted incorrectly.
2736 Severity : enhancement
2738 Description: lfs setstripe use optional parameters instead of postional
2741 Severity : enhancement
2743 Description: Nanosecond timestamp support for ldiskfs
2744 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2745 resolution timestamps. There is not yet support for this at
2746 the Lustre filesystem level.
2749 Frequency : during server recovery
2751 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2752 Details : request with rq_no_resend flag not awake l_wait_event if they get
2756 Frequency : nfs export on patchless client
2758 Description: connectathon hang when test nfs export over patchless client
2759 Details : Disconnected dentry cannot be found with lookup, so we do not need
2760 to unhash it or make it invalid
2763 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2764 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2765 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2770 Description: open req refcounting wrong on reconnect
2771 Details : If reconnect happened between getting open reply from server and
2772 call to mdc_set_replay_data in ll_file_open, we will schedule
2773 replay for unreferenced request that we are about to free.
2774 Subsequent close will crash in variety of ways.
2775 Check that request is still eligible for replay in
2776 mdc_set_replay_data().
2781 Description: disable writes to filesystem when reading health_check file
2782 Details : the default for reading the health_check proc file has changed
2783 to NOT do a journal transaction and write to disk, because this
2784 can cause reads of the /proc file to hang and block HA state
2785 checking on a healthy but otherwise heavily loaded system. It
2786 is possible to return to the previous behaviour during configure
2787 with --enable-health-write.
2789 Severity : enhancement
2791 Description: 64-bit inode version
2792 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2793 to the inode. This will be required for version-based recovery.
2798 Description: MDS fails to start if a duplicate client export is detected
2799 Details : in some rare cases it was possible for a client to connect to
2800 an MDS multiple times. Upon recovery the MDS would detect this
2801 and fail during startup. Handle this more gracefully.
2803 Severity : enhancement
2805 Description: Add -o localflock option to simulate old noflock
2807 Details : This will achieve local-only flock/fcntl locks
2813 Description: log_commit_thread vs filter_destroy race leads to crash
2814 Details : Take import reference before releasing llog record semaphore
2819 Description: Wrong request locking in request set processing
2820 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2821 imp_delayed_list, in this place should be used imp_lock.
2824 Frequency : when reconnection
2826 Description: Grant Leak when osc reconnect to OST
2827 Details : When osc reconnect ost, OST(filter) should check whether it
2828 should grant more space to client by comparing fed_grant and
2829 cl_avail_grant, and return the granted space to client instead
2830 of "new granted" space, because client will call osc_init_grant
2831 to update the client grant space info.
2834 Frequency : when client reconnect to OST
2836 Description: Grant Leak when osc do resend and replay bulk write
2837 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2838 bulk write request, because the grant info while be sync between
2839 OSC and OST when reconnect, and we should ignore the grant info
2840 these of resend/replay write req.
2845 Description: Grant space more than avaiable left space sometimes.
2846 Details : When then OST is about to be full, if two bulk writing from
2847 different clients came to OST. Accord the avaliable space of the
2848 OST, the first req should be permitted, and the second one
2849 should be denied by ENOSPC. But if the seconde arrived before
2850 the first one is commited. The OST might wrongly permit second
2851 writing, which will cause grant space > avaiable space.
2854 Frequency : when client is evicted
2856 Description: Grant might be wrongly erased when osc is evicted by OST
2857 Details : when the import is evicted by server, it will fork another
2858 thread ptlrpc_invalidate_import_thread to invalidate the
2859 import, where the grant will be set to 0. While the original
2860 thread will update the grant it got when connecting. So if
2861 the former happened latter, the grant will be wrongly errased
2862 because of this race.
2867 Description: Checking Stale with correct fid
2868 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2869 to check whether it is stale, because sometimes, we want the
2870 enqueue happened anyway, and op_data.fid2 will not be initialized.
2872 Severity : enhancement
2874 Description: update patchless client
2875 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2878 Frequency : only with 2.4 kernel
2880 Description: random memory corruption
2881 Details : size of struct ll_inode_info is to big for union inode.u and this
2882 can be cause of random memory corruption.
2887 Description: Memory leak in recovery
2888 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2889 It should also check obd_fail before fsfilt_start, otherwise if
2890 fsfilt_start return -EROFS,(failover mds during mds recovery).
2891 then the req will return with repmsg->transno = 0 and rc = EROFS.
2892 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2893 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2894 be freed no matter whether fsfilt_commit success or not.
2897 Frequency : only with huge count clients
2899 Description: Prevents from taking the superblock lock in llap_from_page for
2901 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2902 for llap_from_page call in ll_removepage prevents from taking the
2903 superblock lock for a soon died page.
2908 Description: Not check open intent error before release open handle
2909 Details : in some rare cases, the open intent error is not checked before
2910 release open handle, which may cause
2911 ASSERTION(open_req->rq_transno != 0), because it tries to release
2912 the failed open handle.
2917 Description: Set cat log bitmap only after create log success.
2918 Details : in some rare cases, the cat log bitmap is set too early. and it
2919 should be set only after create log success.
2923 Description: Accessing a block bevice can re-enable I/O when Lustre is
2924 tearing down a device.
2925 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2930 Description: service threads may hog cpus when there are a lot of requests
2931 Details : Insert cond_resched to give other threads a chance to use some CPU
2936 Description: the cat log was not initialized in recovery
2937 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2938 unlink log on mds will not be initialized until mds post
2939 recovery. And also in mds post recovery, the unlink log will
2940 initialization will be done asynchronausly, so there will be race
2941 between add unlink log and unlink log initialization.
2945 Description: brw_stats were being printed incorrectly
2946 Details : brw_stats were being printed as log2 but all of them were not
2947 recorded as log2. Also remove some code duplication arising from
2948 filter_tally_{read,write}.
2952 Frequency : rare, only in recovery.
2953 Description: ASSERTION(req->rq_type != LI_POISON) failed
2954 Details : imp_lock should be held while iterating over imp_sending_list for
2955 prevent destroy request after get timeout in ptlrpc_queue_wait.
2959 Description: replay-single.sh test 52 fails
2960 Details : A lock's skiplist need to be cleanup when it being unlinked
2961 from its resource list.
2965 Description: Short directio read returns full requested size rather than
2967 Details : Direct I/O operations should return actual amount of bytes
2968 transferred rather than requested size.
2970 Severity : enhancement
2972 Description: metadata RPC reduction (e.g. for rm performance)
2973 Details : decrease the amount of synchronous RPC between clients and servers
2974 by canceling conflicing lock before the operation on the client
2975 and packing thier handles into the main operation RPC to server.
2977 Severity : enhancement
2979 Description: Async OSC create to avoid the blocking unnecessarily.
2980 Details : If a OST has no remain object, system will block on the creating
2981 when need to create a new object on this OST. Now, ways use
2982 pre-created objects when available, instead of blocking on an
2983 empty osc while others are not empty. If we must block, we block
2984 for the shortest possible period of time.
2988 Description: improve handling recoverable errors
2989 Details : if request processig with error which can be recoverable on server
2990 request should be resend, otherwise page released from cache and
2993 Severity : enhancement
2995 Description: refine locking for avoid write wrong info into lov_objid file
2996 Details : fix possible races with add new target and write/update data in
2999 --------------------------------------------------------------------------------
3001 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3006 Frequency : on some architectures
3008 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3009 Details : Uninitialized flags sometimes cause configuration commands to
3013 Frequency : patchless clients only
3015 Description: missing __iget() symbol export
3016 Details : The __iget() symbol export is missing. To avoid the need for
3017 this on patchless clients the deathrow inode reaper is turned
3018 off, and we depend on the VM to clean up old inodes. This
3019 dependency was during via the fix for bug 12181.
3021 --------------------------------------------------------------------------------
3023 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3025 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3026 INTEROPERATE with older versions automatically. In many cases a
3027 special upgrade step is needed. Please read the
3028 user documentation before upgrading any part of a live system.
3029 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3030 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3031 * WARNING: Lustre configuration and startup changes are required with
3032 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3036 Severity : enhancement
3038 Description: MountConf
3039 Details : Lustre configuration is now managed via mkfs and mount
3040 commands instead of lmc and lconf. New obd types (MGS, MGC)
3041 are added for dynamic configuration management. See
3042 https://mail.clusterfs.com/wikis/lustre/MountConf for
3045 Severity : enhancement
3047 Description: dynamic OST addition
3048 Details : OSTs can now be added to a live filesystem
3050 Severity : enhancement
3052 Description: startup order invariance
3053 Details : MDTs and OSTs can be started in any order. Clients only
3054 require the MDT to complete startup.
3056 Severity : enhancement
3058 Description: parallel, asynchronous orphan cleanup
3059 Details : orphan cleanup is now performed in separate threads for each
3060 OST, allowing parallel non-blocking operation.
3062 Severity : enhancement
3064 Description: optimized stripe assignment
3065 Details : stripe assignments are now made based on ost space available,
3066 ost previous usage, and OSS previous usage, in order to try
3067 to optimize storage space and networking resources.
3069 Severity : enhancement
3071 Description: Permanently set tunables
3072 Details : All writable /proc/fs/lustre tunables can now be permanently
3073 set on a per-server basis, at mkfs time or on a live system.
3075 Severity : enhancement
3077 Description: Lustre message v2
3078 Details : Add lustre message format v2.
3080 Severity : enhancement
3082 Description: client OST exclusion list
3083 Details : Clients can be started with a list of OSTs that should be
3084 declared "inactive" for known non-responsive OSTs.
3088 Description: SPEC SFS validation failure on NFS v2 over lustre.
3089 Details : Changes the blocksize for regular files to be 2x RPC size,
3090 and not depend on stripe size.
3092 Severity : enhancement
3094 Description: Multiple MD RPCs in flight.
3095 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3096 To avoid overly-overloading MDS, introduce a limit on number of
3097 MDS RPCs in flight for a single client and add /proc controls
3098 to adjust this limit.
3100 Severity : enhancement
3102 Description: client read/write statistics
3103 Details : Add client read/write call usage stats for performance
3104 analysis of user processes.
3105 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3106 file access. extents_stats shows chunk size distribution.
3107 extents_stats_per_process show chunk size distribution per
3110 Severity : enhancement
3112 Description: mds statistics
3113 Details : Add detailed mds operations statistics in
3114 /proc/fs/lustre/mds/*/stats.
3118 Description: Failure of copying files with lustre special EAs.
3119 Details : Client side always return success for setxattr call for lustre
3120 special xattr (currently only "trusted.lov").
3125 Description: Refcount LNET uuids
3126 Details : The global LNET uuid list grew linearly with every startup;
3127 refcount repeated list entries instead of always adding to
3130 Severity : enhancement
3132 Description: Dynamic service threads
3133 Details : Within a small range, start extra service threads
3134 automatically when the request queue builds up.
3137 Frequency : mixed-endian client/server environments
3139 Description: mixed-endian crashes
3140 Details : The new msg_v2 system had some failures in mixed-endian
3143 Severity : enhancement
3145 Description: Easy OST removal
3146 Details : OSTs can be permanently deactivated with e.g. 'lctl
3147 conf_param lustre-OST0001.osc.active=0'
3149 Severity : enhancement
3151 Description: MGS proc entries
3152 Details : Added basic proc entries for the MGS showing what filesystems
3155 Severity : enhancement
3157 Description: provide MGS failover
3158 Details : Added config lock reacquisition after MGS server failover.
3160 Severity : enhancement
3162 Description: add Linux 2.4 support
3163 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3167 Description: a large application tries to do I/O to the same resource and dies
3168 in the middle of it.
3169 Details : Check the req->rq_arrival time after the call to
3170 ost_brw_lock_get(), but before we do anything about
3171 processing it & sending the BULK transfer request. This
3172 should help move old stale pending locks off the queue as
3173 quickly as obd_timeout.
3176 Frequency : when an incorrect nid is specified during startup
3178 Description: ptlrpc connect to non-existant node causes kernel crash
3179 Details : LNET can't be re-entered from an event callback, which
3180 happened when we expire a message after the export has been
3181 cleaned up. Instead, hand the zombie cleanup off to another
3184 Severity : enhancement
3186 Description: plain/inodebits lock performance improvement
3187 Details : Grouping plain/inodebits in granted list by their request modes
3188 and bits policy, thus improving the performance of search through
3192 Frequency : only if OST filesystem is corrupted
3194 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3195 Details : for a short time RPCs with bulk IO are in the replay list,
3196 but replay of bulk IOs is unimplemented. If the OST filesystem
3197 is corrupted due to disk cache incoherency and then replay is
3198 started it is possible to trip an assertion. Avoid putting
3199 committed RPCs into the replay list at all to avoid this issue.
3202 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3204 Description: System hang on startup
3205 Details : This bug allowed the liblustre (e.g. catamount) client to
3206 return to the app before handling all startup RPCs. This
3207 could leave the node unresponsive to lustre network traffic
3208 and manifested as a server ptllnd timeout.
3210 Severity : enhancement
3212 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3213 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3214 boolean to control whether to print peer debug info when a
3215 client's RPC times out.
3218 Frequency : only for kernels with patches from Lustre below 1.4.3
3220 Description: Remove old rdonly API
3221 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3224 Frequency : only for devices with external journals
3226 Description: Set external device read-only also
3227 Details : During a commanded failover stop, we set the disk device
3228 read-only while the server shuts down. We now also set any
3229 external journal device read-only at the same time.
3232 Frequency : when upgrading from 1.4 while trying to change parameters
3234 Description: The wrong (new) MDC name was used when setting parameters for
3235 upgraded MDT's. Also allows changing of OSC (and MDC)
3236 parameters if --writeconf is specified at tunefs upgrade time.
3239 Frequency : when setting specific ost indicies
3241 Description: QOS code breaks on skipped indicies
3242 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3243 created with --index need not be sequential.
3245 Severity : enhancement
3247 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3248 Details : The uninit_groups feature works in conjunction with the kernel
3249 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3250 up the pass1 processing of e2fsck. This is a read-only feature
3251 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3252 mount filesystems that have had this feature enabled.
3254 Severity : enhancement
3256 Description: Improve multi-block allocation algorithm to avoid fragmentation
3257 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3258 allocation locality and avoid filesystem fragmentation.
3260 ------------------------------------------------------------------------------
3262 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3264 * Support for kernels:
3265 2.4.21-47.0.1.EL (RHEL 3)
3266 2.6.5-7.283 (SLES 9)
3267 2.6.9-42.0.10.EL (RHEL 4)
3268 2.6.12.6 vanilla (kernel.org)
3269 2.6.16.27-0.9 (SLES 10)
3270 * Recommended e2fsprogs version: 1.39.cfs5
3272 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3276 Frequency : occasional, depends on client load and configuration
3277 Bugzilla : 12181, 12203
3278 Description: data loss for recently-modified files
3280 Details : In some cases it is possible that recently written or created
3281 files may not be written to disk in a timely manner (this should
3282 normally be within 30s unless client IO load is very high).
3283 The problem appears as zero-length files or files that are a
3284 multiple of 1MB in size after a client crash or client eviction
3285 that are missing data at the end of the file.
3287 This problem is more likely to be hit on clients where files are
3288 repeatedly created and unlinked in the same directory, clients
3289 have a large amount of RAM, have many CPUs, the filesystem has
3290 many OSTs, the clients are rebooted frequently, and/or the files
3291 are not accessed by other nodes after being written.
3293 The presence of the problem can be detected by looking at
3294 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3295 smaller than the second (nr_unused) then dirty files will not
3296 be flushed automatically to disk. "sync; sleep 10" should be
3297 run several times on the node before unmounting it to update
3298 Lustre (this is also safe to run on nodes without this problem).
3300 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3301 that can cause this same problem, so customers using that kernel
3302 also need to update the kernel in addition to Lustre. In order
3303 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3305 It is normal that files written just before a client crash (less
3306 than 30s) may not yet have been flushed to disk, even for local
3310 Frequency : frequent on thin XT3 nodes
3312 Description: UUID collision on thin XT3 Linux nodes
3313 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3314 randomly, since we relied on an insufficiently-seeded PRNG.
3319 Description: OSS hangs after "All ost request buffers busy"
3320 Details : A deadlock between quota and journal operations caused OSS
3321 hangs after printing "All ost request buffers busy."
3324 Frequency : always on liblustre builds
3326 Description: Cleanup compiler warnings on liblustre
3329 Frequency : always on liblustre builds on XT3
3331 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3335 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3336 > 32000 subdirectories
3337 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3338 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3339 subdirectory count crosses 32000. This will aid e2fsck to
3340 correctly handle more than 32000 subdirectories.
3343 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3345 Description: System hang on startup
3346 Details : This bug allowed the liblustre (e.g. catamount) client to
3347 return to the app before handling all startup RPCs. This
3348 could leave the node unresponsive to lustre network traffic
3349 and manifested as a server ptllnd timeout.
3351 Severity : enhancement
3353 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3354 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3355 boolean to control whether to print peer debug info when a
3356 client's RPC times out.
3361 Description: make O_SYNC working on 2.6 kernels
3362 Details : 2.6 kernels use different method for mark pages for write,
3363 so need add a code to lustre for O_SYNC work.
3368 Description: Failure to close file and release space on NFS
3369 Details : Put inode details into lock acquired in ll_intent_file_open.
3370 Use mdc_intent_lock in ll_intent_open to properly
3371 detect all kind of errors unhandled by mdc_enqueue.
3376 Description: proc file read during shutdown sometimes raced obd removal,
3378 Details : Add lock to prevent obd access after proc file removal.
3381 Frequency : Only for files larger than 4GB on 32-bit clients.
3383 Description: improperly doing page alignment of locks
3384 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3385 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3390 Description: RPCs being resent when they shouldn't be
3391 Details : Some RPCs that should not be resent are being resent. This
3392 can cause inconsistencies in the RPC state machine. Do not
3393 resend such requests.
3396 Frequency : rare, only with NFS export
3398 Description: Crash on NFS re-export node
3399 Details : under very unusual load conditions an assertion is hit in
3400 ll_intent_file_open()
3403 Frequency : only if OST filesystem is corrupted
3405 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3406 Details : for a short time RPCs with bulk IO are in the replay list,
3407 but replay of bulk IOs is unimplemented. If the OST filesystem
3408 is corrupted due to disk cache incoherency and then replay is
3409 started it is possible to trip an assertion. Avoid putting
3410 committed RPCs into the replay list at all to avoid this issue.
3415 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3416 Details : Large single O_DIRECT read and write calls can fail to allocate
3417 a sufficiently large buffer to process the request. In case of
3418 allocation failure the allocation is retried with a smaller
3419 buffer and broken into smaller requests.
3421 Severity : enhancement
3423 Description: Add -o localflock option to simulate old noflock behaviour.
3424 Details : This will achieve local-only flock/fcntl locks coherentness.
3429 Description: versioning check is incomplete
3430 Details : Checking the version difference of client vs. server, report
3431 error if the gap is too big.
3436 Description: add support PG_writeback bit
3437 Details : add support for PG_writeback bit for Lustre, for more carefull
3438 work with page cache in 2.6 kernel. This also fix some deadlocks
3439 and remove hack for work O_SYNC with 2.6 kernel.
3441 Severity : enhancement
3443 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3444 Details : The uninit_groups feature works in conjunction with the kernel
3445 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3446 up the pass1 processing of e2fsck. This is a read-only feature
3447 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3448 mount filesystems that have had this feature enabled.
3450 Severity : enhancement
3452 Description: Improve multi-block allocation algorithm to avoid fragmentation
3453 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3454 allocation locality and avoid filesystem fragmentation.
3456 ------------------------------------------------------------------------------
3458 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3460 * Support for kernels:
3461 2.6.9-42.0.3.EL (RHEL 4)
3462 2.6.5-7.276 (SLES 9)
3463 2.4.21-47.0.1.EL (RHEL 3)
3464 2.6.12.6 vanilla (kernel.org)
3465 2.6.16.21-0.8 (SLES10)
3466 * Recommended e2fsprogs version: 1.39.cfs2-0
3468 * The backwards-compatible /proc/sys/portals symlink has been removed
3469 in this release. Before upgrading, please ensure that you change
3470 any configuration scripts or /etc/sysctl.conf files that access
3471 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3472 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3473 in advance of the upgrade on any system running Lustre 1.4.6 or
3474 newer, since /proc/sys/lnet was added in that version.
3475 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3479 Frequency : only when quota is used
3481 Description: avoid scanning export list for quota master
3482 Details : Change the algorithms to avoid scanning export list in order
3483 to improve the efficiency.
3486 Frequency : MDS failover only, very rarely
3488 Description: "went back in time" messages on mds failover
3489 Details : The greatest transno may be lost when the current operation
3490 finishes with an error (transno==0) and the client's last_rcvd
3491 record is over-written. Save the greatest transno in the
3492 mds_last_transno for this case.
3495 Frequency : always for specific kernels and striping counts
3497 Description: client may get "Matching packet too big" without ACL support
3498 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3499 when trying to access files in certain configurations. The
3500 clients should in fact be denied when mounting because they do
3501 not understand ACLs.
3504 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3506 Description: many clients connecting with IO in progress causes connect timeouts
3507 Details : Avoid synchronous journal commits to avoid delays caused by many
3508 clients connecting/disconnecting when bulk IO is in progress.
3509 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3510 OST_IO_PORTAL to avoid delays behind potentially many pending
3514 Frequency : occasionally with multiple writers to a single file
3516 Description: shared writes to file may result in wrong size reported by stat()
3517 Details : Allow growing of kms when extent lock is cancelled
3520 Frequency : always with random mmap IO to multi-striped file
3522 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3523 Details : Only if the hole is at the end of OST object so that kms is too
3524 small. Fix is to increase kms accordingly in ll_nopage.
3527 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3529 Description: writes to a missing object would leak memory on the OST
3530 Details : If there is an inconsistency between the MDS and OST filesystems,
3531 such that the MDS references an object that doesn't exist, writes
3532 to that object will leak memory due to incorrect cleanup in the
3533 error handling path, eventually running out of memory on the OST.
3538 Description: Creating too long symlink causes lustre errors
3539 Details : Check symlink and name lengths before sending requests to MDS.
3542 Frequency : only if flock is enabled (not on by default)
3544 Description: posix locks not released on fd closure on 2.6.9+
3545 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3546 kernels, this caused such locks not to be released on fd close and
3547 then assertions on fs unmount about still used locks.
3550 Frequency : MDS failover only, very rarely
3552 Description: clients may get ASSERTION(granted_lock != NULL)
3553 Details : When request was taking a long time, and a client was resending
3554 a getattr by name lock request. The were multiple lock requests
3555 with the same client lock handle and
3556 mds_getattr_name->fixup_handle_for_resent_request found one of the
3557 lock handles but later failed with ASSERTION(granted_lock != NULL).
3562 Description: handle->h_buffer_credits > 0, assertion failure
3563 Details : h_buffer_credits is zero after truncate, causing assertion
3564 failure. This patch extends the transaction or creates a new
3568 Frequency : NFS re-export or patchless client
3569 Bugzilla : 11179, 10796
3570 Description: Crash on NFS re-export node (__d_move)
3571 Details : We do not want to hash the dentry if we don't have a lock.
3572 But if this dentry is later used in d_move, we'd hit uninitialised
3573 list head d_hash, so we just do this to init d_hash field but
3574 leave dentry unhashed.
3577 Frequency : NFS re-export or patchless client
3579 Description: NFS exports has problem with symbolic link
3580 Details : lustre client didn't properly install dentry when re-exported
3581 to NFS or running patchless client.
3584 Frequency : NFS re-export or patchless client
3586 Description: Various nfs/patchless fixes.
3587 Details : fixes reuse disconected alias for lookup process - this fixes
3588 warning "find_exported_dentry: npd != pd",
3589 fix permission error with open files at nfs.
3590 fix apply umask when do revalidate.
3593 Frequency : occasional
3595 Description: Crash on NFS re-export node
3596 Details : calling clear_page() on the wrong pointer triggered oops in
3597 generic_mapping_read().
3600 Frequency : rarely, using O_DIRECT IO
3602 Description: unaligned directio crashes client with LASSERT
3603 Details : check for unaligned buffers before trying any requests.
3606 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3608 Description: stale data returned from RAID cache
3609 Details : If only a small amount of IO is done to the RAID device before
3610 reading it again it is possible to get stale data from the RAID
3611 cache instead of reading it from disk.
3614 Frequency : always for sles10 kernel
3616 Description: sles10 support
3617 Details : ll_follow_link: compile fixes and using of nd_set_link
3618 under newer kernels.
3621 Frequency : depends on arch, kernel and compiler version, always on sles10
3624 Description: recursive or deep enough symlinks cause stack overflow
3625 Details : getting rid of large stack-allocated variable in
3629 Frequency : depends on hardware
3631 Description: lustre write performance loss in the SLES10 kernel
3632 Details : the performance loss is caused by using of write barriers in the
3633 ext3 code. The SLES10 kernel turns barrier support on by
3634 default. The fix is to undo that change for ldiskfs.
3636 ------------------------------------------------------------------------------
3638 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3640 * Support for kernels:
3641 2.6.9-42.0.3EL (RHEL 4)
3642 2.6.5-7.276 (SLES 9)
3643 2.4.21-47.0.1.EL (RHEL 3)
3644 2.6.12.6 vanilla (kernel.org)
3648 Frequency : quota enabled and large files being deleted
3650 Description: releasing more than 4GB of quota at once hangs OST
3651 Details : If a user deletes more than 4GB of files on a single OST it
3652 will cause the OST to spin in an infinite loop. Release
3653 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3658 Description: statfs data retrieved from /proc may be stale or zero
3659 Details : When reading per-device statfs data from /proc, in the
3660 {kbytes,files}_{total,free,avail} files, it may appear
3661 as zero or be out of date.
3664 Frequency : systems with MD RAID1 external journal devices
3666 Description: lconf's call to blkid is confused by RAID1 journal devices
3667 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3668 of returning all block devices that match the journal UUID.
3671 Frequency : always, for aggregate stripe size over 4GB
3673 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3674 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3675 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3678 Frequency : always if "lfs find" used on a local file/directory
3680 Description: "lfs find" segfaults if used on a local file/directory
3681 Details : The case where a directory component was not specified wasn't
3682 handled correctly. Handle this properly.
3685 Frequency : always on ppc64
3687 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3688 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3689 perform endianness conversion.
3692 Frequency : rarely (truncate to non-zero file size after write under load)
3693 Bugzilla : 10730, 10687
3694 Description: Files padded with zeros to next 4K multiple
3695 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3696 it is possible that files that are truncated to a non-zero size
3697 immediately after being written are filled with zero bytes beyond
3698 the truncated size. No file data is lost.
3700 Severity : enhancement
3702 Description: Allow recovery/failover for liblustre clients.
3703 Details : liblustre clients were unaware of failover configurations until
3706 Severity : enhancement
3708 Description: user file locks should fail when not mounting with flock option
3709 Details : Set up an error-returning stub in ll_file_operations.lock field
3710 to prevent incorrect behaviour when client is mounted without
3711 flock option. Also, set up properly f_op->flock field for
3715 Frequency : always on ia64
3717 Description: "lfs df" loops on printing out MDS statfs information
3718 Details : The obd_ioctl_data was not initialized and in some systems
3719 this caused a failure during the ioctl that did not return
3720 an error. Initialize the struct and return an error on failure.
3723 Frequency : SLES 9 only
3725 Description: Error of copying files with lustre special EAs as root
3726 Details : Client side always return success for setxattr call for lustre
3727 special xattr (currently only "trusted.lov").
3730 Frequency : rarely on clusters with both ia64+i386 clients
3732 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3733 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3734 clients are doing concurrent writes to the same file it is
3735 possible that the ia64 clients may LASSERT because the OST
3736 extent locks are not PAGE_SIZE aligned. Ensure that grown
3737 locks are always aligned on the request boundary.
3740 Frequency : specific use, occasional
3742 Description: Overwriting in use executable truncates on-disk binary image
3743 Details : If one node attempts to overwrite an executable in use by
3744 another node, we now correctly return ETXTBSY instead of
3745 truncating the file.
3747 Severity : enhancement
3749 Description: Async OSC create to avoid the blocking unnecessarily.
3750 Details : If a OST has no remain object, system will block on the creating
3751 when need to create a new object on this OST. Now, ways use
3752 pre-created objects when available, instead of blocking on an
3753 empty osc while others are not empty. If we must block, we block
3754 for the shortest possible period of time.
3759 Description: chmod on Lustre root is propagated to other clients
3760 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3761 invalid by the follow_mount time.
3764 Frequency : liblustre clients only
3766 Description: Race in 'instant cancel' lock handling could lead to such locks
3767 never to be granted in case of SMP MDS
3768 Details : Do not destroy not yet granted but cbpending locks in
3772 Frequency : replay/resend of open
3774 Description: non null lock assetion failure in mds_intent_policy
3775 Details : Trying to replay/resend lockless open requests resulted in
3776 mds_open() returning 0 with no lock. Now it sets a flag if
3777 a lock is going to be returned.
3779 Severity : enhancement
3781 Description: Checksum enhancements
3782 Details : New checksum enhancements allow for resending RPCs that failed
3785 Severity : enhancement
3787 Description: Tunables on number of dirty pages in cacche
3788 Details : Allow to set limit on number of dirty pages cached.
3793 Description: client crash on unmount - lock still has references
3794 Details : In some error handling cases it was possible to leak a lock
3795 reference on a client while accessing a file. This was not
3796 harmful to the client during operation, but would cause the
3797 client to crash when the filesystem is unmounted.
3800 Frequency : specific case, rare
3802 Description: ETXTBSY on mds though file not in use by client
3803 Details : ETXTBSY is no longer incorrectly returned when attempting to
3804 chmod or chown a directory that the user previously tried to
3805 execute or a currently-executing binary.
3808 Frequency : extremely rare except on liblustre-based clients
3810 Description: Lustre space not freed when files are deleted
3811 Details : Clean up open-unlinked files after client eviction. Previously
3812 the unlink was skipped and the files remained as orphans.
3817 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3818 Details : In some cases it was possible to send a blocking callback to
3819 a client doing a glimpse, even though that client didn't get
3820 a lock granted. When the glimpse lock is cancelled on the OST
3821 the freed lock is left on the waiting list and corrupted the list.
3824 Frequency : all core dumps
3826 Description: Broke core dumps to lustre
3827 Details : Negative dentry may be unhashed if parent does not have UPDATE
3828 lock, but some callers, e.g. do_coredump, expect dentry to be
3829 hashed after successful create, hash it in ll_create_it.
3831 ------------------------------------------------------------------------------
3833 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3835 * Support for kernels:
3836 2.6.9-42.0.2.EL (RHEL 4)
3837 2.6.5-7.276 (SLES 9)
3838 2.4.21-40.EL (RHEL 3)
3839 2.6.12.6 vanilla (kernel.org)
3843 Frequency : always on RHEL 3
3845 Description: Number of open files grows over time
3846 Details : The number of open files grows over time, whether or not
3847 Lustre is started. This was due to a filp leak introduced
3848 by one of our kernel patches.
3850 ------------------------------------------------------------------------------
3852 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3854 * Support for kernels:
3855 2.6.9-42.EL (RHEL 4)
3856 2.6.5-7.276 (SLES 9)
3857 2.4.21-40.EL (RHEL 3)
3858 2.6.12.6 vanilla (kernel.org)
3863 Bugzilla : 5719, 9635, 9792, 9684
3864 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3865 Details : If a server is under heavy load and cannot reply to new
3866 connection requests before the client resends the (re)connect,
3867 the connection handling code can behave badly if two service
3868 threads are concurrently handing separate (re)connections from
3869 the same client. Add better locking to the connection handling
3870 code, and ensure that only a single connection will be processed
3871 for a given client UUID, even if the lock is dropped.
3873 Severity : enhancement
3875 Description: add TCP zero-copy support to kernel
3876 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3877 sends if the hardware supports scatter-gather and checksumming.
3878 This allows socklnd to do client-write and server-read more
3879 efficiently and reduce CPU utilization from skbuf copying.
3882 Frequency : only if NFS exporting from client
3884 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3885 Details : If setting attributes on a file created under NFS that had
3886 never been opened it would be possible to oops the client
3887 if the file had no objects.
3890 Frequency : always for liblustre
3892 Description: liblustre client does MDS+OSTs setattr RPC for each write
3893 Details : When doing a write from a liblustre client, the client
3894 incorrectly issued an RPC to the MDS and each OST the file was
3895 striped over in order to update the timestamps. When writing
3896 with small chunks and many clients this could overwhelm the MDS
3897 with RPCs. In all cases it would slow down the write because
3898 these RPCs are unnecessary.
3900 Severity : enhancement
3902 Description: allow number of MDS service threads to be changed at module load
3903 Details : It is now possible to change the number of MDS service threads
3904 running. Adding "options mds mds_num_threads={N}" to the MDS's
3905 /etc/modprobe.conf will set the number of threads for the next
3906 time Lustre is restarted (assuming the "mds" module is also
3907 reloaded at that time). The default number of threads will
3908 stay the same, 32 for most systems.
3913 Description: OST crash if filesystem is unformatted or corrupt
3914 Details : If an OST is started on a device that has never been formatted
3915 or if the filesystem is corrupt and cannot even mount then the
3916 error handling cleanup routines would dereference a NULL pointer.
3921 Description: NULL pointer deref in llap_from_page.
3922 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3923 incorrect) mapping if the page was truncated/reclaimed while it was
3924 searched for. Check for this condition and skip such pages when
3925 doing readahead. Introduce extra check to llap_from_page() to
3926 verify page->mapping->host is non-NULL (so page is not anonymous).
3929 Frequency : Sometimes when using sys_sendfile
3931 Description: "page not covered by a lock" warnings from ll_readpage
3932 Details : sendfile called ll_readpage without right page locks present.
3933 Now we introduced ll_file_sendfile that does necessary locking
3934 around call to generic_file_sendfile() much like we do in
3938 Frequency : with certain MDS communication failures at client mount time
3940 Description: NULL pointer deref after failed client mount
3941 Details : a client connection request may delayed by the network layer
3942 and not be sent until after the PTLRPC layer has timed out the
3943 request. If the client fails the mount immediately it will try
3944 to clean up before the network times out the request. Add a
3945 reference from the request import to the obd device and delay
3946 the cleanup until the network drops the request.
3949 Frequency : occasionally during client (re)connect
3951 Description: assertion failure during client (re)connect
3952 Details : processing a client connection request may be delayed by the
3953 client or server longer than the client connect timeout. This
3954 causes the client to resend the connection request. If the
3955 original connection request is replied in this interval, the
3956 client may trip an assertion failure in ptlrpc_connect_interpret()
3957 which thought it would be the only running connect process.
3960 Frequency : only with obd_echo servers and clients that are rebooted
3962 Description: kernel BUG accessing uninitialized data structure
3963 Details : When running an obd_echo server it did not start the ping_evictor
3964 thread, and when a client was evicted an uninitialized data
3965 structure was accessed. Start the ping_evictor in the RPC
3966 service startup instead of the OBD startup.
3968 Severity : enhancement
3969 Bugzilla : 10193 (patchless)
3970 Description: Remove dependency on various unexported kernel interfaces.
3971 Details : No longer need reparent_to_init, exit_mm, exit_files,
3972 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3975 Frequency : rare (only users of deprecated and unsupported LDAP config)
3977 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3978 Details : LDAP apparently contains 'lustreName' attributes instead of
3979 'name'. A simple remapping of the name is sufficient.
3982 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3984 Description: waiting_locks_callback trips kernel BUG if client is evicted
3985 Details : Running with the dump_on_timeout debug flag turned on makes
3986 it possible that the waiting_locks_callback() can try to dump
3987 the Lustre kernel debug logs from an interrupt handler. Defer
3988 this log dumping to the expired_lock_main() thread.
3990 Severity : enhancement
3992 Description: Support NFS exporting on 2.6 kernels.
3993 Details : Implement non-rawops metadata methods for NFS server to use without
3994 changing NFS server code.
3997 Frequency : very rare (synthetic metadata workload only)
3999 Description: two racing renames might cause an MDS thread to deadlock
4000 Details : Running the "racer" program may cause one MDS thread to rename
4001 a file from being the source of a rename to being the target of
4002 a rename at exactly the same time that another thread is doing
4003 so, and the second thread has already enqueued these locks after
4004 doing a lookup of the target and is trying to relock them in
4005 order. Ensure that we don't try to re-lock the same resource.
4008 Frequency : only very large systems with liblustre clients
4010 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4011 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4012 all OSTs in parallel. This allows the eviction of stale liblustre
4013 clients to proceed much faster than if they were done in series,
4014 and also offers similar improvements for other set_info RPCs.
4019 Description: excessive CPU usage during initial read phase on client
4020 Details : During the initial read phase on a client, it would agressively
4021 retry readahead on the file, consuming too much CPU and impacting
4022 performance (since 1.4.5.8). Improve the readahead algorithm
4023 to avoid this, and also improve some other common cases (read
4024 of small files in particular, where "small" is files smaller than
4025 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4030 Description: MDS crash when receiving packet with unknown intent.
4031 Details : Do not LBUG in unknown intent case, just return -EFAULT
4033 Severity : enhancement
4034 Bugzilla : 9293, 9385
4035 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4036 Details : Do not serialize getattr (non-intent version) and statfs.
4039 Frequency : occasional, when OST network is overloaded/intermittent
4041 Description: client evicted by OST after bulk IO timeout
4042 Details : If a client sends a bulk IO request (read or write) the OST
4043 may evict the client if it is unresposive to its data GET/PUT
4044 request. This is incorrect if the network is overloaded (takes
4045 too long to transfer the RPC data) or dropped the OST GET/PUT
4046 request. There is no need to evict the client at all, since
4047 the pinger and/or lock callbacks will handle this, and the
4048 client can restart the bulk request.
4051 Frequency : Always when mmapping file with no objects
4053 Description: client crashes when mmapping file with no objects
4054 Details : Check that we actually have objects in a file before doing any
4055 operations on objects in ll_vm_open, ll_vm_close and
4061 Description: Request leak when working with deleted CWD
4062 Details : Introduce advanced request refcount tracking for requests
4063 referenced from lustre intent.
4065 Severity : Enhancement
4067 Description: Cache open file handles on client.
4068 Details : MDS now will return special lock along with openhandle, if
4069 requested and client is allowed to hold openhandle, even if unused,
4070 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4071 closing files for every read/write openration.
4073 Severity : Enhancement
4075 Description: Cache open negative dentries on client when possible.
4076 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4077 negative dentries on lock revocation.
4082 Description: Remounting a client read-only wasn't possible with a zconf mount
4083 Details : It wasn't possible to remount a client read-only with llmount.
4085 Severity : enhancement
4086 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4087 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4088 MPICH is included to add Lustre-specific ADIO interfaces.
4089 This is based closely on the UFS ADIO layer and only differs
4090 in file creation, in order to allow the OST striping to be set.
4091 This is user-contributed code and not supported by CFS.
4096 Description: extended inode attributes (immutable, append-only) work improperly
4097 when 2.4 and 2.6 kernels are used on client/server or vice versa
4098 Details : Introduce kernel-independent values for these flags.
4100 Severity : enhancement
4103 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4104 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4105 granularity. Now, Fractional MB granularity is be supported,
4106 this is very useful for low memory system.
4108 Severity : enhancement
4110 Description: Getattr by fid
4111 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4112 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4113 have locks and avoids getattr rpc requests for it on every stat.
4116 Frequency : Always, for filesystems larger than 2TB
4118 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4119 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4120 filesystems larger than 2TB to be mounted with mballoc enabled.
4123 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4125 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4126 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4127 filesystems larger than 2TB is mounted then the kernel will
4128 silently corrupt the start of the filesystem. CONFIG_LBD is
4129 enabled for all CFS-supported kernels, but the possibility of
4130 this happening with a modified kernel config exists.
4132 Severity : enhancement
4134 Description: add client O_DIRECT support for 2.6 kernels
4135 Details : It is now possible to do O_DIRECT reads and writes to files
4136 in the Lustre client mountpoint on 2.6 kernel clients.
4138 Severity : enhancement
4140 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4141 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4142 parallel, not waiting for response from every OST before sending
4143 a rpc to the next OST.
4148 Description: setattr vs write race when updating file timestamps
4149 Details : Client processes that update a file timestamp into the past
4150 right after writing to the file (e.g. tar) it is possible that
4151 the updated file modification time can be reset to the current
4152 time due to a race between processing the setattr and write RPC.
4154 Severity : enhancement
4156 Description: Bring 'lfs find' closer in line with regular Linux find.
4157 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4158 -print0 options and obtains all the needed info through the lustre
4161 Severity : enhancement
4163 Description: support up to 1024 configured devices on one node
4164 Details : change obd_dev array from statically allocated to dynamically
4165 allocated structs as they are first used to reduce memory usage
4170 Description: Flush dirty partially truncated pages during truncate
4171 Details : Immediatelly flush partially truncated pages in filter_setattr,
4172 this way we completely avoid having any pages in page cache on OST
4173 and can retire ugly workarounds during writes to flush such pages.
4178 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4179 Details : protect inode from truncation within vfs_unlink() context
4180 just take a reference before calling vfs_unlink() and release it
4181 when parent's i_sem is free.
4186 Description: last_id value checked outside lock on OST caused LASSERT failure
4187 Details : If there were multiple MDS->OST object precreate requests in
4188 flight, it was possible that the OST's last object id was checked
4189 outside a lock and incorrectly tripped an assertion. Move checks
4190 inside locks, and discard old precreate requests.
4193 Frequency : always, if extents are used on OSTs
4195 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4196 Details : OSTs using the extents format would not zero the high 16 bits of
4197 the index physical block number. This is not a problem for any
4198 OST filesystems smaller than 16TB, and no kernels support ext3
4199 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4200 new/modified files) and can be fixed for existing filesystems
4201 with e2fsprogs-1.39-cfs1.
4206 Description: import connection selection may be incorrect if timer wraps
4207 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4208 import connections to be ignored if the 32-bit jiffies counter
4209 wraps. Use a 64-bit jiffies counter.
4212 Frequency : very large clusters immediately after boot
4214 Description: LNET request buffers exhausted under heavy short-term load
4215 Details : If a large number of client requests are generated on a service
4216 that has previously never seen so many requests it is possible
4217 that the request buffer growth cannot keep up with the spike in
4218 demand. Instead of dropping incoming requests, they are held in
4219 the LND until the RPC service can accept more requests.
4222 Frequency : Sometimes during replay
4224 Description: Assertion failure in ll_local_open after replay.
4225 Details : If replay happened on an open request reply before we were able
4226 to set replay handler, reply will become not swabbed tripping the
4227 assertion in ll_local_open. Now we set the handler right after
4228 recognising of open request
4231 Frequency : very rare
4233 Description: kernel reports "badness in vsnprintf"
4234 Details : Reading from the "recovery_status" /proc file in small chunks
4235 may cause a negative length in lprocfs_obd_rd_recovery_status()
4236 call to vsnprintf() (which is otherwise harmless). Exit early
4237 if there is no more space in the output buffer.
4239 Severity : enhancement
4241 Description: clear OBD RPC statistics by writing to them
4242 Details : It is now possible to clear the OBD RPC statistics by writing
4243 to the "stats" file.
4248 Description: Client mtime is not the same on different clients after utimes
4249 Details : In some cases, the client was using the utimes() syscall on
4250 a file cached on another node. The clients now validate the
4251 ctime from the MDS + OSTs to determine which one is right.
4256 Description: Inability to activate failout mode
4257 Details : lconf script incorrectly assumed that in pythong string's numeric
4258 value is used in comparisons.
4261 Frequency : always with multiple stripes per file
4263 Description: Inefficient object allocation for mutli-stripe files
4264 Details : When selecting which OSTs to stripe files over, for files with
4265 a stripe count that divides evenly into the number of OSTs,
4266 the MDS is always picking the same starting OST for each file.
4267 Return the OST selection heuristic to the original design.
4272 Description: mount failures may take full timeout to return an error
4273 Details : Under some heavy load conditions it is possible that a
4274 failed mount can wait for the full obd_timeout interval,
4275 possibly several minutes, before reporting an error.
4276 Instead return an error as soon as the status is known.
4278 Frequency : quota enabled and large files being deleted
4280 Description: releasing more than 4GB of quota at once hangs OST
4281 Details : If a user deletes more than 4GB of files on a single OST it
4282 will cause the OST to spin in an infinite loop. Release
4283 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4288 Description: statfs data retrieved from /proc may be stale or zero
4289 Details : When reading per-device statfs data from /proc, in the
4290 {kbytes,files}_{total,free,avail} files, it may appear
4291 as zero or be out of date.
4294 Frequency : systems with MD RAID1 external journal devices
4296 Description: lconf's call to blkid is confused by RAID1 journal devices
4297 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4298 of returning all block devices that match the journal UUID.
4301 Frequency : always, for aggregate stripe size over 4GB
4303 Description: assertion fails when trying to use 4GB stripe size
4304 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4305 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4308 Frequency : always on ppc64
4310 Description: the first write on an ext3 filesystem with mballoc got stuck
4311 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4312 perform endianness conversion.
4314 ------------------------------------------------------------------------------
4316 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4318 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4319 INTEROPERATE with older versions automatically. Please read the
4320 user documentation before upgrading any part of a live system.
4321 * WARNING: Lustre networking configuration changes are required with
4322 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4325 * Support for kernels:
4326 2.6.9-22.0.2.EL (RHEL 4)
4327 2.6.5-7.244 (SLES 9)
4328 2.6.12.6 vanilla (kernel.org)
4331 Severity : enhancement
4332 Bugzilla : 7981/8208
4333 Description: Introduced Lustre Networking (LNET)
4334 Details : LNET is new networking infrastructure for Lustre, it includes
4335 a reorganized network configuration mode (see the user
4336 documentation for full details) as well as support for routing
4337 between different network fabrics. Lustre Networking Devices
4338 (LNDs) for the supported network fabrics have also been
4339 created for this new infrastructure.
4341 Severity : enhancement
4342 Description: Introduced Access control lists
4343 Details : clients can set ACLs on files and directories in order to have
4344 more fine-grained permissions than the standard Unix UGO+RWX.
4345 The MDS must be started with the "-o acl" mount option.
4347 Severity : enhancement
4348 Description: Introduced filesystem quotas
4349 Details : Administrators may now establish per-user quotas on the
4352 Severity : enhancement
4354 Description: Configuration change for the XT3
4355 The PTLLND is now used to run Lustre over Portals on the XT3
4356 The configure option(s) --with-cray-portals are no longer used.
4357 Rather --with-portals=<path-to-portals-includes> is used to
4358 enable building on the XT3. In addition to enable XT3 specific
4359 features the option --enable-cray-xt3 must be used.
4364 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4365 Details : A race between allocating a new llcd and re-getting the llcd_lock
4366 allowed another thread to grab newly-allocated llcd.
4368 Severity : enhancement
4370 Description: 2.6 OST async journal commit and locking fix to improve performance
4371 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4372 2.6 kernels are now async as they already were in 2.4 kernels so
4373 that they can commit concurrently with the network bulk transfer.
4374 For block-allocated files the filter allocation semaphore is held
4375 to avoid filesystem fragmentation during allocation. BKL lock
4376 removed for 2.6 xattr operations where it is no longer needed.
4381 Description: lconf incorrectly determined whether two IP networks could talk
4382 Details : In some more complicated routing and multiple-network
4383 configurations, lconf will avoid trying to make a network
4384 connection to a disjoint part of the IP space. It was doing the
4385 math incorrectly for one set of cases.
4390 Description: Fix for potential infinite loop processing records in an llog.
4391 Details : If an llog record is corrupted/zeroed, it is possible to loop
4392 forever in llog_process(). Validate the llog record length
4393 and skip the remainder of the block on error.
4396 Frequency : occasional (liblustre only)
4398 Description: liblustre could not open files whose last component is a symlink
4399 Details : sysio_path_walk() would incorrectly pass the open intent to
4400 intermediate path components.
4403 Frequency : rare (liblustre only with non-standard tuning)
4404 Bugzilla : 7201 (7350)
4405 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4406 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4407 after it has been released, only do this for referenced locks
4409 Severity : enhancement
4411 Description: specifying an (invalid) directory default stripe_size of -1
4412 would reset the directory default striping
4413 Details : stripe_size -1 was used internally to signal directory stripe
4414 removal, now use "all default" to signal dir stripe removal
4415 as a directory striping of "all default" is not useful
4418 Frequency : common for large clusters running liblustre clients
4420 Description: doing an ls when liblustre clients are running is slow
4421 Details : sending a glimpse AST to a liblustre client waits for every AST
4422 to time out, as liblustre clients will not respond. Since they
4423 cannot cache data we refresh the OST lock LVB from disk instead.
4425 Severity : enhancement
4427 Description: doing an ls at the same time as file IO can be slow
4428 Details : enqueue and other "small" requests can be blocked behind many
4429 large IO requests. Create a new OST IO portal for non-IO
4430 requests so they can be processed faster.
4433 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4435 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4436 created file with mode 000
4437 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4438 allow an ftruncate() as a non-root user to a file with mode 000.
4439 Lustre now respects this flag to disable mode checks when
4440 truncating a file owned by the user
4443 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4446 Description: Revoking locks from clients that went dead or catatonic might take
4448 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4449 cancellation of such locks instant on servers without waiting for
4450 any reply from clients. Clients drops these locks when cancel
4451 notification from server is received without replying.
4454 Frequency : liblustre-only, when liblustre client dies or becomes busy
4456 Description: Doing ls on Linux clients can take a long time with active
4458 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4459 granting such locks to it in the first place if possible. Locks
4460 are taken by proxy on the OST during the read or write and
4461 dropped immediately afterward. Add connect flags handling, do
4462 not grant locks to liblustre clients for glimpse ASTs.
4464 Severity : enhancement
4466 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4467 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4468 sequential reads and behaves badly with applications doing random
4469 reads. Improve it by reading ahead at least read region, and
4470 avoiding excessive large RPC for small reads.
4472 Severity : enhancement
4474 Description: Creating more than 1000 files for a single job may cause a load
4475 imbalance on the OSTs if there are also a large number of OSTs.
4476 Details : qos_prep_create() uses an OST index reseed value that is an
4477 even multiple of the number of available OSTs so that if the
4478 reseed happens in the middle of the object allocation it will
4479 still utilize the OSTs as uniformly as possible.
4484 Description: OST or MDS may oops in ping_evictor_main()
4485 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4486 but doesn't restart at beginning of obd_exports_timed list
4489 Severity : enhancement
4491 Description: improve by-nid export eviction on the MDS and OST
4492 Details : allow multiple exports with the same NID to be evicted at one
4493 time without re-searching the exports list.
4496 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4498 Description: MDS may oops in groups_free()
4499 Details : in rare race conditions a newly allocated group_info struct is
4500 freed again, and this can be NULL. The 2.4 compatibility code
4501 for groups_free() checked for a NULL pointer, but 2.6 did not.
4504 Frequency : common for liblustre clients doing little filesystem IO
4505 Bugzilla : 9352, 7313
4506 Description: server may evict liblustre clients accessing contended locks
4507 Details : if a client is granted a lock or receives a completion AST
4508 with a blocking AST already set it would not reply to the AST
4509 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4512 Frequency : lfs setstripe, only systems with more than 160 OSTs
4514 Description: unable to set striping with a starting offset beyond OST 160
4515 Details : llapi_create_file() incorrectly limited the starting stripe
4516 index to the maximum single-file stripe count.
4519 Frequency : LDAP users only
4521 Description: lconf did not handle in-kernel recovery with LDAP properly
4522 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4524 Severity : enhancement
4526 Description: bind OST threads to NUMA nodes to improve performance
4527 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4528 node and do their allocations there to localize memory access
4530 Severity : enhancement
4532 Description: llmount can determine client NID directly from Myrinet (GM)
4533 Details : the client NID code from gmnalnid was moved directly into
4534 llmount, removing the need to use this or specifying the
4535 client NID explicitly when mounting GM clients with zeroconf
4538 Frequency : if client is started with down MDS
4540 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4541 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4542 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4543 ptlrpc_import_delayed_req() because we didn't check if the
4544 request was interrupted, and we also didn't break out of the
4545 event loop if there was no timeout
4550 Description: data loss during non-page-aligned writes to a single file from
4551 both multiple nodes and multiple threads on one node at same time
4552 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4553 inconsistency led to false short-reads, that were cached and later
4554 used by ->prepare_write() to fill in partially written page,
4555 leading to data loss.
4558 Frequency : always, if lconf --abort_recovery used
4560 Description: lconf --abort_recovery fails with 'Operation not supported'
4561 Details : lconf was attempting to abort recovery on the MDT device and not
4564 Severity : enhancement
4566 Description: remove cleanup logs
4567 Details : replace lconf-generated cleanup logs with lustre internal
4568 cleanup routines. Eliminates the need for client-cleanup and
4571 Severity : enhancement
4573 Description: add support for EAs (user and system) on lustre filesystems
4574 Details : it is now possible to store extended attributes in the Lustre
4575 client filesystem, and with the user_xattr mount option it
4576 is possible to allow users to store EAs on their files also
4578 Severity : enhancement
4580 Description: Add possibility (config option) to show minimal available OST free
4582 Details : When compiled with --enable-mindf configure option, statfs(2)
4583 (and so, df) will return least minimal free space available from
4584 all OSTs as amount of free space on FS, instead of summary of
4585 free spaces of all OSTs.
4587 Severity : enhancement
4589 Description: do not expand extent locks acquired on OST-side
4590 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4591 by server: they are not cached anyway.
4594 Frequency : when mmap is used/binaries executed from Lustre
4596 Description: Unmmap pages before throwing them away from read cache.
4597 Details : llap_shrink cache now attempts to unmap pages before discarding
4598 them (if unmapping failed - do not discard). SLES9 kernel has
4599 extra checks that trigger if this unmapping is not done first.
4604 Description: lconf didn't resolve symlinks before checking to see whether a
4605 given mountpoint was already in use
4608 Frequency : when migrating failover services
4609 Bugzilla : 6395, 9514
4610 Description: When migrating a subset of services from a node (e.g. failback
4611 from a failover service node) the remaining services would
4612 time out and evict clients.
4613 Details : lconf --force (implied by --failover) sets the global obd_timeout
4614 to 5 seconds in order to quickly disconnect, but this caused
4615 other RPCs to time out too quickly. Do not change the global
4616 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4618 Severity : enhancement
4619 Frequency : if MDS is started with down OST
4620 Bugzilla : 9439,5706
4621 Description: Allow startup/shutdown of an MDS without depending on the
4622 availability of the OSTs.
4623 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4624 Add appropriate locking and lov-osc refcounts for safe
4625 cleaning. Add osc abort_inflight calls in case the
4626 synchronize never started.
4629 Frequency : occasional (Cray XT3 only)
4631 Description: root not authorized to access files in CRAY_PORTALS environment
4632 Details : The client process capabilities were not honoured on the MDS in
4633 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4634 been accessed by an authorized user then root was able to access
4635 the file on the local client also. The root user capabilities
4636 are now allowed on the MDS, as this environment has secure UID.
4639 Frequency : occasional
4641 Description: ldiskfs "too long searching" message happens too often
4642 Details : A debugging message (otherwise harmless) prints too often on
4643 the OST console. This has been reduced to only happen when
4644 there are fragmentation problems on the filesystem.
4649 Description: Division by zero in statfs when all OSCs are inactive
4650 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4651 lov_statfs divides by value returned by lov_get_stripecnt().
4655 Bugzilla : 9489, 3273
4656 Description: First write from each client to each OST was only 4kB in size,
4657 to initialize client writeback cache, which caused sub-optimal
4658 RPCs and poor layout on disk for the first writen file.
4659 Details : Clients now request an initial cache grant at (re)connect time
4660 and so that they can start streaming writes to the cache right
4661 away and always do full-sized RPCs if there is enough data.
4662 If the OST is rebooted the client also re-establishes its grant
4663 so that client cached writes will be honoured under the grant.
4668 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4669 Details : Now I/O RPCs go to different portal number and (presumably) fast
4670 lock requests (and glimses) and other RPCs get their own service
4671 threads pool that should be able to service those RPCs
4674 Severity : enhancement
4676 Description: Ability to exchange lustre version between client and servers and
4677 issue warnings at client side if client is too old. Also for
4678 liblustre clients there is ability to refuse connection of too old
4680 Details : New 'version' field is added to connect data structure that is
4681 filled with version info. That info is later checked by server and
4685 Frequency : rare, liblustre only.
4686 Bugzilla : 9296, 9581
4687 Description: Two simultaneous writes from liblustre at offset within same page
4688 might proceed at the same time overwriting eachother with stale
4690 Details : I/O lock withing llu_file_prwv was released too early, before data
4691 actually was hitting the wire. Extended lock-holding time until
4692 server acknowledges receiving data.
4695 Frequency : extremely rare. Never observed in practice.
4697 Description: avoid generating lustre_handle cookie of 0.
4698 Details : class_handle_hash() generates handle cookies by incrementing
4699 global counter, and can hit 0 occasionaly (this is unlikely, but
4700 not impossible, because initial value of cookie counter is
4701 selected randonly). Value of 0 is used as a sentinel meaning
4702 "unassigned handle" --- avoid it. Also coalesce two critical
4703 sections in this function into one.
4705 Severity : enhancement
4707 Description: allow liblustre clients to delegate truncate locking to OST
4708 Details : To avoid overhead of locking, liblustre client instructs OST to
4709 take extent lock in ost_punch() on client's behalf. New connection
4710 flag is added to handle backward compatibility.
4712 Severity : enhancement
4713 Bugzilla : 4928, 7341, 9758
4714 Description: allow number of OST service threads to be specified
4715 Details : a module parameter allows the number of OST service threads
4716 to be specified via "options ost ost_num_threads={N}" in the
4717 OSS's /etc/modules.conf or /etc/modprobe.conf.
4721 Bugzilla : 6146, 9635, 9895
4722 Description: servers crash with bad pointer in target_handle_connect()
4723 Details : In rare cases when a client is reconnecting it was possible that
4724 the connection request was the last reference for that export.
4725 We would temporarily drop the export reference and get a new
4726 one, but this may have been the last reference and the export
4727 was just destroyed. Get new reference before dropping old one.
4729 Severity : enhancement
4730 Frequency : if client is started with failover MDS
4732 Description: Allow multiple MDS hostnames in the mount command
4733 Details : Try to read the configuration from all specified MDS
4734 hostnames during a client mount in case the "primary"
4737 Severity : enhancement
4739 Description: Stop sending data to evicted clients as soon as possible.
4740 Details : Check if the client we are about to send or are sending data to
4741 was evicted already. (Check is done every second of waiting,
4742 for which l_wait_event interface was extended to allow checking
4743 of exit condition at specified intervals).
4746 Frequency : rare, normally only when NFS exporting is done from client
4748 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4750 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4753 Frequency : always, if srand() is called before liblustre initialization
4755 Description: Liblustre uses system PRNG disturbing its usage by user application
4756 Details : Introduce internal to lustre fast and high-quality PRNG for
4757 lustre usage and make liblustre and some other places in generic
4758 lustre code to use it.
4760 Severity : enhancement
4761 Bugzilla : 9477, 9557, 9870
4762 Description: Verify that the MDS configuration logs are updated when xml is
4763 Details : Check if the .xml configuration logs are newer than the config
4764 logs stored on the MDS and report an error if this is the case.
4765 Request --write-conf, or allow starting with --old_conf.
4767 Severity : enhancement
4769 Description: Handle symlinks in the path when checking if Lustre is mounted.
4770 Details : Resolve intermediate symlinks when checking if a client has
4771 mounted a filesystem to avoid duplicate client mounts.
4776 Description: lconf can hit an error exception but still return success.
4777 Details : The lconf command catches the Command error exception at the top
4778 level script context and will exit with the associated exit
4779 status, but doesn't ensure that this exit status is non-zero.
4784 Description: failure of ptlrpc thread startup can cause oops
4785 Details : Starting a ptlrpc service thread can fail if there are a large
4786 number of threads or the server memory is very fragmented.
4787 Handle this without oopsing.
4790 Frequency : always, only if liblustre and non-default acceptor port was used
4792 Description: liblustre cannot connect to servers with non-default acceptor port
4793 Details : tcpnal_set_default_params() was not called and was therefore
4794 ignoring the environment varaible TCPNAL_PORT, as well as other
4795 TCPNAL_ environment variables
4800 Description: two objects could be created on the same OST for a single file
4801 Details : If an OST is down, in some cases it was possible to create two
4802 objects on a single OST for a single file. No problems other
4803 than potential performance impact and spurious error messages.
4807 Bugzilla : 5681, 9562
4808 Description: Client may oops in ll_unhash_aliases
4809 Details : Client dcache may become inconsistent in race condition.
4810 In some cases "getcwd" can fail if the current directory is
4816 Description: Inode refcounting problems in NFS export code
4817 Details : link_raw functions used to call d_instantiate without obtaining
4818 extra inode reference first.
4822 Bugzilla : 9942, 9903
4823 Description: Referencing freed requests leading to crash, memleaks with NFS.
4824 Details : We used to require that call to ll_revalidate_it was always
4825 followed by ll_lookup_it. Also with revalidate_special() it is
4826 possible to call ll_revalidate_it() twice for the same dentry
4827 even if first occurence returned success. This fix changes semantic
4828 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4829 reference on a request referred from the intent.
4830 ll_intent_release() then releases such a request.
4833 Frequency : rare, normally benchmark loads only
4835 Description: unlinked inodes were kept in memory on the client
4836 Details : If a client is repeatedly creating and unlinking files it
4837 can accumulate a lot of stale inodes in the inode slab cache.
4838 If there is no other client load running this can cause the
4839 client node to run out of memory. Instead flush old inodes
4840 from client cache that have the same inode number as a new inode.
4843 Frequency : SLES9 2.6.5 kernel and long filenames only
4844 Bugzilla : 9969, 10379
4845 Description: utime reports stale NFS file handle
4846 Details : SLES9 uses out-of-dentry names in some cases, which confused
4847 the lustre dentry revalidation. Change it to always use the
4851 Frequency : rare, unless heavy write-truncate concurrency is continuous
4852 Bugzilla : 4180, 6984, 7171, 9963, 9331
4853 Description: OST becomes very slow and/or deadlocked during object unlink
4854 Details : filter_destroy() was holding onto the parent directory lock
4855 while truncating+unlinking objects. For very large objects this
4856 may block other threads for a long time and slow overall OST
4857 responsiveness. It may also be possible to get a lock ordering
4858 deadlock in this case, or run out of journal credits because of
4859 the combined truncate+unlink. Solution is to do object truncate
4860 first in one transaction without parent lock, and then do the
4861 final unlink in a new transaction with the parent lock. This
4862 reduces the lock hold time dramatically.
4865 Frequency : rare, 2.4 kernels only
4867 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4868 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4869 before unmounting their filesystem in order to avoid deadlock.
4870 The kernel_locked() function in 2.4 kernels only checks whether
4871 the kernel lock is held, not whether it is this process that is
4872 holding it as 2.6 kernels do.
4877 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4878 Details : The client ptlrpc code may be trying to reconnect to a down
4879 server before a previous connection attempt has timed out.
4880 Increase the reconnect interval to be longer than the connection
4881 timeout interval to avoid sending duplicate connections to
4885 Frequency : echo_client brw_test command
4887 Description: fix echo_client to work with OST preallocated code
4888 Details : OST preallocation code (5137) didn't take echo_client IO path
4889 into account: echo_client calls filter methods outside of any
4890 OST thread and, hence, there is no per-thread preallocated
4891 pages and buffers to use. Solution: hijack pga pages for IO. As
4892 a byproduct, this avoids unnecessary data copying.
4896 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4897 Description: Client can oops in mdc_commit_close() after open replay
4898 Details : It was possible for the MDS to return an open request with no
4899 transaction number in mds_finish_transno() if the client was
4900 evicted, but without actually returning an error. Clients
4901 would later try to replay that open and may trip an assertion
4902 Simplify the client close codepath, and always return an error
4903 from the MDS in case the open is not successful.
4906 Frequency : rare, 2.6 OSTs only
4908 Description: OST may deadlock under high load on fragmented files
4909 Details : If there was a heavy load and highly-fragmented OST filesystems
4910 it was possible to have all the OST threads deadlock waiting on
4911 allocation of biovecs, because the biovecs were not released
4912 until the entire RPC IO was completed. Instead, release biovecs
4913 as soon as they are complete to ensure forward IO progress.
4915 Severity : enhancement
4917 Description: Support for specifying external journal device at mount
4918 Details : If an OST or MDS device is formatted with an external journal
4919 device, this device major/minor is stored in the ext3 superblock
4920 and may not be valid for failover. Allow detecting and
4921 specifying the external journal at mount time.
4926 Description: Mounting an MDS with pending unlinked files may cause oops
4927 Details : target_finish_recovery() calls mds_postrecov() which returned
4928 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4929 considers this an error and immediately begins cleaning up the
4930 lov, just after starting the mds_lov process
4932 Severity : enhancement
4934 Description: Implement 'lfs df' to report actual free space on per-OST basis
4935 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4936 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4937 usage of objects; '-h' to report in human readable format.
4939 ------------------------------------------------------------------------------
4941 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4948 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4949 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4950 at startup during orphan unlinking. Instead, skip buddy update
4951 before setup, as it will be regenerated after recovery is complete.
4956 Description: If an OST is inactive, its locks might reference stale inodes.
4957 Details : lov_change_cbdata() must iterate over all namespaces, even if
4958 they are inactive to clear inode references from the lock.
4960 Severity : enhancement
4961 Frequency : occasional, if non-standard max_dirty_mb used
4963 Description: Client will block write RPCs if not enough grant
4964 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4965 then the client will block writes while waiting for another RPC
4966 to complete instead of consuming its dirty limit. With change
4967 we get improved performance when max_dirty_mb is small.
4969 Severity : enhancement
4970 Bugzilla : 3389, 6253
4971 Description: Add support for supplementary groups on the MDS.
4972 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4973 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4974 lookups for user supplementary groups into a cache.
4978 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4979 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4980 for liblustre O_EXCL is enforced by MDS.
4984 Description: readdir never returns NULL in liblustre.
4985 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4986 dentry in struct dirent.
4990 Frequency : liblustre only, depends on application IO pattern
4991 Description: liblustre clients evicted if not contacting servers
4992 Details : Don't put liblustre clients into the ping_evictor list, so
4993 they will not be evicted by the pinger ever.
4995 Severity : enhancement
4997 Description: Add ability to evict clients by NID from MDS.
4998 Details : By echoing "nid:$NID" string into
4999 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5000 $NID would be instantly evicted from this MDS and from all active
5001 OSTs connected to it.
5005 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5006 Details : lookup_it_finish() used to query file size from OSTs that was not
5011 Description: service threads change working directory to that of init
5012 Details : Starting lustre service threads may pin the working directory
5013 of the parent thread, making that filesystem busy. Threads
5014 now change to the working directory of init to avoid this.
5018 Frequency : during shutdown only
5019 Description: shutdown with a failed MDS or OST can cause unmount to hang
5020 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5025 Frequency : 2.6 only
5026 Description: chmod/chown may include an extra supplementary group
5027 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5028 supplementary group and if none is specified this is used.
5031 Bugzilla : 5479 (6816)
5033 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5034 Details : If lookup is in progress on a file that is unlinked we might try
5035 to revalidate the inode and fail in revalidate after lookup is
5036 complete and ll_file_open() enqueues the open again but
5037 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5040 Frequency : always, if lconf --abort_recovery used
5042 Description: lconf --abort_recovery fails with 'Operation not supported'
5043 Details : lconf was attempting to abort recovery on the MDT device and not
5046 ------------------------------------------------------------------------------
5048 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5053 Frequency : rare (only unsupported configurations with a node running as an
5055 Bugzilla : 6514, 5137
5056 Description: Mounting a Lustre file system on a node running as an OST could
5058 Details : OSTs now preallocates memory needed to write out data at
5059 startup, instead of when needed, to avoid having to
5060 allocate memory in possibly low memory situations.
5061 Specifically, if the file system is mounted on on OST,
5062 memory pressure could force it to try to write out data,
5063 which it needed to allocate memory to do. Due to the low
5064 memory, it would be unable to do so and the node would
5065 become unresponsive.
5067 Severity : enhancement
5069 Description: Addition of lconf --service command line option
5070 Details : lconf now accepts a '--service <arg>' option, which is
5071 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5073 Severity : enhancement
5075 Description: Failover mode is now the default for OSTs.
5076 Details : By default, OSTs will now run in failover mode. To return to
5077 the old behaviour, add '--failout' to the lmc line for OSTs.
5079 Severity : enhancement
5081 Description: Health checks are now provided for MDS and OSTs
5082 Details : Additional detailed health check information on MSD and OSTs
5083 is now provided through the procfs health_check value.
5086 Frequency : occasional, depends on IO load
5088 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5089 after numerous create/delete cycles
5090 Details : The ext3 inode allocation policy would not allocate new inodes
5091 very well on the OSTs because there are no new directories
5092 being created. Instead we look for groups with free space if
5093 the parent directories are nearly full.
5098 Description: Network or server problems during mount may cause partially
5099 mounted clients instead of returning an error.
5100 Details : The config llog parsing code may overwrite the error return
5101 code during mount error handling, returning success instead
5107 Description: MDS can fail to allocate large reply buffers
5108 Details : After long uptimes the MDS can fail to allocate large reply
5109 buffers (e.g. zconf client mount config records) due to memory
5110 fragmentation or consumption by the buffer cache. Preallocate
5111 some large reply buffers so that these replies can be sent even
5112 under memory pressure.
5116 Frequency : rare (liblustre)
5117 Description: fsx running with liblustre complained that using truncate() to
5118 extend the file doesn't work. This patch corrects that issue.
5119 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5120 ATTR_SIZE and lsm use in llu_setattr_raw.
5124 Frequency : rare, only 2.6 kernels
5125 Description: Unusual file access patterns on the MDS may result in inode
5126 data being lost in very rare circumstances.
5127 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5128 code in the 2.6 kernel caused the inode and/or EA data not to
5129 be read from disk, causing single-file corruption.
5133 Frequency : rare, only 2.6 filesystems using extents
5134 Description: Heavy concurrent write and delete load may cause data corruption.
5135 Details : It was possible under high-load situations to have an extent
5136 metadata block in the block device cache from a just-unlinked
5137 file overwrite a newly-allocated data block. We now unmap any
5138 metadata buffers that alias just-allocated data blocks.
5142 Frequency : filesystems with default stripe_count larger than 77
5143 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5144 Details : lconf specifies an inode size of 4096 bytes when the default
5145 stripe_count is larger than 77. This conflicts with the default
5146 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5147 this case to avoid pinning too much memory for large EAs.
5149 ------------------------------------------------------------------------------
5151 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5156 Frequency : rare (extremely heavy IO load with hundreds of clients)
5158 Description: Client is evicted, gets IO error writing to file
5159 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5160 caused clients to be evicted for AST timeouts. The fixes in
5161 bug 5192 mean we no longer need to have such short AST timeouts
5162 so ldlm_timeout has been increased.
5165 Frequency : occasional during --force or --failover shutdown under load
5166 Bugzilla : 5949, 4834
5167 Description: Server oops/LBUG if stopped with --force or --failover under load
5168 Details : a collection of import/export refcount and cleanup ordering
5169 issues fixed for safer force cleanup
5172 Frequency : only filesystems larger than 120 OSTs
5173 Bugzilla : 5990, 6223
5174 Description: lfs getstripe would oops on a very large filesystem
5175 Details : lov_getconfig used kfree on vmalloc'd memory
5178 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5179 Bugzilla : 6242, 6243
5180 Description: reading from files that had been truncated to a non-zero size
5181 but never opened returned no data
5182 Details : ll_file_read() reads zeros from no-object files to EOF
5187 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5188 Details : There's little harm in aborting MDS/OSS recovery and letting it
5189 try again, so I removed the LASSERT and return an error instead.
5191 Severity : enhancement
5193 Description: New debugging infrastructure for tracking down data corruption
5194 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5195 (b) cover more of the client-side code path, and (c) try to narrow
5196 down where problems occurred
5200 Bugzilla : 3819, 4364, 4397, 6313
5201 Description: Racing close and eviction MDS could cause assertion in mds_close
5202 Details : It was possible to get multiple mfd references during close and
5203 client eviction, leading to one thread referencing a freed mfd.
5205 Severity: : enhancement
5206 Bugzilla : 3262, 6359
5207 Description: Attempts to reconnect to servers are now more aggressive.
5208 Details : This builds on the enhanced upcall-less recovery that was added
5209 in 1.4.2. When trying to reconnect to servers, clients will
5210 now try each server in the failover group every 10 seconds. By
5211 default, clients would previously try one server every 25 seconds.
5216 Description: After recovery, certain operations trigger a failed
5217 assertion on a client.
5218 Details : Failing over an mds, using lconf -d --failover, while a
5219 client was doing a readdir() call would cause the client to
5220 LBUG after recovery completed and the readdir() was resent.
5222 Severity : enhancement
5224 Description: Default groups are now added by lconf
5225 Details : You can now run lconf --group <servicename> without having to
5226 manually add groups with lmc.
5229 Frequency : occasional
5231 Description: Nodes with an elan id of 0 trigger a failed assertion
5234 Frequency : always when accessing e.g. tty/console device nodes
5236 Description: tty and some other devices nodes cannot be used on lustre
5237 Details : file's private_data field is used by device data and lustre
5238 values in there got lost. New field was added to struct file to
5239 store fs-specific private data.
5242 Frequency : when exporting Lustre via NFS
5244 Description: NFSD failed occasionally when looking up a path component
5245 Details : NFSD is looking up ".." which was broken in ext3 directories
5246 that had grown large enough to become hashed.
5249 Frequency : Clusters with multiple interfaces not on the same subnet
5251 Description: Nodes will repeatedly try to reconnect to an interface which it
5252 cannot reach and report an error to the log.
5253 Details : Extra peer list entries will be created by lconf with some peers
5254 unreachable. lconf now validates the peer before adding it.
5257 Frequency : Only if a default stripe is set on the filesystem root.
5259 Description: Setting a default stripe on the filesystem root prevented the
5260 filesystem from being remounted.
5261 Details : The client was sending extra request flags in the root getattr
5262 request and did not allocate a reply buffer for the dir EA.
5265 Frequency : occasional, higher if lots of files are accessed by one client
5266 Bugzilla : 6159, 6097
5267 Description: Client trips assertion regarding lsm mismatch/magic
5268 Details : While revalidating inodes the VFS looks up inodes with ifind()
5269 and in rare cases can find an inode that is being freed.
5270 The ll_test_inode() code will free the lsm during ifind()
5271 when it finds an existing inode and then the VFS later attaches
5272 this free lsm to a new inode.
5276 Bugzilla : 6422, 7030
5277 Description: MDS deadlock between mkdir and client eviction
5278 Details : Creating a new file via mkdir or mknod (starting a transaction
5279 and getting the ns lock) can deadlock with client eviction
5280 (gets ns lock and trying to finish a synchronous transaction).
5283 Frequency : occasional
5284 Description: While starting a server, the fsfilt_ext3 module could not be
5286 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5287 kernels. Previously, lconf would still use the ext3 name
5288 when trying to load modules. Now, it will correctly use
5289 ext3 on 2.4 and ldiskfs on 2.6.
5291 Severity : enhancement
5292 Description: The default stripe count has been changed to 1
5293 Details : The interpretation of the default stripe count (0, to lfs
5294 or lmc) has been changed to mean striping across a single
5295 OST, rather than all available. For general usage we have
5296 found a stripe count of 1 or 2 works best.
5298 Severity : enhancement
5299 Description: Add support for compiling against Cray portals.
5300 Details : Conditional compiling for some areas that are different
5304 Frequency : occasional
5305 Bugzilla : 6409, 6834
5306 Description: Creating files with an explicit stripe count may lead to
5307 a failed assertion on the MDS
5308 Details : If some OSTs are full or unavailable, creating files may
5309 trigger a failed assertion on the MDS. Now, Lustre will
5310 try to use other servers or return an error to the
5314 Frequency : occasional
5316 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5317 caused lock timeout during readahead (since 1.4.2).
5318 Details : Processes doing readahead might match a lock that hasn't been
5319 granted yet if there are overlapping and conflicting lock
5320 requests. The readahead process waits on ungranted lock
5321 (original lock is CBPENDING), while OST waits for that process
5322 to cancel CBPENDING read lock and eventually evicts client.
5324 Severity : enhancement
5326 Description: Initial enabling of flock support for clients
5327 Details : Implements fcntl advisory locking and file status functions.
5328 This feature is provided as an optional mount flag (default
5329 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5330 locking are implemented yet, and those that are are not guaranteed
5331 to be completely correct in production environments.
5332 mount -t lustre -o [flock|noflock] ...
5335 Frequency : occasional
5337 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5338 assertion in the ext3 JBD (journaling) layer.
5339 Details : The b_committed_data struct is protected by the big kernel lock
5340 in 2.4 kernels, serializing journal_commit_transaction() and
5341 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5342 access to this struct. In 2.6 kernels there is finer grained
5343 locking to improve SMP performance of the JBD layer.
5347 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5352 Description: merge and cleanup kernel patches.
5355 -----------------------------------------------------------------------------
5357 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5359 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5360 versions of Lustre. Please update all servers and clients to
5361 version 1.4.2 or later at the same time. You must also run
5362 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5363 to update the configuration logs.
5365 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5366 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5367 - swab llog records of type '0' so we get proper header size/idx (5861)
5368 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5369 - fix rename of one directory over another leaking an inode (5953)
5370 - avoid SetPageDirty on 2.6 (5981)
5371 - don't re-add just-being-destroyed locks to the waiting list (5653)
5372 - when creating new directories, inherit the parent's custom
5373 striping settings if present parent (3048)
5374 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5375 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5376 - don't hold client locks on temporary worklist from l_lru (5666)
5377 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5378 - automatically evict dead clients (5921)
5379 - Update file size properly in create+truncate+fstat case (6196)
5380 - Do not unhash mountpoint dentries, do not allow removal of
5382 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5383 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5384 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5385 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5386 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5387 - Correctly handle reads of files with no objects (6243)
5388 - lctl recover will also mark a device active if deactivate used (5933)
5390 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5391 - allow --write-conf on an MDS with different nettype than client (5619)
5392 - don't write config llogs to MDS for mounts not from that MDS (5617)
5393 - lconf should create multiple TCP connections from a client (5201)
5394 - init scripts are now turned off by default; run chkconfig --on
5395 lustre and chkconfig --on lustrefs to use them
5396 - upcalls are no longer needed for clients to recover to failover
5398 - add --abort-recovery option to lconf to abort recovery on device
5400 - add support for an arbitrary number of OSTs (3026)
5401 - Quota support protocol changes.
5402 - forward compatibility changes to wire structs (6007)
5403 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5404 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5405 - improved "lustre status" script
5406 - initialize blocksize for non-regular files (6062)
5407 - added --disable-server and --disable-client configure options (5782)
5408 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5409 - Vanilla 2.4.29 support
5410 - increase maximum number of obd devices to 520 (6242)
5411 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5412 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5413 SLES 9 2.6.5 kernels
5414 - limit stripes per file to 160 (the maximum EA size) (6093)
5416 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5419 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5420 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5421 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5422 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5423 - small ext3 extents cleanups and fixes (5733)
5424 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5425 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5426 - avoid needless client->OST connect, fix handle mismatch (5317)
5427 - fix DLM error path that led to out-of-sync client, long delays (5779)
5428 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5429 - fix several locking issues related to i_size (5492,5624,5654,5672)
5430 - don't move pending lock onto export if it is already evicted (5683)
5431 - fix kernel oops when creating .foo in unlinked directory (5548)
5432 - fix deadlock in obdfilter statistics vs. object create (5811)
5433 - use time_{before,after} to avoid timer jiffies wrap (5882)
5434 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5435 - Do not leak request if server was not able to process it (5154)
5436 - If mds_open unable to find parent dir, make that negative lookup(5154)
5437 - don't create new directories with extent-mapping (5909, 5936)
5439 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5440 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5441 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5442 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5443 - support for RHEL4 kernel on i686 (5773)
5444 - provide error messages when incompatible logs are encountered (5898)
5446 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5447 * version 1.4.0.10 (1.4.1 release candidate 1)
5449 - don't keep a lock reference when lock is not granted (4238)
5450 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5451 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5452 - drop import inflight refcount on signal_completed_replay error (5255)
5453 - unlock page after async write error during send (3677)
5454 - handle missing objects in filter_preprw_read properly (5265)
5455 - no transno return for symlink open, don't save no-trasno open (3440)
5456 - don't try to complete elan receive that already failed (4012)
5457 - free RPC server reply state on error (5406)
5458 - clean up thread from ptlrpc_start_thread() on error (5160)
5459 - readahead could read extra page into cache that wasn't ejected (5388)
5460 - prevent races in class_attach/setup/cleanup/detach (5260)
5461 - don't dereference de->d_inode after l_dput of de (5458)
5462 - use "int" for stripe value returned from lock_to_stripe (5544)
5463 - mballoc allocation and error-checking fixes in 2.6 (5504)
5464 - block device patches to fix I/O request sizes in 2.6 (5482)
5465 - look up hostnames for IB nals (5602)
5466 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5467 - don't start multiple acceptors for the same port (5277)
5468 - fix incorrect LASSERT in mds_getattr_name (5635)
5469 - export a proc file for general "ping" checking (5628)
5470 - fix "lfs check" to not block when the MDS is down (5628)
5472 - service request history (4965)
5473 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5474 - create an "evict_client" /proc entry on OSTs, like the MDS has
5475 - fix mount usage message, return errors per mount(8) (5168)
5476 - change grep [] to grep "[]" in tests so they work in more UMLs
5477 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5478 - remove ancient LOV_MAGIC_V0 EA support (5047)
5479 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5480 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5481 - allow readahead allocations to fail when low on memory (5383)
5482 - mmap locking landed again, after considerable improvement (2828)
5483 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5485 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5488 - send OST transaction number in read/write reply to free req (4966)
5489 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5490 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5491 - the watchdog thread now runs as interruptible (5246)
5492 - flock/lockf fixes (but it's still disabled, pending 5135)
5493 - don't use EXT3 constants in llite code (5094)
5494 - memory shortage at startup could cause assertion (5176)
5496 - reorganization of lov code
5497 - single portals codebase
5499 - add extents/mballoc support (5025)
5500 - direct I/O reads in the obdfilter (4048)
5501 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5503 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5506 - send OST transaction number in read/write reply to free req (4966)
5507 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5508 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5509 - the watchdog thread now runs as interruptible (5246)
5510 - handle missing objects in filter_preprw_read properly (5265)
5511 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5512 - drop import inflight refcount on signal_completed_replay error (5255)
5513 - unlock page after async write error during send (3677)
5514 - return original error code on reconstructed replies (3761)
5515 - no transno return for symlink open, don't save no-trasno open (3440)
5517 - add pid to ldlm debugging output (4922)
5518 - bump the watchdog timeouts -- we can't handle 30sec yet
5519 - extra debugging for orphan dentry/inode bug (5259)
5521 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5524 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5525 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5526 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5527 - fix NULL dereference in /proc/sys/portals/routes (4827)
5528 - allow failed mdc_close() operations to be interrupted (4561)
5529 - stop precreate on OST before MDS would time out on it (4778)
5530 - don't send partial-page writes before EOF from client (4410)
5531 - discard client grant for sub-page writes on large-page clients (4520)
5532 - don't free dentries not owned by NFS code, check generation (4806)
5533 - fix lsm leak if mds_create_objects() fails (4801)
5534 - limit debug_daemon file size, always print CERROR messages (4789)
5535 - use transno after validating reply (3892)
5536 - process timed out requests if import state changes (3754)
5537 - update mtime on OST during writes, return in glimpse (4829)
5538 - add mkfsoptions to LDAP (4679)
5539 - use ->max_readahead method instead of zapping global ra (5039)
5540 - don't interrupt __l_wait_event() during strace
5542 - add software watchdogs to catch hung threads quickly (4941)
5543 - make lustrefs init script start after nfs is mounted
5544 - fix CWARN/ERROR duplication (4930)
5545 - return async write errors to application if possible (2248)
5546 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5547 - print NAL number in %x format (4645)
5548 - update barely-supported suse-2.4.21-171 series (4842)
5549 - support for sles 9 %post scripts
5550 - support for building 2.6 kernel-source packages
5551 - support for sles km_* packages
5553 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5556 - ignore -ENOENT errors in osc_destroy (3639)
5557 - notify osc create thread that OSC is being cleaned up (4600)
5558 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5559 - reconstruct ost_handle() like mds_handle() (4657)
5560 - create a new thread to do import eviction to avoid deadlock (3969)
5561 - let lconf resolve symlinked-to devices (4629)
5562 - don't unlink "objects" from directory with default EA (4554)
5563 - hold socknal file ref over connect in case target is down (4394)
5564 - allow more than 32000 subdirectories in a single directory (3244)
5565 - fix blocks count for O_DIRECT writes (3751)
5566 - OST returns ENOSPC from object create when no space left (4539)
5567 - don't send truncate RPC if file size isn't changing (4410)
5568 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5569 - bind to privileged port in socknal and tcpnal (3689)
5571 - rate limit CERROR/CWARN console message to avoid overload (4519)
5572 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5573 - basic mmap support (3918)
5574 - kernel patch series update from b1_4 (4711)
5576 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5579 - avoid crash during MDS cleanup with OST shut down (2775)
5580 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5581 - don't use bad inodes on the MDS (3744)
5582 - dynamic object preallocation to improve recovery speed (4236)
5583 - don't hold spinlock over lock dumping or change debug flags (4401)
5584 - don't zero obd_dev when it is force cleaned (3651)
5585 - print grants to console if they go negative (4431)
5586 - "lctl deactivate" will stop automatic recovery attempts (3406)
5587 - look for existing locks in ldlm_handle_enqueue() (3764)
5588 - don't resolve lock handle twice in recovery avoiding race (4401)
5589 - revalidate should check working dir is a directory (4134)
5591 - don't always mark "slow" obdfilter messages as errors (4418)
5593 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5596 - don't close LustreDB during write_conf until it is done (3860)
5597 - fix typo in lconf for_each_profile (3821)
5598 - allow dumping logs from multiple threads at one time (3820)
5599 - don't allow multiple threads in OSC recovery (3812)
5600 - fix debug_size parameters (3864)
5601 - fix mds_postrecov to initialize import for llog ctxt (3121)
5602 - replace config semaphore with spinlock (3306)
5603 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5604 - don't allow enqueue to complete on a destroyed export (3822)
5605 - down write_lock before checking llog header bitmap (3825)
5606 - recover from lock replay timeout (3764)
5607 - up llog sem before sending rpc (3652)
5608 - reduce ns lock hold times when setting kms (3267)
5609 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5610 - fix NULL deref and obd_dev leak on setup error (3312)
5611 - replace some LBUG about llog ops with error handling (3841)
5612 - don't match INVALID dentries from d_lookup and spin (3784)
5613 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5614 - fix invalid assertion in ptlrpc_set_wait (3880)
5616 - add libwrap support for the TCP acceptor (3996)
5617 - add /proc/sys/portals/routes for non-root route listing (3994)
5618 - allow setting MDS UUID in .xml (2580)
5619 - print the stack of a process that LBUGs (4228)
5621 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5624 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5625 - make sure to unset replay flag from failed open requests (3440)
5626 - if default stripe count is 0, use OST count for inode size (3636)
5627 - update parent mtime/ctime on client for create/unlink (2611)
5628 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5629 - free recovery state on server during a forced cleanup (3571)
5630 - unregister_reply for resent reqs (3063)
5631 - loop back devices mounting and status check on 2.6 (3563)
5632 - fix resource-creation race that can provoke i_size == 0 (3513)
5633 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5634 - more debugging for page-accounting assertion (3746)
5635 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5636 - avoid deadlock after precreation failure (3758)
5637 - fix race and lock order deadlock in orphan handling (3450, 3750)
5638 - add validity checks when grabbing inodes from l_ast_data (3599)
5640 - add /proc/.../recovery_status to obdfilter (3428)
5641 - lightweight CDEBUG infrastructure, debug daemon (3668)
5642 - change default OSC RPC parameters to be better on small clusters
5643 - turn off OST read cache for files smaller than 32MB
5644 - install man pages and include them in rpms (3100)
5645 - add new init script for (un)mounting lustre filesystems (2593)
5646 - run chkconfig in %post for init scripts (3701)
5647 - drop scimac NAL (unmaintained)
5649 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5652 - clean kiobufs before and after use (3485)
5653 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5654 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5655 - change init script to fail more clearly if not run as root (1528)
5656 - allow clients to reconnect during replay (1742)
5657 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5658 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5659 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5660 - fix cond_resched (3554)
5661 - only evict unfinished clients after recovery (3515)
5662 - allow bulk resend, prevent data loss (3570)
5663 - dynamic ptlrpc request buffer allocation (2102)
5664 - don't allow unlinking open directory if it isn't empty (2904)
5665 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5666 - remove extraneous obd dereference causing LASSERT failure (3334)
5667 - don't use get_cycles() when creating temp. files on the mds (3156)
5668 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5669 - handle EEXIST for set-stripe, set proper directory name (3336)
5671 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5672 - fix ksocknal_fmb_callback() error messages (2918)
5674 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5677 - don't copy lvb into (possibly NULL) reply on error (2983)
5678 - don't deref dentry after dput, don't free lvb on error (2922)
5679 - use the kms to determine writeback rpc length (2947)
5680 - increment oti_logcookies when osc is inactive (2948)
5681 - update client's i_blocks count via lvb messages (2543)
5682 - handle intent open/close of special files properly (1557)
5683 - mount MDS with errors=remount-ro, like obdfilter (2009)
5684 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5685 - don't use cancelling-locks' kms values (2947)
5686 - use highest lock extent for kms, not last one (2925)
5687 - don't dereference ERR_PTR() dentry in error handling path (3107)
5688 - fix thread race in portals_debug_dumplog() (3122)
5689 - create lprocfs device entries at setup instead of at attach (1519)
5690 - common AST error handler, don't evict client on completion race (3145)
5691 - zero nameidata in detach_mnt in 2.6 (3118)
5692 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5693 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5694 - initialize RPC timeout timer earlier for 2.6 (3219)
5695 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5696 - print nal/nid for unknown nid (3258)
5697 - additional checks for oscc recovery before doing precreate (3284)
5698 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5699 - don't crash in mdc_close for bad permissions on open (3285)
5700 - zero i_rdev for non-device files (3147)
5701 - clear page->private before handing to FS, better assertion (3119)
5702 - tune the read pipeline (3236)
5703 - fix incorrect decref of invalidated dentry (2350)
5704 - provide read-ahead stats and refine rpc in flight stats (3328)
5705 - don't hold journal transaction open across create RPC (3313)
5706 - update atime on MDS at close time (3265)
5707 - close LDAP connection when recovering to avoid server load (3315)
5708 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5709 - don't leak open file on MDS after open resend (3325)
5710 - serialize filter_precreate and filter_destroy_precreated (3329)
5711 - loop device shouldn't call sync_dev() for nul device (3092)
5712 - clear page cache after eviction (2766)
5713 - resynchronize MDS->OST in background (2824)
5714 - refuse to mount the same filesystem twice on same mountpoint (3394)
5715 - allow llmount to create routes for mounting behind routers (3320)
5716 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5717 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5718 - fix rare problem with rename on htree directories (3417)
5720 - allow default OST striping configuration per directory (1414)
5721 - fix compilation for qswnal for 2.6 kernels (3125)
5722 - increase maximum number of MDS request buffers for large systems
5723 - change liblustreapi to be useful for external progs like lfsck (3098)
5724 - increase local configuration timeout for slow disks (3353)
5725 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5727 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5730 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5731 - don't overwrite extent policy data in reply if lock was blocked (2901)
5732 - drop filter export grants atomically with removal from device (2663)
5733 - del obd_self_export from work_list in class_disconnect_exports (2908)
5734 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5735 - swab reply message in mdc_close, other PPC fixes (2464)
5736 - fix destroying of named logs (2325)
5737 - overwrite old logs when running lconf --write_conf (2264)
5738 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5739 - fix race in target_handle_connect (2898)
5740 - mds_reint_create() should take same inode create lock (2926)
5741 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5742 - don't close files for self_export to avoid uninitialized obd (2936)
5743 - allow MDS with the same name as client node (2939)
5744 - hold dentry reference for closed log files for unlink (2325)
5745 - reserve space for all logs during transactions (2059)
5746 - don't evict page beyond end of stripe extent (2925)
5747 - don't oops on a deleted current working directory (2399)
5748 - handle hard links to targets without a parent properly (2517)
5749 - don't dereference NULL lock when racing during eviction (2867)
5750 - don't grow lock extents when lots of conflicting locks (2919)
5752 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5755 - account for cache space usage on clients to avoid data loss (974)
5756 - lfsck support in lustre kernel code (2349)
5757 - reduce journal credits needed for BRW writes (2370)
5758 - orphan handling to avoid losing space on client/server crashes
5759 - ptlrpcd can be blocked, stopping ALL progress (2477)
5760 - use lock value blocks to assist in proper KMS, faster stat (1021)
5761 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5762 - recovery for initial connections (2355)
5763 - fixes for mds_cleanup_orphans (1934)
5764 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5765 - block all file creations until orphan recovery completes (1901)
5766 - client remove rq_connection from request struct (2423)
5767 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5768 - recovery timer race (2670)
5769 - mdc_close recovey bug (2532)
5770 - ptlrpc cleanup bug (2710)
5771 - mds timeout on local locks (2588)
5772 - namespace lock held during RPCs (2431)
5773 - handle interrupted sync write properly (2503)
5774 - don't try to handle a message that hasn't been replied to (2699)
5775 - client assert failure during cleanup after abort recovery (2701)
5776 - leak mdc device after failed mount (2712)
5777 - ptlrpc_check_set allows timedout requests to complete (2714)
5778 - wait for inflight reqs when ptlrpcd finishes (2710)
5779 - make sure unregistered services are removed from the srv_list
5780 - reset bulk XID's when resending them (caught by 1138 test)
5781 - unregister_bulk after timeout
5782 - fix lconf error (2694)
5783 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5784 - readahead locks pages, leaves pending causing memory pressure (2673)
5785 - increase OST request buffers to 4096 on large machines (2729)
5786 - fix up permission of existing directories in simple_mkdir (2661)
5787 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5788 - don't assign transno to errored transactions (2742)
5789 - don't delete objects on OST if given a bogus objid from MDS (2751)
5790 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5791 - if rq_no_resend, then timeout request after recovery (2432)
5792 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5793 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5794 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5795 - MDS recovery shouldn't delete objects that it has given out (2730)
5796 - if enqueue arrives after completion, don't clobber LVB (2819)
5797 - don't unlock pages twice when trigger_group_io returns error (2814)
5798 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5799 - don't write pages to disk if there was an error (1450)
5800 - don't ping imports that have recovery disabled (2676)
5801 - take buffered bytes into account when balancing socknal conn (2817)
5802 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5803 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5804 - remove little-endian swabbing of llog records (1987)
5805 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5806 - retry reposting request buffers if they fail (1191)
5807 - grow extent at grant time to avoid granting a revoked lock (2809)
5808 - lock revoke doesn't evict page if covered by a second lock (2765)
5809 - disable VM readahead to avoid reading outside lock extents (2805)
5811 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5812 - updated kernel patches for hp-2.4.20 kernel (2681)
5814 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5817 - fix truncated write corruption (2366)
5818 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5820 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5821 - ensure len(uuid) < 37 in lmc (1171)
5822 - fix ia64 OOPS in llog_test (2255)
5823 - zero end of page at obdfilter for partial page writes (2648)
5824 - don't leave stale dentries around after renames (bug 2428)
5825 - fix timeouts when evicting a client with a single lock held (2642)
5826 - set deadline for the initial HELLO message to drain (2634)
5827 - print out dotted-quad IP addresses in the socknal (2302)
5829 - additional debugging for MDS client eviction problem (2443)
5830 - fix mkfsoptions support for osts (2603, 2604)
5832 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5835 - add series for the vanilla 2.6.0 kernel
5836 - add series for the vanilla 2.4.24 kernel
5837 - add series for a cray x86/64 UL kernel drop
5838 - fix xattr patches for the vanilla 2.4.19 series
5840 - generate true UUIDs in lmc (1171)
5841 - have portals stack dumping break in UML (2466)
5842 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5843 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5844 - wake callback waiting threads on client eviction (2460)
5845 - Add --ptldebug and --subsystem to lmc (1719)
5846 - update assertion to allow safe interrupt allocation
5847 - set rq_no_resend for cancel requests (2432)
5848 - recalculate ptlrpcd timeout after resend (2494)
5849 - call vfs_rmdir when removing pending directories (2368)
5850 - fix renaming a file to itself (2429)
5851 - lmc creates a default one-stripe lov (2454)
5852 - expand procfs space to handle large clusters (2326)
5853 - increase UML stack to avoid overflow
5854 - update lconf's list of debug and subsystem masks
5855 - fix lfs find --obd (2510)
5856 - /proc tunable for disabling filter read caching (2591)
5857 - stop rpm packages from altering slapd.conf (2301)
5858 - disable nagle in the socknal under 0conf (2578)
5859 - choose mds inode size based on stripe count (2572)
5860 - fix kernel-source rpm problems (2516)
5862 - add --disable-doc to avoid pdf generation (2421)
5863 - update documentation, tests, type-os, comments
5864 - avoid format warnings on ia64
5865 - remove the TOE NAL
5866 - tiny code cleanups by removing unused fields
5868 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5871 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5872 - remove the most blatant lies from BUILDING file (2371)
5873 - change default debug level to reasonable production setting
5874 - reduce client side cache size to reduce cache flush time
5875 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5876 - make TCP zerocopy and pinger support enabled by default (2476)
5877 - sync writes completed after process exits caused crashes (2319)
5878 - maintain correct mount count on the MDS (2356)
5879 - backout 1557, because 2316 wasn't really fixed
5880 - better file I/O statistics gathering in /proc
5881 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5882 - another kernel patch to fix zero-copy TCP function export
5883 - don't take duplicate lock when processing re-sent getattr (2420)
5884 - lctl uses obd_self_export instead of creating new conn (2353)
5885 - MDS/OST recovery case which requires object creation asserted (2425)
5886 - move lfs from /usr/sbin to /usr/bin in packages
5887 - fix race between mds_client_add and mds_client_free (2417)
5888 - use kmalloc instead of slabs in portals (2430)
5889 - don't create duplicate records when a failover MDS is present (2442)
5890 - remove unnecessary mount age check (2332)
5891 - don't remove directory inodes from locks prematurely (2451)
5892 - don't break if MDS service name is the same as hostname (2103)
5893 - fix races in client write RPC generation when cache full (2482)
5895 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5898 - remove now-unused request->rq_obd (278)
5899 - if an allocation fails, print out how much memory we've used (1933)
5900 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5901 - add the "configurable stack size" patch to most series files (1256)
5902 - ability to write large log records, for 100+ OST configs (2306)
5903 - fix NULL deref when filter_prep fails (2314)
5904 - fix operator precedence error in filter_sync
5905 - dynamic allocation of socknal TX descriptors (2315)
5906 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5907 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5908 - ensure that CWARN messages in Portals always get to the syslog
5909 - __init/__exit are not for prototype decls (ldlm_init/exit)
5910 - x86-64 compile warning fixes
5911 - fix gateway LMC keyword conflict (2318)
5912 - fix MDS lock inversions in getattr/reint paths (1844)
5913 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5914 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5915 - DLM race condition prevented some lock evictions (2328)
5916 - ENOMEM detection and retry on socknal sends (2230)
5917 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5918 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5919 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5920 - better error messages when a client is rejected during recovery (1505)
5921 - avoid cancelling locks which were never granted, after failure (2330)
5922 - fix i_sem/journal inversion in mds_client_add (2333)
5923 - fix truncate/getattr lock cycle deadlock (2334)
5924 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5925 - fix two rare exit paths which could leak an l_lock() ref (2321)
5926 - fencepost error in MDS/OST orphan recovery (2226)
5927 - make log record alignment 8 bytes (1988)
5928 - lstripe now fails when requested offset > ost_count (2237)
5929 - ensure that all kernel series have a complete list.h (1607)
5930 - fix crashes in special-file operations (2316)
5931 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5933 - allow configurable automake binary, for testing new versions
5934 - small update to the lfs documentation
5936 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5938 * fix negative export reference count in fsfilt_sync (2312)
5940 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5941 * release candidate 0.9.1
5943 - orphans are moved into the PENDING directory for possible recovery
5944 - replayed opens now open by fid for orphan/rename safety (1042)
5945 - last close of an orphan inode generates a transno (683)
5946 - chdir() and mount() now pin the directory entry (1020)
5947 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5948 - discard very old requests without processing them (1502)
5949 - remove some common, well-understood CERRORs (1505)
5950 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5951 - clear "grant" flags in OST replies until OST grant code lands (1644)
5952 - fix read performance by not clobbering i_blksize on client (1598)
5953 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5954 - make LRU size a /proc tunable, clears locks when reduced (707)
5955 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5956 - server locks take references on exports now (1558)
5957 - build fixes for 2.4.20-rh trees (1663)
5958 - return an error from lov_create if all OSCs are inactive (1751)
5959 - fix import levels when a reconnect happens without a timeout (1597)
5960 - exit early from mds_open if we get a lookup error (1749)
5961 - partial page read at EOF wouldn't wait for disk before sending (1642)
5962 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5963 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5964 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5965 - write/truncate lock inversion (1639)
5966 - Don't auto-load obdclass, portals modules during cleanup (1495)
5967 - fix timestamps from jumping to "now" (1763)
5968 - extra journal assertions (1648)
5969 - add an extra multiunlink test (1771)
5970 - fix read_record/write_record API (1776)
5971 - fix leak of offset_extent, possible incorrect i_size later (1772)
5972 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5973 - Debugging for the kqswnal_get_idle_tx problems (1820)
5974 - Allow recovery to be attempted multiple times (1536)
5975 - Write out MDS last_rcvd file after it is first created (1600)
5976 - Fix tx_descriptor leak in failed transmit situations (1827)
5977 - ext3 journaling fixes for assertion failure after IO error (1871)
5978 - class_export_put() on freed export after completion AST error (1896)
5979 - Fix revalidate looping in VFS (1322)
5980 - Don't access a freed export during MDS_REINT timeout (1521)
5981 - Add open-unlink recovery support on the MDS (1673,1764)
5982 - Return an error if no MDS data was read from last_rcvd (1946)
5983 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5984 - Don't setup a disk device that doesn't match exported UUID (317)
5985 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5986 - avoid committing NULL handle in force close
5987 - local.sh is now a one-stripe LOV configuration
5988 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5989 - fix ext3 htree duplicate directory entry corruption (1516)
5990 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5991 - update to new LOV EA format (2097)
5992 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5994 2003-06-15 Phil Schwan <phil@clusterfs.com>
5997 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5998 - per-import/export recovery handling (958, 931, 959)
5999 - multiple last-rcvd slots, for serving multiple FSes (949)
6000 - connections are again shared between multiple imp/exports (963, 964)
6001 - "umount -f" would hang if any requests needed to be sent (393, 978)
6002 - avoid pinning large req buffer by copying for queued messages (989)
6003 - add "uuid" to "lctl device" command to help upcalls (991)
6004 - "open" RPCs with transnos would confuse recovery counters (1037)
6005 - do proper endian conversion of all wire messages (288, 340, 891)
6006 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6007 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6008 - fix ptlrpc_connection leak in target_handle_connect (1174)
6009 - fix import refcounting bug in OST and MDS cleanup (1134)
6010 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6011 - fix very unlikely obd_types race condition (501)
6012 - remove osc_open hack for echo_client (1187)
6013 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6014 - a failure in read_inode2 leads to deadlock (1139)
6015 - cancel ack-locks as soon as transaction is committed (1072)
6016 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6017 - make sure to commitrw after any preprw to avoid deadlock (1162)
6018 - failing to execute a file in a lustre FS would lock inode (1203)
6019 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6020 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6021 - obd_uuid2tgt would walk off the end of the list (1255)
6022 - on IA64 the osc would give portals incorrect bulk size (1258)
6023 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6024 - fix lock inversion caused by new llite matching code (1282)
6025 - limit the number of dirty pages on a client to 10MB (1286)
6026 - timed out locks were not being corrected cancelled (1289)
6027 - fix O_DIRECT above 4GB on IA-32 (1292)
6028 * major user-visible changes
6029 - fail out/fail over policy now controlled by the upcall (993)
6031 - add OBD_PING to check server availability and failure (954)
6032 - lustre messages are now sent in sending host order (288, 340, 891)
6033 - add eadatalen to MDS getattr reply (340)
6034 - OST read replies may contain second buffer, with per-page status (593)
6036 2003-03-11 Phil Schwan <phil@clusterfs.com>
6039 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6040 - failed open()s could cause deadlock; fixed (867, 869)
6041 - stop cancelling OST locks when files are closed (481)
6042 - overlapping XID spaces caused network corruption (851, 853)
6043 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6044 - setattr_raw added, to do single-RPC, server-side setattrs
6045 - lmc/lconf syntax change for OST UUIDs
6046 - fix crashy race condition between ptlrpc_free_req and osc_close
6047 - don't use request in mdc_enqueue if we hit a timeout (889)
6048 - don't set the inode i_size for regular files from the MDS (896)
6049 - handle out of order completion AST (842)
6050 - don't LBUG if a lock request times out after receiving AST (913)
6051 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6052 - if a bad lock AST arrives, send an error instead of dropping entirely
6053 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6054 - fix leak in bulk IO when only partially completed (899, 900, 926)
6055 - fix O_DIRECT for ia64 (55)
6056 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6057 - C-z after timeout could hang a process forever; fixed (977)
6059 - client-side I/O cache (678, 924, 929, 941, 970)
6061 - READPAGE and SETATTRs which don't take server-side locks get
6064 2003-02-11 Phil Schwan <phil@clusterfs.com>
6067 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6068 - Fix fsfilt_extN_readpage() to read a full page of directory
6069 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6070 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6071 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6072 - Most intent operations were removed, in favour of a new RPC mode
6073 that does a single RPC to the server and bypasses most of the VFS
6074 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6075 - Aggressively cancel local locks on DLM servers
6076 - mds_reint_unlink sends EA to the client if it's the last nlink.
6077 client uses that EA to unlink OST objects.
6078 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6079 - recursive symlinks were fixed (439)
6080 - fixed NULL deref in DEBUG_REQ
6081 - filter_update_lastobjid no longer calls sync, which annoyed extN
6082 - fixed multi-client small-writes to a single file problem (445)
6083 - fixed mtime updates during file writes (607)
6084 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6085 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6086 - fixed Linux OST size reporting problem (444, 656)
6087 - OST now updates object mtime with writes or setattr (607, 619)
6088 - client verifies file size before zeroing page past EOF (445)
6089 - OST now writes last allocated objid to disk with allocation (108)
6090 - LOV on echo now works (409)
6092 - mds_reint_unlink sends a new buffer, with the EA included. this
6093 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6094 set if a regular file was being unlinked, and it was the last link
6095 - use PtlGet from the target for bulk writes (315)
6096 - OST now updates object mtime with writes or setattr (607, 619)
6097 - LDLM now has a grant-time callback to revalidate locked items, if
6099 - Many MDS operations were reorganized to combat race conditions
6101 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6102 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6104 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6107 - Fully reactivate OST imports after reconnection (512, others)
6108 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6109 - More graceful error handling for truncating on dead OST (515)
6110 - Don't error out unless we're actually accessing dead stripes (474)
6111 - Fix garbage sizes when stripes are missing (410)
6112 - LRU counters were broken, causing constant lock purge (433, 432)
6113 - garbage on read from stripes with failed OSTs (441)
6114 - mark OSCs as active before reconnecting during recovery (438)
6115 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6116 - lfind did not preserve OST order in output (443)
6117 - symlinks cause hung clients, incorrect data (439)
6118 - stop dereferencing request after dropping refcount (457)
6119 - don't LASSERT(spin_is_locked) on non-SMP (455)
6120 - fixes for many rename() bugs
6121 - fstat didn't correctly synchronize attributes (399)
6122 - server must handle lock cancellation during blocking AST prep (487)
6123 - bulk descriptors were free()d too soon (511)
6124 - fix paths in lconf, which would load incorrect modules (451, 507)
6125 - fix confusing lconf 'host not found' error message (386)
6126 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6127 - fix race condition in mdc_blocking_ast() for inode access (526)
6128 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6129 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6130 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6131 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6132 - fix lstripe in several ways (kernel side) (527)
6133 - fix request leak in ldlm_cli_enqueue (262)
6134 - incorrect OSC was marked inactive after OST failure
6135 - call mds_fs_cleanup before unmounting filesystem (524)
6136 - fix races between taking ns_lock and ldlm_lock_change_resource
6137 - fix races updating LOV export open file list
6138 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6139 - fix recovery NULL deref in ldlm_cli_cancel_unused
6140 - fix some DLM races by using new hash table for lock handles (419)
6141 - permit the client to specify desired inodes, at replay
6142 - duplicate requests when we queue them for replay reintegration
6143 - fix last_rcvd offset calculation
6144 - sync after each recovered transaction, so we always make progress
6145 - never, not always, ERESTART requests without transnos
6146 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6148 - skip replay if the MDS says that the client is already connected
6149 - don't check for a recovery-enabled export to match lctl's UUID
6150 - don't INC_USE_COUNT for phantom exports
6151 - don't crash when cleaning up phantom exports (567)
6152 - don't double-finish or set replay data for errored mdc_open requests
6153 - abort requests when they time out, so we don't get old replies
6154 - send/receive replies for AST messages again
6155 - if the client says that it doesn't have the lock, cancel it on the
6157 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6158 mark it as destroyed, it will all work out when decref is called
6159 - fix module use counts (22, 581)
6161 - ASTs now expect a reply (server cancels lock on error reply)
6163 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6166 - fix many simultaneous client startup (392)
6167 - fix dentry->d_it clobbering
6168 - credentials weren't being shipped for readdir/getattr operations
6169 - remove invalid assertions triggered during some concurrent MD
6171 - proper Lustre versions added (336, 389)
6172 - fix memory leak for create error case (398)
6173 - fix LOV locking bug that would get cli/srv out of sync
6174 - fix echo client over LOV (409)
6175 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6176 - fix double-O_EXCL intent crash (424)
6177 - avoid sending multiple lock CANCELs (352)
6179 - MDS can do multi-client recovery (modulo bugs in new code)
6181 - many updates, edits, cleanups
6183 2002-11-18 Phil Schwan <phil@clusterfs.com>
6186 - fix null d_it dereference (346)
6187 - fix full OST/dbench hang (333)
6188 - fix permission problem with file removal (286)
6189 - fix removal of OSCs from LOV when they fail
6190 - fix NULL deref during bulk timeout (214)
6191 - fix problems related to multiple filesystems on one MDS (241)
6192 - fixed serious subtle metadata locking bugs
6193 - free locks on clients when inodes are removed due to memory
6195 - fix inode pointer in lock data (285)
6196 - partial support for multiple MDS on a single host (241)
6197 - data locks weren't cancelled at clear_inode time (290, 311)
6198 - intent locks could lead to unbounded lock growth (205)
6199 - added a maximum lock count, an LRU list, and a flusher
6200 - fix multiple rename (365)
6201 - properly abstracted the echo client
6202 - OSC locked 1 byte too many; fixed
6203 - rewrote brw callback code:
6204 - fixed recovery bugs related to LOVs (306)
6205 - fixed too-many-pages-in-one-write crash (191)
6206 - fixed (again) crash in sync_io_timeout (214)
6207 - probably fixed callback-related race (385)
6209 - Add capability to MDS protocol
6210 - LDLM cancellations and callbacks on different portals
6212 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6215 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6216 - defer open object destruction to close time (601981, 138)
6217 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6218 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6219 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6220 - don't call obd_setattr in ll_file_release for destroyed objects
6221 * protocol change to lustre_msg: move |version| and add |flags|
6222 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6223 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6224 * added replay of create, unlink, link and rename operations during
6225 MDS failover; recovery should be much more robust now
6226 * remove failed OSCs from LOVs (only lov_create uses this so far)
6227 * the lustre-HOWTO was brought (more) up to date (582544)
6229 2002-10-23 Phil Schwan <phil@clusterfs.com>
6232 - in-use dentries weren't being reused properly (617851)
6233 - prevent multiple LDLM setup (599178)
6234 - fix LOV size calculations for truncate (617853)
6235 - fix client handling of MDS intent errors (POSIX)
6236 - fix permission bug in lovstripe.c test (624321)
6237 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6238 - truncate past end of file could corrupt data
6239 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6240 - a race in recovery could return ETIMEDOUT to apps (623947)
6241 - building outside the source directory was fixed
6242 * the lustre-HOWTO was brought (more) up to date (582544)
6243 * major progress was made on recovery functionality
6245 2002-10-10 Phil Schwan <phil@clusterfs.com>
6248 - recovery deadlock fix
6249 - rm -rf causes LBUG fix (617817)
6250 - file open by multiple tasks fix (618962)
6251 - directory permissions bugs (602707 and 620007)
6252 - journal_stop fixed with locking (611313)
6253 - O_APPEND failures resolved (618273, perhaps 614459)
6254 - lconf PATH fix (619770)
6255 - IA64 build fix (621450)
6256 - RPC buffer sizes scale with amount of memory
6258 2002-10-01 Phil Schwan <phil@clusterfs.com>
6261 - locks would be cancelled without throwing away data pages,
6262 resulting in inconsistent data (605627)
6263 - inode attributes were not always being refreshed (605627, 612449)
6264 - lconf now continues to cleanup after lctl reports an error
6265 - MDS now enforces user permissions (602707)
6266 - lprocfs cleanup fixed, but not yet enabled (614157)
6267 - fixed infinite server hang, should a client not respond to an AST
6268 - avoid going into recovery if user calls readlink() with a buffer
6269 that's too small (613941)
6270 - AST RPCs no longer require replies (614867) -- this may be changed
6271 - don't crash server if client sends an IOV that's too big (611336)
6272 - fixed lock conversion deadlock (611892)
6273 - fixed the following of symlinks (614622)
6274 * recovery: the server can remove locks from a client that dies, other
6275 clients can make progress
6276 * more extN patch fixes
6277 * compile-time configurable ptlrpc buffer allocations
6279 - collaborative read cache document
6280 - Lustre Lite Performance CDR document-in-progress
6282 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6285 - fix typo in patch-2.4.18
6287 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6290 - clear ptlrpc request each time in handle_incoming_request()
6291 - unlink of files now destroys the object on the OST
6293 2002-09-19 Peter Braam <braam@clusterfs.com>
6295 * add hard link support
6296 * change obdfile creation method
6297 * kernel patch changed
6299 2002-09-19 Peter Braam <braam@clusterfs.com>
6302 - stack overflow bug in extN fixed
6304 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6306 * documentation updates
6307 - add man pages for config tools
6308 - update tests/README to describe testing with new config tools
6309 - finish metadata API descriptions
6310 * bug fixes and cleanups
6311 - statfs workaround for 16TB limit
6312 - LOV stripe allocation improved, can stripe on subset of OSTs
6313 - LOV file size/IO offset was wrong for files > 4GB in size
6314 - object EA data was being dropped, caused files to be unreadable
6315 - memory overflow with non-LOV OST caused memory corruption
6316 - fixed regression tests to work with new config tools, obdfilter
6317 - fixed bug when directory size became larger than 1 block
6318 - fixed bug (for single client case) when PWD was deleted
6319 - invalidate local directory pages when doing intent-based ops
6320 - avoid LDLM oops when lock callback contained bad data
6322 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6324 * documentation updates
6325 * bug fixes and cleanups
6326 - configuration tools
6329 - 64-bit compile warnings
6330 - 64-bit internal statfs data
6332 * test_brw on persistent OST devices
6334 * lprocfs (disabled)
6336 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6338 * documentation updates
6339 * bug fixes and cleanups
6340 * configuration tools
6342 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6345 * many small fixes to 0_5_4
6346 * io/network handling
6347 * thinkos in MDS operations
6349 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6352 * crucial basic fixes to 0.5.3
6353 * IOR, Iozone work over Elan
6356 2002-08-07 Phil Schwan <phil@clusterfs.com>
6357 * version 0_5_3, our first alpha
6358 * we use the new Portals iovs
6359 * documentation updates
6360 * bug fixes and cleanups
6361 * small changes in the DLM wire protocol
6363 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6364 * version 0_5_1 with some initial stability,
6365 * locking on MD and file I/O.
6366 * documentation updates
6367 * several bug fixes since 0.5.0
6368 * small changes in wire protocol
6370 2002-07-18 Phil Schwan <phil@clusterfs.com>
6372 * delivered as Lustre Light Alpha
6373 * fixed a crash after handling invalid MDS requests
6374 * fixed directory pages for architectures with non-4k pages sizes
6376 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6377 * release version v0_4_4
6378 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6380 * Fixes a number of interruption problems with OST operations.
6381 * Update documentation for portals header changes
6382 * Move all wire protocol structs/defines to lustre_idl.h
6383 * Fixes symlink length bug.
6384 * Add tcpdump to repository.
6386 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6387 * release version v0_4_3
6388 * Fixes statfs for inodes on extN.
6389 * Fixes bug in runtests which would delete /etc/hosts.
6390 * Use 64-bit object IDs wherever possible (not into VFS though)
6391 Remove ost_get_info, which is unused by lustre, and out of date.
6393 2002-07-03 Peter Braam <braam@clusterfs.com>
6394 * release version v0_4_2 Fixes a lookup error (type not passed)
6395 * move forward to head of Portals
6396 * move forward to latest Lustre kernel
6398 2002-06-25 Peter Braam <braam@clusterfs.com>
6399 * release version v0_4_1. Hopefully stable on single node use.